Development of a Compton Camera for Biological and Medical Imaging (SCoKa)

Project description

The proposed Compton camera is designed to reconstruct source distributions in the MeV energy range. At these energies, photons primarily undergo Compton scattering, which makes imaging with conventional devices challenging. The approach in this project is to
develop a Compton camera capable of detecting both the scattered photon and the recoil electron simultaneously.
The imaging system consists of two layers: a scatter layer and an absorber layer. In this setup, PMMA is used as the scattering material, and scintillators are used in the absorption layer. In medical imaging, a radioactive material is introduced into the subject’s body at a target location. The emitted gamma rays interact with the detector material, and the resulting data are used to determine the source location of the gamma particle.
When a gamma photon interacts with the PMMA scattering plane, it produces a Cherenkov electron and a scattered low-energy gamma ray. During this Compton scattering process, part of the photon’s energy is transferred to a free electron, releasing it. As the electron travels through the medium faster than the phase velocity of light, it emits Cherenkov photons forming a conical pattern (Cherenkov cone). The scattered gamma rays are absorbed in the scintillator-based absorption layer and detected using SiPMs.
The Cherenkov photons detected by the SiPMs form circular or elliptical patterns corresponding to conic sections. Each event is processed using traditional edge detection algorithms such as the Hough Transform or RANSAC, as well as machine learning models like CNNs, to extract ellipse parameters such as center coordinates, major and minor axes, and orientation. Using these parameters, the electron trajectory is traced back to the interaction point (P1). The algorithms are adapted to the detector’s spatial resolution, and physical effects such as multiple scattering and partial conic sections are taken into account. The methods are evaluated using Monte Carlo–simulated data and GEANT4 simulations to model realistic detector behavior.
Using the reconstructed scatter point (P1), the absorption point (P2), and the measured scattered gamma energy (E1), a complete Compton cone can be projected.
Traditional reconstruction algorithms such as Simple Back Projection (SBP), Maximum Likelihood Expectation Maximization (MLEM), and Stochastic Origin Ensemble (SOE) are used to determine the source distribution and location of the gamma emitter. In addition, ML/DL-based models are explored for point source reconstruction. The impact of uncertainties arising from detector spatial and energy resolution is analyzed and incorporated into the reconstruction process. These algorithms are evaluated using ideal, simulated, and GEANT4 datasets.

Bullet points for focus points of the project:

● Development of a Compton camera for MeV-range gamma imaging
● Algorithm and hardware co-design for real-time event detection
● Detection and analysis of Cherenkov events in the scatter layer
● Application of traditional and ML-based algorithms for data processing
● Real-time edge and ellipse detection using CNN models
● Image reconstruction using classical and data-driven approaches
● FPGA-based acceleration for low-latency processing

The project is funded by the DFG.