Predictive Value of Baseline CT Radiomics for Jianpibushen Prescription Efficacy in Femoral Head Necrosis

Abstract Objectives To explore the predictive value of baseline CT radiomics for the 6-month and 12-month treatment efficacy of the Jianpibushen Prescription in femoral head necrosis (FHN), with the goal of optimizing treatment strategies. Methods Retrospectively, ARCO stage 2—4 FHN patients who underwent hip joint CT scans before receiving Jianpibushen Prescription treatment from September 2016 to December 2023 were collected. 315 patients (M/F = 210/105, median age 39.0 years) were included. A total of 1928 radiomics features were extracted, downscaled and filtered. Finally, features were selected to construct the radiomics predictive model of the efficacy at 6 and 12 months. Results For predicting the treatment efficacy at 6 months, eight features were selected to build model using Bootstrap Aggregating Decision Tree (Bagging). The model attained an AUC of 0.999 (0.997—1.0) in the training set and 0.736 (0.638—0.834) in the validation set. For predicting the 12-month treatment efficacy, a comparable radiomics model was constructed with Random Forest, with AUCs of 0.995 (0.991—0.999) in the training set and 0.783 (0.676—0.89) in the validation set. Conclusion Baseline CT radiomics features can relatively accurately predict the 6-month and 12-month efficacy of Jianpibushen Prescription, thus facilitating individualized and precise clinical treatment. ADVANCES IN KNOWLEDGE For the first time, this study established a relatively accurate prediction model for the 6-month and 12-month efficacy of the Jianpibushen Prescription on FHN, based on baseline CT radiomics features, thus optimizing treatment strategies.