A process analytical technology for characterising the physical properties of crystals

Final Report Summary - CRYSTAL-VIS (A process analytical technology for characterising the physical properties of crystals)

Executive Summary:
An active pharmaceutical ingredient (API) is the substance in a pharmaceutical drug product that has the pharmacological activity or effect on the diagnosis, cure, mitigation, treatment or prevention of disease, or in restoring, correcting or modifying physiological functions (World Health Organization, 2011). Crystallisation is a critical step in the API manufacturing process, for delivering “product” in its final form, and some 70% of API’s go through a crystallization step at some point in their manufacture. During crystallisation, crystals are formed in a super saturated solution and grow into the final form. The point of nucleation - where the first crystal molecules come out of solution is critical. Very often the correct - or incorrect - crystal form (polymorph) is determined at nucleation. Often an external “seed” is required to start the crystallisation process, but once started it may proceed rapidly. Once nucleation has been achieved, crystal growth usually dominates and is the process, which leads to the evolution of embryonic crystals into a crystal form of defined size and shape.
Both the size and the shape of crystals are important to the quality and safety of the product and variation during crystal growth can have a significant effect on final product behaviour such as blend homogeneity, drug absorption rates, product robustness, shelve life, etc. Controlling the final shape and size of crystals is a major challenge to manufacturers, particularly as it is often difficult to measure in-line and the crystallisation process is often nonlinear in behaviour, particularly if multi components, physico-chemical properties, and/or phases are present in the system. Moreover, API molecules are becoming increasingly complex, tending to reduce aqueous solubility, in turn reducing bioavailability and often leading to difficulties in crystal growth and control.
The ability to monitor in detail and in real time the nucleation and crystal growth properties would be a breakthrough for API manufacturers and would allow for control of the crystal phase, in terms of purity, particle size and shape, for high quality, safe products. Moreover, the crystal structure during crystal growth is a direct indicator of process end point and if more closely tracked could lead to leaner process development and enhanced commercial process monitoring and control. Moreover, the real time tracking of crystallisation is fundamental for reducing the incidences of failure in the development of new, especially expensive and novel molecular entities.
In response the CrystalVis project has developed a technology that is an imaging-based physical characterisation technology that is probe based, in-line for real time control of a manufacturing process. The R&D effort centred on prototype engineering design, software/algorithm design, user interface design, integration and validation within a crystallisation vessel in the API industry. The benefits for the API sector are increased crystallisation process knowledge/control, increased process quality, increased batch yields, reduced downtime, reduced investigations due to deviations and reduced risk of non-supply of products and greater assurance for regulatory bodies.

Project Context and Objectives:
An active pharmaceutical ingredient (API) is the substance in a pharmaceutical drug that confers pharmacological activity. The manufacture of pharmaceutical products based on small-molecule API remains critical to global healthcare.
The ultimate efficacy of a drug molecule depends on its interactions with the appropriate target in the human body at the molecular level. However, the delivery of the drug in a safe and economical way partly depends on the properties of its solid-state, at least in those cases involving a solid dosage form. API manufacturing is a complex process and may consist of many steps such as, hydrogenation, crystallisation, filtration, drying, milling etc., all of which need to be tightly controlled in order to assure product quality. Crystallinity confers various advantages during isolation, processing and storage of the drug, such as better impurity rejection, improved handling characteristics, such as sticking and flow and, in the majority of cases, better physical and chemical stability. These factors are particularly important in defining a robust processing platform and storage conditions so that a stable product can be delivered to patients. The crystallisation process is a critical step in the API manufacturing process and for delivering “product” in its final form, with some 70 % of products going through a crystallization step at some point in their manufacture.
During crystallization, crystals are formed in a super saturated solution and built into the final form. The point of nucleation; where the first crystal molecules come out of solution is critical. Very often the correct - or incorrect - crystal form (polymorph) is determined at nucleation. Often a “seed” is required to start the crystallisation process, but once started may proceed rapidly. Once nucleation has been achieved, crystal growth dominates and is the process, which leads to the evolution of embryonic crystals into a crystal form of defined size and shape. The size, shape and form of the crystals are important to the quality and safety of the product, and variation during crystal growth can have a significant effect on final product behaviour such as blend homogeneity, drug absorption rates, product robustness, etc. Understanding intermolecular interactions within solutions, in the solid-state and between phases is vital for full control.

The aim of the CrystalVis project was to develop an imaging-based physical characterisation device that would enable the API manufacturers to understand, monitor and control their crystallisation processes. As despite its importance, controlling the final shape and size of crystals remains a major challenge to manufacturers, particularly as it is often difficult to measure on line and the crystallisation process is often nonlinear in behaviour. Physical properties can vary widely between manufacturing batches of API. Sometimes this can be attributed to a difference in crystalline form, or mixtures of amorphous and crystalline forms. Usually, this issue is first noticed as a failed dissolution test or a poor filtration. In later phase processes, variances in crystal size distribution and crystal habit are the most common issues encountered, and this suggests that crystallisations at scale are still under inadequate process control. Indeed, crystallisation has been described as one of the most difficult unit operations to control. The many processes occurring during crystallisation are not understood at a level that enables the creation of materials with pre-defined functionalities. To add to the challenge, API molecules are becoming increasingly more complex, tending to reduce aqueous solubility, in turn reducing bioavailability and often leading to difficulties in crystallisation.
The ability to understand the crystallisation process would be a breakthrough for API manufacturers. Such an understanding would allow for control of the crystal phase, in terms of purity, crystal size and shape, for high quality, safe products. Moreover, the crystal structure during crystal growth is a direct indicator of process end point and if more closely tracked could lead to leaner process development and process monitoring -thereby reducing energy requirements during the crystallisation process. An understanding of crystallisation is fundamental for reducing the incidences of failure in the development of new molecular entities (NMEs).
The CrystalVis technology would be capable of providing all the physical characteristics (such as Feret Diameter, size, circularity, solidity, aspect ratio) of a crystal in-line in a crystallisation process. In order to achieve this level of information both novel optical approaches and pattern recognition methods were required to be developed in order to enable accurate monitoring of crystallisation parameters such as size and shape distributions. The R&D effort centred on engineering design, software/algorithm design, user interface design; and integration and validation in API manufacturing facilities.
To this end, over the course of the two years of this project, the aim was to develop a device that could identify and characterise the physical properties of particles / crystals within a liquid environment. This effort included the design, fabrication and testing of the technology.
The methodology combines fast on-line image capturing techniques and contrast enhancement microscopy methods for creating an information rich image. Applying novel pattern recognition methods was required to enable detecting and characterizing crystals in-line with high accuracy.
In order to realise the above aim, the scientific and technological objectives and corresponding Performance Indicators, were identified and are outlined as follows:
1. To implement an industry driven approach, the needs and specifications of companies from the API manufacturing industry were consulted, and all regulatory aspects were considered. These findings were used to guide the definition of the specifications for the CrystalVis technology.
Performance Indicator: A report on the findings of the bottom-up research was required to be delivered and the industrial specifications for the CrystalVis technology were required to be drawn up and documented in a report that would be available for project use in order to guide the evolution of the R&D work to ensure that the CrystalVis technology met with market needs.
2. To build a laboratory scale test-rig technology in order to facilitate testing under both ex-situ and in-situ conditions. The test-rig was required to be able to monitor the crystal development and the different crystal forms.
Performance Indicator: The CrystalVis lab based test-rig technology was required to be designed, built and integrated. The device needed to enable crystal morphological information to be obtained with one single setup. Based on the user requirements and performance criteria a number of approaches to image analysis were required to be tested. Stabilization tests were also required to be performed and feedback between image analysis and stabilization control procedures in agglomeration and/or crystallization phases were required to be analysed, along with tackling polymorphism issues, whereby compounds adopt more than one crystal structure during their solid state, and while chemically identical can have dramatically altered properties. The sensitivity of test-rig system for the detection and quantification of crystal characteristics in ex-situ conditions were required to be proven. From this, the scale-up parameters and in-situ monitoring requirements for developing a precompetitive CrystalVis prototype for industrial-scale crystallisation monitoring purposes were required to have been defined.
3. To evaluate the effectiveness of the lab test-rig technology to characterise the morphological features and size distribution of various API crystals in-situ.
Performance Indicators: To evaluate the effectiveness of the lab test rig technology a protocol for the analysis of crystal structures with various crystal size distributions and shapes was required to be defined. The protocol was required to be executed and a report generated determining the efficacy of using the test-rig under both ex-situ and in-situ conditions to monitor crystal properties. The limits of detection for the approach were required to be defined via correlations with reference methods. Micro imaging was required to be employed to visually detail the physical structure of the crystal. The test-rig was required to be be evaluated in-situ in a laboratory scale crystalliser.
4. To design the precompetitive image-based prototype for the identification and characterisation of the physical properties of crystals within a liquid environment in keeping with the pre-defined industry specifications, as well as the defined laboratory parameters and to assemble the system hardware.
Performance Indicator: Completion of a design document for the CrystalVis technology including a design report and P&ID drawings.
The design stage was required to focus on ensuring that the CrystalVis system was capable of offering the following features:
- Method robustness and reliability- The technology would be based upon knowledge of lab prototype which would provide a strong starting point for the envisaged scaled-up prototype.
- Adheres to industry standards, and meets with regulation and regulatory body expectations. The technology would be based upon imaging and light and therefore would be relatively nonintrusive in nature.
- Consideration of integration challenges- installing the technology inside the crystallizer. Sanitary design features would be considered along with approaches to cope with the harsh operational conditions experienced during in-situ monitoring of crystallisation processes.
- Easy to use and interpret- The envisaged technology would be, intuitive, user friendly providing size and shape information in near real time.
- Cost effective- The envisaged technology would be highly competitive through placement and pricing in the marketplace.
5. To develop the general software for the operation of the CrystalVis technology, including the synchronized control of the different subsystems, the user application and the analytical software that will process the data obtained by the technology. This analytical software will integrate the algorithms developed for the test rig, in order to characterize the physical properties of the crystals under analysis and readily display the results via an ergonomic User Interface.
Performance Indicator: The CrystalVis Control Software and the CrystalVis User Interface (UI) and database will be developed and tested to ensure it meets pre-determined requirements.
6. To integrate the system hardware, software and User Interface in order to provide a pre-competitive prototype that can be validated in industry and to carry out pre-validation tests with the system to ensure its proper functioning before shipping to industry test-sites.
Performance Indicator: The CrystalVis technology prototype, including hardware and software would be fully developed, fabricated and integrated together to form the pre-competitive prototype.
7. To test and evaluate the CrystalVis technology in commercial API manufacturing facility sites in order to assess the efficacy of the developed technology to effectively provide all the physical characteristics of a crystal in-line in a crystallisation process.
Performance Indicator: The CrystalVis technology pre-competitive prototype would be, challenged, tested and subsequently validated in industry.
8. To carry out optimisation work on the CrystalVis technology pre-competitive prototype based on learnings and industry feedback from the field trials.
Performance Indicator: A report would be generated identifying the optimisation work required to be carried out on the technology, based on the learnings and the results of the industrial trials.
9. To facilitate the uptake of the CrystalVis results by the participating SMEs, as well as a wider audience of API manufacturers, by carrying out a comprehensive series of knowledge transfer and training activities to on the one hand show the validity for the system for reliable, in-line crystal characterisation in commercial API production facilities, and on the other hand to capacitate end-users about its usability, performance and functionality, and to outline its benefits for facilitating quality control of the crystallization process, improved process control and understand, increased reliability and safety of the final product, etc.
Performance Indicator: Training materials and documentation would be generated, and training would be delivered and evaluated. The training materials and documentation would be revised based on the outcomes of the evaluations.

