European Commission logo
italiano italiano
CORDIS - Risultati della ricerca dell’UE

Two Stage Rapid Biological Surveillance and Alarm System for Airborne Pathogenic Threats

Final Report Summary - TWOBIAS (Two Stage Rapid Biological Surveillance and Alarm System for Airborne Pathogenic Threats)

Executive Summary:

The TWOBIAS project has developed a demonstrable, modular and “close-to-market” demonstrator of a stationary, rapid, reliable, vehicle-portable, Two Stage, Rapid Biological Surveillance and Alarm System for Airborne Threats (TWOBIAS) with extremely low false alarm rates. TWOBIAS is able to provide reliable information to command and control systems (CCS) and first responders within seconds, enhancing security related to biological threats at high profile public sites (Figure 1.1).

The TWOBIAS system includes both detection (BDU – biological detection unit) and identification (BIU – biological identification unit) schemes:
i) Stage ONE: First alarm based on best-in-use optimized optical BDU (detect-to-warn).
ii) Stage TWO: Second alarm based on highly automated microfluidic-based platform with a molecular BIU (detect-to-treat).

In order to develop a more specific and low false alarm rate BDU three operative biological detection systems, using complementary and orthogonal techniques, were used and run in parallel. An alarm algorithm was developed to utilize signals from all three BDUs before the first alarm (detect-to-warn) signalled the BIU to start identification of a sample collected by an air sampler.

The main activities of TWOBIAS project have been:
1. Simulation and modelling of the TWOBIAS system architecture and the environment of the real-life test site Prague interchange metro station “Muzeum” and the mitigation actions of an alarm to develop a Command and Control System software.
2. Comparing and characterising three operative BDUs and a BIU in parallel in a closed aerosol chamber where relevant biological simulants for biological threat agents are released.
3. Improving the quality and classification of identified data from the BDUs in order to develop improved alarm algorithms for individual and coherent response.
4. Development of a BIU for an automatic, sensitive and fast identification. Integrate the BIU and BDU into an automated two stage alarm system.
5. Prepare and carry out a final test and demonstration of the TWOBIAS system at a real-life test site (Prague Metro interchange station “Muzeum”) regarded as potential target for a bioterrorist attack.

The final tests and the demonstration of the TWOBIAS system was conducted at the Prague interchange metro station “Muzeum” at night hours for five consecutive nights. During these tests a harmless and relevant biological simulant was dispersed while train sets were running at rush hour frequency to create a realistic air flow pattern and background environment. Invited guests from EDA, stakeholders and end users were able to attend at the demonstration during the fifth and last night of testing.

Project Context and Objectives:
Recent bioterrorist events have emphasized the need for necessary national and global actions to enhance preparedness and response against biological threat agents. Biological preparedness includes implementing safety measurements to reduce the number of individuals exposed to biological pathogenic threat agents, whether they are caused by a bioterrorist attack, an accidental release of biological threat agents or by naturally occurring disease-causing pathogens (incl. pandemics). Safety actions also include early warning detection systems, surveillance systems, identification schemes, response and recovery. First responders need to be rapidly informed about as much as possible about a bioterrorist attack, such as the time and place of the attack, concentration levels of the biological threat agent, contaminated areas, type of threat agent and exposure time. In particular, the concentration levels, exposure time and type of agent set a baseline for the first responders. This knowledge controls timescales for isolation and/or evacuation of exposed individuals as well as initiation of correct medical treatment in order to save lives.

It is highly likely that a deliberate release of biological threat agents would take place at locations where people are gathered (crowds) in order to maximize exposure (such as at airports, train stations, arenas, large buildings etc.). There are no reliable biological detectors currently installed at such places in Europe, proving the existence of security gaps that can be exploited by a bioterrorist.

The TWOBIAS Consortium consisted of nine partners from six nations, including users (end users and Associated User Representative): Ministere de la Defence, France and Dycor Global Solutions, Cyprus and Totalforsvarets forskningsinstitutt, Sweden and Q-Linea AB, Sweden and Statni Ustav Jaderne, Chemicke A Bilogicke Ochrany, Czech Republic and Thales SA and Thales Communication and Security, France and Uppsala Universitet, Sweden. The project has been coordinated by the ninth partner, Norwegian Defence Research Establishment FFI, Norway.

