Research on kernel reconstruction methods for all-digital PET image reconstruction
Positron Emission Tomography (PET) is an advanced medical imaging technique widely used in clinical diagnosis and medical research. As an upgraded version of the traditional PET system, the all-digital PET system has significant hardware advantages. Its core strength is that it can provide accurate and complete single-event information, thereby enabling convenient filtering of data from different energy windows and laying a foundation for high-quality imaging. However, traditional PET image reconstruction algorithms have obvious limitations and struggle to achieve a balanced optimization between noise and contrast. To address this issue, the Kernel Expectation Maximization (KEM) algorithm has been proposed, inspired by kernel methods in machine learning, which can effectively overcome this technical challenge. Therefore, conducting research on the KEM algorithm based on the all-digital PET system is of great significance for fully exploiting the hardware advantages of the system and further improving PET imaging quality. This paper investigates the implementation of kernel reconstruction methods in the all-digital PET system, focusing on their performance under multiple energy windows and various scan durations, and compares the differences among the Ordered Subsets Expectation Maximization (OSEM) algorithm, the KEM algorithm, and the KEM algorithm with a quadratic penalty term. Simulation results demonstrate that in two typical scenarios of all-digital PET—different energy windows and different scan durations—adopting the KEM algorithm and the KEM algorithm with a quadratic penalty term effectively optimizes key indicators such as the voxel standard deviation (STD) and contrast recovery coefficient (CRC) in the region of interest (ROI), and significantly improves the quality of reconstructed images. The results are expected to provide technical ideas and practical references for the innovative application of all-digital PET in low-count imaging scenarios.