Development and Validation of an Artificial Intelligence Model Using 3D Convolutional Neural Networks for CBCT-Based Diagnosis of Jaw Cysts and Tumors

Background: Accurate differentiation of jaw cysts and tumors is essential for diagnosis, treatment planning, and prognosis. Cone Beam Computed Tomography (CBCT) provides three-dimensional visualization of jaw lesions; however, interpretation remains dependent on observer experience. Artificial Intelligence (AI), particularly Three-Dimensional Convolutional Neural Networks (3D CNNs), has emerged as a promising tool for automated analysis of volumetric medical imaging data.

Materials and Methods: This retrospective study included CBCT scans of 100 histopathologically confirmed jaw lesions, comprising 50 cysts and 50 tumors. The volumetric datasets were preprocessed and standardized prior to analysis. Regions of interest corresponding to the lesions were identified and used for model development. A 3D CNN-based AI model was trained and validated to classify lesions as cysts or tumors using volumetric CBCT data. Model performance was evaluated using accuracy, sensitivity, specificity, precision, F1-score, and area under the receiver operating characteristic curve (AUC).

Results: The 3D CNN model successfully differentiated jaw cysts from tumors using CBCT volumetric data and demonstrated high classification performance across the evaluated metrics. The model showed reliable discrimination between the two lesion categories and maintained consistent diagnostic performance during validation.

Conclusion: AI-based analysis using 3D CNNs demonstrated effectiveness in the classification of jaw cysts and tumors on CBCT images. The proposed approach has the potential to serve as a supportive diagnostic tool for oral and maxillofacial radiologists by enhancing diagnostic accuracy, consistency, and efficiency in the evaluation of jaw lesions.