The partners represent a mixture of:
• Research institutes (civil and defence) (Partners FFI, DGA, FOI, and SCB)
• SMEs (Partners DGS and QL)
• Large companies (Partners TRT and TCS)
• University (Partner UoU)
• Users (Prague Public Transit Company Inc PRM, French Ministry of Interior FMOI
and Directorate for Civil Protection and Emergency Planning DSB)

Based on the absence of a reliable biological detection system the main objective of the TWOBIAS project has been to develop a demonstrable, modular and “close-to-market” demonstrator of a stationary, rapid, reliable, vehicle-portable Two Stage, Rapid Biological Surveillance and Alarm System for Airborne Threats (TWOBIAS) with extremely low false alarm rates.

The project contained seven work packages (WP) in which WP1 and WP7 were devoted to project management, dissemination and exploitation of foreground, while the scientific and technical results/foregrounds are described in WP2-6. WP2-5 and WP6 were defined as RTD and DEM activities, respectively.

Project Results:
Development of a dispersion simulation tool for the real-life test site:

To simulate as close as possible the real environment of the Prague Metro interchange station “Muzeum”, a specific simulation model was developed which take into account the effects of train traffic on aerosol dispersal of biological particles. It was clearly shown that the effect of trains on the dispersed aerosol fluctuated with a period corresponding to the period of the train (320 seconds). Simulations with dispersal of BG and moving trains at “Muzeum” were able to calculate the BG concentration (ACPLA, Agent Containing Particles per Liter Air) as function of time at possible sensor locations. These simulations were used to optimize the source and sensor placement during the real-life tests and demonstration at the Prague Metro interchange station “Muzeum”.

The results of the dispersal simulation showed a good match with real-life measurements and reference data from the two real-life test and demonstrated that dispersal of biological agents can be simulated in large indoor infrastructures. The simulation also showed that it was required to have biological sensors with high sensitivity (close to 1 ACPLA) if the aim was to raise alarm in less than 5 minutes with a small number of sensors. At least 2 sensors were required to cover all possible dispersal locations.

Developments by the TWOBIAS partners on their own device/system and the development of the Command and Control System by TCS (Thales Communications & Security) procured to the consortium a software integrated system.

Development of- and improvement of the LIBS BDU:

LIBS (Laser Induced Breakdown Spectrometry) have been a relatively proven technique for analysis of macroscopic objects, but LIBS system for practical use as bioaerosol detectors have not been commercially available. Thus it was concluded that in order to use a LIBS system in TWOBAS, some development was needed to reach a demonstrable level. Much effort was consumed on improving the detection capabilities of the LIBS, in which the DGA chamber trial and real-life tests in the Prague Metro interchange station “Muzeum” provided valuable results and knowledge. These development efforts resulted in a number of improvements:

• Construction of a new vacuum chamber, completely reconfiguring the whole system.
• The system was fitted with a TSI CSVI mini XMX particle concentrator.
• Making the system ruggedized and developed into a transportable measurement system.
• Air flows in the measurement chamber was better controlled, thereby keeping the particles time of flight constant which resulted in a significantly higher hit rate.
• A double trigging system was fitted to measure time of flight of particles between trigging point and the point where the plasma was created.
• Improved signal-to-noise ratio in the trigger signal reduction also made it possible to trig at smaller particles.
• Measurement frequency was increased by reducing the read out time from the camera. This could be done by decreasing resolution and/or limiting the wavelength area of interest.

At the time of the 2nd real-life trial and demonstration at the Prague Metro station “Muzeum” LIBS had a sampling rate of 3 Hz.

Development of a miniaturised platform for sample processing and identification:

In order to reduce time of identification after a positive alarm signal from the BDUs, a highly automated on-site microfluidic-based identification system was developed. The BIU component included sample collection and automated sample processing based on multiplexed and simultaneous amplified single molecule detection analyses on different types and entities of biological agents on a microfluidic platform.

- Selecting an appropriate air-sampler
During the chamber trial at DGA seven different air-samplers available among TWOBIAS partners were tested and evaluated. Among these the SASS 2300 wetted-wall cyclone air sampler was the only air sampler which offered full integration with the BIU. It offered the possibility to sample continuously for hours, while a sample of desired volume could be pumped from the SASS 2300 at any time during air sampling into the automated sample processing robot for downstream processing by the BIU, while the remaining sample could be pumped into an on-board sample vial for subsequent identification. SASS 2300 also had an automatic cleaning program which could also be controlled by the BIU.

