View in EDS

TransFArchNet: Predicting dental arch curves based on facial point clouds in personalized panoramic X-ray imaging.

Saved in:
Bibliographic Details
Title: TransFArchNet: Predicting dental arch curves based on facial point clouds in personalized panoramic X-ray imaging.
Authors: Lin, Guoye1,2,3 (AUTHOR) linguoye123@i.smu.edu.cn, Yang, Shuo1,4 (AUTHOR) alex2005191007@smu.edu.cn, Chen, Yangfan1,2,3 (AUTHOR), Xu, Qing1,2,3 (AUTHOR), Yap, Pew-Thian5 (AUTHOR), Yun, Zhaoqiang1,2,3 (AUTHOR) yun1982@smu.edu.cn, Feng, Qianjin1,2,3 (AUTHOR) fengqj99@smu.edu.cn
Source: Expert Systems with Applications. Apr2025, Vol. 270, pN.PAG-N.PAG. 1p.
Subjects: Dental arch, Cone beam computed tomography, X-ray imaging, Point cloud, Feature selection
Abstract: Traditional panoramic X-ray imaging, constrained by a fixed scanning path aligned with a preset dental arch curve, often results in image distortions such as blurring, magnification, or shrinking, particularly when structures lie outside the focal layer, leading to diagnostic inaccuracies. To address this, we propose TransFArchNet, a novel personalized panoramic scanning solution that predicts individualized dental arch curves from scanned facial point clouds. Utilizing multi-scale transformer layers and feature selection modules, the network is able to learn both global and local facial geometric. Our method is robust, incorporating a noise-filtering step, and an auxiliary confidence network that emphasizes stable facial points to improve curve-fitting accuracy, particularly for jawbone-located dental arches, distant from the soft-tissue facial point clouds. Experimental validation using cone-beam computed tomography (CBCT)-derived facial point clouds shows that TransFArchNet achieves a mean 2D prediction error of 1.50 mm, outperforming PointNet++'s error of 1.67 mm and smaller than the dimensions of a small tooth. Besides, the promising results on real facial predictions suggest potential clinical applications. Our code is available at: https://github.com/lancesye/TransFArchNet. • A novel solution for personalized panoramic imaging. • Transformer layers are utilized to learn individual dental arches from facial point clouds. • Include noise-filtering and confidence networks for robust curve fitting. • Achieves a mean 2D dental arch prediction error of 1.50 mm. [ABSTRACT FROM AUTHOR]
Copyright of Expert Systems with Applications is the property of Pergamon Press - An Imprint of Elsevier Science and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:Traditional panoramic X-ray imaging, constrained by a fixed scanning path aligned with a preset dental arch curve, often results in image distortions such as blurring, magnification, or shrinking, particularly when structures lie outside the focal layer, leading to diagnostic inaccuracies. To address this, we propose TransFArchNet, a novel personalized panoramic scanning solution that predicts individualized dental arch curves from scanned facial point clouds. Utilizing multi-scale transformer layers and feature selection modules, the network is able to learn both global and local facial geometric. Our method is robust, incorporating a noise-filtering step, and an auxiliary confidence network that emphasizes stable facial points to improve curve-fitting accuracy, particularly for jawbone-located dental arches, distant from the soft-tissue facial point clouds. Experimental validation using cone-beam computed tomography (CBCT)-derived facial point clouds shows that TransFArchNet achieves a mean 2D prediction error of 1.50 mm, outperforming PointNet++'s error of 1.67 mm and smaller than the dimensions of a small tooth. Besides, the promising results on real facial predictions suggest potential clinical applications. Our code is available at: https://github.com/lancesye/TransFArchNet. • A novel solution for personalized panoramic imaging. • Transformer layers are utilized to learn individual dental arches from facial point clouds. • Include noise-filtering and confidence networks for robust curve fitting. • Achieves a mean 2D dental arch prediction error of 1.50 mm. [ABSTRACT FROM AUTHOR]
ISSN:09574174
DOI:10.1016/j.eswa.2025.126577