Periodic Reporting for period 1 - HIPPOCRATE (Hybrid Imaging of PET and PrOmpt gamma for preCision RAnge- and biological- guidance in proton ThErapy)
Reporting period: 2017-06-01 to 2019-05-31
HIPPOCRATE’s main objective was the design optimization of an innovative hybrid medical imaging device that combines Positron Emission Tomography (PET) and prompt gamma imaging in the scope of high-precision proton therapy range monitoring. The development of such device contributes to the primary European policy goal in the scope of Health which is the promotion of longer and healthier lives for its citizens through innovative and revolutionary imaging and therapeutic approaches for cancer with minimal side effects.
The main problem being addressed in HIPPOCRATE is the necessity of the Compton sensitivity matrix, i.e. the probability that a gamma-ray undergoes Compton scattering between the scatterer and the absorber detector, as an indispensable set of values that corrects the Compton image reconstruction voxel by voxel and poses a requirement for conducting research on the design of a Compton and/or hybrid medical imaging device (‘γ-PET’).
HIPPOCRATE’s foremost priority was therefore given to the crucial task of developing an innovative method to calculate the Compton sensitivity matrix in a precise and time efficient manner. The validation of the novel method is presented in the final report, along with the technological quality of the method which is also supported by the successful image reconstruction of laboratory data using the Compton sensitivity matrix as calculated from the presented method.
The main problem being addressed in HIPPOCRATE is the necessity of the Compton sensitivity matrix, i.e. the probability that a gamma-ray undergoes Compton scattering between the scatterer and the absorber detector, as an indispensable set of values that corrects the Compton image reconstruction voxel by voxel and poses a requirement for conducting research on the design of a Compton and/or hybrid medical imaging device (‘γ-PET’).
HIPPOCRATE’s foremost priority was therefore given to the crucial task of developing an innovative method to calculate the Compton sensitivity matrix in a precise and time efficient manner. The validation of the novel method is presented in the final report, along with the technological quality of the method which is also supported by the successful image reconstruction of laboratory data using the Compton sensitivity matrix as calculated from the presented method.
At the starting point of the project, the initial task was the familiarization with the MEGAlib software (Medium-Energy Gamma-Ray Astronomy Library) for the modelling of the detectors and their characteristics, the performance of Monte Carlo simulations, the implementation of Compton event classification algorithms and Compton image reconstruction.
Due to the fact that MEGAlib was initially only designed to perform Compton image reconstruction, in order to being able to combine and make most of the two technologies, i.e. ‘PET’ and prompt gamma imaging, the MEGAlib software was extended for its ability to distinguish, reconstruct and image gamma-rays coming from Compton, PET and/or Compton-PET events. At that stage, extensive Monte Carlo simulations eventually took place for the optimization and characterization of the proposed ‘γ-PET’ design in terms of distances and for the decision on the set-up and prediction of the detectors performance of an in-beam experiment using light ions1, conducted at the GSI Helmholtzzentrum für Schwerionenforschung in synergy to the EU-funded project BARB (https://www.gsi.de/work/forschung/biophysik/erc_barb(opens in new window)).
In parallel with the above mentioned simulation tasks, the image reconstruction of data acquired using radioactive point sources (Na-22) and a single Compton Arm system initially built at the laboratory of the host institute showed a clear need for the application of an efficiency matrix in order to correct for the images (elongation of source and position misalignment) and directed us to the conclusion that the essential requirement for the image reconstruction of the ‘γ-PET’ is the calculation of the Compton detection efficiency.
Following the above mentioned conclusion the focus was given on the calculation of the Compton sensitivity matrix and its application during image reconstruction. The first approach to tackle the sensitivity matrix problem was done via a model approximation, which was however proven to be inappropriate to describe our system. Due to time limitations for studying a different model, a novel method was developed that was based on a selected number of Monte Carlo simulations and not on the full calculation. This method was proven able to successfully extrapolate the sensitivity values for the omitted Monte Carlo simulations: the extrapolated sensitivity values reached to an accuracy in the range of 99-100% of the sensitivity values calculated from the simulations, validating the method. The technical quality of the presented method to calculate the Compton sensitivity matrix was also supported by the successful image reconstruction for the experimental radioactive source data acquired with the above mentioned single Compton arm system: the misalignment of the source as well as the elongations observed in the image reconstruction of the source without a matrix implementation were removed when applying the developed sensitivity matrix (please see attached figure).
The preliminary results of this project were communicated to Prof. Dr. Magdalena Rafecas and members of her group at the University of Lübeck as part of a collaboration approach on Compton camera development between the research groups. The preliminary results of this project were also communicated at a seminar talk of the Chair of Medical Physics at the host institute. There have not been any publications submitted yet since the final step, which was the image reconstruction after applying the produced Compton sensitivity matrix, was only completed a few weeks before the submission of the final report. However, exploitation of the results and their dissemination is planned for the next months.
