Artificial Intelligence for Automated Detection and Segmentation of Pulp Stones in Cone Beam Computed Tomography Scans.

Abstract Objective : This study aimed to automatically detect the human tooth pulp cavity and pulp stones on cone-beam computed tomography (CBCT) using deep learning-based semantic segmentation. Methodology : 523 annotated images of teeth from 112 CBCT examinations were used to train and evaluate four established deep learning architectures (U-Net, Attention U-Net, U-Net++, and DeepLabV3+) in segmenting the background, pulp cavity, and pulp stone(s). Subsequently a novel architecture was developed (DACA-DeepLab) and evaluated through an additive study. DACA-DeepLab incorporates DeepLabV3+ with a pyramid channel-spatial attention (PCSA) module, a dual-axis cross-attention (DACA) decoder, and a Sobel-guided boundary-refinement (BFR) head. Results : DeepLabV3+ achieved the best overall Dice performance among the four baseline architectures (mean, 0.775; pulp cavity, 0.838; pulp stone, 0.711). In the second study, combining PCSA and DACA yielded the highest performance (mean Dice, 0.733). Adding the boundary-refinement head reduced performance, suggesting sensitivity to the limited training set. Conclusion : Deep learning-based semantic segmentation can accurately identify human tooth pulp cavity and pulp stone(s) on CBCT. These findings establish a systematic benchmark for CBCT-based pulp segmentation and provide a foundation for AI-assisted endodontic diagnosis and pre-treatment planning.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11554539/v1
Primary Topic
Dental Radiography and Imaging
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Artificial Intelligence for Automated Detection and Segmentation of Pulp Stones in Cone Beam Computed Tomography Scans.

Sally Mansour, Mohamed El-Sharkawy, Sabah Sobhy*, Yara Helaly et al.
2 links
Research Square
Dental Radiography and Imaging
preprint

Artificial Intelligence for Automated Detection and Segmentation of Pulp Stones in Cone Beam Computed Tomography Scans.

Sally Mansour, Mohamed El-Sharkawy, Sabah Sobhy*, Yara Helaly, William Scarfe, Hanaa Mansy
preprint en

Abstract

Abstract Objective : This study aimed to automatically detect the human tooth pulp cavity and pulp stones on cone-beam computed tomography (CBCT) using deep learning-based semantic segmentation. Methodology : 523 annotated images of teeth from 112 CBCT examinations were used to train and evaluate four established deep learning architectures (U-Net, Attention U-Net, U-Net++, and DeepLabV3+) in segmenting the background, pulp cavity, and pulp stone(s). Subsequently a novel architecture was developed (DACA-DeepLab) and evaluated through an additive study. DACA-DeepLab incorporates DeepLabV3+ with a pyramid channel-spatial attention (PCSA) module, a dual-axis cross-attention (DACA) decoder, and a Sobel-guided boundary-refinement (BFR) head. Results : DeepLabV3+ achieved the best overall Dice performance among the four baseline architectures (mean, 0.775; pulp cavity, 0.838; pulp stone, 0.711). In the second study, combining PCSA and DACA yielded the highest performance (mean Dice, 0.733). Adding the boundary-refinement head reduced performance, suggesting sensitivity to the limited training set. Conclusion : Deep learning-based semantic segmentation can accurately identify human tooth pulp cavity and pulp stone(s) on CBCT. These findings establish a systematic benchmark for CBCT-based pulp segmentation and provide a foundation for AI-assisted endodontic diagnosis and pre-treatment planning.

Research Square
University of Louisville (US), Cairo University (EG), Al-Azhar University (EG), Al Azhar University (QA)
Good health and well-being
Dental Radiography and Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon