Development and Validation of a Multi-Modal Ensemble Model for Predicting Progression in Idiopathic Scoliosis

Area of Science:

• Orthopedics
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate prediction of idiopathic scoliosis progression is crucial for timely clinical intervention.
• Previous deep learning models showed promise using frontal radiographs alone.
• Improving prediction accuracy requires integrating diverse data sources.

Purpose of the Study:

• To develop and validate a multimodal ensemble model for enhanced prediction of idiopathic scoliosis curve progression.
• To integrate frontal radiographs, lateral radiographs, and clinical data for improved predictive performance.
• To compare the performance of the multimodal model against single-modality prediction models.

Main Methods:

• A retrospective cohort study involving 471 patients with idiopathic scoliosis.
• Input data included initial whole-spine frontal and lateral radiographs and clinical variables (age, sex, Risser sign, baseline Cobb angle).
• A weighted ensemble model combined predictions from multiple deep learning and machine learning models, evaluated using area under the receiver operating characteristic curve (AUC).

Main Results:

• The multimodal ensemble model achieved the highest predictive performance with an AUC of 0.819.
• This integrated model significantly outperformed models using only frontal radiographs (AUC 0.791), lateral radiographs (AUC 0.767), or clinical features (AUC 0.721).
• The ensemble approach demonstrated superior accuracy in predicting scoliosis progression.

Conclusions:

• A multimodal ensemble model integrating radiographic and clinical data significantly improves the prediction of idiopathic scoliosis progression.
• This advanced model offers a more accurate tool for clinical decision-making at the initial patient visit.
• Future research should explore further refinement of multimodal approaches for pediatric orthopedic conditions.