Skip to main content

European NoVel Imaging Systems for ION therapy

Final Report Summary - ENVISION (European NoVel Imaging Systems for ION therapy)

Executive Summary:
ENVISION (European NoVel Imaging Systems for ION therapy, http://cern.ch/envision) is an R&D consortium of sixteen leading European research centres and one industrial partner.
The project was launched in February 2010 with the aim of implementing cutting edge imaging solutions to ensure maximal safety and effectiveness of hadron therapy, an advanced form of radiation therapy that uses protons or other ions to destroy cancer cells.

In fact, hadron therapy demands absolute control of the amount and location of the dose deposited in human tissues: errors in the dose delivery can lead to incomplete destruction of the tumour cells and/or to damage to the surrounding healthy tissues. Sophisticated medical imaging tools (both hardware and software) are needed to monitor the dose in real time, i.e. while the patient is being treated.

The hadron therapy community has been exploring in-beam imaging techniques for quality assurance in dose delivery since the early 2000, when GSI first used in-beam PET in clinical routine. ENVISION made considerable progress in refining existing techniques already used in clinical settings (in-beam PET), in bringing innovative technologies close to clinical deployment (in-beam single photon tomography), and in proof-of-principle testing of new concepts (in-beam single charged particle tomography).
Members of the consortium have been working on the development of a new knife-edge slit prompt gamma camera concept and secured access to this technology by means of patent applications. A full-size prototype knife-edge slit camera was built and will now be evaluated with clinical partners. Upon successful evaluation, it is hoped that this camera will provide to the radiotherapy community an effective solution to improve the quality of treatments.


The consortium also studied in detail motion monitoring and correction techniques, developing new, promising approaches. Two methods for the automated evaluation of the results from treatment monitoring have been developed, evaluated and compared. Different aspects of detector design, imaging protocol optimisation, and treatment monitoring have been investigated using Monte Carlo methods. The existing Monte Carlo models have been evaluated in details, and enhanced models for various treatment scenarios have been developed.

ENVISION also serves as a training platform for the Marie Curie Initial Training Network ENTERVISION (http://cern.ch/entervision).
ENVISION and ENTERVISION are under the umbrella of ENLIGHT (http://cern.ch/enlight) the European Network for Light Ion Hadron Therapy.

Project Context and Objectives:
The EC-funded project ENVISION (http://cern.ch/envision) aims at improving the safety, outcome, and cost of a highly advanced technique of cancer radiation therapy (RT) called hadron therapy (also known as particle therapy), by developing cutting-edge imaging solutions. In order to achieve its goals, ENVISION pursues the development of innovative detector solutions, motion monitoring techniques, and simulation studies, and aims at integrating the dosimetric information from these tools into the treatment planning system.

ENVISION is an R&D consortium of sixteen leading European research centres and one industrial partner: the European Organisation for Nuclear Research (CERN - Switzerland), who co-ordinates the project; Technische Universität Dresden (TUD - Germany); Medizinische Universität Wien (MUW – Austria); Université Claude Bernard Lyon 1 (UCBL – France); Oxford University (UOXF – UK); Ion Beam Applications SA (IBA – Belgium); Istituto Nazionale di Fisica Nucleare (INFN – Italy); Fondazione per Adroterapia Oncologica (TERA – Italy); University Hospital of Heidelberg (UKL-HD – Germany); GSI Helmholtzzentrum für Schwerionenforschung GmbH (GSI – Germany); Maastro Clinic (MAASTRO – The Netherlands); Centre National de la Recherche Scientifique (CNRS – France); Consejo Superior de Investigaciones Científicas (CSIC – Spain); Ghent University (UGENT – Belgium); Politecnico di Milano (POLIMI – Italy); Universidad de Castilla La Mancha (UCLM – Spain).

The project is supported by the European Network for Light Ion Hadron Therapy (ENLIGHT, http://cern.ch/enlight).

All radiation therapies use various forms of ionizing radiation to treat cancer: beams of photons, protons, or ions are produced by accelerators, aimed at the tumour, and deposit energy as they pass through the human body. It is the deposition of energy that damages the DNA of cells, both malignant and healthy. The effectiveness of radiation oncology lies in the capability to target only the cancerous mass, while sparing the healthy tissues. Conventional radiation therapy uses photons (X rays), while hadron therapy uses protons and other ions.

At present, the mainstay of radiation therapy is photon therapy: this has become highly sophisticated, with methods like image-guided RT, intensity-modulated RT, stereotactic radiosurgery. No matter how advanced RT techniques get, the underlying dose deposition mechanism will always be the same: the deposited energy falls off exponentially as the photon beam traverses the body, damaging all cells on its pathway – cancerous or not. This makes it difficult to protect neighbouring healthy tissues during treatment, which is an issue for deep lying tumours, tumours in/near critical organs, and paediatric tumours.

The clinical interest of hadron therapy resides in the fact that it delivers precision treatment of tumours, exploiting the characteristic shape of the dose deposition as a function of the depth of matter traversed. In fact, protons and ions deposit almost all their energy in a sharp peak – the Bragg peak – at the very end of their path. The Bragg peak makes it possible to target a well-defined cancerous region at a depth in the body that can be tuned by adjusting the energy of the incident particle beam, with reduced damage to the surrounding healthy tissues. The dose deposition is so sharp that new techniques had to be developed in order to treat the whole target.

The latest generation of hadron therapy treatments aims at achieving the level of sophistication of intensity modulated (conventional) radiation therapy: IMPT (Intensity Modulated Particle Therapy) would allow to adjust the beam dose and depth and hence to “paint” the tumour more accurately.

These refined methods require state-of-the-art medical imaging techniques, not only prior to the treatment but ultimately also in real time, while the dose is being delivered, in order to provide fast feedback to the treatment planning system. In fact, uncertainties in the actual range of the particle beam inside the patient and factors related to the patient set-up or dose calculation may lessen the inherent accuracy of particle therapy. Advanced imaging is also key to address the issue of organ motion, such as in lung cancer, and to redefine the target volume as the tumour shrinks with treatment, or the patient’s anatomy changes due, for example, to weight fluctuations.

