Exploring the Potential of No-Code Artificial Intelligence in Identifying Periapical Radiolucent Lesions

ABSTRACT EXPLORING THE POTENTIAL OF NO-CODE ARTIFICIAL INTELLIGENCE IN IDENTIFYING PERIAPICAL RADIOLUCENT LESIONS Marguerite Miller, D.M.D. Marquette University, 2025 Title: Exploring the Potential of No-Code Artificial Intelligence in Identifying Periapical Radiolucent Lesions. Objective: To employ a no-code platform for the development of an AI model designed to identify periapical radiolucent lesions using periapical radiographs. Methods: Three hundred twelve periapical radiographs were retrospectively collected, along with corroborating cone-beam computed tomography (CBCT) volumes and radiology reports obtained within six months. Radiology reports from board-certified oral and maxillofacial radiologists, were used as the reference standard for labeling the periapical radiographs. Among these, 180 images displayed at least one periapical radiolucent lesion, while 132 images without lesions were included as controls. The AI model was developed using LandingLens, a cloud-based, no-code platform. Radiographic images were uploaded, and lesions were manually annotated using bounding boxes. The dataset was divided into three subsets: training (70%), validation (20%), and testing (10%). Diagnostic performance metrics (sensitivity, specificity, precision, accuracy, and F1 score) were calculated both at tooth level. Results: The model demonstrated promising performance at the tooth level with 87.79% sensitivity, 92.73% specificity, 75.46% precision, 91.72% accuracy, and an F1 score of 0.812%. The root-level metrics were similar, showing 89.190% sensitivity, 94.0% specificity, 77.98% precision, 93.0% accuracy, and an F1 score of 0.83.% Conclusions: The developed no-code model was able to accurately identify periapical radiolucencies using periapical radiographs. Additional studies are required to validate the utilization of no-code AI in dental imaging. Keywords Artificial intelligence, computer vision, no-code, periapical radiolucency, deep learning