Developing Machine Learning-Based Predictive Models for Hallux Valgus Recurrence Based on Measurements From...

Area of Science:

• Orthopedic surgery
• Medical imaging
• Machine learning

Background:

• Hallux valgus (HV) recurrence after surgery is a clinical concern.
• Predicting HV recurrence is crucial for surgical success and patient outcomes.
• Machine learning (ML) offers potential for prognostic modeling in orthopedic conditions.

Purpose of the Study:

• To develop and validate ML models for predicting hallux valgus recurrence post-surgery.
• To identify key predictors associated with hallux valgus recurrence.
• To enhance surgical decision-making by identifying patients at higher risk of recurrence.

Main Methods:

• Retrospective analysis of 198 feet undergoing chevron osteotomy and distal soft tissue procedure.
• Development and external validation of 9 ML models using preoperative and postoperative radiographic data.
• Feature selection via bootstrap validation to identify significant predictors of recurrence.

Main Results:

• The Support Vector Machine (SVM) model achieved the highest predictive accuracy (AUC 0.88, accuracy 75.6%).
• Key predictors identified include preoperative hallux valgus angle, tibial sesamoid position, and postoperative intermetatarsal angle.
• ML models successfully predicted HV recurrence at 1-year follow-up.

Conclusions:

• ML classifiers, particularly SVM, can effectively predict hallux valgus recurrence.
• The study identifies crucial radiographic predictors for HV recurrence.
• This predictive capability can assist foot and ankle surgeons in managing patients at risk of post-surgical HV recurrence.