The application of deep learning in early enamel demineralization detection DOI Creative Commons

Ketai He,

Rongxiu Zhang,

Muchun Liang

и другие.

PeerJ, Год журнала: 2025, Номер 13, С. e18593 - e18593

Опубликована: Янв. 2, 2025

Objective The study aims to develop a diagnostic model using intraoral photographs accurately detect and classify early detection of enamel demineralization on tooth surfaces. Methods A retrospective analysis was conducted with 208 patients aged 14 44. total 624 high-quality digital images captured under standardized conditions were used construct deep learning based the Mask region-based convolutional neural network (Mask R-CNN). trained automate demineralization. Its performance compared two junior dentists’ abilities. Results achieved an F1-score 0.856 for detecting demineralized teeth validation set, metric that reflects comprehensive performance, demonstrating close senior dentists. With model’s assistance, average F1-scores improved significantly—from 0.713 0.689 0.897 0.949, respectively ( p < 0.05). segmented surfaces detected areas, allowing precise areas monitoring lesion progression. Conclusion Deep can segment contours, enhancing precision, accuracy, efficiency diagnosis area delineation.

Язык: Английский

The application of deep learning in early enamel demineralization detection DOI Creative Commons

Ketai He,

Rongxiu Zhang,

Muchun Liang

и другие.

PeerJ, Год журнала: 2025, Номер 13, С. e18593 - e18593

Опубликована: Янв. 2, 2025

Objective The study aims to develop a diagnostic model using intraoral photographs accurately detect and classify early detection of enamel demineralization on tooth surfaces. Methods A retrospective analysis was conducted with 208 patients aged 14 44. total 624 high-quality digital images captured under standardized conditions were used construct deep learning based the Mask region-based convolutional neural network (Mask R-CNN). trained automate demineralization. Its performance compared two junior dentists’ abilities. Results achieved an F1-score 0.856 for detecting demineralized teeth validation set, metric that reflects comprehensive performance, demonstrating close senior dentists. With model’s assistance, average F1-scores improved significantly—from 0.713 0.689 0.897 0.949, respectively ( p < 0.05). segmented surfaces detected areas, allowing precise areas monitoring lesion progression. Conclusion Deep can segment contours, enhancing precision, accuracy, efficiency diagnosis area delineation.

Язык: Английский

Процитировано

0