ENVISION addresses the crucial challenge of quality assurance during hadron therapy: making sure that the therapeutic beam is delivered at exactly the right place, at the right moment. The project aims at developing solutions for real-time non-invasive monitoring and response to moving organs, quantitative imaging, precise determination of delivered dose, and fast feedback for optimal treatment planning. In order to achieve its goals, ENVISION pursues the development of innovative detector solutions, motion monitoring techniques, and simulation studies, and aims at integrating the dosimetric information from these tools into the treatment planning system.

The first aim of ENVISION is to accurately monitor in-beam (i.e. during irradiation) the actual dose delivered to the patient. A method first attempted at LBNL (Berkeley, USA) and then clinically implemented at GSI (Darmstadt, Germany) makes use of Positron Emission Tomography (PET).
PET is routinely used in hospitals to obtain 3-D images of the functional processes inside the body. In this standard PET implementation, a radioactive tracer is injected into the patient, concentrates in the tissues of interest, and undergoes positron emission (i.e. beta+) decay; the positron annihilates in the body, and the PET system detects the two photons emerging from this process, thus allowing to reconstruct the annihilation vertexes. The distribution of the vertexes gives us an image of the tracer distribution in the tissues.

In-beam PET for hadron therapy monitoring relies on the Beta+ activity induced by the hadrons themselves, with no need for an injected radioactive tracer. The quality of the images suffers from the lower counting rate with respect to conventional PET, and from the open geometry needed to accommodate the beam nozzle: instead of a ring of photon detectors surrounding the body, the set-up will be composed of two detector “heads”, leading to dead areas and hence to imaging artefacts. ENVISION aims at improving the image quality by limiting the region of interest with a measurement of the time difference between the photons (Time-of-Flight Positron Emission Tomography, TOF-PET). The achievable image improvement depends on the accuracy of the TOF determination. The ENVISION project intends to compare different detector technologies (scintillating crystals and Resistive Plate Chambers, RPCs), to design, build, and test dual-head demonstrators including read-out electronics and data acquisition systems, to simulate a full TOF-PET set-up and to develop fast image reconstruction algorithms exploiting TOF for in-beam PET.

A recent approach to in-beam monitoring of the delivered dose is based on the detection of prompt radiation emitted immediately following the nuclear reactions induced by the therapeutic hadron beam. ENVISION investigates the feasibility of using prompt photons or protons for online quality assurance during hadron therapy, an innovative approach called in-beam single particle tomography (SPAT). In-beam SPAT is not influenced by metabolic processes, and thus should provide more precise information on the deposited dose distribution with respect to in-beam PET.

The intensity of prompt photons emitted orthogonally to the beam direction exhibits a peak structure, which is correlated with the Bragg peak. Single photon emission tomography (SPECT) is widely used in nuclear medicine to obtain 3D images through the injection of a gamma-emitting radioisotope, but the challenge for the ENVISION project is to design and build a system adapted to the special requirements of in-beam applications in particle therapy. In fact, the typical device used for SPECT imaging is the so-called gamma camera; however, the gamma cameras for nuclear medicine are not suited for the detection of photons at the energies relevant in hadron therapy, mainly between 0 and 7 MeV. The ENVISION consortium is therefore comparing different detection technologies, and developing different prototypes being tested on suitable hadron beams. One passively collimating gamma-camera has reached a high development status that allows for starting the clinical evaluation of this system and the clinical potential of prompt amma-ray imaging in general for the quality assurance of particle therapy.

In-beam dose monitoring with prompt charged particles, in particular protons, has not been attempted so far. Fragmentation measurements performed at GSI have shown that light charged particles can be recorded outside the patient's body, with a maximum intensity in the forward direction, i.e. in the direction of the beam. ENVISION is investigating the feasibility of this approach, building and testing various detector systems.

In parallel to the development of non-invasive, real-time monitoring systems that will be able to provide an accurate in-vivo measurement of the actually delivered dose, ENVISION is tackling the issue of effectively and safely irradiating tumours subject to physiological motion.

Hadron therapy, due to the high dose at the Bragg peak, is particularly sensitive to movements and changes in the anatomy, as the dose could be accidentally delivered outside of the target; this could have severe consequences especially when tumours are close to critical organs that must be preserved. State-of-the-art beam delivery “paints” the tumour in 3D, i.e. covers the entire tumour region in subsequent passes of the beam, and therefore poses the challenge of delivering a moving beam to a moving anatomy.

ENVISION explores software and hardware solutions that would improve the quality and reliability of hadron therapy in the case of moving targets. This means first of all moving from 3D to 4D monitoring, including time evolution in addition to the spatial dimensions. Within ENVISION, for the first time, the potential of different implementations of in-vivo PET imaging for a motion-mitigated delivery of hadron beams is assessed. This will allow to identify the specific requirements for 4D in-beam PET imaging for hadron therapy.

Also, since most of the upcoming European hadron therapy facilities will not be directly equipped with ibPET instrumentation, but will have availability of nearby nuclear medicine PET/CT scanners, the potential of post-irradiation 4D PET/CT imaging is also being investigated. In parallel, ENVISION explores enhanced non-invasive 4D motion monitoring techniques, which could be applied to all the different dosimetric technologies being developed within the project or routinely available.

All of these developments will be meaningful only if they can be integrated in the clinical setting. For this reason, ENVISION also focuses on developing and fine-tuning the techniques required to automatically integrate dosimetric information from ibPET and ibSPAT into the treatment planning and the whole beam delivery workflow. Such tools will facilitate the prompt detection of treatment delivery errors and permit adaptive radiotherapy, meaning that the treatment plan can be modified to account for changes in patient anatomy and tissue dosimetry. As an example, when using PET for dosimetry checks, the verification of treatment plans is ensured by the comparison between a measured and a simulated β+-activity distribution, where the latter is acquired from the treatment plan. This comparison is routinely performed by well-trained observers: this is a subjective, time consuming, and expensive procedure. An automated comparison would allow a more efficient and cost saving treatment workflow.

