Clinical Report: Transitioning from Standard Full-Spine X-Rays to Risser Stratification
Overview
This study presents a deep-learning model for automatic Risser stratification in adolescent idiopathic scoliosis (AIS) using full-spine radiographs. The model demonstrated high accuracy and improved reading efficiency, potentially enhancing clinical decision-making in AIS management.
Background
Accurate assessment of skeletal maturity is vital for managing adolescent idiopathic scoliosis (AIS), as it influences treatment decisions and risk stratification. Risser staging, a common method for evaluating skeletal maturity, often suffers from reproducibility issues in clinical practice. This study addresses these challenges by employing a deep-learning approach to improve the reliability and efficiency of Risser stratification.
Data Highlights
Metric
Value
AUC
0.938
Accuracy
0.875
Cohen's Kappa
0.748
Mean Reading Time (unaided)
34.7–43.2 s
Mean Reading Time (aided)
25.0–31.8 s
Key Findings
The deep-learning model achieved an AUC of 0.938 for Risser stratification.
Accuracy of the model was reported at 87.5% with a Cohen's kappa of 0.748.
Model assistance reduced reading time by 9.7–11.8 seconds per case.
Improved agreement with expert reference was noted, especially among junior surgeons.
The deep-learning model provides a reliable and efficient method for Risser stratification, which can enhance clinical decision-making in AIS management. By reducing reading time and improving accuracy, this approach may facilitate timely interventions and better patient outcomes.
Conclusion
The study demonstrates that an interpretable deep-learning model can effectively stratify skeletal maturity in AIS using standard radiographs, potentially transforming routine clinical practice.
