Objective: To construct a 3.0T MRI multi-sequence radiomics model for non?invasive prediction of median nerve function recovery in patients with carpal tunnel syndrome (CTS) after surgery. Methods: A total of 47 CTS patients who underwent surgical treatment from March 2023 to November 2025 were retrospectively enrolled. Preoperative 3.0T MRI multi sequence examinations were performed, including axial proton density weighted imaging fat suppression sequence (PD-FS), fast spin echo T2 weighted imaging (FSE-T2WI), coronal PD-FS, fast spin echo T1 weighted imaging (FSE-T1WI), and sagittal PD-FS. Manually delineate the region of interest (ROI) of the median nerve within the carpal tunnel, extract radiomics features, and construct a multivariate logistic regression prediction model after stability screening (ICC>0.80), low variance screening (variance>10-4), and LASSO regression (P<0.05). The Boston carpal tunnel syndrome score (BCTQ) was defined as good recovery one month after surgery, and the model efficacy was evaluated through ROC curve, calibration curve, and decision curve analysis.. Model performance was evaluated using ROC curves, calibration curves, and decision-curve analysis. Results: A total of 6 optimal features were selected. Multivariate analysis showed that wavelet-transformed GLCM correlation, first-order entropy, GLRLM long-run emphasis, and median nerve swelling ratio were independent predictive factors (P<0.05). The constructed model achieved an AUC of 0.892. Conclusion: The 3.0T MRI multi-sequence radiomics model can effectively predict postoperative median nerve function recovery in CTS, demonstrating high clinical value.