ENVISION also aims at developing dedicated simulation tools for hadron therapy applications, in order to reliably predict and understand all possible signals and effects that could be crucial for treatment monitoring. At present, the most reliable calculation models in the context of in-vivo dosimetry are based on Monte Carlo simulations. This approach can model the interactions of the hadron beam with matter, and the resulting production of radioactive nuclides and secondary particles. However, at this time, nuclear interactions cannot be described by well-established calculable models, and therefore different phenomenological models have been proposed. These phenomenological models depend on parameters that have to be set using experimental data as a reference, and one of the challenges for ENVISION is to review the status of available data and to make a thorough comparison of the different existing models with the data.

The ENVISION consortium is also working on the accurate description of the detection of emitted secondary particles and the production of beta+ emitters, and is putting together all the tools and procedures needed to describe realistic patient treatments in all their steps.

All of the developments within ENVISION are meant to find application in the clinical setting, and to provide solutions to some of the crucial issues in state-of-the-art hadron therapy.

Project Results:
ENVISION is organised into five research Work Packages (WP). The activities and progress during project are summarised here.

List of abbreviations and definitiions

ADC Analogue to Digital Converter
CTR Coincidence Time Resolution
DAQ Data Acquisition system
DOI Deph of Interaction
FPGA Fully Programmable Gate Array
ib in-beam
LabVIEW National Instruments Scientific Software
MC Monte Carlo
MLEM Maximum Likelihood Expectation Maximization
MMPC Multi-Pixel Photon Counter
MRPC Multi-gap Resistive Plate Chambers
PET Positron Emission Tomography
PGI Prompt Ggamma-ray Imaging
PM Photomultiplier
RPC Resistive Plate Chambers
SiPM Silicon Photomultiplier
SPAT Single Particle Tomography
SPECT Single Photon Emission Computed Tomography
TOF Time of Flight
WP Work Package


Work Package 2: Time-of-flight in-beam PET

Imaging of positron emitting isotopes (PET) produced during therapeutic irradiation with protons or other ions is currently the only clinically applied in-vivo imaging technique for the verification of the amount and location of the delivered dose. WP2 focuses on in-beam PET, and aims at improving image quality limiting the region of interest with a measurement of the time difference between the photons emitted by positron annihilations in the body (Time-of-Flight Positron Emission Tomography, TOF-PET). Contrary to diagnostic PET, in-beam PET faces additional challenges, such as biological wash out, very low emitted count rates and open ring geometries, required to allow space for beam entry and patient bed. The hypothesis behind this WP is that TOF information should help to mitigate the artefacts resulting from open-ring configuration and increase image quality.

The recent developments have achieved TOF resolutions approaching 200 ps, corresponding to a few cm resolutions in the body, with substantial improvements in image quality, Compton scatter rejection and artefacts reduction.

Different in conception and in the technical realization, two detector development lines (CRYSTAL-PET and RPC-PET) have been pursued, sharing within the project a common background for recording the time intervals and for the image reconstruction algorithms and artefacts reduction.

The first approach, following closely the mainstream implementation of commercial PET instrumentation, makes use of arrays of high-atomic number scintillating crystals coupled to fast scintillation photon detectors and recording electronics. In the initial studies, several alternative choices have been tested and analysed to select the more promising crystals in terms of light yield and fast response. Described in detail in the various collaboration documents, the study has restricted the choice to Lanthanum-Yttrium Orthosilicate (LYSO) and Cerium-doped LYSO, read-out with a Silicon Photomultiplier solid state sensors.
A thorough optimization work to maximize the light yield and the response of multi-crystal arrays has identified the best mechanical assembly making use of reflectors and wrappings around the component crystals. The outcomes of the optimization have been exploited for the construction of the demonstrators, a pair of devices used to assess the coincidence TOF resolution (CTR) in the detection of the pair of 511 keV photons emitted by a 22Na source.

Less conventional, the RPC approach capitalises on the development of a family of ultra-fast gaseous detectors, the Multi-Gap Resistive Plate Chambers (MRPC), widely used in High Energy Physics for sub-ns timing of charged prongs. The application for detection and timing of 511 keV photons has required the development of dedicated simulation programs to assess the detection efficiency, as well as detailed comparison with experimental results; prototype detectors satisfying the geometrical requirements of maximum active detection area in order to cover large solid angles have been built and tested. Limited in performance by limitations in the number of electronics channels available, a twin-head demonstrator has provided a CTR of 550 ps FWHM. This is short by a factor of two of the proposed target, but compares with existing crystal-based commercial instruments. The MRPC major advantage lays in the possibility to realize at low cost large angular coverage; prototypes 30 cm long have been built. For long detectors, the longitudinal coordinate can be deduced from the measured time differences between the two sides and corresponds to 3.5 mm rms.

Several electronics systems for recording and digitization of the fast detector signals have been implemented, either completely original or based on existing instrumentation in the various laboratories.
The hardware developments have been guided and interactively adapted to the results of extensive model calculations and simulations describing the performances of complete arrays of detectors, both for the Crystal and the RPC approaches.

Comparisons of the image reconstruction features of the different systems have been made with dedicated simulation programs that use as source real patient data, simulate the detector response and the coincidence events adding time and energy resolutions and blurring.
As golden standard reconstruction technique, the Maximum Likelihood Expectation Maximization algorithm was implemented for list-mode data from each set-up. TOF information was included by weighting each line of response by a Gaussian distribution.
Extensive simulation work, both with Monte Carlo generated therapeutic beams and with real patient data have permitted to find the most effective image reconstruction algorithms to take full advantage of the sub-ns Time of Flight information in removing artefacts and improve the knowledge of the exposure dosimetry.

