(PR078) Comparative Effectiveness of Endodontists Versus AI Software in the Detection of Periapical Lesions

Convolutional Neural Networks (CNNs), a form of artificial intelligence (AI), have shown promise in dental diagnostics, including in the early and accurate detection of periapical lesions. CNNs, such as Adravision AI, consist of multiple layers that automatically and hierarchically learn to detect features from input images.

Objectives: This study evaluates the effectiveness of Adravision AI in detecting periapical lesions on periapical radiographs compared to endodontists, using CBCT as the gold standard.

Methods: This study was conducted using imaging data from 300 teeth in 99 patients aged 18-65 years. Periapical radiographs were analyzed independently by 3 endodontic evaluators and by the Adravision AI software. The AI was trained on over 10,000 annotated x-rays. CBCT images were used as the reference standard for detecting periapical lesions. Sensitivity, specificy, and accuracy were statistically analyzed to compare the diagnostic performance of Adravision AI and endodontists.

Results: Adravision AI outperformed the endodontists, with higher sensitivity (90.1% vs 80.5%, SD=1.47), specificity (94.5% vs 90.8%, SD=1.45), and overall accuracy (92.3% vs 85.7%, SD=1.33). The Adravision AI demonstrated consistent and objective analysis, reducing variability and subjectivity associated with human evaluators.

Conclusion: The integration of AI represents a significant advancement in the field of endodontics, enhancing diagnostic precision and efficiency. Adravision AI shows superior performance in detecting periapical lesions, suggesting its potential as a reliable tool in clinical practice, complementing traditional diagnostic methods. These findings align with previous research indicating the robustness of AI in dental diagnostics.