The Study On The Technology Of Intelligent Recognition And Classification System Of Acetabular Fracture

Objective:To independently develop an intelligent recognition and classification system for acetabular fractures based on X-ray images,and to evaluate its diagnostic accuracy and efficacy,as well as its clinical application value.Method:The imaging data of 600 samples containing 470 patients with acetabular fracture who were hospitalized from January 1,2011 to May 31,2022 in the experimental group and 130 normal controls were collected.All the imaging data were checked and classified with Judet classification of acetabular fractures by the chief physician of the imaging department and the chief orthopaedic physician.Artificial intelligence learning system model was built,using Tensorflow2 and MobileNetV3 at the Python3.8 platform.All the data were randomly divided into machine learning group and manual detection group,and the machine learning group was further divided into training set and test set randomly according to a certain proportion.The data of the machine learning group was trained and tested by the established artificial intelligence identification and diagnosis system.In the manual detection group,doctors at different levels(resident physicians,chief physicians)made diagnosis alone or with the assistance of the system under the double-blind condition.The diagnosis results,time required for diagnosis and diagnosis coincidence rate were recorded respectively.Difference analysis was conducted with the paired Chi-square test and rank sum test.Results:1)The diagnostic accuracy of the machine learning group in the test set was 86.5%,and the diagnostic accuracy of the intelligent diagnostic system was 85%,with 200 seconds diagnostic time.In the manual test,the diagnostic accuracy of the resident doctor was 52%(the diagnostic time was 792 seconds),and the diagnostic accuracy of the chief physician was 69%(the diagnostic time was 710 seconds).With the assistance of the intelligent diagnostic system,the diagnostic accuracy of resident physicians and chief physicians was 76%and 89%respectively,and the diagnostic time was 590 seconds and 569 seconds,respectively.2)The diagnostic accuracy of acetabulum Judet classification by doctors at all levels was significantly lower than that of artificial intelligence diagnosis system(P<0.05).The diagnostic accuracy of doctors at all levels with the aid of the system was much better than that of manual single diagnosis(P<0.05),and the time required for diagnosis was also less than individual diagnosis significantly.(P<0.05).3)Among all subtypes of Judet acetabulum fractures,the diagnostic accuracy of posterior column fracture and posterior wall fracture was significantly higher than that of other types of fractures,and the diagnostic accuracy of doctors and systems at all levels was more than 80%.The diagnostic accuracy of anterior column with posterior transverse fracture and transverse with posterior wall fracture is lower than that of other fracture types,and the diagnostic accuracy of all levels of physicians is less than 50%.Conclusions:The artificial intelligence classification diagnosis system of acetabulum fracture based on deep learning algorithm can conduct automatic classification prediction and auxiliary diagnosis of acetabulum fracture Judet classification.With the aid of artificial intelligence diagnosis system,the accuracy of acetabular fracture classification is improved significantly,which has certain clinical application value.