Work Package 3: In-beam single particle tomography

The objective of this ENVISION WP was to investigate the feasibility of applying irradiation components that promptly follow nuclear reactions between the particles of therapeutic beams (protons or light ions) and the atomic nuclei of the irradiated tissue for real time in-vivo dosimetry in proton and ion therapy. These radiation components comprise gamma rays (photons), neutrons, and light charged particles (e.g. protons and α-particles). The work package focuses mainly on the application of prompt gamma rays for in-vivo dosimetry, with the goal to establish a single photon emission computed tomography (SPECT) to be integrated into treatment sites at proton and ion therapy facilities.

This method is generally referred to as prompt gamma-ray imaging (PGI), or in-beam SPECT (ibSPECT), if tomographic methods are applied for.
Since PGI and ibSPECT are expected to have a realistic chance to become clinically applicable, the following five topics have been comprehensively studied under WP3:

- the feasibility of single photon imaging for in-vivo dosimetry in ion therapy;
- development of a system for single gamma-ray imaging;
- a fast position sensitive beam monitor;
- tomographic reconstruction for ibSPECT;
- simulation of the imaging process.

As an alternative to the gamma-ray detection, the potential of secondary charged light particles for in-vivo dosimetry has been investigated within a 6th sub-workpackage.

The main results delivered by this workpackage are the following:

1. On the basis of the prompt gamma rays following nuclear interactions of therapeutic proton and ion beams prompt gamma-ray imaging techniques and algorithms have been developed, which are capable of deducing primary particle ranges in-vivo during therapeutic irradiation with an accuracy of a few millimetres. Thus prompt gamma-ray imaging meets the requirements of particle therapy for controlling the particle range.

2. Both passively and electronically collimating gamma cameras are feasible for high-energy prompt gammas. Effective background suppression methods, like time-of-flight techniques or geometric localisation of the beam by a hodoscope in front of the patient, is mandatory for prompt gamma-ray imaging at an acceptable signal-to-noise ratio.

3. One passively collimating amma-camera has reached a high development status that allows for starting the clinical evaluation of this system and the clinical potential of prompt gamma-ray imaging in general for the quality assurance of particle therapy.

4. The clinical evaluation of prompt gamma-ray imaging data requires a comparison of measured data (images or range profiles) with reference data, which have to be predicted from the treatment plan by means of precise Monte Carlo modelling.

5. For therapeutic light ion beams, secondary protons with a range being larger than the primary ion range are produced, giving the feasibility of deducing the range of the primary ions with clinically required accuracy via the method of interaction vertex imaging.


Work Package 4: Particle therapy in-vivo dosimetry and moving target volumes

Positron emission tomography (PET) still represents the only clinically available technique for in-vivo non-invasive monitoring of the ion beam delivery and range during or shortly after radiotherapy treatment. Due to the not straightforward correlation between nuclear-induced ß+-activation and electromagnetic energy deposition, information on the treatment delivery is inferred from the comparison of the irradiation-induced PET activity with an expectation based on the planned treatment and actual time course of the beam delivery and imaging.

So far, promising application of PET monitoring to in-vivo range verification has been limited to tumour entities in stationary anatomical sites, or with considerably reduced or mitigated motion amplitude. However, methods for in-vivo treatment verification would be urgently of need for tumour entities subject to unrestrained physiological motion, to confirm safe and conformal application of a moving ion beam (for state-of-the-art scanning delivery) to a moving anatomy.

The ENVISION WP4 aimed at assessing the feasibility and enabling optimal performance of time-resolved in-vivo dosimetric imaging for validation of motion-mitigated ion beam delivery to moving targets. Seven beneficiaries (GSI, IBA, PoliMI, TUD, UCBL, UCLM, UKL-HD) contributed to the research activities, concentrated in two synergetic sub-workpackages:

- WP4.1: Experimental validation of the feasibility and potential of (conventional) PET monitoring of motion-compensated dose delivery to moving targets

- WP4.2: Monitoring of moving targets volumes and impact on the accuracy of novel in-vivo dosimetric systems for QA of motion-compensated dose delivery

The research activities carried out in WP4.1 could experimentally confirm the potential of motion-compensated PET, not only in terms of lateral field position and beam range confirmation, but also with respect to the possibility of identifying motion-induced interplay effects, achieving an equivalent similarity of measurement and simulation as observed in the static reference measurements.

Similar performances of in-beam and offline PET imaging could be observed in the simplified phantom scenarios of comparably high counting statistics, no washout, and regular phantom movement. However, analysis of the data at more realistic low counting scenarios already indicated the need of noise-robust reconstruction and counting statistics optimization strategies.

A dedicated 4D MLEM algorithm has been developed and demonstrated for the dedicated in-beam PET scanner at GSI. For the commercial offline PET/CT scanner at HIT, the initial clinical investigations had to rely on the only available reconstruction strategy of gated PET, where the already poor counting statistics is further subdivided into 4 – 8 motion phases.

Despite the successful efforts to optimise the image reconstruction parameters with respect to the original recommendations, derived from high counting statistics scenarios in nuclear medicine, the clinical investigation for a small cohort of patients receiving carbon ion scanned irradiation in the liver confirmed the expected challenges due to a reduced number of true counts emerging from a less confined region of activation, irregular respiratory motion and activity washout due to biological processes.

Nevertheless, the feasibility of 4D offline PET-based treatment verification under clinical conditions could be shown. In particular, the benefit of taking into account the target motion could be particularly demonstrated for patient cases with comparably high numbers of true counts (above 600,000) and superior-inferior motion amplitudes in the order of 10 mm. Still, improvements have mainly been noticed on the side of the calculated prediction of the induced activity, while a gain of knowledge by 4D motion compensated PET imaging could not be finally concluded due to the underlying uncertainties in the 4D gated PET image reconstruction, despite already using an optimised reconstruction scheme.

