Evaluation of Artificial Intelligent Systems Based Analysis in Dental Periapical Lesions - A Radiological Study.

BACKGROUND

This study evaluated AI-based analysis of dental periapical lesions using CBCT scans, conducted at the Department of Oral Medicine and Radiology, New Horizon Dental College, Bilaspur, Chhattisgarh.

METHOD: A total of 500 CBCT scans were analyzed, with 400 scans used to train AI software and 100 scans assessed by two radiologists to test the software's performance. The AI classified lesions into periapical cysts, abscesses, or granulomas. Sensitivity, specificity, and accuracy were calculated.

RESULT: Cysts were the largest lesions, with regular margins (99.09%) and significant cortical expansion (93.36%), causing teeth displacement (66.36%). Abscesses and granulomas predominantly affected the maxilla, showing moderate hypodensity (100%) with minimal structural changes. Radiologists achieved perfect agreement (0.98, P < .001) in 77% of scans. Manual machine learning AI achieved 100% accuracy, while deep learning AI demonstrated 84.62% accuracy, with moderate to substantial agreement for lesion dimensions.

CONCLUSION

Manual machine learning AI showed superior accuracy compared to deep learning AI, demonstrating its potential for radiographic diagnosis of periapical lesions.