Future prediction of children using medical images in pediatric dentistry

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K12725
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 90130:Medical systems-related
• Research Institution: Osaka University
• Principal Investigator: Kokomoto Kazuma 大阪大学, 歯学部附属病院, 特任助教(常勤) (00803107)
• Co-Investigator(Kenkyū-buntansha): 野崎 一徳 大阪大学, 歯学部附属病院, 准教授 (40379110) ; 大川 玲奈 大阪大学, 大学院歯学研究科, 准教授 (80437384)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 小児歯科 / 成長予測 / 人工知能 / 敵対的生成ネットワーク

Outline of Final Research Achievements

First, by combining Scaled-YOLOv4 and EfficientNet V2-M, we successfully detected tooth germs and identified their developmental stages from panoramic radiographs, thus achieving automated dental age calculation. Next, using PGGAN for intraoral image generation, we found that the generated images with resolutions of 512×512 pixels or lower were of such high quality that it was difficult to distinguish them from real images. We explored the properties of the latent space and demonstrated smooth transitions in the generation of images representing the primary, mixed and permanent dentition stages. Finally, by using StyleGAN-XL for growth prediction based on real images, we were able to smoothly reproduce the development process from primary to permanent teeth, suggesting its potential usefulness in predicting tooth development stages and eruption.

Free Research Field

小児歯科

Academic Significance and Societal Importance of the Research Achievements

本研究の成果から、小児歯科診療における診断と治療計画の精度および効率をAIによって大幅に向上させられる可能性が示された。これらのAIにより、専門医の少ない地域や一般の歯科医師でも高品質な小児歯科診療が提供できるようになると考えられる。また、歯科領域における画像生成はほとんど研究されていなかったため、当該領域の可能性を広げるとともに、医療での活用法についても新たな知見も得ることができ、他の医療分野への応用も期待される。