Multi-Radiomics Features of Super-Resolution MRI for Predicting Minor Amputation in Wagner 3 Diabetic Foot Ulcers

This study aimed to investigate effectiveness of multi-radiomics features (RFs) extracted from super-resolution MRI for predicting minor amputation in patients with Wagner 3 diabetic foot ulcers.

275 eligible patients with Wagner 3 ulcers were retrospectively included and randomly divided into a training set (n = 192) and a test set (n = 83). A deep-transfer learning network was utilized to process coronal proton density weighted images (PDWI) and generate super-resolution PDWI (SRPDWI). RFs were extracted separately from soft tissues and phalanges. Random forest models were then developed to predict minor amputation.

The minor amputation rate was 36.4% for the left foot and 32.2% for the right foot. All random forest models demonstrated excellent performance in the training set, achieving AUC ranging from 0.91 to 0.99. In the test set, the integrated radiomics model (combining RFs extracted from both soft tissue and phalanges) exhibited superior performance, with an AUC of 0.83, sensitivity of 0.71, and specificity of 0.79. Additionally, the inclusion of clinical indicators did not enhance the predictive performance of the integrated radiomics model.

The multi-RFs extracted from SRPDWI demonstrated promising potential in predicting the risk of minor amputation for Wagner 3 diabetic foot ulcers.