Simulation of a pixelated Compton camera for actinium-225 imaging.

Research area

Nuclear physics applied to medicine.

Background

Actinium-225 is a radionuclide used in targeted alpha therapy. Imaging the gamma rays emitted by its decay daughters, particularly at 218 and 440 keV, provides information about their distribution within the patient and contributes to estimating the doses absorbed by targeted lesions and healthy tissues. The very low activities involved require imaging systems that combine high detection efficiency with accurate localization.

In a two-stage Compton camera, a usable event may consist of an initial Compton scattering in a detector called the scatterer, followed by the deposition of all the photon’s remaining energy in a second detector called the absorber. The measured energies allow the scattering angle to be calculated using Compton kinematics. The positions of the first interactions in the scatterer and absorber define the photon’s direction between the two detectors and hence the axis of the reconstruction cone. Each event therefore constrains the possible source positions to the surface of a cone, whose apex is the first interaction point in the scatterer and whose opening half-angle is the scattering angle. Combining many such cones allows the spatial distribution of the sources to be reconstructed.

Accurate interaction positioning is therefore essential. In particular, the photon may undergo several Compton scatterings in the absorber before being fully absorbed. If these energy deposits are not resolved, the reconstructed position may represent an average position rather than the first interaction point in the absorber. The cone axis is then incorrectly estimated, even when the photon’s full energy is correctly measured. This error can broaden the distribution of the Angular Resolution Measure (ARM), defined as the difference between the angle derived from the measured energies and the geometrically expected angle for a known source position. It can also increase the tails of this distribution and degrade the reconstructed image. The challenge is therefore to identify the relevant first interaction position in each detector, particularly in the absorber when several successive interactions occur.

The MACACO III camera uses monolithic scintillators coupled to silicon photomultiplier (SiPM) arrays. The interaction position is estimated from the distribution of light collected by the SiPM array. When several energy deposits occur, the superposition of their light signals does not, by itself, allow the first interaction position to be identified unambiguously. A pixelated architecture could provide additional information by resolving energy deposits associated with successive interactions in different detector elements. However, it remains to be determined whether pixelization actually facilitates first-interaction identification and whether the benefits outweigh potential detection losses and the associated readout complexity.

This internship will investigate this alternative for actinium-225 imaging. The relationship between segmentation, energy resolution, spatial information and interaction classification will be studied to compare the performance of a pixelated camera with that of a MACACO III-type monolithic reference architecture.

This internship is part of the European AIDER project. It will contribute to the study of detection and reconstruction approaches suited to actinium-225 imaging, helping to better characterize the capabilities and limitations of pixelated architectures for this application.

The simulations will be carried out using GATE 10 (https://github.com/OpenGATE/opengate) within the TOMORADIO team at CREATIS, which has extensive expertise in using this tool for medical physics simulations.

Description of the proposed work

The main objective is to assess the conditions under which detector pixelization facilitates identification of the first interaction site and improves the trade-off between angular resolution, detection efficiency and electronic simplicity. The work will primarily involve Monte Carlo simulations and analysis of detected events. The planned tasks include:

- Literature review and definition of the comparison: become familiar with the principles of Compton imaging and the characteristics of MACACO III, then define comparable pixelated configurations. The influence of pixelization will initially be investigated using configurations close to the monolithic reference system.
- Modeling and analysis of the available information: simulate the interactions of 218 and 440 keV photons for different pixelation configurations and detector types (scintillators or semiconductors). Events containing one or more energy deposits will be analyzed to determine which information can be extracted from the detector signals.
- First-interaction identification: formulate the classification problem and develop methods to associate the measured signals with the first interaction position. Simple physics-based criteria will be investigated first; more advanced classification methods, including deep learning, may be considered as needed. Performance will be evaluated using the interaction histories provided by the simulations, distinguishing correctly classified, misclassified and rejected ambiguous events.
- Comparative performance evaluation: reconstruct Compton cones from the positions and energies assigned to the interactions after classification, then compare the results with the monolithic reference. The study will examine the ARM distribution, including its width, bias and tails, as well as useful detection efficiency, defined as the fraction of emitted photons yielding a usable event. Electronic simplicity will be assessed through the number of readout channels and the information required for measurement and classification. The impact on reconstructed images may provide a complementary assessment.

The internship should identify the conditions under which a pixelated architecture could offer an advantage over a MACACO III-type monolithic architecture. Particular attention will be paid to the trade-off between improved ARM, the fraction of retained events and instrumental complexity.

Location

Centre Léon Bérard, Lyon, France — CREATIS, TOMORADIO team

Desired skills and experience

- Academic background in nuclear physics applied to medicine.
- Experience with Geant4/GATE would be appreciated.
- Experience in detector development would be appreciated.
- Practical experience with Python for simulation and data analysis.
- Familiarity with TensorFlow or PyTorch would be an asset.

Application

Interested applicants are invited to send the following documents to the contacts below:

- CV;
- cover letter;
- most recent academic transcript.

Maxime JACQUET: maxime.jacquet@creatis.insa-lyon.fr

Zaina HURANI: zaina.hurani@creatis.insa-lyon.fr

David SARRUT: david.sarrut@creatis.insa-lyon.fr

Ane ETXEBESTE: ane.etxebeste@creatis.insa-lyon.fr

Internship allowance

€4.50 per hour, equivalent to approximately €693 per month.

References

[1] System characterization and performance studies with MACACO III Compton camera, Radiation Physics and Chemistry, 208 (2023), 110922. https://doi.org/10.1016/j.radphyschem.2023.110922

[2] S. Liprandi et al., Compton Camera Arrangement With a Monolithic LaBr3(Ce) Scintillator and Pixelated GAGG Detector for Medical Imaging, IEEE Transactions on Radiation and Plasma Medical Sciences, 7 (2023), 764–774. https://doi.org/10.1109/TRPMS.2023.3282838

[3] B. Das et al., Compton Imaging of Ac-225 in Preclinical Phantoms With a 3D-Positioning CZT Camera, IEEE Transactions on Radiation and Plasma Medical Sciences, 10 (2026), 112–125, first published online in 2025. https://doi.org/10.1109/TRPMS.2025.3577212