Project Results:
Research and technological development activities did form the core of this project, and was for the most part be carried out by the 3 RTD performers, VTT, DIT and IRIS. The RTD Performers guided and supported by the participating SMEs and OTHER (INNOPHARMA, TOPCHEM, NUTRA RESEARCH, J&M ANALYTIK and LABIANA) to ensure that the research performed complied with their expectations. To this end, the participating SMEs did provide inputs and contributions to the research activities when possible and in keeping with their ability and expertise.
The research and technological development activities were detailed in a number of work packages delivered throughout the project phase:
Industrial Specifications (WP1) –
Objectives
• To ensure a clear understanding of the technological needs of European API manufacturers in terms of crystallisation control.
• To review the existing state-of-the-art and patents, to ensure that the R&D work is up to date on its starting point (in view of the time lapse between project preparation and project implementation)
• To a review and analysis of regulatory guidelines
To use the results of the research to drive the definition of the industrial specifications for the CRYSTAL-VIS technology
This WP was led by INNOPHARMA, and they were heavily assisted by IRIS, VTT and DIT. The rest of the SMEs (J&M ANALYTIK, TOPCHEM, NUTRA) and Other Participant (LABIANA) collaborated actively in order to ensure that they communicated their industrial needs in order to drive the R&D effort
Critical to the success of the project was confirming that we had correctly identified a market need, and to ensure that the specifications for the technology aligned with the industry needs.
Within the project schedule the reporting of the market requirements and the industrial specification for the Crystal-Vis device were captured in Deliverable 1.1.
In order to accurately identify the need and specifications for the Crystal-Vis technology, a bottom-up approach was mobilised by consulting with industry stakeholders, and carrying out research to obtain a clear understanding of the technological needs and specifications of the European and US API industry. These consultations served to understand and detail the current needs and limitations of process monitoring technologies. In addition, they guided the future research and development work to be in line with the process analytical technology (PAT) initiative that aims to transform approaches in the API industry to quality assurance, leading to better process control and ultimately improved drug production and quality.

Market requirements were determined through three different approaches, firstly through visits to pharmaceutical companies within Europe, secondly through an online survey conducted across the pharmaceutical sector by the circulation of a questionnaire developed to identify the need for the technology and appropriate specifications for the technology, and finally through participation in a conference on Process Analytics and Control Technology.
Pharmaceutical company visits
Discussion took place with pharmaceutical industry representatives during site visits to Applied Process Company, Topchem, Labiana, Nutra Research, Helsinn and Pfizer.
Applied Process Company (APC), Dublin, IE: APC delivers chemical engineering solutions and technologies to enable streamlined development, optimization and supply of new and existing chemical and biological entities. These engineering solutions, delivered via APC's proprietary platforms, reduce the time, costs and risk associated with the development and supply of pharmaceuticals.
This site visit included a review of typical lab scale crystallisation processes. The manufacturing process flow was reviewed and a number of lab scale processes (100 – 2,000 ml) which were running during the visit were observed. This information helped the RTDs to understand the challenges associated with in-line measurement of these materials such as the density and turbidity of the solution and the movement of particles during the crystallisation phase of the process. APC’s experience with existing PAT instruments such as Lasentec’s Focussed Beam Reflectance Measurement (FBRM), for in line particle size measurement and Lasentec’s Particle Vision & Measurement (PVM) for in line imaging was discussed. The main problems APC encountered with these systems were identified as probe blinding or fouling and the impact of these issues on results was discussed. Other requirements of new technologies discussed were identified as accuracy and robustness of the computation algorithms, the appropriate presentation of data in real time, and the requirement that the number of measured crystals delivers results that are representative of the overall process.
Topchem, Sligo, IE: Topchem was established in 2007 and develops and manufactures APIs for supply to the generic pharmaceutical industry in the US, Canada, EU and India. The Company specialises in niche low volume high value products and generally develops products to specific customer requirements. Their products are used in a wide variety of dosage forms (injectable, oral, topical) and therapeutic areas. Products are prepared by multi step chemical synthesis. The majority of their products are purified by crystallisation. Typical medium-scale crystallisation processes running during the visit (20 L) were observed. The requirement for the Crystal-Vis technology was discussed and its benefits to the industry were confirmed. This visit helped the RTDs to understand the challenges for process analytical technology integration with small scale manufacturing equipment.
Labiana, ES: Labiana Pharmaceuticals was founded in 1958 and offers third party manufacturing and research services for the human and veterinary pharmaceutical industry. They work in strict compliance with EU GMP standards. They hold the manufacturing licence for one of their high value API’s which is manufactured by a CMO under the supervision of Labiana staff. The manufacturing process was discussed which identified the criticality of particle size to the stability of the crystals. As Labiana also produces the secondary pharmaceutical product with this API they appreciate the requirement for consistent quality API in order to achieve a consistent drug product.
Nutra research, ES: Nutra research 2011 S.L. is a scientific consulting company created by professionals with experience in the pharmaceutical industry, specializing in the development of pharmaceutical specialties, food supplements (nutraceutical and dietetics), veterinary products (premixes, complementary feed), hygiene products and cosmetics. The Nutra research team, consisting of individuals with greater than 40 years industrial experience shared their vast knowledge of the pharmaceutical industry and PAT applications in the industry. The requirements for PAT to be user friendly, intuitive, accurate, robust were discussed. Also the requirement that the instrument is capable of measuring a representative sample in real time during the process was discussed.
Helsinn Birex, Dublin, IE: Since 1997 Helsinn Birex Pharmaceuticals has operated from a new state of art finished dosage facility in Damastown, Dublin, Ireland. Helsinn supplies its partners globally with ready to sell products on an exclusive basis, which are manufactured on site and/or using a network of high quality external manufacturing partners (CMO’s) located around the world. As a manufacturer of solid dose drug products Helsinn view the ability of the API manufacturer to supply API of a consistent quality and safety of paramount importance to their ability to supply quality drug products. The particle size distribution of the API is a critical quality attribute that impacts such elements of the final drug product as blend homogeneity, drug absorption rates, product robustness, and process robustness.
Pfizer, Cork, IE: Pfizer at Ringaskiddy exports bulk pharmaceuticals, the active ingredients in Pfizer's medication for both human and animal healthcare to its’ drug product manufacturing sites worldwide. Pfizer set up its first Irish production facility in Ringaskiddy in 1969. In 1972, the first production plant, Organic Synthesis Plant 1 (OSP1), was constructed to produce bulk pharmaceutical products. During the visit it was stated that Pfizer would gain a significant competitive advantage through the introduction and use of a PAT device such as Crystal-Vis in their development activity. They are acutely aware of the potential for the use of physical monitoring instrumentation for in-line understanding and control. They particularly endorse the idea of gaining both image and size information simultaneously through one probe. Pfizer’s significant experience with PAT was discussed. Interestingly they have an FBRM probe in line for particle size monitoring. The process analytical system is integrated and the data being generated can be monitored in their control room. This gives them the option to monitor the process parameters and the particle size data simultaneously from their control room.

Online survey
The online survey was carried out through Google Forms. A questionnaire was developed by the consortium members and industry experts were invited to participate in the survey. The objective of the questionnaire was to determine current technologies in use in the industry (in-line & off-line), the industry’s needs for the Crystal-Vis technology, regulatory requirements for the technology, operating ranges & operating platforms for the technology, and the process conditions in which the technology would be used.
The responses to the requirement for the Crystal-Vis technology were very positive indicating a definite need within the industry for such a device. Particle size distribution was identified as the most important characteristic followed closely by crystal shape and then in order of decreasing importance, tracking changes over time, crystal concentration and surface/morphology details. While crystals in the size range 0-50 µm were identified as the most important, crystals >150 µm were also identified as important for 12% of respondents. All respondents utilised some crystal characterisation technologies, whether in-line or off-line. FBRM and Lasentec PVM were identified as the main in-line technologies used. However, neither system provides both images and size information which will give the Crystal-Vis technology a competitive advantage.
The survey also indicated that customers would find OPC useful. Other valuable information gained from the survey included process specific requirements such as; solvent classification used by manufacturers, Crystallisation temperature ranges, ATEX rating requirements, reaction vessel volumes.

Conference on Process Analytics and Control Technology
Innopharma labs exhibited at EUROPACT 2014. EUROPACT 2014 was the third European Conference on Process Analytics and Control Technology. The conference covered new technologies in process analytics, the implementation of these technologies in various fields and the transformation of data into knowledge. The conference was supported by an exhibition of instrumentation, applications and data evaluation tools. EUROPACT 2014 provided a meeting and a discussion forum for scientists and users of process analytics from academia and industry.
This conference provided Innopharma with the opportunity to assess current industry trends and applications as it brought together in one place academia, the pharmaceutical and drug product companies, analytical solutions providers and equipment suppliers. This forum allowed for updates and presentations of various avenues of research and development. A visit to this event was very insightful and also some of the publications were presented at this conference for the first time.

Key findings during WP 1 reveald a very definite need among API manufacturers to gain a greater understanding of their crystallisation processes. This understanding is necessary to ensure the development of reliable robust manufacturing processes, which help in the assurance of supply of quality products to the market place. Regulatory expectations have also shifted: regulatory authorities are now proposing a continuous verification of process performance using scientific data generated at all phases of the process from initial development through to commercial manufacture. Scientific data generated in-line in real time enables the manufacturer to control the manufacturing process thus ensuring a repeatable and robust manufacturing process. This is clearly outlined in the most recent FDA and EMA guidelines on process validation.

The research in the Crystal-Vis project was based on and guided by these results in order to ensure that the technology would be useful and relevant for its uptake in the API manufacturing sector. Therefore, the specifications of the Crystal-Vis technology were given in this deliverable in line with the needs and requirements of the stakeholders.

Laboratory Bench Top Building and Trials (WP 2)
Objectives
• To design and build a laboratory scale test rig image analysis sensor
• To optimize optical designs for high resolution imaging of dense suspensions of large particles.
• To develop image analysis algorithms to identify, segregate and classify crystal structure
• To design basic element to control the measurements with the sensor.
To carry out the necessary tests to assess the functionality of the developed system

This WP was led by VTT with support by DIT and IRIS. The rest of the SMEs (INNOPHARMA, J&M ANALYTIK, TOPCHEM, NUTRA,) and Other Participant (LABIANA) were involved to follow and steer the research. IRIS follow this WP to prepare for work in WP4

The following contrast enhancing optical principles were considered during the design of the test rig: Differential interference contrast (DIC) based on a Nomarski prism, PlasDIC, a simplified DIC principle and Phase contrast (PC) based on spatial modulation of the illumination and collection.
DIC microscope study
Two optical designs were created for studying the effectiveness of optical modelling tools in DIC and PlasDIC constructions.
In difference interference contrast (DIC) construction a specimen is illuminated through a polariser and a Nomarski prism and after passing through the specimen, the light beam is reconstructed using another Nomarski prism. If the specimen changes the polarisation angle of light rays very close to each other, it will pass through the second polariser and a change in contrast can be detected.
An optical design was created and tested using a virtual specimen of glass spheres immersed in water.
In PlasDIC construction, a specimen is illuminated with a light beam with a limited set of angles and after passing through the specimen, the light beam is polarised and passes through a Nomarski prism and second polariser. If the specimen changes the polarisation angle of the light rays with nearly identical angles, it will pass through the second polariser and a change in contrast can be detected.
Both designs were ray trace simulated, but with a relatively small ray count as the designs were slow to process and the lack of prior experience on simulating a DIC design left the reliability of the simulated results open. Designs appeared to be working as both of them could create visual images from specimens with variable refractive index.
The optical construction of a DIC microscope is complex and expensive as it requires a Nomarski prism as a part of the illumination optics. Due to high temperatures in some of the crystallisation processes, this would most likely require a custom design and build prism or cooling with temperature isolation.
PlasDIC can be applied using standard components as the contrast enhancing part of the optics can be located outside of the process vessel.
For preparing to design a microscope, a breadboard approximation of an early version of the optical design was build. The purpose of the microscope was to verify that the practical aspects of DIC and PlasDIC theories were understood correctly (Figure 1).
A flow cell was designed and build to enable imaging of existing crystals in a flowing solution of varying concentration and speed (Figure 2).