This makes the development of further strategies of noise-robust reconstruction and counting statistics optimization of utmost importance. At reduced numbers of true counts (about 400,000), image noise was found to hinder a reliable detection of motion-induced interplay patterns. Additional challenges were imposed by inaccuracies in the modelling of activity washout, which is especially pronounced in the considered offline imaging implementation. All these drawbacks of low counting statistics and severe biological clearance could be overcome by future dedicated in-beam PET prototypes, resulting from the main achievements of WP2.

In order to pave the way towards improved performances of in-vivo verification, new reconstruction solutions have also been proposed and tested on the clinical offline PET data, overcoming the restrictions due to the confidentiality of vendor-specific information at the commercial scanner with proper reverse engineering strategies.
Moreover, pioneering work has been devoted to the integration of a prototype ultrasound tracking system in the PET imaging and dose delivery workflows, promising the possibility to provide in future a more thorough and dose-free information on the internal tumour motion than currently used external motion sensors.

As far as the activities in the framework of WP4.2 are concerned, algorithm development activities have been directed towards the implementation and testing on phantom and clinical datasets of a motion-aware 4D Maximum Likelihood (ML) reconstruction strategy for PET-based treatment verification in ion beam therapy, providing a three-dimensional (3D) comparison between the measured PET and the expected PET, which is derived from a Monte Carlo (MC) simulation of treatment delivery based on the treatment plan and the time course of irradiation and imaging. The aim was to provide higher robustness and accuracy, compared to the mono-dimensional (1D) analysis based on the analysis of distal fall-offs of PET activity profiles along the beam direction.

The biomechanical approach for motion compensated PET image reconstruction based on tetrahedral meshes and finite element modelling accounting for motion and deformation has been further developed and tested on clinical data sets (lung and liver cases). Results provided evidence of comparable reconstruction accuracy for static targets and increased correlation with respect to reference MC activities when small and large motion and deformation were added in the simulations.

Energy minimizing active model based on geometrical and morphological information were also synthesized for tracking a set of organs delineated by regions of interest (ROI). The deformable segmentation model was applied to prostate and lung cancer with good results.

Translational research activities focused on prompt gamma imaging for range uncertainties quantitative assessment in presence of motion revealed a potential significant role of prompt gamma detection in the frame of a clinical scenario for the online range verification, by comparing the measured prompt gamma profiles to the simulated profiles on the corresponding breathing phase estimated by a breathing monitoring device, on the use of the prompt gamma signal itself to estimate the phase of the breathing and on range monitoring leading to potential treatment pauses for large or unexpected range variations on basis of a classifier (go/no-go decision).

Work Package 5: The combination of in-vivo dosimetry, treatment planning, and clinical relevance

Semi-Automated analysis of PET or ibSPAT based in-vivo monitoring in ion beam therapy

Positron emission tomography (PET) is the only clinically applied technique to monitor particle therapy in vivo. In order to facilitate the comparison of predicted and measured β+ distributions, a software for the automated detection of treatment delivery deviations was implemented. This provides the basis for adaptive ion beam therapy, i.e. to modify the treatment plan for subsequent treatment fractions to ensure that the dose delivery is as close as possible to what was prescribed.

At the first clinical installation at GSI, the range assessment and comparison of β+-distributions was performed by visual inspection, a rather time consuming procedure requiring well-trained personnel. For an automated comparison of PET images, two approaches were developed in parallel within this work package of the ENVISION project. The underlying idea of both methods is that, although β+-activity distributions are highly influenced by stochastic fluctuations, the distribution at the end of the track of the particles depends mainly on the ion type and energy and the composition of matter within the track. Further systematic errors caused by the treatment planning processes, or random errors due to anatomical changes also cause changes in the dose deposition due to modification of primary particle ranges.

The first approach was based on the analysis of the distal fall-off of the activity distribution, which contains all necessary information for a full 3D evaluation of the in-beam PET data. The second method for the semi-automated comparison was based on statistical assessments in a predefined region of interest (ROI), using Pearson’s correlation coefficient (PCC) to detect ion beam range deviations.

Both algorithms were validated on 12 in-beam PET studies at GSI of skull-base tumour patients. Seven artificial beam range modifications were introduced for each study (0, 4, 6, and 10 mm water equivalent path length (WEPL) in positive and negative beam directions). Sensitivity and specificity of both methods were evaluated and found to provide similar results in terms of sensitivity and specificity compared to visual inspections of in-beam PET data.

Both methods are thus promising and effective approaches to improve the accuracy and efficiency in the clinical workflow for PT-PET. They can be considered to be complementary. The first one may be advantageous for initial evaluation, while the second is more appropriate for in-depth analysis. In addition, the PCC based PET image comparison was successfully tested for detecting patient setup errors in carbon-ion therapy.

Finally, two customized phantoms were developed and irradiated with 12C beams at the GSI facility and respective PT-PET images were recorded by means of the PET system for objective testing and verifying the algorithms for an automated comparison of PET images. The feasibility and accuracy of the detection of range differences for the complete PT-PET imaging chain was evaluated by means of a PMMA phantom designed to simulate simple range differences. Cavity structures as e.g. found in the human head were, were simulated with a second phantom. In this experimental setting it was possible to distinguish PT-PET images for a pre-set range deviation down to about 2 mm with both image comparison methods. A change in the cavity status (filled cavity versus hollow cavity) could be detected if the inner diameter of the cavity was at least 5 mm. Here the range comparison method yielded slightly better results, as in some cases also a filling status changes of cavities with 3 mm diameter were detected.

Since SPECT methods have not been applied clinically so far, the software-based methods for image comparison could not be benchmarked or validated for SPECT under clinical conditions. However, preliminary tests performed indicated that the two methods described above can also be applied for automated comparisons of SPECT images.

