Weight-Adjusted SBP-Based LM-MLEM Algorithm for a High-Resolution Prototype Compton Camera

Objective: This study aims to address the challenges associated with the traditional image reconstruction algorithms such as the numerous artifacts and poor spatial resolution for the simple back-projection (SBP) method, as well as the lengthy iteration times and extensive noise levels for the list mode maximum likelihood expectation maximization (LM-MLEM) method. A specialized reconstruction algorithm was developed and evaluated using data from a prototype high-precision Compton camera. Methods: The Compton camera incorporates two layers of detectors. The scatter detector consists of an array of $21 \times 21$ GAGG crystals of $0.68 \times 0.68 \times 2 \mathrm{~mm}^{3}$ single-ended readout by an $8 \times 8$ SiPM array with an active pixel area of $2 \times 2 \mathrm{~mm}^{2}$. The absorber detector consists of an array of $23 \times 23$ LYSO crystals of $1.0 \times 1.0 \times 20 \mathrm{~mm}^{3}$ dual-ended readout by $8 \times 8$ SiPM arrays with an active pixel area of $3 \times 3 \mathrm{~mm}^{2}$. The proposed image reconstruction algorithm utilizes an inverse projection technique to incorporate prior image data and employs parallel computing along with a weight adjustment strategy. It is named the weight-adjusted SBP-based LM-MLEM algorithm (WA-SBP-MELM). This facilitates the computation and quantification of the weights of Compton projection rings at varying positions, thereby supporting the iterative calculations required for MLEM. Data acquired with the prototype Compton camera using a 0.3 mm diameter ${ }^{22} \mathrm{Na}$ point source will be used to compare different algorithms. Results: The proposed WA-SBP-MELM algorithm resulted in improvements in spatial resolution, noise performance, and reconstruction speed compared to previous algorithms. A spatial resolution of $\sim 0.9 \mathrm{~mm}$ was achieved for the point source. In the future, data of multiple point sources and phantoms will be acquired to further evaluate the proposed algorithm.