Figure 1 The breadboard microscope with LED flash illumination.

Figure 2 First tests in the breadboard microscope with the printed flow cell using Thiamine as the sample.
For measuring the size and the shape of a crystal, a bidirectional illumination and imaging principle was developed. In this, a crystal would pass through a channel that would be looked at from two directions resulting in two images of the crystal in the channel. Using two side profiles should give better information on the shape and the volume of a crystal than a single side profile image. The imaging could be done using only one camera, but with two cameras the optical construction would be more manageable.
As the channel was to be an active part of the optical design, Imaging individual crystals would require several optical designs to cover the required crystal size range.
On review of this design potential difficulties were identified, related to handling fragile crystals individually including the size variation related to potential polymorphic forms. Sampling required for the liquid channel was one of the major issues and in particular for the end user validation. To keep the risk of crystal damage to the minimum, the sample would need to be diluted inside the process vessel and all material related to the measurement transported out of the vessel. It was agreed that the prototype should work in a crystallization process inside the vessel without any assistance to the flow of the crystals and only minimally disturb the process.
The system was modified in order that a probe that would be immersed in the solution where crystals would be growing was designed. The probe had a long tubular part that could be immersed into a liquid and a compartment for sensitive optical and electrical parts that could not easily be adapted to the process environment or would not fit into the tubular part for mechanical or usability reasons. A sketch of the test rig is shown in Figure 3.

Figure 3 An early 42 mm diameter sketch featuring a 10mm free flow-through sampling and a relay optic to extend the tube and get sensitive DIC optics out of the process.
As there was some uncertainty to what kind of flow-through construction would remain clear the longest and be easy to clean, a hole through the probe present in the sketch was replaced with a gap design shown in Figure 4.

Figure 4 Free flow-through solution.
The LED had to be placed outside the process with DIC optics and other sensitive parts in order to protect them from the harsh environment experienced inside a reaction vessel. An optical fibre was used to transfer the light to the tip of the probe. To connect the LED to the fibre a mirror was designed. As the optical fibres suitable for transferring the light beam to the tip of the probe had minimum bending radiuses of 300 and 450 mm, these could not be bent inside the probe to connect the beam to the condenser optics. Two solutions were studied: a waveguide and mirrors. As the mirror design appeared to give more even distribution of light a dual mirror solution was selected. Mirrors for the both ends of the fibre can be seen in Figure 5.

Figure 5 On the left a single-block aluminium parabolic dual mirror construction to connect the light from the LED to the fibre and on the right a similar structure to connect the light from the fibre to the condenser optics.
At the time when the contrast enhancement methods to be implemented were to be selected there was no measured data available to back any of the methods. As time was limited and the Zeiss DIC/PlasDIC equipped microscope (Figure 6) finally arrived, the microscope was developed from DIC prisms as the design could be adapted to use the Zeiss DIC part and the specification and order of bespoke Nomarski prisms could be skipped saving both time and money.
A dual camera construction passed through from the initial liquid channel design, because simultaneous imaging with two different optical configurations would allow comparative data to be recorded from live samples. The direct optical path would be more manageable and therefore preferred option for contrast enhanced imaging. Reflected path would be used for bright field techniques, that are better known and less effected by manufacturing tolerances. Details of PlasDIC optics can be seen in Figure 7 and dual camera construction in Figure 8.
After completion of the optical design the mechanical design was initiated.

Figure 6 The Zeiss Axio microscope with LED transmission illumination (left) and one of the included Nomarski prisms (right).

Figure 7 Sketch of the (Plas)DIC accessory cubes.

Figure 8 Dual camera imaging optomechanics. The optical elements and rays have been displaced to the right for visibility.
The test rig was assembled with PlasDIC optics using the 3D printed beamsplitter mechanics. The bright field optics could not be installed as the mechanical part was not ready, but bright field measurements could be made by taking out the DIC prism and adjusting the rotatable polarizer.
On completion of the Test rig assembly it was tested for image quality. Both bright field and PlasDIC configuration produced acceptable images. For detecting possible leaks, the probe was submerged firstly in isopropyl alcohol and then in deionized water. No leaks were detected.

Software
A camera SDK example was modified into a camera user interface for dual camera operation and automatic configuration. For crystal detection a separate executable was developed. This executable receives the images captured by the camera user interface, detects crystals, calculates crystal dimensions and optionally stores calculated results and images. The camera interface program controls cameras and captures images. This software is based on an example from the SDK supplied by the camera manufacturer.
When the program is executed, it looks for cameras present as devices in the operating system and creates a connection to up to two cameras. A preconfigured set of parameters is uploaded to the connected cameras. After the cameras are configured, the program creates a shared memory allocation for each camera, that is used for transferring the images to the crystal detection program. The modified program retains the original functionality of camera parameter adjustments, zooming into the image and saving of images and video.
The program must be running for the crystal detection program to work. A screen capture of the program is shown in Figure 9.

Figure 9 Camera interface. Figure 10 Crystal detection program.
The crystal detection program was developed to host the crystal detection algorithm. In addition, it contains some basic functionality for controlling the acquisition process and visualization. The images are shown in the user interface with cursor for reading pixel values. The user can zoom into the image and move the zoomed image. The image processing can be paused if there is a period during which no data needs to be collected. If the distribution of light is not satisfactory, the user has an option to measure reference, which is used to correct uneven lightning. The user can choose if images or algorithm results are to be stored and the location for these files. A screen capture of the program is shown in Figure 10.
Images are acquired from the shared memory created by the camera interface program. The program can read images from up to two shared memory areas. After an image has been acquired, an internal thread scheduling pool is queried for processing capacity. If the pool determines there is capacity available, a thread is launched that will detect the crystals from that image. If there is not enough capacity available, the query is met with a negative reply and the image will not be processed. The processing load depends on the number of detected crystals and a high count will slow down the processing. If the user activates the reference option, one image is captured and stored to be used as a reference image for normalizing the illumination across the image. The reference image processing is activated if a reference image is found in the storage folder. The reference image processing is executed on colour images. Thresholding is a process where each pixel of an image is compared to a value and assigned a new value of null or one depending on the result. Typically, the value of one is assigned to the pixel belonging to the target and the value of null to the background. Before thresholding the pixel format is converted from three colour values to a single intensity value. The crystals are detected by a multi-thread optimized blob detection library. For each detected crystal the following features are calculated: location, area, and minor and major axes, lengths of an ellipsoid that would cover the crystal. The result contains both features per crystal and size distribution for all the crystals in an image.

WP 3 Laboratory testing of the system for measuring the physical characteristics of crystals
Objectives
• To assess the efficacy of the lab test-rig to monitor crystal shape and size under ex situ conditions.
• To test the efficacy of the system to operate in-situ within a lab scale crystalliser for seeded crystals of know properties and concentrations
• To test the system for monitoring crystal growth
• To compare the system to off-line reference image analysis techniques and in-line FBRM measurements
To define the limits of detection for the lab test-rig system
This WP was lead by DIT while the end-user SMEs (TOPCHEM, NUTRA) and Other Participant (LABIANA) shared their industrial knowledge with DIT. J&M ANALYTIK, and INNOPHARMA followed the work in order to absorb knowledge. VTT assisted DIT and IRIS followed the research in order to carry knowledge over to WP4.
The following chemicals were crystallized in the lab using standard protocols: benzoic acid, paracetamol, glutamic acid, lysozyme, taurine, glycine, mannitol, potassium sulfate. An image of Lysozyme is presented in Figure 11

Figure 11: Examples of lysozyme crystals grown in our lab.

Images of crystals were also acquired using DIC objectives on the BH2 Olympus microscope in house. They were compared with the normal objective (example below, Figure 12).

Figure 12: Comparison of DIC objective (40X) with the normal objective using lysozyme crystals. Left: Normal image. Right: DIC image.
The images generated were used for algorithm development

Image processing: brief introduction and overview
Many programs are available to use such as Rgui, ImageJ/Fiji, Icy, etc. The required image pre-treatment tools can be used: image conversion (8 bit, RGB, binary, etc), thresholding, noise reduction (despeckle, outliers, erode, dilate, Gaussian blur and smoothing, etc). Furthermore, basic segmentation and watershed can be applied with different methods and algorithms. Finally scaling and image evaluation can be performed for particle analysis of shape and size. All steps are then recorded as a macro and applied on batch mode for images of similar mode.
Successful image processing usually starts with the engineering: good details and contrast. Sampling is also important, for example overlapping particles, very concentrated samples, etc are extremely hard to process. Light distribution is essential. Particles are observed due to that light, and to the microscopy technique that is used, which provides more or less sharp contrast and edges.
For reliable image processing, particles have to be in the same focal plane, and as less blurred as possible, to preserve their characteristics and shape/size information. Light distribution within the particle is also important, to avoid information loss and program mistakes. This happens when intensity differences within crystals are important due to light diffraction and transparency/shadows, etc. The program or any algorithm will hardly distinguish objects sticking together and reliably separate them: edge detection and watershed of irregular features will be very challenging.
II- Image processing, part 1
The images processed in this section were obtained from VTT, Finland, using a bench top system developed to simultaneously record images in the PlasDIC/DIC mode as well as with the cross polarizer mode (at 45°). The PlasDIC images were ~5.5-5.8 MB (Figure 13) while the Phase contrast images were ~2.5-2.6 MB. From a first look, some images had two dark spots (located specifically at the upper right and the bottom left corners) (Figure 13). These spots might be due to the early stage engineering, bench top tests, or partially to a stack/crowd of crystals in a specific location.

Figure 13: Example of image execution when opened with ImageJ/Fiji.

General image preprocessing
The general work was performed with the open source ImageJ Fiji, which hosts a vast number of algorithms and scripts, and can be used and edited in both graphical and command modes.
a- Enhanced local contrast mode

Figure 14: Application of enhanced local contrast. a: original image. b: treated image.

The CLAHE algorithm function (Enhanced local contrast) in fast mode was used, with the following parameters: Blocksize: 150; Histogram pins: 100; Maximum slope: 5; Mask: No.

b- Background subtraction
The background was subtracted from the image above, using different criteria (Figure 15). In this case, the two big spots in the corners are eliminated whenever such abnormalities are present.

Figure 15: Example of the background noise removed using the background subtraction algorithm.
The upper right and lower left spots from the previous figures were isolated in the background, as well as other spots which might be due to the reflection or shadow by some crystals. When this function is coupled with the enhanced contrast method for this crystal image example, both the unwanted spots are removed while the contrast is enhanced (Figure 15). When the background is first removed, then the contrast enhanced, the resulting image which is better than if the opposite was performed.
The best criteria for background subtraction in ImageJ Fiji for our case are strictly when the “light background” and “Separate colors” options are set ON, while the “Create background”, “Sliding paraboloid” and “Disable smoothing” are all set OFF. The rolling ball radius should be 80 pixels.

c- Gaussian blur and Smoothing filters
At any stage, the Gaussian blur filter as well as the smoothing filters can be applied. These filters help reduce the noise and some details. They also help to enhance the scales of main objects. The settings for the Gaussian blur filter should be 1 for the Sigma (Radius).

Figure 16: Applying the Gaussian blur and the smoothing filters. The images are shown for “before” (left) and “after” (right) the application of filters.
d- Contrast change
This function is a classic contrast change, and is different from the function in “a” (enhanced local contrast). It can be applied to get a better signal to background ratio. The settings should be 1 % of saturated pixels.