Purpose built phantoms for ion beam studies and inter-comparison

After detailed literature searches, special phantoms have been constructed for detailed ion beam dosimetry studies.
Some contain tissue inhomogeneities (air and several different soft and bone tissues), moving targets, regions to place a Roos ionisation chamber and/or a stack of Gafchromic EBT3 films for 3-D dose distribution analysis and can be used for scanned ion beams.

A further phantom contains a movable optical system including a camera and an organic scintillator screen in water, where light emission in real time can be imaged. Despite being more suitable for scattered beams (it has been bench tested at Clatterbridge Cancer centre, UK) this phantom can also be employed to sample scanned proton or ion beams.

These beams have been well characterised by Fluka Monte Carlo simulations and in some instances simulated dose distributions have been compared to dose distributions from available ion treatment planning systems delivering plans in inhomogeneous phantoms. Similar comparisons can be used as a reliable source of patient treatment verification at ion beam therapy centres.

Clinical Significance of deviations assessed by Radiobiological Modelling

These studies have shown that particle therapy is not as robust as photon based therapy if severe deviations of dose placement and inappropriate allocation of relative biological effect (RBE) occur.

A new, much simpler, model was developed: this links linear energy transfer (LET), which reflects the increased density of radiation in particle beam Bragg peaks, with linear quadratic model radio-sensitivities (α and ß) and RBE, expressed as RBEmax (the ratio of the α parameter for control and particle radiation) and RBEmin (the ratio of the square roots of the  parameter for control and particle radiation) which can be used within biological effective dose (BED) equations.
This is shown to fit classical ion beam data sets very well, and can be used for tentative radiobiological modelling as well as a convenient check on existing international systems.

Methods for correcting errors in treatment delivery, using changes in LET, radio-sensitivities and RBE within BED equations have been used in worked examples of errors. These are naturally more complex than photon based corrections, but allow such corrections to be made rather than apply what could be very erroneous corrections by physical dose with complete disregard to the radiobiological consequences. These issues have important implications for clinical trial design and informed patient consent.

Analysis of the ‘Rococo’ data base at Maastro of prostate cancer patients with inclusion of prostate motion between fractions has been extended to patients re-planned for best photon, proton and carbon ion beams, with reasonably good predictive accuracy for known patient biochemical control outcomes. The results include derivation of fitted radiobiological parameters such as RBEmax and RBEmin, while also suggesting that dose re-calculation is not required for iso-centre realignments of up to 11mm in intensity modulated proton therapy of the prostate. Stability of treatment is improved by image guidance and daily dual field treatment is predicted to be superior to daily single field treatment in intensity modulate forms of proton and carbon ion therapy.

Work Package 6: Monte Carlo simulation of in-vivo dosimetry

Monte Carlo (MC) simulation is an essential tool in particle therapy and in imaging-based particle therapy monitoring. MC tools can describe the complex physics of hadronic interactions and, in general, all aspects of interaction of radiations with matter. At the same time, they are able to consider complex geometries (and interfaces between rather different materials), provide accurate multi-dimensional transport and include fully detailed descriptions of the patient anatomy.

The management of tissue inhomogeneities is where the use of MC simulations reveals its full strength for accurate kerma, fluence and absorbed dose calculations. MC particle transport codes can predict the production of radionuclides, such as β+ emitters, and the propagation and detection of secondary radiation, such as prompt photons. These predictions can be compared with measured data produced concomitantly with the treatment so as to assess the conformity of the delivered treatment with the treatment plan.

Within WP6, various ENVISION research groups have used MC simulations to investigate different aspects of detector design, imaging protocol optimisation, and treatment monitoring. The main aim has been that of developing dedicated simulation tools appropriate for use in hadron therapy applications so to reliably predict and understand all possible signals useful for the treatment monitoring. This activity capitalised on pre-existing developments by different groups, mostly working with GEANT4/Gate and MCNPX simulation codes.

The ENVISION work programme on MC simulation tools focused on three main aspects. First, the accuracy of the physics models included in MC codes and involved in particle therapy applications has been investigated. Second, codes features have been enhanced so as to improve their accuracy, provide software supporting the modelling of the complex geometries involved in hadron therapy and allows for the joint modelling of a particle therapy treatment with an imaging acquisition within the same framework. Third, the use of the code has been optimized so to become more and more suitable for the direct application to actual clinical cases and to match, as far as possible, the requirements in terms of reliability, ease and speed of operation.

WP6 has planned the activity in annual steps. The first one was to define the physics models appropriate to characterize the interactions of interest in hadrontherapy in order to develop complete simulations of ibPET and ibSPECT experiments, from the treatment beam to the imaging data acquisition.

Therefore, during the first year, simulation work consisted in modeling simple experiments relevant in hadrontherapy (ion or proton beam hitting a simple target) and in comparing observables that could be measured experimentally with the corresponding observables obtained from various simulation tools.

The main results were:
- the description of the physical models available to describe the interactions of interest in hadrontherapy.
- the results of the comparison between available observables obtained experimentally and corresponding observables obtained from various simulation tools.
- the identification of the shortcomings of available physical models for hadrontherapy, and recommendations regarding the reliability and uncertainties of the physics models available in the selected codes.

As outcome, we have obtained a unique compilation of results obtained using Fluka, Geant4, and MCNPX codes for the modeling of the physics involved in hadrontherapy experiments, providing a reference document for all those involved in Monte Carlo simulations of hadrontherapy experiments.

The second year of work in WP6 has been dedicated to the continuation of the development of the physics models contained in our reference codes and to assess the MC simulation tools capable of simulating a complete in-beam ET experiment, from the beam to the reconstructed images. For this purpose, we have considered a set of phantoms irradiated by both proton and ion beams and we learned how to manage the detections of emitted secondary particles and the production of beta+ emitters, understanding the differences between different codes and optimizing the selection of physical variables necessary to improve the agreement between simulated and experimental data. These results are now considered also by the experimental WPs in ENVISION.

