Gu Lab 
Imaging science lab

Our paper has been published in "Physics in Medicine & Biology" on April 27, 2023.
Posted onMay 04,2023

Wen He's latest paper A CNN-based four-layer DOI encoding detector using LYSO and BGO scintillators for small animal PET imaging was published in Physics in Medicine & Biology. This work develops a four-layer PET detector using LYSO and BGO scintillators, with the first two layers decoding 0.99 mm crystal pixels and the third and fourth layers decoding 1.53 mm crystal pixels, achieving a total crystal thickness of 24 mm. The detector successfully distinguishes the depth of interaction (DOI) of gamma photons using the pulse signal characteristics of scintillator crystals and Convolutional Neural Network (CNN) methods, laying a core technological foundation for the development of next-generation high spatial resolution and high sensitivity small animal PET systems.


图片3.png


We propose a CNN-based design for a four-layer high-resolution, high-sensitivity PET detector. The structure is illustrated in Figure 1, with the first two layers utilizing 0.99 mm crystal pixels for enhanced spatial resolution, while the third and fourth layers utilize 1.53 mm crystal pixels. The middle layers employ different scintillator light distribution characteristics to achieve high DOI resolution, ensuring a total crystal thickness of 24 mm for high detection efficiency.


Experimental results showed that the CNN achieves an accuracy of 91% for LYSO events and 81% for BGO events. The trained CNN predicts data from a top-incident radiation source, resulting in clear crystal decoding images and energy spectra. By utilizing the pulse signal characteristics of the crystals and CNN methods, we successfully identifies events from all four crystal layers as well as CLCS events.


This research realizes an innovative four-layer DOI detector based on CNN classification, laying a core technological foundation for the development of next-generation high spatial resolution and high sensitivity small animal PET systems. This study is supported by the National Natural Science Foundation, Guangdong Provincial Basic and Applied Basic Research Fund, and Shenzhen Bay Laboratory.


Source Information:

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