Figure 17: Applying a general contrast function.
e- Image treatment evaluation
At this stage, the image looks much better than the starting point (Figure 17). Colors and brightness/contrast levels can be evaluated using LUT functions (Look Up Table), which can also show specific color channels as intensities or energies. Images have to be converted to 8-32 greyscale bit before starting. More than 20 LUTs exist, but in this section only those of significance were used. Union Jack is also an important LUT but wasn’t shown since it is similar to the Phase LUT.
1
2
3
4
5
Figure 18: Evaluation of the intensities and color channels using LUT functions. Left: Image before processing. Right: Image after processing using the algorithms “a” to “d”. 1: Images in the 32 bit greyscale mode. 2: Fire. 3: Spectrum. 4: Phase. 5: Thermal.

Thresholding
a- Basic thresholding
Once the previous steps have been done, thresholding can now be performed with higher confidence. To proceed to thresholding, either the “Threshold” function can be used, or the “Brightness/Contrast” algorithm can be first applied, to adjust the minimum-maximum color (0 to 255) and the brightness and contrast levels (Figure 19). It is always better to start with a greyscale image before adjusting the parameters.
The brightness/contrast is a four parameter setting, which show the intensities/amplitude for each channel (0 to 255) as a curve.

Figure 19: Applying the Brightness/Contrast parameters. The parameters should be adjusted until a good threshold is reached, while conserving the particle size and excess noise. Up: The original image before and after thresholding. Bottom: The processed image using the previous methods, before and after thresholding.

As for the thresholding settings, there are already methods that can be used to process the image (Figure 20). The full description and references for each method has been listed here: http://fiji.sc/Auto_Threshold#Try_all

Figure 20: Thresholding methods for data segmentation.

Thresholding from 8 bit image, at 88 % saturation:

Figure 21: Thresholding comparison of 8 bit and RGB images. Left: 8 bit image used for the start, with thresholding at 88 % signal saturation. Right: RGB image used for the start, with thresholding at 100 to 255 brightness intensity, 0 to 70 saturation intensity and full Hue levels.
b- Post-thresholding treatment
Many filters and polishing effects can be applied, once the image was subjected to thresholding and binary transformation.

Figure 22: Removing outliers from the binary image. Left: Original binary image. Right: Outliers removed, at 10 pixel settings

Smoothing filter can then be applied at many rounds. In this case the image was subjected to five rounds. A Fill holes (Figure 23) function is then helpful for most cases when the particles have missing information inside them due to different stages of processing or from the light or images capturing stage.
This function is important before applying wastershed to separate connecting objects, because some holes will be recognized as a spot for object separation. Watershed algorithm is a very tricky function because the program will not know exactly where to separate particles. Therefore settings have to be adjusted carefully, but also advanced algorithms used such as the Watershed algorithm from EPFL, CH, or BioVoxxel algorithm for controlling erosion cycles and know how sensitive the program has to be for the separation of connected objects.

Figure 23: Fill holes function. Left: Original image. Right: Image after applying the function.

Figure 24: Watershed application. Left: Original image. Right: Image after applying the function. In this case an advanced plugin from BioVoxxel was used, to adjust the Erosion cycle number (set as 3).

Macro editing
All steps can then be summarized once know, in the exact order and settings, into a macro. For example for the steps show above, the following macro was written. It has to be checked always and tested many times.
run("Subtract Background...", "rolling=80 light separate");
run("Enhance Local Contrast (CLAHE)", "blocksize=149 histogram=100 maximum=5 mask=*None* fast_(less_accurate)");
run("Smooth");
run("Smooth");
run("Gaussian Blur...", "sigma=1");
run("Enhance Contrast...", "saturated=1");
run("8-bit");
setAutoThreshold("Default dark");
//run("Threshold...");
setThreshold(91, 255);
run("Make Binary", "thresholded remaining black");
run("Remove Outliers...", "radius=10 threshold=50 which=Bright");
run("Smooth");
run("Smooth");
run("Smooth");
run("Smooth");
run("Make Binary");
run("Remove Outliers...", "radius=5 threshold=50 which=Bright");
run("Fill Holes");
saveAs("Tiff", "C:\\Users\\Administrator\\Dropbox\\1 CrystalVis\\Misc\\Material\\Images Lauri 270614\\output.tif");

In this study, the macro was tested on other images of PlasDIC. The results are shown below for before (left) and after processing (right). All images can be used for automatic particle analysis: geometrical statistics. The results looked great because both small and big particles were detected. The sensitivity of particle detection can be adjusted even more depending on the experiment and general API crystal size.

Figure 25: Macro testing using different images.

Scale and statistics
Adding the scale bar with full customization is very simple. Calibration can be made using samples or references of known size, so that the corresponding distance in terms of pixels is also determined (Figure 26).

Figure 26: Setting the scale bar and analyzing particles. The scale bar (left) allows the program to calculate parameters (right) based on a known distance and unit of measurement.

III- Image processing, part 2
For more extreme cases, such as more sensitive focal levels, overlapping particles, transparent big crystals, connected crystals, concentrated samples, some images of lysozyme were taken and tested. Three examples have been shown below.

Example 1: Big crystals with strong internal diffraction
When the image is processed normally, as in part 1 above, the resulting image show what is detected: the parts that are different from the background. These could be some of the edges of a single crystal, connected parts, some reflecting spots inside the crystal, etc (Figure 27).

Figure 27: The result of a normal processing of a crystal image containing a lot of internal diffraction, image after standard processing, an orange heat filter showing the areas that are detected by standard algorithms and a Binary image of the result.
The image processed this way shows empty geometries. When these objects are filled for example, particles, especially those in the bottom right of the image, become all one big cloud which is totally useless for particle characterization. Therefore, (1) more dilute samples, (2) manual measurement, (3) an additional filter to take into consideration particular geometries that have to be taught to the program, or, (4) an advanced edge detection before binary conversion has to be used.

Example 2: Edge detection using Canny algorithm
In this example, the Canny edge detection tool was applied (Figure 28). The following settings have been used during the processing of the image: Gaussian kernel radius: 5; Low threshold: 0.1; High threshold: 5; No contrast normalization. These were the best settings. Some objects and shapes were successfully identified after applying many additional filters such as smoothing, dilatation, and then Fill holes. Statistics may be obtained to show the acceptable range of size to be taken into consideration.

Figure 28: Canny edge detection before binary conversion. An 8 bit image was used. 1st Gaussian kernel radius: 5; Low threshold: 0.1; High threshold: 5. 2nd Gaussian kernel radius: 10; Low threshold: 0.1; High threshold: 5. The 3rd image was more useful and resulted in the identification of some crystals as shown below it.

Using another image, the algorithm was successful to pick up most crystals. This Canny edge detection algorithm was the most successful for customization and application among many others that were tested.

Figure 29 a, b, c, d: Canny edge detection before binary conversion.
The image in “a” was processed and converted to an 8 bit image. b: Gaussian kernel radius: 5; Low threshold: 0.1; High threshold: 5. c: Gaussian kernel radius: 10; Low threshold: 0.1; High threshold: 5. d: Conversion of the image in “b” after smoothing, dilatation and filling holes.
An additional filter was then applied to remove sides (unwated particles) of particular thickness, before applying the watershed algorithms.

Example 3: Processing of overlapped transparent particles
Additional images of lysozyme were tested below, starting with both normal and DIC images. Images were processed as usual (part 1), the converted into binary image (Figure 30).

Figure 30: Image processing comparison of normal and DIC images.

In this example, no major difference or advantage was observed for the processing of images acquired with one technique over another. The program can be taught to complete edges, and separate objects at a particular angle and distance.

General comments
These examples and studies show that the main difficulty to reach a total and successful image processing starts from the method that is used. Advanced optics may help solve many problems in visualizing particles with great contrast and aspect ratio. The most important parameter is to measure the length and count of particles. The maximum length can be obtained even if only one side of a crystal or object is observed. Counting is also possible, however there is no way around overlapping particles, because all methods and algorithms will recognize them as one particle, or depends on the shadows, as separate ones, but probably not very close to how our mind perceives the image.
Furthermore, engineering can also find solutions related to concentrated solutions, and help visualizing dilute particles, or images where there is little overlap and transparency. However, there are too many variations of samples and particle size and shape, indicating no universal solution.
As for image processing itself, crystals of high transparency and overlap, which have a very similar color in solution as that of the background, are hard to process, and require additional steps and particular algorithms “case/image -dependent”. For example, thresholding cannot be successful for most particles which have different shadow and brightness intensities in different parts within the same particle/crystal. The answer is simple: we see it in a certain way, but is the information really there for the program to detect and use? Particles in different focal planes and thus blurred, cannot be treated in the same way as those in focus. Wastershed/segmentation function for those particles, and those overlapping like above, will be very limited.
It is hoped that the cross polarized image in the current prototype of CrystalVis under development, coupled (overlayed) with the PlasDIC, as well as used in parallel with the latter, will be a big advantage for aspect ratio.

WP4 Design and building of the CrystalVis prototype
Objectives
• To design, build and calibrate the CRYSTAL-VIS system, in keeping with the results and parameters from WP2 and WP3 and taking into account the requirements defined by end-users of the system in WP1. It is envisaged that as many commercially available components as possible will be used in order to assemble the precompetitive prototype.
• To design the control interfaces and basic process controls for the CRYSTAL-VIS subsystems.
To carry out the necessary tests to assess the functionality of the developed subsystems

IRIS lead this WP. They were assisted by VTT to ensure a smooth scale up from the lab. DIT also assisted to bring their knowledge especially to task 4.4. The participating SMEs (INNOPHARMA, J&M ANALYTIK, TOPCHEM, NUTRA,) and Other Participant (LABIANA) were involved to steer the design work and to ensure it met with their specifications.

The general concept of the Crystalvis system, as shown on Figure 31, is a probe that can be introduced into the reaction vessel, attached to an external enclosure housing all the electronics and sensitive elements, which is itself connected to a laptop with process control, data analysis, data storage and user interface all packaged into the Crystalvis software, described in deliverable 5.1.

Figure 31. Main elements and dimensions of the Crystalvis prototype.
(1) external housing, (2) probe, (3) sampling gap, (4) prism position adjustment
The main features of the Crystalvis system can be seen on the block diagram Figure 32. It consists of three main function blocks: optics, electronics and software, enclosed within a robust enclosure consisting of a probe part and an external housing. A regular user doesn’t need access inside this enclosure as all relevant controls are external, mainly accessible from the user interface. Nevertheless there is easy access to the content of the external housing containing the control board and cameras. In order to access the optics inside the probe part, the prototype needs to be dismantled. This is partly because of the need for robustness of the probe part that is placed inside the reaction vessel but also because the settings of the internal components should be factory set and the user not allowed to change them.

Figure 32. Block diagram of the Crystalvis system.
Optics
Following laboratory trials, the imaging system developed for the test-rig performed extremely well therefore most of the optical elements were re-used for the pre-competitive prototype developed during WP2. The analysis of the images acquired showed that both brightfield and plasDIC microscopy gave valuable information. Depending on the application one technique might be more suitable than the other, allowing the system to be simplified and resulting in a smaller and cheaper device. However in order to demonstrate the full capabilities of the Crystalvis system, both microscopy techniques are integrated in the final prototype. Images of the same crystals are acquired at the same time so they can be displayed and analysed in real time, thus extracting information simultaneously from both brightfield and plasDIC microscopy.
General concepts
In addition to regular brightfield microscopy, three different contrast enhancing techniques were investigated as part of the project: phase contrast (PC), differential interference contrast (DIC), and a modification of the DIC technique using the same principles (plasDIC). In brightfield microscopy contrast arises from the difference in the amplitude of electromagnetic field only, while these contrast enhancing techniques are also sensitive to the change of phase. The issue here is that changes in phase are not only due to the specimens of interest (API crystals) but can also be induced by changes in refractive index of the solution they are in.

