Outcomes and Predictive Modeling in Helmet Therapy for Plagiocephaly

Area of Science:

• Pediatric orthopedics
• Medical artificial intelligence
• Computational biology

Background:

• Deformational plagiocephaly (DP) and deformational brachycephaly (DB) are common in infants.
• Cranial orthosis therapy is a standard treatment for DP and DB.
• Predicting treatment outcomes can optimize therapy and patient management.

Purpose of the Study:

• To develop and validate a machine learning (ML) model.
• To forecast treatment-induced changes in cranial vault asymmetry index (CVAI) and cranial index (CI).
• To aid clinical decision-making for infants receiving cranial orthosis.

Main Methods:

• Retrospective analysis of 6,694 infants with DP and/or DB.
• Utilized serial CVAI and CI measurements from cranial orthosis therapy.
• Employed Extreme Gradient Boosting (XGBoost) for predictive modeling.

Main Results:

• XGBoost models achieved high accuracy in predicting final CVAI (R²=0.794) and CI (R²=0.864).
• Models explained significant variance in treatment outcomes.
• Mean absolute errors were 0.756 for CVAI and 0.936 for CI.

Conclusions:

• A robust ML model can forecast cranial shape changes during orthotic therapy.
• This predictive tool has potential clinical applications.
• Further research is needed to integrate the model into clinical practice and parental counseling.