- Sample preparation
A sample preparation protocol for protein-based identification which was possible to implement in an on-site automated sample preparation station was adapted and placed as an intermediate between the sample collector and the core sample processing unit.

The sample preparation module chosen for the on-site system was built around Cavro RSP9000 robotic arm (Cavro Scientific Instruments, San Jose, USA). The robotic arm could, when integrated with pumps, valves and tubes function as a complete liquid handling system. It was controlled from an external host and any integrated valve and pump were also controlled through it. The robotic arm was operational along three axes, X/Y/Z, each with its own motor. It could be designed to move to specific positions specified by the external host.

- BIU Identification and detection
Identification was accomplished by applying a multiplex mixture of PLA probes for protein-based, immunological, identification of the target molecules. Reacted probes were then amplified according to Q-linea’s Amplified Single Molecule Detection system, which uses a single end-point measurement using a TWOBIAS-specific adaptation of Q-linea’s proprietary Sample Processing Module (SPM). The SPM enabled high sample throughput and was designed for autonomous operation. Enumeration of RCP (Rolling Circle Products) formed in the SPM by the detection module was the final step of the BIU process. This was performed by Q-linea’s Aquila 400 detector. In the TWOBIAS on-site system, the detector operated in integrated mode together with the SPM.

Chamber and real-life tests:

During two weeks (12-23 March 2012) of testing a total of 29 dispersions were performed in the 1400 m3 chamber located at DGA CBRN centre. During these trials data from the three BDUs was collected from background aerosol, bioaerosol test simulants Bacillus atrophaeus (BG) and Escherichia coli (EC) dispersed with and without natural aerosol background.

In order to obtain BDU data from a real-life test site, trials were performed at the Prague Metro interchange station “Muzeum” in the period October 7th - 12th 2012 and 6th – 11th 2013. The area of the metro station was reserved for the TWOBIAS project at night, when the transport of passengers ended and the station was closed for the public, in the time window from 00:30 to 04:00 AM.

During the real-life a total of 16 trials of dispersing of the biosimulant Bacillus atrophaeus (BG) spores into air were successfully carried out. Trials were performed both with and without trains running and allowed data collection from all three BDUs in a real-life environment. During the last night of real-life tests a demonstration for selected and invited visitors representing Users and the public society was held.

The final demonstration was conducted with two dispersions of biological simulant, the first one without trains running and the second one with trains running at morning rush hour frequency. Results showed, as expected, that detecting the bioaerosol (stageONE) was far more difficult with the higher background particle concentration created by train movement. The identification step (stageTWO) was able to identify the dispersed biosimulant also with the particle background created by trains running.

Results from the demonstration showed:

1. During the first trial, without trains running, the BG cloud reached the demonstrator with a concentration above 40 ACPLA and was detected by all three BDUs.
2. As part of the second demonstration trial trains entered the station at 01:36. This had an immediate and major impact on the variation of the internal parameters. No BDU alarms were raised, despite a BG concentration above 40 ACPLA.
3. The BIU was able to identify the dispersed BG aerosol during both trials within the time frame of just above one hour after the dispersion (figure 1.4). As a total during the October 2013 real-life tests, 8 out of 10 biosimulant dispersions were identified. The 2 unidentified dispersions were both due to air sampler and air sampler integration problems.
The real-life tests and demonstration of a biological surveillance and alarm system for airborne threats directly in an important metro station under its routine operational conditions was regarded as a unique experiment. The bio detection and identification systems developed within the project was tested against the real mixture of biosimulant and natural particle background at the metro station, and any comparable experiment have never before taken place in Europe.

BDU and fusion data analysis:

- BDU data analysis
The CFLAPS Gated Fluorescence Ratio (GFR), which was used as 1st level data output, did not generate a signal strong enough to raise an alarm at the low sensitivity settings used in the Prague Metro real-life tests and demonstration. However, the CFLAPS detection alarm can be configured depending on the desired response. Configuring for higher detection sensitivity can be done at the expense of increased of false alarm frequency. When increasing the GFR sensitivity in post experimental analysis, the CFLAPS was able to raise an alarm also at the time of the second dispersion with trains running. The TWOBIAS fusion software can be allowed to check and verify any CFLAPS alarms against the LIBS and MAB signal outputs and thus alleviate any false alarm indications.