Figure 33. Elements and principle of plasDIC technique.
For this reason plasDIC (Figure 33) was found to be the most promising technique because it is implemented without polarisation dependent elements affecting the light travelling through the specimen. Instead a small shear is induced by a coherence diaphragm, separating the input light into two wavefronts. Travelling through the specimen can cause a phase shift between the two wavefronts, which are then recombined using a Nomarski prism. This phase shift is then converted to a change in intensity by the combination of a polariser and an analyser on each side of the prism, thus enhancing the contrast in the image. While the amount of light reaching the camera is greatly reduced, some crystals barely visible in brightfield can be seen in plasDIC much more clearly. In addition what can appear as blurriness in brightfield, appears as a relief effect, which may be easier to interpret during image analysis.

Optical design
Dual microscopy system
The schematics of the dual microscopy system are shown on Figure 34. All electronics are confined to the external housing part of the prototype, protecting them from high temperatures. The light emitted by the LED source is couple to a fibre using mirrors, the fibre then carries the light to the tip of the probe where it is coupled out of the fibre and into the condensing optics via mirrors. Light travels through the condensing optics and coherence diaphragm before illuminating the specimen flowing between two windows. Magnification is provided by a x10 microscope objective and light is carried through the remaining length of the probe via two relay lenses. Once reaching the polarising beamsplitter there, light is separated into two arms. The reflected one is used for brightfield microscopy and goes through an adjustable aperture and imaging optics before reaching the camera, while the transmitted arm is used for plasDIC microscopy. For plasDIC the light polarised by the beamsplitter goes through a Nomarski prism and a polariser (analyser) before travelling through the adjustable aperture and imaging optics and reaching the second camera.

Figure 34. Schematics of the Crystalvis optical system.
Illumination
The light source used is a high power white LED (Cree XLamp XM-L2 in U2 on star, see 35). It is placed outside the reaction vessel in the external housing to protect it from high temperatures, thus optimising its operating conditions and avoiding thermal damage. LEDs are reliable sources with a small footprint, allowing us to easily integrate it into the system.

Figure 35. Light source specifications.
In order to bring the light to the condenser optics placed at the bottom of the probe, a large core multimode optical fibre is used (Thorlabs FT1500UMT). Its large core, 1500mm diameter, and large numerical aperture, 0.39NA make it suitable to couple and transport a wide spectrum of wavelength. Light is coupled from the LED to the fibre and to the fibre to the condenser optic using dual parabolic reflectors. The optical design is shown on Figure 36, while the corresponding final parts used for the prototype on Figure 37.

Figure 36. Parabolic reflectors design (TracePro)

Figure 37. Manufactured parts from the prototype. Parabolic reflector pairs used for coupling LED output to fibre input (left) and fibre output to condenser optics (right).

Detection
Two identical cameras are used for image detection. More details on acquisition and data analysis can be found in D5.1 software development. The cameras are Point Grey Grasshopper 3 / GS3-U3-91S6C-C and their specifications can be seen on Figure 38. They feature a high sensitivity, high frame rates as well as high resolution 9.1M pixels colour CCD image sensor and were extensively tested during WP3.

Figure 38. Specifications of the Crystalvis cameras.
Opto-mechanical assembly
Full details of the opto-mechanical assembly can be found in the 3D SolidWorks. In order to ensure optical alignment, the assembly consists of a combination of standard tube systems and cage systems from Thorlabs, which are integrated within the prototype using custom made parts. Details of each part of the prototype are also shown, illumination on Figure 39, microscope optics on Figure 40 and plasDIC path in outer housing on Figure 41.

Figure 39. Detailed illumination optics assembly.

Figure 40. Detailed microscope optics assembly.

Figure 41. Detailed plasDIC path assembly in outer housing.
Electronics
The control board is based on a common device manufactured by VTT and offering a number of configuration variations. The electronics block diagram is shown in Figure 42. The control board features one LED pulsing unit (LPU) and one LED driver unit (LDU). The LPU and LDU use +24V power supply and the two Point Grey Grasshopper 3 cameras are powered through USB 3.0 cables. The +24V power supply is a plugin power adaptor EMSA240125 from CUI Inc. The LPU is a Atmel AtMega128 based CPU card, which synchronizes the LED pulsing and camera trigger signal to get LED flash images. The driving pulses delivered to the LED have a frequency of 2Hz and a length of 10μs.

Figure 42. Crystalvis system control board.
Mechanics
As mentioned above most of the optical elements were re-used in the final prototype and therefore most of the optomechanical elements they were mounted on were also re-used. Changes in the probe part, more specifically optimisation of the sampling gap, were made in accordance with the results obtained during task 3.3 (in-situ crystalliser testing). An external housing was also designed and built to protect all the elements outside the probe part of the prototype, while providing secure electrical connections. Finally an external system allowing control of the Nomarski prism position was added to the prototype.
Sampling gap
A new measurement gap part (Figure 43), was designed and built following investigation during previous work packages. It was observed that a reduction in the length of the sampling gap, i.e. the distance between its two windows, allowed for acquisition of better quality images, as can be seen on Figure 44. It was therefore decided to reduce the gap from 6mm to 4mm. However the total length of the part was kept unchanged in order to maintain the integrity of the optical system inside the probe. Because larger or heavier crystals tend to pass through the gap close to the bottom sampling window, its position was raised by 2mm bringing it closer to the focal plane of the system. On the other hand the position of the upper window remained unchanged to avoid clogging, which could become a problem with a shorter gap length and increased crystal concentration. As the probe will be immersed in a solution where crystals will be growing, all its parts were built in stainless steel and with no sharpened edges for an easy cleaning process.

Figure 43. New sampling gap part.

Figure 44. Image quality improvement shown for in-situ air bubbles when raising bottom window by 2mm
External housing
The external housing is made from a standard stainless steel box (Figure 45) and mechanical drawings in Annex I) and it houses two cameras, an LED, the control board, a beamsplitter and the following two sets of optical elements corresponding to the separate brightfield and pasDIC optical paths.

Figure 45. Standard (left) and adapted (right) cabinet

This enclosure protects all the elements inside, ensuring it is dust proof and splash proof by using custom gaskets. At the same time as it makes the system user friendly by means of an easy to open and close door which gives access to the optics and electronics inside. In order to provide secure connections for all leads coming out of the external housing without compromising the protection of the internal elements, connections are protected by using special IP65 cable glands and connectors (see Figure 46).

Figure 46. Electrical connections

1.1 Adjustable Position of plasDIC prism
The function of the Nomarski prism is to recombine the two optical wavefronts after they have travelled through the specimen. Its optimum position for contrast enhancement is determined both by the shear introduced with the coherence diaphragm and by the optical properties of the specimen. Since different coherence diaphragms can potentially be used for the same system and since the optical properties of the crystals vary for each type of crystallization reaction, the position of the prism needs to be adjustable. The standard Zeiss microscopy slider it is mounted on allows translation of the prism through a thumb screw, however its position is nestled inside the optomechanical assembly of the plasDIC optical path inside the external housing. Therefore, a regulation system accessible by the user from outside the housing was designed for the Nomarski prism slider (see Figure 47).
This regulation systems consists in a bar with a groove for the screwdriver on one end and a holding screw on the other end which attaches the bar to the Nomarski prism regulation screw. This bar is fixed to a custom stainless steel piece (by using an o-ring which also makes the system splash and dustproof) which adapts the 45 degree position of the Nomarski prism to the position of the enclosure. Moreover, a position indicator nut was added so an approximate adjustment can quickly be made by positioning the screw respect to the grooves of the nut.

Figure 47. Adjustment system for Nomarski prism position.
Custom reaction vessel
Following discussions with one of our industry partners, Labiana, it was decided to design a small reaction vessel that could accommodate both their research and development activities for API crystallisation and the Crystalvis system. Design and feasibility were further discussed with the glass manufacturer Linealab in Spain and the customised reaction vessel, shown on Figure 48 manufactured and sent to Labiana.

Figure 48. Custom reaction vessel (front view).

Software Development, system integration & optimisation (WP 5)
Objectives
• To develop the general software for the operation of the CRYSTAL-VIS system, including the user application and the analytical software that will process the data obtained by the system. This analytical software will use the algorithms and spectral databases developed during WP2 and WP3.
• To design and develop the user application according to the requirements established during WP1.
To integrate the system hardware, software and User Interface
IRIS lead this WP. DIT assisted to ensure the effective use of the knowledge generated in WP3. VTT also assisted, especially with the integration of the software. All the other SMEPs (INNOPHARMA, J&M ANALYTIK TOPCHEM, NUTRA,) and Other Participant (LABIANA) were involved, especially to steer the work to ensure usability and ergonomics met with their requirements.

Operating System
Operating system used for CrystalVis project is 64bit Microsoft Windows 7, as the most widely used and stable OS from Microsoft. More than 55% of all desktop PCs in the world have it installed and it is more popular and stable than its successor Windows 8 or 10.

Framework
Microsoft .Net Framework 4.5 was used. It includes a large class library known as Framework Class Library (FCL) and provides language interoperability, means that each language can use code written in other languages. Programs written for .NET Framework execute in a software environment, known as Common Language Run-time (CLR), an application virtual machine that provides services such as security, memory management, and exception handling.

Integrated Development Environment (IDE)
Microsoft Visual Studio 2013 was used for programming CrystalVis software. It is an Integrated Development Environment (IDE) that provides many features for authoring, modifying, compiling, deploying and debugging software. Source is also compatible with Microsoft Visual Studio 2015.

Programming Language

CrystalVis software is desktop application made in C# programming language with Windows Presentation Foundation (WPF) and XAML combination. This made a rich presentation system for building Windows desktop applications with visually stunning user experiences that incorporate UI, media, and complex business models because it has a comprehensive set of features like controls, data binding, animation, styles, templates and more.

Source Control

Git was used as source code management system for version control and code sharing during development process.
Data Storage and Analysis

Database

To store data about image capture, analysis and experiments in CrystalVis, a combination of MariaDB database engine and custom file-system storage system is used. It is the symbiosis of two underlying software components that works together in order to create, read, update and delete data needed during the program usage.
MariaDB is a community-developed fork of the MySQL relational database management system intended to remain free under the GNU GP License. Database server is located in local machine, it has a schema named “crystalvis” and the table named “experiments” keeps all the data about what has been done. That table is visible and editable in Data-log screen and it has following fields:
• ID: is primary index of the table, it is not visible in user interface
• DATE: time-stamp showing when the experiment was executed, it is read-only data
• USER: name of the user who has created the experiment, it is maintained by the system
• SUPPLIER, BATCH, API, COMMENT: information about the material analyzed, can be edited
• PATH: keeping a link to a local folder where all work files are stored (read only)
• B/P: short information about the technique used (1-Brightfield, 2-PlasDIC, 3-Both)

Files and folders

Files and folders used and generated by CrystalVis software are located at “crystalvisData” directory in the system root disk (C:\crystalvisData).

When user starts an experiment, a new folder is created in “C:\crystalvisData\Experiments\” folder with a name that contains user, batch, date and time when the experiment was made. It is automatically given by the system, stored into database and shown in data-log table.

Likewise, images captured from cameras are stored in “source” folder, under the experiments root, following the simple naming convention, “cap-cameraName-date-time.bmp”. When analysis is done, for each image in “source” folder, the corresponding Comma Separated Values file (CSV) with the same name is created in analysis directory, as well as the image of the histogram/trend-line graph.
This naming convention keeps all the data in efficient, human-friendly format, well sorted and organized. CSV files could be easily opened by programs for tabular calculations like Excel or imported to some other database for further analysis.

