Predicting orthognathic surgery results as postoperative lateral cephalograms using graph neural networks and...

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Oral and maxillofacial surgery

Background:

• Orthognathic surgery corrects dentofacial deformities and is increasingly used for cosmetic enhancement.
• Accurate prediction of surgical outcomes is crucial for treatment planning and patient satisfaction.

Purpose of the Study:

• To introduce GPOSC-Net, a generative AI model for predicting post-operative outcomes in orthognathic surgery.
• To synthesize realistic post-operative lateral cephalograms from pre-operative data.

Main Methods:

• GPOSC-Net utilizes a two-component architecture: a landmark prediction model and a latent diffusion model.
• The model predicts post-surgical cephalometric changes and generates synthetic cephalograms based on landmarks and profile lines.

Main Results:

• The model demonstrated accurate prediction of cephalometric landmark positions.
• High-fidelity synthesized post-operative lateral cephalogram images were generated.
• Validation included diverse patient datasets, a visual Turing test, and simulation studies.

Conclusions:

• GPOSC-Net offers a valuable tool for enhancing the accuracy and visualization of orthognathic surgery planning.
• The model has the potential to improve clinical decision-making and patient communication regarding surgical outcomes.