Accuracy Of An Ai-Assisted Dental Prototype In Radiographic Detection

Abstract

The current study aimed to evaluate the accuracy of an artificial intelligence-assisted dental prototype in automated radiographic detection from digital panoramic radiographs. 300 panoramic radiographs collected between January 2020-2024 were analysed by two trained and calibrated specialist evaluators. The diagnostic consensus, “ground truth,” was labelled as IT: Impacted teeth, RR: Retained roots, C: Caries, F: Filling, P: Prosthesis, PR: Periapical radiolucency, I: Implants, PC: Post-Core, and RF: Root fillings. The radiographs were uploaded to the dental prototype, and the results were compared. Sensitivity, specificity, positive, and negative predictive values were calculated using Stata version 15.0 (StataCorp). The sensitivity, specificity, positive, and negative predictive values, respectively, in percentages for each outcome, were IT (82.43, 99.74, 82.99, 99.72), RR (89.18, 99.42, 87.73, 99.50), C (96.72, 94.52, 60.36, 99.70), F (95.05, 97.35, 81.82, 99.37), P (85.85, 99.37, 93.03, 98.63), PR (93.06, 98.06, 66.9, 99.70), I (100, 100, 100, 100), PC (90.74, 99.33, 43.36, 99.95), and, RF (98.07, 98.63, 84.42, 99.85). The key errors identified in the qualitative analysis were tooth identification errors, missed recurrent caries under fillings and crowns, and identification of extensive fillings as crowns. The dental prototype demonstrated high sensitivity and specificity in identifying dental pathology. Accuracy in identifying teeth that have migrated, secondary caries, and differentiating extensive fillings from crowns requires further improvement.

Ähnliche Arbeiten