Analysis software

Image analysis is done using ImageJ macros that are created during the software execution. For each captured image the custom macro file is made using templates stored in “C:\crystalvisData\Macros\fiji\Fiji.scripts\” folder. Files named “macroBfX.txt” and “macroPdX.txt” are templates, they have several variables that are replaced by real values before the file is saved as a macro and executed.
Custom image analysis can be done easily by editing those macro template files with adding new image analysis steps, changing existing values etc. According to that templates are text files, any text editor program can be used.

User interface

Main window

The CrystalVis user interface consists of two windows, the Main window and Data-log window. Once the program has been started the main window will appear in the middle of the screen, as shown in Figure 56: Main window.
The CrystalVis main window is divided in five parts according to its functionality: header, button bar, two cameras preview and analysis sections and parameter section, as shown in Figure 57: User interface sections

These five sections of the Main window are a part of the program that controls the cameras and allows user to configure different parameters for both devices, Brightfield and PlasDIC. These configurable parameters will be explained more deeply later in this document. In the Main window user is able to visualize the particle image, their number, size and shape in histograms, to zoom-in and out captured images, to see the process of crystals growing trend-lines in real time etc.

Header

The CrystalVis name and logo are shown in the header, also it contains “off” button that will close the application. Application can be closed at any moment and all the information acquired during the session which has not been saved will be lost. It is recommended to stop the analysis before closing the application.

Button bar

The common bar has buttons for turning on and off camera capturing processes for both systems, Start/Stop button for manual and automatic analysis, switch for simulation mode and button for Data-log page.
Camera preview and analysis section

There are two separate displays that are showing the current output from cameras. Images are captured in real time by each camera. There is the possibility of enlarging the preview image by clicking on the arrows button. This action will open a new window where the image can be previewed at its original size. In the lower-right corner is the set of commands for image preview, scroll, zoom, select etc. in order to let user to have closer look at what is going on with captured sample.

The triangle/pause symbol on the left part of the display has the same functionality as the button above the image, it starts/stops image capture on corresponding camera. During the analysis process it is not possible to enable, disable or reconfigure input devices by pressing those buttons.
Using the mouse, the user can select and zoom the particular part of the image to examine closely the area of interest and particle captured. During that process, the command buttons will allow to the user to go back in previous zoom level, to move across the image, or to easily view the whole picture. The window will be closed by clicking on right mouse button outside of the image area.

Parameters section

This part of the Main window is used to set up the parameters for images capture and analysis, as shown in Figure 60: Acquisition and analysis parameters

This part of the Main window is used to set up parameters before the capture/analyze is started. One of the most important parameters is Acquisition mode drop-down menu, that is changing from Manual to Automatic acquisition mode and vice versa. This option is changing the overall program behavior.

Analysis

Overview

Bottom part of this section is reserved for analysis results of both captured images or image sequence. Each graph is showing two graphs: histogram and trend-line.
Histogram is a bar-chart made of 5 ranges of feret diameter size [μm]. Y-axis is displaying number of particles detected in a sequence of images defined by “Images per histogram” settings parameter. Histogram is saved, then reset after that number of grabbed images.

On the other hand, trend-line is defined by date-time on X-axis and value of a selected parameter on Y-axis. It will not be reset during all the process of analysis. Feret is selected by default, user is able to change the selection to Size, Circularity, etc. in the small drop-down under the graph.

Manual mode

Manual mode was made for one image analysis, with one or two cameras. When the “Save and Analyze” button is pressed, program will create new experiment, make a capture, analyze the image, save the analysis table and finally display the histogram. In this case, trend-line is shown as a single dot.

Auto mode

On the other hand, Automatic mode serves for continuous analysis of the sample, using one or both cameras. Before start, user can change default parameters that will define the process of sampling, like duration, sampling rate, or number of images when the histogram will be reset. Changing one value will automatically calculate the other values, as well as a short summary. It is approximately calculating the disk space that will be occupied by captured images, according to the parameters entered. Summary text will change the color if the memory needed become too big.
There are some fields that are the same for both modes, like Batch, Supplier and API that contain information about the sample observed. “Max Particles” parameter is limiting the number of analyzed particles per image, “Min Particle Size” set up the lower limit of feret diameter of a particle to be taken into account for analysis. That means that all particles smaller than (by default) 5μm will be ignored.
There are different parameters whose default values are defined in the settings file. Those “global” program default settings can be changed manually by editing the configuration text file located in the execute folder of the program. File name is “WpfApp_CrystalVis.exe.config” and it is shown in Figure 63: CrystalVis configuration file

Camera settings

The small photo-camera symbol is used to reset all camera parameters, for example: brightness, shutter speed, exposure, flash and triggers settings to the default. Those settings are configured in a way to achieve the best quality of a captured image from both cameras.

By clicking on the wrench icon, user will be able to change the parameters of camera capture before it starts. As it is shown in Figure 64: Camera settings, there are a lot of settings that can be configured depending of experiment requirements. Settings are saved in cameras internal memory.

Integration with hardware

Drivers and Packages

Point Grey Camera Drivers

To be able to control the cameras, the PointGrey software suite has to be installed on the computer, FlyCapture_2.7.3.18_x64 version is used with prototype, but it is also compatible with new version (FlyCapture_2.8.3.1_x64).

DLL References

CrystalVis C# application is using two dll's as a reference: “FlyCap2CameraControl_v100.dll” and “FlyCapture2Managed_v100.dll”. If the new version is used, application has to be recompiled and built in Microsoft Studio with corresponding dll's from the installation package (FlyCap2CameraControl_v110.dll and FlyCapture2Managed_v110.dll).

USB Connection

Cameras are connected via two USB3 ports and to avoid data-transfer bottle necks they has to be connected through separate ports, not using the usb-hub or similar hardware for connector sharing. Drivers for usb-ports are supplied by operating system.

Database Installation

Database package installed in CrystalVis prototype is MariaDB, version mariadb-10.0.21-winx64 in localhost. It is fully compatible with MySQL, so for database management was used MySQL Workbench, version mysql-workbench-community-6.3.4-winx64.

Analysis Software

Software package used to make analysis of the captured images is ImageJ, that is an open source image processing program for multidimensional image data with a focus on scientific imaging. Version Fiji, that is installed in CrystalVis prototype, is using very powerful plugin framework and scripting language for macros.

Fiji is an image processing package, distribution of ImageJ and ImageJ2, bundling Java, Java3D and a lot of plugins organized into a coherent menu structure. The main focus of Fiji is to assist research in life sciences. From the users aspect, Fiji is easy to install and has an automatic update function, bundles a lot of plugins and offers comprehensive documentation. Software is installed into “C:\crystalvisData\Macros\fiji\Fiji.app\” folder.

Camera settings

The small photo-camera symbol is used to reset all camera parameters, for example: brightness, shutter speed, exposure, flash and triggers settings to the default. Those settings are configured in a way to achieve the best quality of a captured image from both cameras.

By clicking on the wrench icon, user will be able to change the parameters of camera capture before it starts. As it is shown in Figure 66: Camera settings, there are a lot of settings that can be configured depending of experiment requirements. Settings are saved in cameras internal memory.

Potential Impact:
The European pharmaceutical industry has an important role to play in ensuring that the people of Europe enjoy a good standard of health. The Active Pharmaceutical Ingredients (API) industry is intrinsic to this, as APIs form the most vital part of every formulated end product, and are an important part of the whole pharmaceutical industry. The European Union strives, therefore, to guarantee broad access to medicinal products, to provide the public with high quality information, and to ensure that the medicinal products manufactured are safe and effective. The single market for pharmaceutical products makes it possible to achieve these aims by increasing the competitiveness of the industry through promoting research and innovation for the benefit of the public.
Process Analytical Technology (PAT) is now the buzzword in the pharmaceutical industry. It is the new framework for better understanding in pharmaceutical processes of which crystallisation is a critical part. Central to PAT is improving final product quality by process design through knowledge of the fundamental scientific principles behind it, and continuous online control of a process.
Without doubt, the impact of providing the industry with a world-first imaging-based physical characterisation device, applying both novel microscopy methods and pattern recognition methods, capable of providing all the physical characteristics of a crystal in-line in a crystallisation process to enable accurate monitoring of crystallisation parameters such as size and shape distributions will be significant for advancing the European API industry. There will also be benefits for improved pharmaceutical product quality and safety, as well as for patient health and safety. Moreover, as the novel system will be designed, manufactured, supplied and serviced from Europe, this will have a positive impact for jobs, growth and innovation.
Ownership of the rights to exploit the project results will afford the participating supply SMEs with a real business opportunity that will generate growth in their companies, and the participating end-user SMEs will enjoy “first client” benefits.
Looking at the former, INNOPHARMA and J&M ANALYTIK will increase their competitiveness by acquiring the IP and knowledge needed to take the pre-competitive footing that they have secured from this project, and fast-tracking the CRYTSAL-VIS system through industrialisation to a commercially exploitable system, which will allow them to serve a multi-million euro opportunity in the API market, which will allow them to increase their business activity, their sales and turnover to grow their companies.
Crystallisation is a critical process during the manufacture of APIs, upon which the end product behaviour, quality and safety is hinged. Once CRYSTAL-VIS is ready to launch in the market, its market potential is expected to be very healthy. Indeed, CRYSTAL-VIS is in line with the pharmaceutical industry's shift in quality focus from post-production inspection to building quality into its products and into the manufacturing process. This only becomes possible via the movement of traditionally laboratory-based determinations of physical properties such as crystal size distribution, to instruments, such as CRYSTAL-VIS on the production floor.
The global API industry is placed at a very promising position with the growth in the global pharma and rising diseases. In fact, the worldwide pharmaceutical market is very lucrative for pharmaceutical producers and also for equipment suppliers, service providers and consultants. The world pharmaceutical market more than doubled in size between 1998 and 2006. In 2009 global pharmaceutical sales were worth approximately 750 billion-760 billion U.S. dollars. The US, however, is the world market leader with a 39.3% share of world production, followed by Europe and Japan. Seven emerging markets- Brazil, China, India, Mexico, Russia, South Korea, and Turkey- contribute nearly 25% of growth worldwide and were expected to grow 12–13% in 2008 to $85–90 billion. To satisfy demand, a large number of capital projects have been initiated and completed worldwide, creating what appears to be the ideal growth market for companies that specialise in the design and delivery of pharmaceutical plants and equipment, and especially suppliers of specialised equipment.
Looking at the worldwide API industry, the overall API market was valued at $101.08 billion in 2010 market size is likely to register a healthy CAGR (compound annual growth rate) of 7.9% during 2011-2016. Within the industry, as was seen, High Potency Active Pharmaceutical Ingredients (HPAPIs) represent the fastest growing segment, with the global HPAPI market was worth USD 9.1 billion in 2011 and is expected to reach USD 17.5 billion in 2018, growing at a CAGR of 9.9% from 2012 to 2018. These compounds are extremely effective in the treatment of cancers, respiratory disorders, and hormonal imbalances and the market for HPAPIs is mostly driven by the growth in the oncology therapeutics market worldwide. Technology such as CRYSTAL-VIS, which will advance production understanding for effective manufacture of these increasing complex APIs, as will assist during R&D work of new APIs, is expected to enjoy good market adoption in the global API industry, put in particular in Europe and the US as manufacturers position themselves to claim the more complex, high value niche manufacturing.
As the first in-line system in the world capable of measuring all physical characteristics of a crystal during the crystallisation, which will be marketed at a highly competitive price of approximately € 100,000/unit (considering that the typical market value for such a PAT technology would be €150,000, which such systems as the Malvern particle characteriser €150K, SisuChem NIR Chemical Imager €150K), the system is estimated to generated over € 40.2 M in its first 5 years of commercialisation. This is based on 310 units being sold in Europe and a further 92 being sold globally (Table 1).
Table 1- number of units of the CRYSTAL-VIS technology sold 2017-2021

