Adravision Perio

Adravision Perio is a semi-automated, machine learning-driven software intended to aid dental professionals in the quantitative assessment of interproximal bone levels by analyzing bitewing and periapical radiographs from adult patients with permanent teeth. The system offers objective measurement suggestions as an adjunctive tool, and its efficacy in accurately identifying key anatomical points and measuring bone levels has been demonstrated through bench testing against established ground truth data.

Adravision Perio is a radiological semi-automated image processing software, categorized as a Software as a Medical Device (SaMD), designed to assist dental professionals like dentists and dental hygienists. Its core function is to measure mesial and distal bone levels associated with each tooth from bitewing and periapical dental radiographs. The software is intended for use with patients in the United States who are 22 years of age or older and have only permanent teeth, as its performance on primary dentition has not been evaluated. Adravision Perio utilizes machine learning technology to analyze radiographs and provide suggested measurements of interproximal bone levels. It supports various image input formats including DICOM, JPEG, JPG, PNG, TIFF, TIF, JFIF, and BMP. Importantly, the device acts as an aid and its recommendations are for reference only; it is not designed to replace a full patient evaluation or the clinician's independent clinical judgment, which should consider all relevant patient information. The performance of Adravision Perio was rigorously assessed through bench testing. An Object Key Similarity (OKS) Assessment was conducted on a dataset of 340 dental radiographs (both bitewing and periapical), evaluating precision and recall against ground truth key points. For bitewing radiographs, it achieved 91.0% precision and 94.0% recall; for periapical radiographs, it showed 84.8% precision and 89.3% recall. In a separate Length Measurement Assessment using 78 dental radiographs, the software's sensitivity and specificity were evaluated against consensus Clinical Attachment Level (CAL) length measurements from three ground truthers. For bitewing images, sensitivity was 90.7% and specificity was 94.3%, with a Mean Absolute Error (MAE) of 0.434 mm. For periapical images, sensitivity was 92.5% and specificity was 86.8%, with an MAE of 0.504 mm. The device also demonstrated high accuracy in tooth number classification, achieving 91.8% in the OKS study and 93.2% in the Length Assessment study.

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 the Adravision Perio software account for anatomical variations or pathologies (e.g., extensive restorations, root proximity, furcation involvement) that might affect the accuracy of automated bone level measurements?; What specific features or algorithms within the machine learning model contribute most significantly to its performance in the Object Key Similarity and Length Measurement assessments, and how robust is it to variations in radiographic image quality (e.g., angulation, contrast, noise)?; Could the system be adapted or trained to assess bone levels around dental implants, given that the current patient population is defined as those with no remaining primary teeth but does not explicitly exclude implants?; What was the demographic and clinical diversity (e.g., severity of periodontal disease, tooth types, presence of calculus) of the patient radiographs included in the 340-image OKS dataset and 78-image Length Measurement dataset used for bench testing?; Beyond suggesting line segments for bone levels, does the software offer any additional quantitative outputs or visual aids, such as bone loss percentage or progression tracking over time, that could further assist in periodontal diagnosis or treatment planning?