Prediction of surgery-first approach orthognathic surgery using deep learning models

Area of Science:

• Orthodontics and Oral Surgery
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• The surgery-first approach (SFA) offers benefits like reduced treatment time and earlier aesthetic improvements in orthognathic surgery.
• Selecting between SFA and the orthodontics-first approach (OFA) is crucial for optimal patient outcomes.

Purpose of the Study:

• To develop and evaluate a deep learning model for predicting the treatment modality (SFA vs. OFA) in orthognathic surgery.
• To assess the clinical accuracy and performance of the deep learning model using various imaging inputs.

Main Methods:

• A supervised deep learning model, employing three convolutional neural networks (CNNs), was trained on lateral cephalograms and 3D dental model scans.
• Data comprised 228 skeletal Class III malocclusion patients (114 SFA, 114 OFA).
• An ablation study analyzed the impact of different imaging inputs (cephalogram, maxilla, mandible, combined).

Main Results:

• The model achieved high average validation metrics (accuracy: 0.978, AUROC: 0.998) and testing metrics (accuracy: 0.906, AUROC: 0.952).
• The maxilla image input demonstrated the highest accuracy.
• The lateral cephalogram alone yielded the lowest accuracy.

Conclusions:

• Deep learning offers a novel, accurate method for predicting orthognathic surgery treatment modalities.
• This AI application can accelerate workflows, support clinical decision-making, and enable personalized treatment planning.
• The model shows significant potential for improving efficiency and outcomes in orthognathic surgery.