1Boscolo D. et. al. (2021). ‘Radioactive Beams on Image-Guided Particle Therapy: The BARB Experiment at GSI. Frontiers in Oncology doi:10.3389.fonc.2021.737050
Due to the fact that MEGAlib was initially only designed to perform Compton image reconstruction, in order to being able to combine and make most of the two technologies, i.e. ‘PET’ and prompt gamma imaging, the MEGAlib software was extended for its ability to distinguish, reconstruct and image gamma-rays coming from Compton, PET and/or Compton-PET events. At that stage, extensive Monte Carlo simulations eventually took place for the optimization and characterization of the proposed ‘γ-PET’ design in terms of distances and for the decision on the set-up and prediction of the detectors performance of an in-beam experiment using light ions1, conducted at the GSI Helmholtzzentrum für Schwerionenforschung in synergy to the EU-funded project BARB (https://www.gsi.de/work/forschung/biophysik/erc_barb(opens in new window)).
In parallel with the above mentioned simulation tasks, the image reconstruction of data acquired using radioactive point sources (Na-22) and a single Compton Arm system initially built at the laboratory of the host institute showed a clear need for the application of an efficiency matrix in order to correct for the images (elongation of source and position misalignment) and directed us to the conclusion that the essential requirement for the image reconstruction of the ‘γ-PET’ is the calculation of the Compton detection efficiency.
Following the above mentioned conclusion the focus was given on the calculation of the Compton sensitivity matrix and its application during image reconstruction. The first approach to tackle the sensitivity matrix problem was done via a model approximation, which was however proven to be inappropriate to describe our system. Due to time limitations for studying a different model, a novel method was developed that was based on a selected number of Monte Carlo simulations and not on the full calculation. This method was proven able to successfully extrapolate the sensitivity values for the omitted Monte Carlo simulations: the extrapolated sensitivity values reached to an accuracy in the range of 99-100% of the sensitivity values calculated from the simulations, validating the method. The technical quality of the presented method to calculate the Compton sensitivity matrix was also supported by the successful image reconstruction for the experimental radioactive source data acquired with the above mentioned single Compton arm system: the misalignment of the source as well as the elongations observed in the image reconstruction of the source without a matrix implementation were removed when applying the developed sensitivity matrix (please see attached figure).
The preliminary results of this project were communicated to Prof. Dr. Magdalena Rafecas and members of her group at the University of Lübeck as part of a collaboration approach on Compton camera development between the research groups. The preliminary results of this project were also communicated at a seminar talk of the Chair of Medical Physics at the host institute. There have not been any publications submitted yet since the final step, which was the image reconstruction after applying the produced Compton sensitivity matrix, was only completed a few weeks before the submission of the final report. However, exploitation of the results and their dissemination is planned for the next months.
1Boscolo D. et. al. (2021). ‘Radioactive Beams on Image-Guided Particle Therapy: The BARB Experiment at GSI. Frontiers in Oncology doi:10.3389.fonc.2021.737050
The developed method to extrapolate the sensitivity values is a novel method which as far as we know has not yet been attempted in the State of the Art. The usual methods to extract the sensitivity are either based on a full Monte Carlo simulation or a model approximation which is less accurate than the Monte Carlo simulations. The presented method, instead, provides accurate results in a short amount of time, allowing to use only a 3% of the simulation sets required for a full Monte Carlo simulation.
We believe that the first attempts to reconstruct the laboratory data using the Compton sensitivity matrix as calculated from the presented method show promising results which support the technological quality of the method developed in the project, and create a basis and a structure for continuing research in the development of a complete Compton camera and more particularly a ‘γ-PET’ system, contributing to the primary European policy goal in the scope of Health which is the promotion of longer and healthier lives for its citizens through innovative and revolutionary imaging and therapeutic approaches for cancer with minimal side effects. In addition to this, the work carried out in this project brings innovation capacity not only in the field of medical imaging but also in other scientific fields that make use of Monte Carlo simulations. The ability of being able to use a selected number of Monte Carlo simulations and thus reducing the computational time and being able to extrapolate the results with high accuracy opens a road for newly developed applications.
We believe that the first attempts to reconstruct the laboratory data using the Compton sensitivity matrix as calculated from the presented method show promising results which support the technological quality of the method developed in the project, and create a basis and a structure for continuing research in the development of a complete Compton camera and more particularly a ‘γ-PET’ system, contributing to the primary European policy goal in the scope of Health which is the promotion of longer and healthier lives for its citizens through innovative and revolutionary imaging and therapeutic approaches for cancer with minimal side effects. In addition to this, the work carried out in this project brings innovation capacity not only in the field of medical imaging but also in other scientific fields that make use of Monte Carlo simulations. The ability of being able to use a selected number of Monte Carlo simulations and thus reducing the computational time and being able to extrapolate the results with high accuracy opens a road for newly developed applications.