Interleaved signal multiplexing readout in depth encoding Prism-PET detectors.

Area of Science:

• Medical Imaging
• Nuclear Instrumentation
• Detector Physics

Background:

• Clinical Positron Emission Tomography (PET) scanners require signal multiplexing to reduce complexity, power consumption, heat, and cost due to a large number of readout pixels.
• Depth-encoding Prism-PET detector modules offer high resolution but can be complex to readout.

Purpose of the Study:

• To introduce an interleaved multiplexing (iMux) scheme for Prism-PET detector modules.
• To leverage the light-sharing pattern of depth-encoding Prism-PET modules for efficient single-ended readout.

Main Methods:

• Developed an iMux readout scheme connecting four anodes from alternating silicon photomultiplier (SiPM) pixels to a single application-specific integrated circuit (ASIC) channel.
• Utilized a 4-to-1 coupled Prism-PET detector module with a 16x16 array of LYSO scintillator crystals and an 8x8 array of SiPM pixels.
• Employed a deep learning-based demultiplexing model to decode energy signals and evaluated spatial, depth of interaction (DOI), and timing resolutions.

Main Results:

• The deep learning demultiplexing architecture achieved perfect crystal identification with negligible decoding error.
• Average energy resolution was 9.6 ± 1.5% (non-multiplexed) and 10.3 ± 1.6% (multiplexed).
• Average DOI resolution was 2.9 ± 0.9 mm (non-multiplexed) and 2.8 ± 0.8 mm (multiplexed).
• Average timing resolution was 266 ± 19 ps (non-multiplexed) and 311 ± 28 ps (multiplexed).

Conclusions:

• The iMux scheme enables 16-to-1 crystal-to-readout multiplexing in Prism-PET detector modules without significant performance degradation.
• The method reduces the number of readout channels and detector capacitance, enhancing cost-effectiveness and maintaining high resolution.
• The integration of deep learning for demultiplexing is effective for recovering encoded energy signals.