In the third year of activity, WP6 started to put together all the tools and procedures to manage realistic patient treatments, thanks to the collaboration of some hadrontherapy centers which provided us with actual treatment planning results. This brought to the production of a set of examples, available to the community, where all the steps, from treatment to secondary particle detection, are accounted for. An important effort has been devoted to assess accuracy and usefulness of these codes, to develop their relevant physics models and ancillary tools, and to make them user-friendly and widely available.

In the final year of activity, WP6 concentrated the efforts in the optimization of codes, mainly in terms of computing time, in order to provide to clinicians and medical physicists better tools for the daily activity in hadron therapy centers.

The most important scientific and technical results achieved in WP6 activity can be summarized as follows:

- Assessment of physics models and settings in FLUKA and GATE/GEANT4 for the simulation of Particle Therapy treatments.

- Intercomparison of the particle production features of the two codes which are of relevance for Particle Therapy, so to provide a reference for MC users.

- Upgrading of physics models for the generation of prompt gammas and Beta+ emitters in the interaction of therapeutic beams for Particle Therapy

- Study and optimization of gamma detection for imaging in Particle Therapy

- Reproduction of the experimental results of experimental test beam measurements of both prompt gamma yield and beta+ activity induced by proton and ion beams in the energy range useful for Particle Therapy

- Demonstration of the capability to fully simulate a true patient case in Particle Therapy reproducing at the same time the operation and results of imaging devices for in-beam monitoring of the treatments.

- A mixed MC/analytic approach for the fast simulation of a photon detection device (IBA, proprietary device)

- Development of strategies for faster simulation in Particle Therapy with FLUKA and GATE/GEANT4:
a) Biasing direction of annihilation photons and prompt de-excitation photons in FLUKA
b) Track-length approach for the simulation of prompt photon production in proton therapy using GATE/GEANT4
c) Development a GATE/GEANT4 version for highly parallelized simulations using GPUs

Potential Impact:

ENVISION has been answering the explicit demand of the medical community for improved hadron therapy. Indeed, the precision of hadron therapy is only as good as the medical imaging guiding it: the attractive features of ions with respect to photons are the conformal dose delivery and the low integral dose to healthy tissue, but these can become a double-edged sword if the target volume is not accurately defined and if changes in the tumour size and position are not taken into account. Hadron therapy is potentially the treatment of choice in case of deep-seated tumours, tumours close to critical organs, or paediatric tumours: all cases where sparing the healthy tissue close to the tumour, while at the same time making sure that the totality of the cancerous mass is irradiated, is essential.

Within the ENVISION project, several methodologies for improving the quality of proton and ion therapy have been explored. Technical developments resulted in several prototypes for different imaging techniques, and covered the software needs before, during and after the irradiation. The techniques developed within ENVISION will help translate the intrinsic physical and radiobiological advantages of ions over photons into improved cancer outcome, as they will enhance the control of where the dose is actually being delivered.

The in-depth evaluation of the existing Monte Carlo models, as well as the enhancements to the simulation software packages used for hadron therapy will be beneficial for all the scientific community involved in the fields: in fact, MC simulations are the basis for prototype design and evaluation, as well as for treatment planning.

The timescale for the clinical deployment of the various imaging prototypes will vary, depending on the maturity of the underlying technique. PET is already used as post-treatment verification, and there is experience also concerning its use in real time. ENVISION has brought advancements not only in the detector technology, but also in motion monitoring techniques that will effectively allow to follow the variations in size and position of the tumour during treatment.

By means of their characteristic secondary emissions, proton beams offer a unique opportunity to perform online in vivo quality control of radiotherapy treatments. Members of the consortium have been working on the development of a new knife-edge slit prompt gamma camera concept and secured access to this technology by means of patent applications. A full-size prototype knife-edge slit camera was built and will now be evaluated with clinical partners. Upon successful evaluation, it is hoped that this camera will provide to the radiotherapy community an effective solution to improve the quality of treatments.

The ENVISION researchers were among the first to explore the potential of imaging prompt charged particles during carbon ion irradiation, proving that the concept is viable and could be applicable clinically, once the current proof-of-principle prototypes will be optimised.

ENVISION also tackled the issue of adaptive ion beam therapy, i.e. how to detect deviations from the prescribed treatment plan and modify the treatment plan for subsequent treatment fractions accordingly. In particular, ENVISION developed new algorithms and software tools to automatically detect treatment delivery.

All of these techniques will ultimately allow not only to deliver the best possible treatment to each patient, but also to optimise the patients’ throughput at any given centre, as the set-up time of each patient will be greatly reduced, without compromising on the safety and quality assurance. In fact, it is easy to associate the high costs of hadron therapy to the size and complexity of the equipment, forgetting that a higher throughput is one of the necessary ingredients to mitigate these costs.

A full exploitation of the hadron therapy potential and a reduction of its costs are all pressing concerns at a time when several new centres are being planned or built, along with an increasing demand for qualified experts in this truly multidisciplinary field. ENVISION has contributed to the establishment of the next generation of young experts in medical imaging tools for hadron therapy: thirteen PhD thesis have been completed during the life time of the project, and several post-docs contributed to the research activities. In addition, ENVISION serves as a training platform for the ENTERVISION project (http://cern.ch/entervision) a Marie-Curie Initial Training Network aimed at educating young researchers in online 3D digital imaging for hadron therapy.


Role of the ENLIGHT network

ENVISION is under the umbrella of the European Network for Light Ion Hadron Therapy (ENLIGHT), which now counts 400 members from more than 20 European countries, coming from all fields of expertise. ENLIGHT is co-ordinated by Manjit Dosanjh from CERN, who is also the ENVISION project co-ordinator. The Network holds annual meetings, which are open to all the hadron therapy community and represent an invaluable occasion to discuss results and future research directions in an informal setting, with clinicians, industrial players, physicists, radiation biologists, engineers and ICT experts.

