Evaluation of an Artificial Intelligence-based software in the detection of caries in Panoramic Radiographs | ||
| Ain Shams Dental Journal | ||
| Volume 38, Issue 2, June 2025, Pages 630-640 PDF (1.2 M) | ||
| Document Type: Original articles | ||
| DOI: 10.21608/asdj.2024.316086.1480 | ||
| Authors | ||
| Yasmin Morsy Saleh* 1; Sahar Mohamed Samir2; Walaa Mohamed Hamed1 | ||
| 1Oral and Maxillofacial Radiology, Faculty of Dentistry, Ain Shams University | ||
| 2Faculty of Dentistry -Ain Shams University, Cairo, Egypt. | ||
| Abstract | ||
| Aim: To evaluate the clinical diagnostic accuracy of WediagnostiX artificial intelligence-based software for helping dental professionals in the automatic detection and identification of dental caries in panoramic radiographs. Materials and methods: A dataset of 325 anonymized panoramic radiographs (PR) were selected. First, the images were manually evaluated by two experienced oral radiologists, where consensus was established by a third evaluator to set the “ground truth”. The evaluators classified their findings as follows: (DC) for dental caries, (M) for missing teeth and all teeth were numbered and labelled with Federation Dentaire Internationale (FDI) nomenclature. The OPGs were then anonymously uploaded and analyzed by the AI-based software (WeDiagnostiX). Caries detection module was operated using the specific confidence threshold of the software. Results were recorded on excel spreadsheets, and a statistical analysis was performed to compare the automated diagnosis of the software to the ground truth in terms of Sensitivity (S), Specificity ( E), Positive Predictive Value (PPV), Negative Predictive Value (NPV), Diagnostic Accuracy (DA), and their presentation in the area under (AUC) the Receiver Operating Characteristic (ROC) curve. Results: Diagnostic metrics for each variable obtained in this study were as follows: (DC) S=50%, E=91.8%, PPV=50.4%, NPV=91.7%, DA= 85.9%, AUC=0.709; (M) S=82.4%, E=93.9%, PPV=72.1%, NPV=96.5%, DA=92%, AUC=0.882; (FDI) S=90.3%, E=87.7%, PPV=69.7%, NPV=96.7, DA=74.6%, AUC=0.89. Conclusion: Results of this study suggest that WeDiagnostiX can provide reliable evaluation for dental caries and other variables on PRs improving diagnostic quality and performance of dental clinicians. | ||
| Keywords | ||
| Deep learning; Convolutional neural network; Machine learning; Diagnostic; medical imaging | ||
|
Statistics Article View: 6 PDF Download: 2 |
||
