Related Experiment Video
Updated: Jun 28, 2026
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Application and performance of artificial intelligence-based models in the detection, segmentation and classification
Ali Alaqla1,2,3,4, Sanjeev B Khanagar2,3,5, Alzahraa Ibrahim Albelaihi6
1Department of Restorative and Prosthetic Dental Sciences, College of Dentistry, King Saud bin Abdulaziz University for Health Sciences, Riyadh, Saudi Arabia.
Background:
Periapical lesions appear as periapical radiolucency on various imaging modalities. The accuracy of dentists in diagnosing periapical radiolucency varies significantly. Recent scientific and technological advancements have enabled the development and evaluation of artificial intelligence (AI) systems for various diagnostic applications in dentistry.
Objectives:
The aim was to report on the application and performance of AI-based models in the detection, segmentation, and classification of periapical lesions.
Methods And Methods:
A systematic effort for data acquisition began with an exploration of a wide range of reputable databases, including PubMed, Scopus, Embase, Cochrane, Web of Science, Google Scholar, and the Saudi Digital Library. Our comprehensive investigation spanned from 1st January 2000 to 31st March 2025.
Results:
Twenty-eight articles fulfilled the eligibility criteria. Among these, 20 (71.4%) applied AI technology for automated detection, 3 (10.7%) for segmentation, 2 (7.2%) for periapical lesion detection and segmentation, and 3 (10.7%) for periapical lesion classification. Thirteen (46.5%) studies in this review utilized dental panoramic radiographs, 8 (28.5%) used intraoral radiographs (periapical and bitewing), and 7 (25%) employed CBCT scans. The AI models demonstrated an accuracy range of 70% to 99.65%, with sensitivity varying from 65% to 100% and specificity ranging from 62% to 100%. The risk of bias assessment using the QUADAS-2 tool, indicated 32.1% of the studies exhibited a significant risk of bias regarding the assessment of bias and applicability issues in the reference standard arm. While the certainty of evidence was evaluated through the GRADE approach, which indicated that the included studies demonstrated a moderate degree of evidence certainty.
Conclusions:
According to the results of the studies presented, AI-based technologies hold significant potential to assist clinicians and enhance the reliability of clinical diagnoses, enabling less experienced clinicians to identify lesions with greater accuracy. However, prospective studies and randomized clinical trials are essential to evaluate the effectiveness and cost-efficiency of deep learning-based lesion detection in real clinical settings.
Systematic Review Registration:
PROSPERO CRD420251011455.
