Mask R-CNN算法在转子间骨折诊断中的应用研究OACSTPCD

Objective To develop a computer aided diagnosis(CAD)tool based on the Mask R-CNN algorithm,aiming to assist less-experienced physicians in the diagnosis of intertrochanteric fractures.Methods A total of 665 cases of intertrochanteric fracture X-ray data were selected as the research subjects.The data were divided into training set,validation set,and testing set in a ratio of 8∶0.5∶1.5.The network model was trained using transfer learning method to develop a CAD tool with capabilities of localization,segmentation,and classification for intertrochanteric fractures.Three resident doctors and three attending physicians were recruited to test the classification performance of the CAD tool.Results The CAD tool achieved an accuracy of 0.867,which was slightly lower compared to the average classification level of attending physicians at 0.888±0.010.With the assistance of the CAD tool,the average accuracy of resident doctors improved from 0.707±0.021 to 0.850±0.015.Although it did not reach the classification level of the attending physicians,the difference was not statistically significant(P=0.179).Conclusion The CAD tools can provide valuable assistance to doctors,aid inexperienced physicians in diagnosis and reduce the occurrence of misdiagnosis by offering effective auxiliary information.