基于DR放射组学预测Ⅰ-Ⅱ期股骨头缺血性坏死的初步研究OA

Objective To predict stageⅠ-Ⅱosteonecrosis of the femoral head(ONFH)through radiomics nomograph based on digital radiography(DR)of the hip joint,so as to expand the application scope of conventional DR in the assessment of phase Ⅰ-Ⅱ ONFH.Methods Sixty-one patients with ONFH and 24 healthy volunteers were selected as the research subjects.All patients and healthy volunteers underwent DR and MRI scans of the hip joint.A total of 1409 radiomics features were extracted from the artificially labeled regions of interest in DR images.Feature selection was carried out using the minimum absolute contraction selection operator regression method to construct a multilayer perceptron(MLP)and support vector machine(SVM).The two machine learning classification models were used for ONFH detection.Combining radiomics scores and independent demographic data,radiomics nomoplots were established through logistic regression analysis,and the diagnostic performance of all models was evaluated by indicators such as the receiver operating characteristic curve and its area under the curve(AUC),accuracy,specificity and sensitivity.Results All the research subjects were randomly divided into the training set(n=58)and the validation set(n=27).In the validation set,the AUC of the MLP and SVM radiomics models was 0.980 and 0.954 respectively,and the AUC of the radiomics nomogram was 0.981.Conclusion Machine learning based on the radiomics characteristics of DR is helpful for screening high-risk populations of stage Ⅰ-Ⅱ ONFH.