TOF MLEM Adaptation for the Total-Body J-PET With a Realistic Analytical System Response Matrix

We report a study of the original image reconstruction algorithm based on the time-of-flight maximum-likelihood expectation-maximization (TOF MLEM), developed for the total-body (TB) Jagiellonian PET (J-PET) scanners. The method is applicable to generic cylindrical or modular multilayer layouts and is extendable to multiphoton imaging. The system response matrix (SRM) is represented as a set of analytical functions, uniquely defined for each pair of plastic scintillator strips used for the detection. A realistic resolution model (RM) in detector space is derived from fitting the Monte Carlo simulated emissions and detections of annihilation photons on oblique transverse planes. Additional kernels embedded in SRM account for time-of-flight (TOF), parallax effect, and axial smearing. The algorithm was tested on datasets, simulated in GATE for the NEMA IEC and static extended cardiac-torso phantoms inside a 24-module 2-layer TB J-PET. Compared to the reference TOF MLEM with none or a shift-invariant RM, an improvement was observed, as evaluated by the analysis of image quality, difference images, and ground-truth metrics. We also reconstructed the data with additive contributions, prefiltered geometrically and with non-TOF scatter correction applied. Despite some deterioration, the obtained results still capitalize on the realistic RM with better edge preservation and superior ground-truth metrics. The envisioned prospects of the TOF MLEM with analytical SRM include its application in multiphoton imaging and further upgrade to account for the noncollinearity, positron range, and other factors.