Automatic Recognition on Region of Interest for Left Ventricular Myocardium in
SPECT Images Based on Deep Learning
YANG Lian1a, JIA Qiang1b, YANG Zhaohui2, SUN Lifeng3, JI Yingcai2, HOU Yansong4
1. a. Department of Equipment; b. Department of Nuclear Medicine, Tianjin Medical University General Hospital, Tianjin 300052,
China; 2. Beijing Novel Medical Equipment Co., LTD., Beijing 102206, China; 3. CNNC High Energy Equipment (Tianjin) Co.,
LTD., Tianjin 300300, China; 4. Suzhou Novel Intelligent Manufacturing Medical Technology Co., LTD.,
Suzhou Jiangsu 215000, China
Abstract：Objective To design and verify a deep learning-based single-photon emission computed tomography (SPECT) imaging
technique for left ventricle region of interest (ROI). Methods A deep learning algorithm based on neural network and automatic
SPECT image was designed to detect the left ventricular ROI. 30 real myocardial perfusion SPECT tomography reconstruction
data were used for evaluation. Results The automatic myocardial recognition technology based on deep learning proposed in this
paper enhanced the contrast between the myocardial tissue and the surrounding background in the cross-section slice, and accurately
identifies the 3D left ventricular ROI, which could consistent with the results of artificial judgment. Conclusion The fully automatic
object-detection algorithm bases on deep learning described in this paper can automatically and effectively identify the 3D region of
interest of myocardium in SPECT images, which also lays a solid foundation for the quantitative analysis of myocardium in SPECT
nuclear medicine images.
杨练1a，贾强1b，杨朝辉2，孙立风3，纪英财2，侯岩松4. 基于深度学习的SPECT图像左心室心肌感兴趣区自动识别技术[J]. 中国医疗设备, 2023, 38(2): 41-46.
YANG Lian1a, JIA Qiang1b, YANG Zhaohui2, SUN Lifeng3, JI Yingcai2, HOU Yansong4. Automatic Recognition on Region of Interest for Left Ventricular Myocardium in
SPECT Images Based on Deep Learning. China Medical Devices, 2023, 38(2): 41-46.