As for the LIBS, at the time of the dispersion without trains it displayed a moving average of 25 classified single shots and indicated for a short period that all the measured particles were classified as BG. During the dispersion with trains running, it did not show any indication of an increased level of BG particles in the air.

As it was not clear to TRT how the internal variables (2nd data level) were used to raise MAB alarms (1st data level), TRT built a new classification system based on the MABs internal variables to produce alarms. A classification system based on minimal Mahalanobis distance classification algorithm was built. The performance of this classification system (MABTRT) exhibited better performance than the original MAB classification system.
This predicted that if the signal to noise ratio during the trial was not too low, the classification algorithm would give satisfactory results in terms of false alarms. After post-processing and analysis of the 3rd trial data to improve classification of MAB, test scenarios exhibited a higher alarm rate. On the other hand it triggered alarms after the BG dispersion with running trains.

- Fusion analysis
A key concept of TWOBIAS was to try to combine the available detectors in order to improve the overall detection performance. Various fusion strategies to combine the BDUs were studied, in particular the combination of the BDUs’ 1st level of data and the combination of the BDUs’ 2nd level of data.

A first finding of this work was that the 1st level of data of the MAB and CFLAPS detectors, i.e. the alarms that they raised, which were the results of proprietary algorithms applied on the detectors’ 2nd level data, had a potential for improvement. Results also showed that fusion of the detectors’ 1st level data didn’t bring significant improvement in terms of performance. Classification and fusion systems based on the detectors 2nd level of data was developed, and exhibited better detection performances than the CFLAPS and the MAB. This showed that the MAB and the CFLAPS did generate discriminative information that it was possible to exploit in order to detect efficiently the presence of a biological threat agent in a real-life environment.

The main conclusion for the first level fusion analysis, concerns the second BG release (during the demonstration) that went undetected by the BDUs, even though the slit samplers confirmed that the cloud had reached the demonstrator point. This indicates at least two aspects: either the sensor sensitivity was too low (in other words the alarm threshold was too high, needing at least an ACPLA concentration over 40 and perhaps during a time interval over 1 minute) or the signal to noise ratio was too low, meaning that the background environment obscures the BG cloud.

Using 2nd data level for the fusion alarm algorithms, a clear alarm was triggered by the fusion system when no BDUs triggered one itself. In conclusion, provided that the right annotation was found, the classification and fusion algorithms can help improve the performances of the detect-to-warn (stage-ONE) level in the TWOBIAS proof-of-concept system.

Overall, post-experimental analysis of the 2nd real-life trial in Prague showed that early biological alarm triggering is possible under certain conditions, provided that a short preliminary calibration stage is performed on the system. When comparing the two real-life experiences, the first observation is that the background variability has a tremendous impact on the accuracy of the system. The sources of this variation are multiple. For instance, atmospheric conditions might vary greatly from one year to another, even from one season to another and sometimes even during the same night. Also, trains circulation periodically moves the entire air mass in the underground and possibly lifts old particles from previous dispersions. Good placement of the TWOBIAS system in the metro station was also an important factor, as established by the simulation in WP2 and proven by the demonstrator. Further improvement could be achieved for instance by adding a redundant TWOBIAS system and temporal analysis to reduce the train blinding effect.

The success of the fusion algorithms depends on the accuracy of data annotation and extensive analysis of the available reference data from slit-samplers in order to establish the best annotation was performed. Availability of reference data is regarded as mandatory to assess algorithms performance in order to interpret isolated alarms as true or false. Algorithms performances depend on operational choices like balancing the importance of false negatives (FN) or false positives (FP). Data fusion improved the detection threshold at the cost of increased false alarm rate.

A more detailed description on the data analysis from the real-life experiments at the real-life test site can be found in Deliverable D4.16 “Data analysis and improvement of classification and fusion tools, based on real life tests at “Muzeum” station in Prague (2nd trial)”.


The results obtained during the trials and presented in the course of the final demonstration, confirmed that the TWOBIAS system had achieved the following results:

• TWOBIAS proof-of-concept
• T&E system for testing at a real life facility
• Improved alarm algorithm for TWOBIAS
• A 2-stage BDU and BIU integrated system
• Detection and identification in an hour
• Sensitive, selective system for fast and reliable identification
• Decision-making system to take actions
• TWOBIAS surveillance system for fast and reliable identification of airborne pathogen(s) within one hour

In the light of these results, further prospects for the development or the use of the TWOBIAS achievements can be considered to increase safety for civilians, especially at high profile public sites regarded as targets for bioterrorist attacks, to increase situational awareness of the bio incident and to improve national and European bio preparedness and response.

