AI Detection Efficiency of Pulmonary Nodules Under Dual-source CT with Different Tube Voltages

Objective: To explore the detection of misdiagnosis and missed pulmonary nodules by artificial intelligence assisted diagnosis system under dual-source CT with different tube voltages. Methods: a retrospective collection of 200 outpatient patients who underwent dual-source CT was conducted. Images were screened and 198 qualified images were finally obtained. The images were post-processed to obtain chest CT images under 100kVp, 120kVp and 140kVp. The number of false positives and false negatives of pulmonary nodules detected by artificial intelligence under different tube voltages were compared according to the size, density and location of the nodules. Results: AI had better resolution ability for ground glass nodule under dual-source CT 100kVp. In the dual-source CT fusion image of 120kVp, the misdiagnosis rate of pulmonary nodules was the highest, but the rate of missed diagnosis was lower. However, the automatic detection of pulmonary nodules was the least effective under dual-source CT 140kVp. Conclusion: The false negative rate of artificial intelligence detection of pulmonary nodules under 120kVp fusion is low, which can reduce the rate of missed diagnosis of pulmonary nodules by physicians.