A CNN-based four-layer DOI encoding detector using LYSO and BGO scintillators for small animal PET imaging

Area of Science:

• Medical Imaging
• Nuclear Physics
• Detector Technology

Background:

• Positron Emission Tomography (PET) is crucial for small animal research.
• High sensitivity and spatial resolution are key for advanced PET imaging.
• Depth-of-Interaction (DOI) encoding improves PET scanner performance.

Purpose of the Study:

• To develop and evaluate a novel four-layer DOI encoding phoswich detector.
• To enhance sensitivity and spatial resolution in small animal PET imaging.
• To utilize LYSO and BGO scintillator crystals for improved performance.

Main Methods:

• Constructed a four-layer detector with alternating LYSO and BGO crystal arrays.
• Employed pulse energy and time-over-threshold for event separation.
• Utilized convolutional neural networks (CNNs) for layer identification.
• Integrated with a multi-pixel photon counter (MPPC) array and TOFPET2 ASIC.

Main Results:

• Successfully identified events from all four detector layers.
• Achieved 91% classification accuracy for LYSO layers and 81% for BGO layers.
• Reported energy resolutions of 12.3-13.1% (LYSO) and 33.9-34.0% (BGO).
• Demonstrated timing resolutions between 328 ps and 2.8 ns for individual layers.

Conclusions:

• The proposed four-layer DOI detector demonstrates high performance.
• This detector is a promising candidate for next-generation small animal PET systems.
• Offers potential for improved sensitivity and spatial resolution in preclinical imaging.