In 2010, ENLIGHT launched the conference “Physics for Health in Europe” (PHE); in 2012, the PHE conference joined forces with the “International Conference on Translational Research in Radio-Oncology” (ICTR), giving birth to the new ICTR-PHE conference series. Held in Geneva in 2012 and 2014, ICTR-PHE has become an important multidisciplinary gathering of experts in the wider field of radiation oncology. ENVISION has been pro-actively participating in the conference: Manjit Dosanjh is co-chair of ICTR-PHE, many of ENVISION’s senior scientists are members of the advisory committee and/or session chairs, a large number of abstracts are regularly submitted by the ENVISION researchers, and the project is always showcased at the conference’s industrial exhibition.

The inclusion within the ENLIGHT network has obviously contributed to maximising the impact and dissemination of ENVISION results.


Main dissemination activities

The ENVISION Project Office designed and deployed a publicly accessible website with general information about ENVISION that may be of interest to the general public and to young researchers, including the list of scientific publications and job openings in the field.

Public deliverables are made available through the CERN Document Server.

Articles for general public and various stakeholders

CERN Bulletin “CERN’s policy in the field of knowledge and technology transfer goes global”, November 2009 (http://cds.cern.ch/record/1221067?ln=en)

CERN Bulletin, “The European hadron therapy community touches base”, September 2010 (http://cds.cern.ch/record/1291611?ln=en)

VTT Salud, “Nuevas tecnologías de imagen médica para evaluar la calidad del tratamiento del cáncer dentro del proyecto europeo ENVISION”, December 2010 (http://www.vitsalud.es/Diseo_de_paginas/Revista.aspx)

CERN Bulletin, “A new EU-funded project for enhanced real-time imaging for radiotherapy”, February 2011 (http://cds.cern.ch/record/1327631?ln=en)
Medical Physics Web, “Calculation uncertainties undermine dose verification”, May 2012
(http://medicalphysicsweb.org/cws/article/opinion/49540)

ENLIGHT Highlights, “ENVISION Mid Term Review”, July 2012 (http://enlight.web.cern.ch/download-the-full-first-issue-of-highlights)

CERN Bulletin, “Medical imaging projects meet at CERN”, January 2013 (http://cds.cern.ch/record/1507178?ln=en)

CERN Update, “Medical imaging network passes midterm review”, January 2013 (http://home.web.cern.ch/about/updates/2013/01/medical-imaging-network-passes- midterm-review)

ENLIGHT Highlights, “Unveiling the hadron beam during treatment”, January 2013 (http://enlight.web.cern.ch/highlights-january-2013)

ENLIGHT Highlights, “ENVISION comes to CERN”, June 2013 (http://enlight.web.cern.ch/sites/enlight.web.cern.ch/files/media/downloads/enlight_hi ghlights_june_v6.pdf)

CERN Knowledge Transfer Group Report 2011, 2012, and 2013 (available from http://cern.ch/knowledgetransferweb)

GSI Scientific Report 2011, 2012

ENVISION was featured in the issue #11 “Atoms for Health” of the magazine 
Accastampato, targeted at high school physics teachers and students (http://www.accastampato.it/en/home/copertina-11/). This special issue in four languages was prepared for the 2013 European Researchers’ Night and distributed through various channels (including during the CERN Open Days, the ENVISION and ENTERVISION courses and meetings, and through the CERN Visits Service) 



Brochures

A general brochure about ENVISION was produced, and is available in electronic format on the website. The brochure has been printed and distributed upon several occasions. ENVISION is also listed in a brochure on medical applications at CERN, produced for the 2013 Open Days of the laboratory, and regularly distributed to visitors, press, VIPs.

Dissemination seminars and events

CERN, TERA:In September 2013, ENVISION was featured in the exhibition on Knowledge Transfer set up within the framework of the CERN Open Days. The exhibition welcomed more than 8000 visitors from all over the world over the Open Days weekend.

CERN:ENVISION is regularly included in presentations to students, teachers, VIP guests, and general public all year round.
ENVISION is one of the supporting organisations for the ICTR-PHE conference series (http://cern.ch/ictr-phe14) and as such its logo is available both on the website and on the printed programme. The project was also represented at the conference technical exhibition, together with ENLIGHT and the related EU-funded projects.

UKL-HD:In October 2013, Katia Parodi gave a public talk at the Deutsches Museum Munich, where ENVISION was also featured.

IBA:ENVISION was featured at the IBA R&D Day on 16 September 2013, with presentations from J. Smeets, G. Janssens, F. Roellinghoff and E. Clementel to an audience of more than 100 people.
ENVISION was also featured in talks by D. Prieels at the IBA User Meeting 2013 (Chicago) and 2014 (Chantilly). The audience consisted of more than a hundred IBA customers from Europe, US, and Asia.

Oxford:Talk by B. Jones Our Big Gaps in high LET Radiobiology Knowledge for Medical Applications in Cancer Therapy at the CERN Brainstorming Meeting on Medical Applications, February 2014.


Multimedia

A video on ENVISION and the related Marie Curie ITN ENTERVISION was published in January 2013 (”ENVISION and ENTERVISION: training and R&D for the cancer therapy of the future” http://cds.cern.ch/record/1507226?ln=en).

An animation on ENVISION was produced in 2013 (“ENVISION, innovative medical imaging tools for radiotherapy” https://cds.cern.ch/record/1611721?ln=en). The animation is routinely included in presentations, and displayed on screens around the laboratory.

High definition images extracted from the animation are being used in several publications (as an example, on the cover of the April 2014 issue of the CERN Courier http://cerncourier.com/cws/archive/cern/54/3).
A new animation is being produced (September 2014), to show the role of ENVISION imaging tools in patient treatment.

List of Websites:
http://cern.ch/envision

Contact:
Manjit Dosanjh
CERN
CH 1211 Geneva 23

Life.Sciences@cern.ch