TWOBIAS may be used as a part of a networking surveillance system linked together with several similar and/or complementary detection units and air sampling units. TWOBIAS may be suitable for detecting and identifying emerging natural infectious diseases (not only B attacks).

Potential Impact:
A significant impact of the TWOBIAS project has been the technology development where emphasis was on improving the biodetector’s (BDUs) performance characteristics and false alarm rates to improve TWOBIAS, for users at high-decision levels, such as central and local governments and security authorities. The overall impact of TWOBIAS has been to increase safety for civilians, in particular at high profile public sites regarded as targets for bioterrorist attacks. Furthermore, TWOBIAS has taken part in increasing the situational awareness of a biological incident and improved both national and European biological preparedness and response systems.

We hope and believe that the results of the TWOBIAS project will be applicable and lead to further developments of warning systems against biological threat agents released deliberately. This will meet the expressed need to protect critical infrastructure against biological threats, in Europe, but also in the US.
We believe the impact of the project, consortium contributions and results have all contributed to the performance of the TWOBIAS system, to warn of biological threats agents at public sites, which will have an impact on:

• The real-life test at a relevant site and the project results are significant contributions to the societies’ and authorities’ efforts to increase safety for civilians, in particular at high profile public sites regarded as targets for bioterrorist attacks

• The integration of BDU and BIU, proof of concept at the TWOBIAS demonstration, has provided valuable know-how as to the ability to detect and classify biological threat agents (within seconds) and to identify biological threat agents (within one hour), to provide accurate alarm signals to command and control operators, and subsequently to first responders. We have contributed to increased understanding of the natural biological background in operational transport interchange areas, and the metrology to capture the crucial parameters.

• The TWOBIAS system supports and increase first responder effectiveness in terms of reducing time needed to take necessary protective actions and medical counter measurements to reduce casualties, and support the elaboration of appropriate countermeasures.

• Having the support of the TWOBIAS system at high profile public sites will increase situational awareness of a biological incident

• The TWOBIAS consortium aimed at improving national and European biological preparedness and response, and we believe the project and dissemination of results have contributed to this.

• We believe the TWOBIAS system could be developed to further integration into future network of systems. The results achieved and proven from this project could make a foundation for future T&E studies and strategies.

• The results are contributing to bridging the gap between user requirements and an operative and reliable biological detection, surveillance and alarm system, building a basis for further approaches.

• We would also claim that the project has contributed to increased mutual understanding between scientists, engineers and user communities of basic and applied science, production and user requirements regarding biological surveillance systems, to benefit EU nations and the relevant industry.

The following have been the main objectives for the dissemination activities:
• Disseminate results by providing a web site of the project
• Disseminate results through End Users and Associated Users
• Disseminate results by presentations at international targeted conferences
• Further dissemination of results to additional Users through the prime Users LMP, PRM and DSB (the prime Users will provide information to other Users by e.g. presentations to public/ governmental/civil agencies/institutes, by providing a link on their web site to the TWOBIAS web page, by presentations at meetings/seminars and conferences and by networking)
• Demonstration of the BDU/BIU/TWOBIAS capability at a real-life test site – the Prague Muzeum metro station
• Through continuous reports, symposia presentations, scientific papers and a workshop
• A web site describing TWOBIAS throughout the project, disseminating non-restricted results and information obtained

The aim of the Dissemination activities has been to disseminate knowledge, experience and benefits from TWOBIAS to expert groups, system providers, End Users and user communities, through publications, trade press, printed material, web site and conferences.
Dissemination activities have focused on project results and implications for the relevant security sectors, and especially on the defence and security sector in northern and southern parts of Europe.

The main dissemination tasks performed have been the TWOBIAS demonstration and the international Dissemination conference

List of Websites:

The contact details for the beneficiaries:

FFI : Janet-Martha Blatny
DGA : Dimitri Cauna
TCS : Laurens Dudragne
TRT: : Florence Aligne
FOI: : Torbjörn Tjärnhage
QL: : Jan Grawé
SCB: : Josef Brinek
UoU : Mats Nilsson
DGS : Antony Roth