Korean journal of orthodontics · 2026
Development of Leading Enhancement Assistive Planning: A three-dimensional malocclusion classification system for orthodontic treatment with clear aligners.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Lee HeeJeong Jasmine, How Reuben Axel Wee Ming, Gan Kok Beng, Ashari Asma, Syed Mohamed Alizae Marny Fadzlin, Megat Abdul Wahab Rohaya
Аннотация
OBJECTIVE: To develop and validate the Leading Enhancement Assistive Planning (LEAP) system, a deep learning-based tool for automated malocclusion classification from three-dimensional (3D) intraoral scans, integrated into orthodontic computer-aided design (CAD) workflows to support clinical diagnosis and treatment planning.
METHODS: This study used 841 anonymized 3D intraoral scans in standard tessellation language (STL) format, annotated by expert orthodontists, comprising 125 unique multilabel combinations across 4 main classes and 11 binary subclasses. The preprocessing pipeline converted STL meshes into standardized voxel grids, which served as inputs to a 3D convolutional neural network based on a modified EfficientNet architecture. A hierarchical classification architecture enabled simultaneous prediction of main-class and subclass malocclusions. Model performance was assessed using accuracy, precision, recall, and F1-score metrics on a held-out test set. Integration into orthodontic CAD software was demonstrated to provide real-time diagnostic feedback.
RESULTS: The LEAP system achieved robust classification performance across 15 orthodontic labels treated as independent binary classification tasks, with a macro-averaged accuracy of 87.6%, precision of 86.6%, recall of 87.5%, and an F1-score of 87.1%. Hierarchical classification yielded clinically interpretable predictions for complex, overlapping malocclusions. The voxel-based pipeline supported resolutions of up to 256 × 256 × 256 with graphics processing unit acceleration for efficient inference. Integration into orthodontic CAD platforms was successfully demonstrated, enabling automated malocclusion classification at the point of care.
CONCLUSIONS: LEAP is an accurate, efficient, and scalable artificial intelligence-assisted system for classifying malocclusions from 3D dental scans. Its integration into orthodontic CAD software offers standardized, real-time diagnostic support, potentially improving workflow efficiency and consistency. The LEAP system may enhance diagnostic accuracy, reduce variability, and serve as a valuable decision-support tool in orthodontic practice.
METHODS: This study used 841 anonymized 3D intraoral scans in standard tessellation language (STL) format, annotated by expert orthodontists, comprising 125 unique multilabel combinations across 4 main classes and 11 binary subclasses. The preprocessing pipeline converted STL meshes into standardized voxel grids, which served as inputs to a 3D convolutional neural network based on a modified EfficientNet architecture. A hierarchical classification architecture enabled simultaneous prediction of main-class and subclass malocclusions. Model performance was assessed using accuracy, precision, recall, and F1-score metrics on a held-out test set. Integration into orthodontic CAD software was demonstrated to provide real-time diagnostic feedback.
RESULTS: The LEAP system achieved robust classification performance across 15 orthodontic labels treated as independent binary classification tasks, with a macro-averaged accuracy of 87.6%, precision of 86.6%, recall of 87.5%, and an F1-score of 87.1%. Hierarchical classification yielded clinically interpretable predictions for complex, overlapping malocclusions. The voxel-based pipeline supported resolutions of up to 256 × 256 × 256 with graphics processing unit acceleration for efficient inference. Integration into orthodontic CAD platforms was successfully demonstrated, enabling automated malocclusion classification at the point of care.
CONCLUSIONS: LEAP is an accurate, efficient, and scalable artificial intelligence-assisted system for classifying malocclusions from 3D dental scans. Its integration into orthodontic CAD software offers standardized, real-time diagnostic support, potentially improving workflow efficiency and consistency. The LEAP system may enhance diagnostic accuracy, reduce variability, and serve as a valuable decision-support tool in orthodontic practice.