BoneView

BoneView is a software tool that utilizes supervised deep learning algorithms to analyze adult radiographs, functioning as a concurrent reading aid for clinicians to improve the detection of fractures across numerous anatomical regions. It processes DICOM images to identify and highlight potential fractures with distinct visual cues based on algorithmic confidence, aiming to enhance diagnostic sensitivity and specificity.

BoneView is a radiological computer-assisted detection software designed to aid clinicians in identifying fractures in adult patients using 2D X-ray images. It employs supervised deep learning algorithms to analyze radiographs across a wide range of anatomical areas, including extremities, pelvis, ribs, and spine. The software integrates with existing radiological infrastructure, such as PACS or X-ray systems, either through on-premise or cloud deployment. It processes DICOM images automatically, identifying potential regions of interest without altering the original radiographs, and generates result files including a summary table and annotated images for concurrent review. The system provides four levels of results, visually indicated by bounding boxes: solid lines for definite fractures (high-specificity) and dotted lines for suspicious or subtle fractures (high-sensitivity), alongside "NO FRACT" or "NOT AVAILABLE" classifications. Standalone bench testing on 8,918 radiographs demonstrated high global sensitivity (0.928 at high-sensitivity point) and specificity (0.932 at high-specificity point). A multi-reader, multi-case clinical study involving 24 readers and 480 cases showed that using BoneView as an aid significantly improved reader diagnostic accuracy, increasing sensitivity by 10.4% (from 0.648 to 0.752) and specificity by 5% (from 0.906 to 0.956) compared to unaided readings.

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 BoneView's performance vary across different X-ray system manufacturers and image acquisition parameters not explicitly detailed in the bench testing?; What is the performance impact on fracture detection for patients with comorbidities or pre-existing bone conditions (e.g., osteoporosis, previous healed fractures) within the validated anatomical regions?; Could the 'DOUBT FRACT' (high-sensitivity) and 'FRACT' (high-specificity) thresholds be adjusted by users, and what is the impact of such adjustments on diagnostic accuracy in specific clinical settings?; What is the average processing time per study, and how might this vary with different deployment models (cloud vs. on-premise) and image set sizes?; How does the software handle images with significant artifacts or suboptimal quality, and what is the false positive/negative rate in such scenarios?