Abstract:Objective To analyze the impact of CT dose on artificial intelligence algorithm. Methods Clinical thoracic CT images of
pulmonary nodules were collected as reference. Different levels of Gaussian noises were added to the reference to simulate low dose
CT images, according to mathematical model. The simulated images were then processed by computer aided detection algorithms,
the output of which was monitored and analyzed. Results The result of the algorithm showed significant fluctuation. The maximum
relative decrease of recall was 73.2%, and the maximum relative decrease of precision was 70.2%. Conclusion The dose during CT
imaging directly impacts image quality and may further lead to significant algorithm degradation. It may be necessary to fix this issue
during research and development and improve the generalizability of AI algorithms.
张超,王浩,孟祥峰,王权,李宁,任海萍. CT成像剂量对人工智能算法性能的影响分析[J]. 中国医疗设备, 2020, 35(4): 68-70.
ZHANG Chao, WANG Hao, MENG Xiangfeng, WANG Quan, LI Ning, REN Haiping. Impact of CT Dose on Performance of Artificial Intelligence Algorithm. China Medical Devices, 2020, 35(4): 68-70.