Back to all papers

Interpretable machine learning model integrating MRI-derived paraspinal muscle parameters for predicting new vertebral compression fractures after vertebral augmentation.

May 15, 2026pubmed logopapers

Authors

Wang C,Gao M,Ye J,Luo Y,Yang Y,Wu R,Xu Z

Affiliations (6)

  • Medical Imaging Center, The First People's Hospital of Foshan (The Affiliated Foshan Hospital of Southern University of Science and Technology), School of Medicine, Southern University of Science and Technology, Guangdong, China.
  • Department of Radiology, Foshan Hospital of Traditional Chinese Medicine, Guangdong, China.
  • The Eighth Clinical Medical College of Guangzhou University of Chinese Medicine, Guangdong, China.
  • Medical Imaging Center, The Second People's Hospital of Foshan, Guangdong, China.
  • Department of Minimally Invasive Spine and Joint Surgery, The First People's Hospital of Foshan (The Affiliated Foshan Hospital of Southern University of Science and Technology), School of Medicine, Southern University of Science and Technology, Guangdong, China.
  • Medical Imaging Center, The First People's Hospital of Foshan (The Affiliated Foshan Hospital of Southern University of Science and Technology), School of Medicine, Southern University of Science and Technology, Guangdong, China. [email protected].

Abstract

To develop and validate interpretable machine learning (ML) models incorporating MRI-derived paraspinal muscle parameters to predict new vertebral compression fractures (NVCF) after vertebral augmentation. This multicenter retrospective study included patients with osteoporotic vertebral compression fractures (OVCF) who underwent vertebral augmentation with a 2-year follow-up. Data from two centers were merged and randomly divided (8:2) into training and test sets, while data from the third center served as an independent external validation set. Clinical variables, radiographic parameters, and MRI-derived muscle parameters-including paraspinal muscle fat infiltration (PMFI) and psoas muscle index (PMI)-were analyzed. Variable selection was performed using LASSO and logistic regression. Seven supervised ML algorithms were trained and compared using the area under the curve (AUC), calibration, and decision curve analyses. SHAP values were used to enhance the clinical interpretability of the ML models by providing transparent explanations of model predictions. A total of 359 patients (median age, 75 years [IQR 69-80]; 77 men) were included, with NVCF occurring in 133 patients (37.0%). Six independent predictors-age, bone mineral density (BMD), PMFI, PMI, kyphotic angle correction rate (KACR), and Genant semiquantitative grade (GSQ)-were identified. The random forest model showed the best performance, with AUCs of 0.963, 0.913, and 0.837 in the training, test, and external validation sets (n = 53), respectively. SHAP analysis indicated that PMI, PMFI, age, and BMD contributed most to model prediction. This interpretable ML model combining MRI-based muscle metrics with conventional factors achieved strong predictive performance and may assist in personalized management after vertebral augmentation. Question New vertebral compression fractures after vertebral augmentation cause substantial morbidity, yet reliable risk stratification is lacking, and the prognostic value of MRI-derived paraspinal muscle remains unclear. Findings An interpretable random forest model incorporating MRI-derived paraspinal muscle fat infiltration and psoas muscle index accurately predicted new vertebral compression fractures after vertebral augmentation. Clinical relevance Incorporating MRI-derived paraspinal muscle quality into machine learning models improves risk stratification for new vertebral compression fractures, supporting personalized clinical decision-making, targeted follow-up, and early preventive strategies after vertebral augmentation.

Topics

Journal Article

Ready to Sharpen Your Edge?

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

We respect your privacy. Unsubscribe at any time.