Segmentation algorithms of dental CT images: A comprehensive review from classical to deep learning trend

Computed tomography (CT) imaging is the most accurate modality for screening and treatment monitoring of oral diseases. Segmentation of anatomical tissues (teeth, root canals, jaws, etc.) in CT data is a prerequisite for treatment planning and computer-aided detection and diagnosis. While the comple...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Expert systems with applications Jg. 275; S. 126853
Hauptverfasser: Wu, Dianhao, Jiang, Jingang, Wang, Jinke, Bi, Zhuming, Yu, Guang
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.05.2025
Schlagworte:
Computed tomography
Deep learning
Dental detection/diagnosis
Image segmentation
Tooth segmentation
Deep learning
Image segmentation
Dental detection/diagnosis
Computed tomography
Tooth segmentation
ISSN:0957-4174
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Computed tomography (CT) imaging is the most accurate modality for screening and treatment monitoring of oral diseases. Segmentation of anatomical tissues (teeth, root canals, jaws, etc.) in CT data is a prerequisite for treatment planning and computer-aided detection and diagnosis. While the complexity of CT imaging characteristics and oral physiology makes it difficult to extract target boundaries. This paper systematic reviews 145 articles on dental CT image segmentation. The segmentation method is categorized into classical and deep learning-based algorithms, then a new classification framework of classical algorithms is presented, which includes morphology feature-based, contour feature-based, and pixel feature-based segmentation algorithms. We perform a literature clustering analysis, dissect the algorithm from the perspective of classical to deep learning trend, and compare the characteristics, performance and applications of various algorithm. The applicability and limitations of classic algorithms are further expounded, and the development direction of the deep learning-based algorithm from several aspects of training sample, network structure, execution method and training strategy. Oral tissue segmentation is an unresolved topic due to the limitations of existing approaches, in which deep learning algorithms, the mainstream future direction, are already comparable to manual segmentation, yet there is still room for improvement. Finally, we identified promising research directions, that is the high-quality samples, multi-algorithm integration, performance balance and multi-modal imaging, and an image application framework based on the Internet of Things (IoT) platform for the whole cycle of dental treatment is proposed. We hope that this review will provide a possible formwork for research in the field of dental CT segmentation.
ISSN:0957-4174
DOI:10.1016/j.eswa.2025.126853