[Development of prediction model for therapeutic efficacy of Jingu Tongluo Granules on acute lateral ankle ligament injuries based on Delta-radiomics].

The efficacy of Jingu Tongluo Granules in treating acute lateral ankle ligament injuries is remarkable, however, accurately assessing its clinical efficacy remains challenging. Therefore, a machine learning-based prediction model for the therapeutic efficacy of Jingu Tongluo Granules in treating acute lateral ankle ligament injuries was developed using Delta-radiomics. The study involved 107 patients with acute anterior talofibular ligament injuries who received treatment for two weeks with topical Jingu Tongluo Granules based on the protection, optimal load, ice, compression, elevation(POLICE) principle. The patients were divided into an excellent efficacy group(71 cases) and a poor efficacy group(36 cases) according to their American Orthopedic Foot and Ankle Society(AOFAS) scores. The data was further divided into a training set and a validation set in 7∶3. Radiomics features strongly correlated with anterior talofibular ligament injury were extracted and selected from transverse ankle magnetic resonance imaging performed before and after treatment. Light gradient boosting machine(LightGBM) classifier was employed to develop the pre-treatment, post-treatment, and Delta-radiomics models. The clinical utility and predictive performance of each model were evaluated using the receiver operating characteristic curve(ROC), decision curve analysis(DCA), and calibration curves. A total of 1 197 radiomics features were extracted, and after feature selection and dimensionality reduction, 4, 13, and 8 radiomics features from pre-treatment, post-treatment, and Delta, respectively, were retained to construct machine learning models using LightGBM. The LightGBM model based on Delta-radiomics features demonstrated the optimal predictive performance, with AUC of 0.917(95%CI[0.856, 0.979]) and 0.863(95%CI[0.729, 0.998]) for the training and validation sets, respectively. Both the DCA and calibration curves indicated favorable prediction performance. The LightGBM machine learning model based on Delta-radiomics is effective in predicting the therapeutic efficacy of Jingu Tongluo Granules in treating acute anterior talofibular ligament injuries.