Back to all papers

Machine learning-based differentiation of congenital and adolescent idiopathic scoliosis using routine radiographic parameters.

November 16, 2026pubmed logopapers

Authors

Ates A,Gulsever CI,Ozkara Bilgili C,Kazanc Afsin A,Oke D,Sahin N

Affiliations (3)

  • Hakkari State Hospital, Hakkari, Turkey.
  • Hakkari State Hospital, Hakkari, Turkey. [email protected].
  • Gaziosmanpaşa Eğitim Ve Araştırma Hastanesi, Istanbul, Turkey.

Abstract

Accurate differentiation of congenital scoliosis (CS) from adolescent idiopathic scoliosis (AIS) is important for appropriate diagnostic evaluation and subsequent imaging work-up. We evaluated machine learning models using routinely available demographic and radiographic variables to distinguish CS from AIS. In this retrospective single-center study, 166 pediatric patients with scoliosis who underwent standing radiography and MRI between 2018 and 2025 were included. Etiological classification was based on clinical and imaging findings, including whole-spine MRI. Demographic and conventional radiographic variables were used to develop Logistic Regression, Support Vector Machine (SVM), Random Forest, and k-Nearest Neighbors (k-NN) models. Model performance was assessed using accuracy, sensitivity, specificity, precision, F1-score, receiver operating characteristic area under the curve (ROC-AUC), calibration analysis, decision curve analysis, and SHAP-based interpretability. Of the 166 patients, 96 (57.8%) had AIS and 70 (42.2%) had congenital scoliosis. Logistic Regression achieved the highest overall accuracy (79.4%), whereas Random Forest demonstrated the highest discriminative performance (ROC-AUC, 0.829). Pairwise comparisons using the DeLong test showed no statistically significant differences between the evaluated models. SHAP analysis identified Cobb angle, age, vertebral level, and Nash-Moe rotation grade as the most influential predictors. Calibration performance varied across models, while decision curve analysis demonstrated the greatest net benefit for Logistic Regression. Machine learning models demonstrated moderate diagnostic performance in differentiating congenital scoliosis from AIS. However, their limited sensitivity and potential for false-negative predictions preclude independent clinical use. Larger studies and external validation are required before their potential role as adjunctive decision-support tools can be established.

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.