Magnetic Resonance Imaging Data Features to Evaluate the Efficacy of Compound Skin Graft for Diabetic Foot

Abstract

This study aimed to analyze the role of magnetic resonance imaging (MRI) data characteristics based on the deep learning algorithm in evaluating the treatment of diabetic foot (DF) with composite skin graft. In this study, 78 patients with DF were randomly rolled into the experimental group (composite skin graft) and control group (autologous skin graft) with 39 patients in each group. MRI scans were performed before and after treatment to compare the changes of experimental observation indicators such as healing time, recurrence rate, and scar score. The results showed that T1-weighted imaging (T1WI) of the scanning sequence was considerably increased in the experimental group after treatment. The signal intensity of fat-suppressed T2-weighted imaging (T2WI) and fat-suppressed T1WI enhancement sequences was considerably decreased $\left(P<0.05\right)$ . In addition, compared with the control group, the recurrence rate, healing time, and scar score in the experimental group were considerably decreased $\left(P<0.05\right)$ . The accuracy, specificity, and sensitivity of MRI imaging information in evaluating the therapeutic effect of DF patients were 85.2%, 89.75%, and 86.47%, respectively. According to the specificity and sensitivity, the subject operating characteristic curve was drawn, and the area under the curve was determined to be 0.838. In summary, MRI image data characteristics based on the deep learning algorithm can provide auxiliary reference information for the efficacy evaluation of compound skin transplantation for DF.