BunkerHill BMD

The BunkerHill BMD is a software-only medical device that utilizes deep-learning computer vision and post-processing algorithms to opportunistically estimate spinal bone mineral density from existing CT scans in adults aged 30 and above, aiming to flag low bone density findings. Validated for safety and effectiveness, this Class II medical device assists clinicians in identifying potential concerns without replacing dedicated BMD tests, demonstrating the application of AI for secondary screening in routine medical imaging.

The BunkerHill BMD device, a bone densitometer developed by BunkerHill Health, has achieved substantial equivalence to existing legally marketed devices, paving its way for market entry. Classified as a Class II medical device, it is subject to general controls including annual registration, device listing, adherence to good manufacturing practices, quality system (QS) regulation for design controls, and the Unique Device Identification (UDI) System rule. The BunkerHill BMD is a software-only medical device (SaMD) that utilizes deep-learning-based computer vision and post-processing algorithms to estimate the average areal bone mineral density of spinal bones from existing CT scans in adults aged 30 and above. It functions as an opportunistic AI tool to assist clinicians in identifying low bone density "findings of interest" and is not intended to replace dedicated BMD tests or serve as the sole basis for clinical decisions. Its safety and effectiveness were validated through a retrospective study of 371 CT scans, demonstrating high sensitivity and specificity, confirming its generalizability, and identifying no new safety or effectiveness concerns compared to predicate devices.

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 deep-learning algorithm specifically differentiate between various types of spinal CT scans (e.g., different slice thicknesses, reconstruction kernels) and what impact do these variations have on BMD estimation accuracy?; What is the clinical utility and impact of identifying "findings of interest" for low bone density, considering the device is not intended for primary clinical decision-making or to replace dedicated BMD tests?; Are there specific patient populations (e.g., those with severe degenerative changes, metallic implants, or prior spinal surgeries) where the device's performance might be limited or require additional caution?; What is the long-term plan for post-market surveillance to monitor the device's performance, especially regarding false positives or negatives, in diverse clinical settings?; How does the device's estimated average areal BMD correlate with gold-standard DXA measurements, and are there specific thresholds or ranges where this correlation is stronger or weaker?