An Automated Dental Bracket Positioning Model: A Tailored AI Solution for Clinical Orthodontic Practice

Orthodontic bracket placement is a crucial step in fixed orthodontic treatments, requiring high precision and significant clinical experience. The manual nature of this procedure makes it time-consuming and subject to inter-operator variability. This thesis presents a deep learning–based framework aimed at automating the prediction of orthodontic bracket installation points directly from three-dimensional dental models. The proposed approach relies on a pipeline that combines semantic segmentation and landmark prediction to identify individual teeth and extract anatomically meaningful features necessary for accurate bracket positioning. The framework operates on patient-specific STL files acquired from a specialized orthodontic studio and leverages expert-provided annotations as ground truth for supervised learning. By exploiting geometric and spatial information from 3D dental meshes, the system is able to predict bracket placement locations in a consistent and reproducible manner across different patients. The architecture is designed to be computationally efficient, enabling fast inference times without compromising prediction quality. Experimental results show that the proposed method achieves reliable performance while maintaining low computational cost, making it suitable for integration into real-world clinical workflows. The overall framework is modular and production-ready, allowing seamless deployment in orthodontic software pipelines. This work demonstrates the potential of deep learning techniques to support orthodontic specialists by reducing manual workload, improving standardization, and enabling scalable automation of bracket placement procedures.