Videa Dental Assist

Videa Dental Assist is an AI-powered software utilizing supervised deep learning to analyze dental radiographs (bitewing, periapical, panoramic) for patients aged 3 and older, aiming to improve diagnostic accuracy by automatically detecting suspected dental findings, historical treatments, and normal anatomy. Its clinical utility as an adjunctive tool for dental professionals reviewing suspected dental findings was demonstrated through studies showing statistically significant enhancement in identifying various dental conditions when the software aided their review.

Videa Dental Assist (VDA) is a cloud-based, AI-powered software designed as a computer-assisted detection (CADe) tool for dental professionals. Its purpose is to analyze various dental radiographs, including bitewing, periapical, and panoramic views, from patients aged 3 years and older. The system, accessed via an API and integrated into dental image viewers, automatically identifies and localizes a comprehensive range of features: suspected dental findings (such as caries, attrition, and periapical radiolucency), historical treatments (like crowns, fillings, and implants), and normal anatomical landmarks. VDA provides a binary indication of the presence or absence of findings, the number of findings by type, and coordinates for bounding boxes around detected areas. For suspected dental findings, VDA serves as an adjunctive tool to aid dentists in their diagnostic review, while identification of historical treatments and normal anatomy is intended to assist the dental practice in understanding radiographic features and patient operations management. The underlying technology utilizes supervised deep learning algorithms. The performance of Videa Dental Assist was evaluated through extensive testing. Bench testing on a dataset of 1,445 radiographs, ground-truthed by US board-certified dentists, assessed the standalone accuracy for all VDA indications. Most indications met their predefined sensitivity and specificity acceptance criteria; caries specificity was an exception in standalone testing but the overall caries indication performance was validated in the clinical study. A pivotal clinical study, employing a fully crossed, randomized, multiple reader multiple case (MRMC) design with 26 US licensed dentists reviewing 378 radiographs, demonstrated VDA's efficacy. Results showed statistically significant improvements in diagnostic accuracy, measured by the Alternative Free-response Receiver Operating Characteristic Figure of Merit (AFROC FOM), when dental professionals used VDA as an aid. For instance, average improvements in AFROC FOM with VDA ranged from 0.024 (4.3% improvement, p=0.0085) for caries detection to 0.236 (36.8% improvement, p=1.9e-9) for dens invaginatus, confirming its value in enhancing clinical assessment of suspected dental findings.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does Videa Dental Assist's performance vary across different X-ray sensor manufacturers not extensively covered in the generalizability analysis, and what is the minimum image quality required for optimal system performance?; What is the specific impact of the substantial proportions of 'unknown' gender (48% in bench, 39% in clinical) and 'unknown' age (18% in bench, 15% in clinical) demographics in the testing datasets on the AI model's generalizability and potential biases, particularly for underrepresented groups within these categories?; Given that caries specificity did not meet its target acceptance criteria in bench testing, what specific features or training elements does Videa Dental Assist employ to help clinicians differentiate true positives from false positives for caries, especially in complex cases?; How does the system differentiate between normal anatomical variations, such as developing or erupting teeth in pediatric patients, and early pathological changes for indications like 'Widened Periodontal Ligament' or 'Periapical Radiolucency' across the full 3 years and older patient age range?; What are the typical processing times for Videa Dental Assist to analyze different types of radiographs (bitewing, periapical, panoramic) via its API, and how does this integrate into a high-volume dental practice workflow without causing significant operational delays?