These unit sales are considered realistic and achievable in view of the revolutionary benefits that the technology offers, the willingness of the sector to invest in equipment, technology and R&D (the context of the cost of the CRYSTAL-VIS in comparison to the cost of for example, researching and developing a new chemical or biological entity, estimated at € 1,059 million in 2006, is highly accessible, especially in view of the benefits that it will bring), as well as the fact that globally there are well over 8,000 API production sites globally (and this is a growing industry) that could invest in the CRYSTAL-VIS technology, and that there is potential for sale of multiple units to many of the manufacturing sites.
Table 2- Revenue generation potential and economic benefits for the supply SMEs

Consequently, INNOPHARMA and J&M ANALYTIK , as the companies that will be supplying the system to serve market demand, will increase their sales and turnover (Table 2), leading to economic growth and job creation (based on average capital intensity in the industrial sector of €120,000 per job per annum).
In relation to the benefits of the results of the project for the end-users of the system, TOPCHEM and LABIANA, both joined this consortium as “first clients” as they are convinced of the competitive advantages that access to a development such as CRYSTAL-VIS could have for their businesses and as such they are keen to experience these benefits ahead of their competitors as early and first adopters. As has been discussed the technological capability of the novel combination of advanced microscopy methods and pattern recognition methods of the CRYSTAL-VIS technology (and that could be readily retrofitted in crystallisers), would equip TOPCHEM, LABIANA and NUTRA to accurately monitor crystallisation parameters such as size and shape distributions. To have access to this information in-line, in real-time as crystallisation is ongoing would allow then to control the critical crystallisation process in a way that has been previously impossible for them to do.
They are confident that the adoption of this technology will better equip them to secure additional contracts for API manufacture and to further align themselves with the high value specialty pharmaceutical industry. Business in the bulk actives is experiencing tough competition from such emerging countries as India and China, who are driving prices downward. Competing in this segment requires producing high volumes at low costs. The ability to secure contracts for niche products, for which demand is too low to interest Indian/Chinese manufacturers, is of strategic interest and vital for future growth. In relation to quantifying the potential direct economic benefits (Table 3) that could be derived through the adoption of the CRYSTAL-VIS technology, a conservative starting point would be a 1 % increments in relation to increased turnover from 2011 figures (with a 5% increment year-on-year), as a result of new contracts secured. It must be noted that these are very conservative estimates, as the ability to land a large contract from a pharmaceutical innovator or speciality pharma company could be worth many millions, and indeed technological capability could be a winning differentiator for decision-makers. Again job creation has been based on average capital intensity in the industrial sector of € 120,000 per job per annum.

Table 3 - Potential economic benefits for the end-user SME and Other Participant

Beyond strengthening the competitiveness of the participating SMEs, this CRYSTAL-VIS project will also contribute to improving industrial competitiveness across the European Union through the commercialisation of the system.
Indeed, the role of API manufacturers in the pharmaceutical industry is rapidly changing and can be expected to evolve even more in the future. Biotechnology, for example, is still in its infancy, and may create fundamental changes both with finished products and the definition of an active ingredient. The pharmaceutical industry is constantly in flux: changes in the complexity of pharmaceuticals, competitive pressures from India and China, and new developments in the dose form markets due to new legislation all influence API manufacturers. As the pharmaceutical industry evolves, the most successful API manufacturers will be those that are able to quickly adapt to industry needs and global pressures. Technological innovation will be intrinsic with the ability to adapt. Moreover, API manufacturers are operating in a complex regulatory framework and they are facing global challenges regarding GMP and regulatory compliance. PAT technologies, such as the proposed CRYTSAL-VIS system can step in to assist.
Generic drugs companies’ willingness to source active ingredients from China and India has been particularly damaging to API manufacturers in Europe. Despite having decades of experience supplying APIs to regulated markets, European API manufacturers continue to lose ground to their lower cost Asian competitors. Unfortunately, the generic drugs environment often places more emphasis on cost than on relationships and reputation.
However, the pendulum is beginning to swing in the opposite direction as more quality issues surface in the Asian pharmaceutical supply chain, such as China’s heparin scandal, whereby in January 2008, Baxter International, Inc. had to make a worldwide recall of heparin after detecting an "unusual increase" in allergic reactions. Heparin is an anticoagulant used during heart surgeries and other procedures to prevent blood clots. The heparin had been produced in China and was contaminated with over-sulphated chondroitin sulphate, a chemical that is structurally similar to heparin but toxic to humans. The contaminated heparin caused the deaths of 81 US citizens and triggered severe allergic reactions in 785 others.
European API manufacturers are in a good position to compete against China and India based on their history of reliability, environmental controls, customer service and speed of delivery, however the quality versus cost battle will probably continue into the future. As such, to counteract low growth in the large-scale production of APIs due to fierce competition from China and India, many European API manufacturers are turning to low volume, high profit niche opportunities such as high potency APIs. Additional islands of opportunity for European API players exist within the manufacturing of peptides, controlled substances and biopharmaceuticals. Post-2015, the majority of APIs coming off-patent will be small volume products within the specialist-driven oncology and orphan drug categories1. Only those European API manufacturers with strong technological capabilities in niche areas will be in an advantageous position to profit from this future opportunity. CRYTSAL-VIS will serve to increase the technological capacity of European API manufacturers.
In relation to new job creation, if a technology such a CRYSTAL-VIS could position the European API manufacturers to secure an additional 2-5% of the growing HPAPIs segment, this would represent an additional €27.4 M- €68.5 M per annum growth in this segment alone which by 2021, would represent with could translate into € 137 M- € 342.5 M per annum, which could generate up to 570 new jobs in the sector over the same period.

The strategy for dissemination of the results, findings and benefits of the Crystal-Vis project followed an agreed pre-determined communication plan. The unique and proprietary nature of the technology involved in the design and building of the device will need to be protected until full exploitation can be realised.
To ensure protection of the early research results, external communication focused on the project website, industry visits and consultations with a specific focus on those companies in the API and related industries who have a genuine interest and need for the inclusion of PAT in typical API crystallisation processes.
The project leaflet and poster were preliminary designed with the printing of the leaflet targeted to include some images and non-proprietary results for WP4, “Design and Building of the Crystal-Vis prototype”.
Note: As the project moved through each of the development phases associated with the construction and testing of a pre-competitive prototype portable test rig, communication and dissemination of the results and the strategy associated with the communication plan, were agreed by all members of the consortium to protect and maximise the opportunity to exploit the Crystal-Vis technology.

