Accuracy of an artificial intelligence-driven tool for the detection of small edentulous regions on cone-beam computed tomography

Abstract

To assess the accuracy of a novel artificial Intelligence (AI)-driven tool for automated detection of small edentulous regions on Cone-Beam Computed Tomography (CBCT) images. A total of 45 CBCT scans from the University Hospitals of Leuven, Leuven, Belgium were selected and used for validation after network training with another 175 CBCT scans. The AI-driven tool (Virtual Patient Creator, Relu, Leuven, Belgium) automatically detected, labeled, and segmented partial edentulous jaws and compared them to the human performance serving as the ground truth. The accuracy and speed of the AI-driven tool to detect and label teeth and edentulous regions in partially edentulous jaws were therefore assessed. The AI-driven tool achieved a general accuracy of 0.98, precision of 0.97, and recall of 0.90 for detection of small edentulous regions. For detection and labeling of teeth, the general accuracy was 0.97, precision and recall were 0.98 and 0.99, respectively. The median time needed for the professional to identify both teeth and edentulous areas was 67 s while the AI performed the same task with a median time of 12 s (p-value < 0.0001). This 12 s time also involved accurate and consistent segmentation of teeth (mean Intersection over Union = 0.97) which was up to 900 times higher for human performance. The AI-Driven tool was fast and highly accurate to detect teeth and small edentulous regions, tooth labeling was up to six times faster and segmentation up to 900 times faster than an experienced dental surgeon.