Detection and Segmentation of Carotid Atheroma Calcification in Dental Panoramic Radiographs Using a Hybrid Deep Learning Model

Objectives To develop a labeled image dataset of suspected carotid artery calcifications (CACs) on dental panoramic radiographs (PR) and to propose an automated deep learning framework for their detection and segmentation. Methods A retrospective dataset of 19,205 PRs obtained from our institution (2015 and 2023) was analyzed. After manual labeling 372 images showing CACs and 322 controls were included. A hybrid architecture combining FasTVit, AttentionNet, and DC-UNet was implemented for classification, detection, and segmentation. Model performance was assessed using accuracy, precision, recall, dice coefficient, Intersection over Union (IoU), and Area Under the Curve (AUC). The study evaluated two scenarios: one where detection and segmentation followed classification, and another where classification was omitted. Results The classification model achieved an accuracy, precision, recall, and specificity of 0.878, 0.878, 0.880, and 0.880 respectively. In terms of detection, the model reached an AP of 0.321 and an AR of 0.475 for medium-sized CACs, with AP at 0.414 and AR at 0.559 at IoU thresholds between 0.50 and 0.95. The segmentation obtained precision of 0.612, a recall of 0.737, an IoU of 0.500, a Dice score of 0.621, and an AUC of 0.805. Incorporating the classification step improved precision from 0.48 to 0.93, accuracy from 0.79 to 0.97, and specificity from 0.75 to 0.98. The AUC for the ROC curve of the final framework reached 0.96. Conclusions The proposed hybrid artificial intelligence (AI) framework showed high diagnostic accuracy in identifying and delineating CACs on PRs. This tool may contribute to the early detection of carotid calcifications in routine dental imaging, enabling timely medical referral and cardiovascular risk assessment. Clinical relevance: The hybrid multistage framework enhances automated detection of CACs on PRs and enable opportunistic cardiovascular screening during routine dental examinations, bridging dental imaging with systemic health prevention.