Machine Learning Effectively Diagnoses Mandibular Deformity Using Three-Dimensional Landmarks.

Area of Science:

• Oral and Maxillofacial Surgery
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Objective diagnosis of jaw deformities is crucial.
• Current cephalometric methods have limitations in accuracy.
• Machine learning shows promise for diagnosing 2D jaw deformities.

Purpose of the Study:

• To develop a 3D machine learning model for diagnosing mandibular retrognathism and prognathism.
• To compare the diagnostic performance of the 3D model against traditional cephalometric techniques.

Main Methods:

• An in-silico study using deidentified retrospective patient data with 3D facial models.
• Development of a multilayer perceptron (MLP) model using 50 3D facial landmarks.
• Comparison of MLP model accuracy against SNB angle, facial angle, and mandibular unit length (MdUL).

Main Results:

• The MLP model achieved a diagnostic accuracy of 85.2%.
• Traditional methods (SNB, facial angle, MdUL) had accuracies of 74.3%, 74.3%, and 75.3%, respectively.
• The MLP model demonstrated significantly better performance (P < .05) and moderate agreement with the gold standard, surpassing fair agreement of traditional methods.

Conclusions:

• A 3D multilayer perceptron model offers superior accuracy for diagnosing jaw deformities compared to conventional cephalometry.
• This AI-driven approach has the potential to enhance diagnostic precision in orthodontics and maxillofacial surgery.
• Further research can explore integrating 3D MLP models into routine clinical practice for improved patient outcomes.