A classification system for developmental dysplasia of the hip based on deep learning of hip radiographs
ZHENG Rongwei
ZHANG Yeyong
WANG Gongteng
LIU Qingsheng
WANG Shengli
MA Yanchao
SUN Huaqian
YU Guanzhen
LI Shufeng
Abstract：Objective To develop a recognition and classification system for developmental dysplasia of the hip(DDH)based on AI algorithms.Methods A total of 541 hip radiographs were collected,and 529 radiographs were finally included and divided into normal radiographs and DDH radiographs.DDH radiographs were divided into types Ⅰ,Ⅱ,and Ⅲ according to Hartofilakidis classification.A new classification framework of DDH was designed,and DDH was diagnosed and classified by DL model training and X-ray recognition.At the same time,a new method for the diagnosis of type Ⅲ by locating key points was defined.The trained model was used for test set evaluation,and the probability value of each image classification result was given.Finally,six models were selected for training and the performance was evaluated on the test set to find the best fit.Results For the classification of normal and DDH,the specificity,sensitivity,positive predictive value and negative predictive value of DenseNet in the test set were 98.7％,97.8％,98.7％and 97.8％,respectively.The specificity,sensitivity,positive pre-dictive value and negative predictive value of DDH Ⅲ were 90.0％,87.0％,89.7％and 87.4％,respectively.For DDH Ⅰ and Ⅱ,the specificity,sensitivity,positive predictive value and negative predictive value were 90.4％,86.7％,90％and 87.1％,respectively.Conclusion The AI algorithm can diagnose DDH and classify it with high accuracy by identifying hip X-ray films,which can reach the level of professional doctors.
Keywords：X-ray radiographyDevelopmental dysplasia of the hipDeep learningClassification model
Publication Date：2026-01-30
Online Publishing Date：2026-03-20(First online date of this platform, not the publication date of the document)
Pages：6( 116-120,129 )