An attempt to analyze facial photographs of patients with jaw deformity using artificial intelligence

Abstract

Purpose

Here, the feasibility of orthodontic surgery was determined by obtaining cephalometric and model analyses. To make predictions easier without using these methods, I examined a plan in which AI judged whether orthodontic surgery was possible.

Methods

This study was approved by the bioethics committee of our hospital (No. I459). We included 1766 patients with suspected jaw deformity among those who visited the outpatient clinic of the Department of Oral and Maxillofacial Surgery in Kanazawa Medical University from August 2004 to October 2019. I used medical records to divide patients into those who underwent orthodontic surgery and those who underwent conservative orthodontic treatment. Side profile photographs of the patients were used to train the AI system, and the untrained data were analyzed by AI.

Results

The AI judgments for surgical orthodontic surgery of jaw deformity in patients aged 16 years or younger, aged 17 years or older, and of all ages were 0.630, 0.668, and 0.768 (ROC curves), respectively.

Conclusion

The AI judgement of orthodontic surgery for jaw deformities in patients has high accuracy. Further improvements will require a larger number of subjects, the use of subject scope, image processing methods, and a loss function system.