Dissemination Activities
• Development and upkeep of the Crystal-Vis website. http://crystalvis.eu/
• Develop a link to the Crystal-Vis website on the Innopharmalabs technology website. http://www.innopharmalabs.com/tech/about-us/research-activities
• Power-point presentation template was developed to facilitate communication of the project results. See appendix 1
• Poster and leaflet layout was finalised and prepared for the inclusion of non-proprietary results obtained during the completion of WP3. See appendix 1
• Manuscript published in “Chemical Engineering Journal” http://www.sciencedirect.com/science?_ob=ArticleListURL&_method=list&_ArticleListID=-999964982&_sort=r&_st=13&view=c&md5=e06caf0f80a346a6b1d175159abe52c5&searchtype=a
• Manuscript published in “Chemical Engineering Science” http://www.sciencedirect.com/science?_ob=ArticleListURL&_method=list&_ArticleListID=-999965591&_sort=r&_st=13&view=c&md5=2fb7429321b440ac1d36ac079c87c9e7&searchtype=a
• Conference talk at the EuPAT, Austria – May 2015
• Conference talk at the IFPAC, USA – Jan 2015
• Visits and communications to API and related industries, Academic and Development facilities across Europe and the US. This activity commenced with the generation of the industrial specifications, Deliverable 1.1. Market requirements were determined through three different approaches: firstly, through visits to pharmaceutical companies within Europe, secondly through an online survey conducted across the pharmaceutical sector by the circulation of a questionnaire developed to identify the need for the technology and appropriate specifications for the technology, and finally through participation in a conference on Process Analytics and Control Technology. As well as visits to consortium members Topchem, Labiana, & Nutra research facilities the team also visited facilities of Pfizer, Helsinn and APC. The feedback from each facility was very positive with Pfizer offering the opportunity to test/validate the technology at their site in Ringaskiddy, Cork. Through the online survey and the conference on Process Analytics and Control Technology, EUROPACT 2014 in Barcelona, we took the opportunity to communicate the benefits of the crystal-vis technology to API and related industries, Academic and Development facilities representatives.
• Innopharmalabs are a member of the Synthesis and Solid State Pharmaceutical Centre (SSPC), a Global Hub of Pharmaceutical Process Innovation and Advanced Manufacturing, funded by Science Foundation Ireland and industry. This is a unique collaboration between 24 industry partners, 9 research performing organisations and12 international academic collaborators. The SSPC transcends company and academic boundaries and is the largest research collaboration in Ireland, and one of the largest globally, within the pharmaceutical area. The role of the SSPC is to link experienced scientists and engineers in academia and the pharmaceutical industry, to address critical research challenges. The SSPC research programme leads the way for next generation drug manufacture and spans the entire pharmaceutical production chain from synthesis of the molecule, to the isolation of the material, and the formulation of the medicine. The aim of SSPC is to deliver relevant solutions that address the manufacturing needs of the pharmaceutical companies and, through this, to build a strong pharmaceutical community and a pharma-friendly environment in Ireland.
• Innopharma Labs are also members of the Pharmaceutical Manufacturing Technology Centre (PMTC). The PMTC is hosted by the University of Limerick with core funding of €1M per annum from the Irish government (Enterprise Ireland and the IDA Ireland). Income is supplemented with co-funding from industry and other public sources. PMTC, established in December 2013, is led by an industry steering board which includes Ian Jones, CEO of Innopharmalabs, with an active research program driven by its industry members. Companies access PMTC to create projects and execute world-beating industry-relevant research in advanced technology solutions to address contemporary manufacturing issues. Current members of the PMTC community include multi-national pharmaceutical companies, service providers, indigenous companies and start-ups, higher education institutes and national research centres.
• Through Innopharmalabs involvement in these centres we regularly attend meetings, workshops, functions etc whereby we have the opportunity to meet and network with leaders in the pharmaceutical and academic fields. We have taken these opportunities to generate awareness of crystal Vis to these leaders and the feedback has been overwhelmingly positive that this is a technology that is long overdue and would greatly benefit both researchers and industry.
In conjunction with these activities members of the consortium attended a number of conferences/forum’s/workshop’s and exhibitions throughout 2015. These events offered a unique opportunity for the team to meet face to face with specific target groups in order to relate the findings, benefits and results of the Crystal-Vis technology. As appropriate, the consortium members took these opportunities to deliver talks/presentations, exhibit (Innopharma Labs & J&M Analytik) and/or network with the specific target groups. The following are the conferences and exhibitions that members attended in 2015:
• The IFPAC (International Foundation of Process Analytical Chemistry) conference took place in Washington DC, USA, on 25-28th Jan 2015. IFPAC has emerged as the pharmaceutical industry's premier event relating to the techniques of Industrial Process Analysis, Process Knowledge and Quality Control. IFPAC attendees come to learn more about how they can boost productivity, streamline implementation processes, enhance operational efficiency, and maximize their resources and therefore fit perfectly into the target audience category for Crystal-Vis. DIT delivered a presentation titled “Developing a new PAT Device and Algorithms for the In-situ Imaging of Crystallisation” highlighting work done and achievements of the project thus far. Both Innopharma Labs and VTT exhibited at the conference. Innopharmalabs delivered three presentations at the event which created awareness of and drove traffic to their exhibition booth. Crystal-Vis awareness was generated through the DIT presentation, direct communication to attendees both at the exhibition booths & networking throughout the conference and through the distribution of project leaflets.
• The Pittcon, Conference & Expo, will take place in New Orleans, LA USA on March 8-10th 2015. Pittcon is the world’s largest annual premier conference and exposition on laboratory science. This dynamic global event offers a unique opportunity to get a hands-on look at the latest innovations and to find solutions to all your laboratory challenges. The robust technical program offers the latest research in more than 2,000 technical presentations covering a diverse selection of methodologies and applications. Pittcon attracts more than 16,000 attendees from industry, academia and government from over 90 countries worldwide. J&M Analytik will exhibit at this conference. Crystal-Vis awareness will be generated through direct communication with attendees both at the exhibition booth & networking throughout the conference and through the distribution of project leaflets.
• Interphex Expo (International Pharmaceutical Expo), took place at the Javits Centre, New York City, USA on April 22-23 2015. INTERPHEX sponsored by PDA (Parenteral Drug Association), is the single source for complete (bio) pharmaceutical development and manufacturing solutions to safely and cost effectively process all dosage forms for life-enhancing drugs with a unique combination of exhibition, education, workshops, partnering opportunities, and networking events. This expo included; 7,000+ global pharmaceutical and biotechnology industry professionals together with 600+ suppliers, 300,000sf of event space offering state-of-the-art innovation and technologies, Pharmaceutical Industry leaders and subject matter experts from more than 48 countries offering the opportunity to share their extensive knowledge of cutting-edge technologies, Direct access to the largest, most innovative biologics and pharmaceutical manufacturers. Innopharma Labs exhibited at this Expo in the Glatt Section of the Excellence United exhibition booth. Excellence United is a strategic alliance of five family owned, German based companies who are all market leaders in manufacturing equipment used for the production of pharmaceutical products. Through this alliance, Bausch+Ströbel, Fette Compacting, Glatt, Harro Höfliger, and Uhlmann, have pooled their expertise to deliver top-quality, technologically advanced solutions spanning the entire value chain for the production of medical goods and pharmaceuticals. Crystal-Vis awareness was generated through direct communication with attendees at the education workshops and networking events and through the distribution of project leaflets.
• The seventh pan-European Science Conference on QbD and PAT Sciences (EuPAT 7) took place in University of Graz, Austria, May 18-19, 2015. This conference brings together pharmaceutical scientists and engineers from industry, academia and regulatory agencies to discuss recent developments and future trends in the field of pharmaceutical product and process development. DIT delivered a presentation on the achievements of the Crystal-Vis project to date. Innopharma Labs exhibited at this conference and also delivered a presentation which created awareness of and drove traffic to the exhibition booth. Crystal-Vis awareness was generated through the DIT presentation direct communication with attendees both at the exhibition booth & networking throughout the conference and through the distribution of project leaflets.
• The ACHEMA Forum took place in Frankfurt, Germany from June 15-19th 2015. ACHEMA boasts attendance figures in excess of 166,000 with approximately 4,000 exhibitors. ACHEMA 2015 focused on three themes: Industrial water technology, Process analytical technology and Bio-based production. The extensive lecture programme provided information on new technological developments and trends. Developers and providers as well as users and plant carriers had opportunities to discuss focal themes and exchange ideas both during the exhibition and the events. J&M Analytik exhibited at this conference and Innopharma Labs attended this conference. Crystal-Vis awareness was generated through direct communication with attendees both at the exhibition booth & networking throughout the conference and through the distribution of project leaflets.
• The 7th International Granulation workshop took place in Sheffield, UK from July 1-3rd 2015. Granulation encompasses a number of processes vital to the food, pharmaceutical, chemical, mineral, metallurgical, fertiliser, waste water treatment and catalysts industries, among others. Growing interest in the field has led to the development of the International Granulation Workshop, which has become one of the world’s largest forum for industry and academia to share and discuss recent developments in this complex and evolving field. Although the focus was mainly granulation Innopharma Labs have attended previous workshops and found that the attendees cover primary as well as secondary pharmaceutical manufacturing. Innopharma Labs exhibited at this conference and included information on the Crystal-Vis project at their exhibition booth.
• The conference on Pharmaceutical Ingredients (CPhI) took place in Madrid, Spain from October 13-15th 2015. CPhI boasts attendance figures of 36,000 from 140 countries and 2,200 exhibitors. CPhI Conferences deliver the latest pharma market insight, in-depth case studies and exceptional networking opportunities through a programme of high-level conferences. The worldwide series of events, spanning four continents, provides the optimum forum for you to learn, make new business connections and identify the latest growth opportunities. Delivered by expert teams in each region, every conference is extensively researched with leading professionals, ensuring that CPhI Conferences tackle the most business critical issues facing the pharma industry today. Labiana exhibited at this conference and included information on the Crystal-Vis project at their exhibition booth. Topchem attended this conference, as they do every year. Crystal-Vis awareness was generated through direct communication with attendees both at the exhibition booth & networking throughout the conference and through the distribution of project leaflets.
• The first Excellence United Symposium in India took place on the 7th, 9th & 11th September 2015 with an interested audience of approx. 250 attendees. The events took place in the cities of Goa, Hyderabad and Ahmadabad and were jointly prepared by the Indian subsidiaries of the three Excellence United partner companies: Fette Compacting, Glatt and Harro Höfliger. Each of them had sent experts to present their best-in-class technologies. Innopharma Lab also participated as a guest presenter of the Excellence United group which gave Innopharmalabs the opportunity to meet with interested parties in a country that is the 3rd largest manufacturer of pharmaceuticals in the world. According to visitors’ feedback they found the topics very beneficial, and the event was very well received. This very satisfying and gratifying experience calls for a repeat! Crystal-Vis awareness was generated through direct communication with attendees throughout the conference and through the distribution of project leaflets.
• Innopharmalabs in conjunction with our partner Glatt held Workshops at Griffith College, Dublin on October 7th 2015 & at the Imperial War Museum in Duxford UK on November 11th 2015. Over 60 invited guests from the pharmaceutical industry attended these workshops and Innopharma labs presented on how using PAT can help in better process development, understanding and control. Crystal-Vis awareness was generated through direct communication with attendees throughout the workshop and through the distribution of project leaflets.
• PMEC, took place in Mumbai, India from Dec 1 – 3. As the pharma industry increasingly looks towards India for high quality, low cost pharma solutions, this event was the perfect place for companies to pick up on the latest trends and innovations the market has to offer. Meet the movers and shakers in India's pharma machinery, technology and ingredients sectors for a competitive advantage that will help grow your business. CPhI India is also co-located with PMEC India, the most influential pharma machinery show in Asia. The show was attended by 33,000 visitors and Innopharmalabs exhibited at it. Crystal-Vis awareness was generated through direct communication with attendees throughout the conference and through the distribution of project leaflets

Planned Dissemination Activities
The following activities are to be planned with consortium agreement for the content and mechanisms for the communication and dissemination plan.
• Development of target groups and communities such as, ISPE, communities of practice, academic forums and networking tools
• Communication sessions specifically targeted at regulatory authorities
• Presentation of project results at IFPAC, in Washington USA in Jan 2016; Powtech, in Nurenburg, Germany in April 2016; Making Pharmaceuticals, in Birmingham UK in April 2016; Interphex, in New York USA in April 2016; EuPAT 8 in Cork Ireland in Oct 2016 and PMEC, in Mumbai India in Nov 2016.
• Further workshops in conjunction with Glatt throughout 2016. (Milan in May 2016 plus 2/3 others to be planned)
• Further process development activities to be completed at Labiana. This will include DOE trials on site with one or more API’s. Further discussion to take place in order to scope out and schedule trials.

The exploitation plan has been represented diagrammatically as follows:

As can be seen from the above diagram, the starting point will centre on a definition of the protected IPR, whereby a verifiable list of all intellectual property rights that have been applied for or registered (such as the application of a European patent) will be provided, as well as a description of all the results that may have commercial or industrial applications. Table 4 below outlines the envisaged project results. The further development work and implied risks involved are detailed in the findings of Task 6.2: Industrial Trials and evaluation of results. This report will be analysed by the partners and specific roles and responsibilities will be agreed among the participating SMEs. The RTDs will assist the SMEs in costing this future development work (and there will be no obligation whatsoever for this future development work to be subcontracted to the RTD performers). It is envisaged that the investment in further development work will be shared among the 4 SMEs and OTHER participants in proportions to be agreed as the IPR unfolds, as they all hope to benefit from the exploitation of its results. If required, the RTDs will offer assistance and guidance in accessing additional sources of funding for this future development work.

Table 4
Following on from the discussion with the patent attorney, it is essential that the unique technological applications and designs for the portable test-rig and proposed commercial designs be closely guarded and continually assessed. This is necessary to protect the IPR and the opportunities for patent filing for the technological application for in-line monitoring of physical characteristics within a crystallisation process.

Foreground Definition and protection routes
The early and clear specification of the expected foreground IPR that will emerge from this project is fundamental for effective knowledge management, whereby issues relating to the assessment of the intellectual property in terms of novelty, patentability and protection can be addressed.
Up to date reviews of the state of the art and patents have been performed to confirm the novelty of the crystal-vis technology. The foreground has been identified as follows:
Crystal-vis precompetitive test rig unit available for further dissemination.
Statistical validation of the crystal-vis system for characterising the physical properties of crystals.
Knowledge of the scale-up and hardware system and the chemometrics models.
Knowledge for the optoelectronic and microelectronic modules of the system.
Table 4 provides a breakdown of the ownership and exploitation opportunities by consortium member, based on the effort and contribution by the RTDs on behalf of each consortium member.
The participating supply SMEs will receive full ownership and exploitation rights of the different elements of all the Foreground results generated by the project. The RTD performers will not share in the ownership of the results.
All the SMEs own the precompetitive prototype system which will be used by the SMEs for future post-project demonstration and further scale-up work. All the SMEs will own the result in relation to the statistical validation of the system for the characterization of the physical properties of crystals, which will prove as a powerful marketing tool for stimulating market pull and technology push. Labiana, will be granted preferential use of these results in their own facilities, whereby they will have early access to the technology (and for free throughout the industrialisation phase and a discount off any commercial technology that they purchase). Innopharma will, on the one hand, own the knowledge in relation the scale-up of the hardware system and the chemometrics models, which they will further industrialise post project, towards a market ready system. Innopharma will provide preferential use to the end-user, Topchem and Labiana. J&M Analytik, on the other hand, will own the knowledge in relation to the scale up of the optoelectronic and microelectronic modules of the system, and they too will provide preferential use of these results to the end-user SMEs.

A review and analysis of the most appropriate IPR tools to protect the different results of this research effort was carried out. It is envisaged that the Foreground will be protected by patents (such as sub-elements of the method, centring on novelty related to the camera technology, the software component that will control the process). In addition, the CRYSTAL-VIS brand and logo that will be designed to distinguish the future industrial system in the course of trade will be protected by Trademark™.
In due course the need for any Industrial Design Registration will be evaluated in order to protect any novel non-functional features of shape, configuration, pattern, ornamentation or composition of lines or colours, which might be applied. Also the need for copyrighting © any work developed, such as models, computer software, etc. will also be assessed. Finally, the need for the use of any other IPR tools, such as utility models, Geographical Indications, Trade Secrets and Undisclosed Information, Competitive Practices in Contractual Licenses, etc. will also be discussed with the patent attorney.

List of Websites:
http://crystalvis.eu/
Contact Details:
Luke Kiernan
Ravenscourt
Sandyford
Dublin 18

Final Report Summary - CRYSTAL-VIS (A process analytical technology for characterising the physical properties of crystals)

Related documents

Share this page

Download