Back to all papers

Prediction of Postoperative Neurological Function Recovery in Cervical Spondylotic Myelopathy Using MRI-based Delta-Radiomics Combined with Interpretable Machine Learning Model.

November 17, 2026pubmed logopapers

Authors

Zhao S,Gao X,Zhang Y,Zhang W,Zhang Z,Zhang L

Affiliations (5)

  • School of Medicine, Tianjin University, Tianjin City, 300072, China.
  • Department of Acupuncture and Rehabilitation, Guanghua Branch, Shijiazhuang Traditional Chinese Medicine Hospital, Shijiazhuang City, 050011, China.
  • Department of Radiology and Nuclear Medicine, The First Hospital of HeBei Medical University, ShijiaZhuang City, 050031, China.
  • Department of Radiology and Nuclear Medicine, The First Hospital of HeBei Medical University, ShijiaZhuang City, 050031, China. [email protected].
  • Department of Radiology and Nuclear Medicine, The First Hospital of HeBei Medical University, ShijiaZhuang City, 050031, China. [email protected].

Abstract

Cervical Spondylotic Myelopathy (CSM) presents progressive neurological deterioration and highly variable postoperative outcomes, while existing prognostic assessment methods have inherent limitations. This dual-center study developed an interpretable MRI-based delta-radiomics model for individualized CSM postoperative prognosis evaluation. A total of 237 CSM patients from Center 1 and 96 patients from Center 2 who underwent decompression surgery were retrospectively enrolled. The Center 1 cohort was stratified by clinical outcomes and randomly divided into training (n = 165) and internal testing (n = 72) cohorts at a 7:3 ratio, while the Center 2 cohort served as an independent external validation set (n = 96). Radiomic features were extracted from pre- and postoperative T2WI MRI images, and delta-radiomic features were calculated. LASSO regression and spearman correlation analysis were applied to screen core radiomic features, and clinical-routine imaging features. Three prediction models were constructed via five machine learning algorithms. Model performance was evaluated using AUC, accuracy, sensitivity, specificity, and F1-score, and pairwise comparisons performed via the DeLong and McNemar tests. Calibration curves and radar charts were generated to determine the optimal algorithm for each model. Decision Curve Analysis (DCA) and Shapley Additive Explanations (SHAP) analysis were performed to assess clinical benefit and model interpretability. Baseline data were well balanced. 16 preoperative radiomic features, 15 delta-radiomic features, and 5 clinical-routine imaging features were finally selected. The XGBoost-based delta-radiomic model achieved the highest numerical AUCs in the internal test (0.899) and external validation (0.802) sets, demonstrating a statistically significant improvement in AUC compared to both the preoperative radiomic model and the clinical-routine imaging model. The model maintained a balanced performance profile across cohorts. This model maintained a higher clinical net benefit across wide threshold ranges. SHAP analysis identified wavelet-domain texture features as risk factors for poor prognosis, while gradient-domain kurtosis features showed bidirectional regulatory effects. Overall, the MRI-based delta-radiomic model effectively predicts long-term neurological recovery in CSM patients. The delta-radiomic model exhibits a more balanced performance profile and statistically better generalizability. By utilizing routine short-term postoperative MRI (within 7 days), this interpretable model holds promise as an objective tool for early prediction of 12-month postoperative neurological recovery.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.