Musculoskeletal radiography covers the cervical spine down to the toes, and most imaging AI splits those regions into separate problems. HOPPR is betting one model can cover the whole skeleton.
The Chicago company released a beta vision-language model, the FS MSK Whole-Body Narrative Model, on October 28. It drafts findings text for radiographs, handling up to three views per study, and was trained across the spine, thorax and chest wall, upper extremity, and pelvis and lower extremity.
674,000 studies, and a focus on abnormality
The training pool held some 674,000 of the company’s own outpatient MSK scans, spread evenly over body parts. Rather than narrating normal anatomy, the model concentrates on flagging what looks wrong. Hardware placement, lesions, bone quality, soft tissue, alignment, degenerative changes in joints and spine, and fractures are the kinds of findings it is measured on across 26 categories.
CEO Khan Siddiqui framed the pitch as a platform argument. Organizations should not need a separate model for every bone and joint, he said, and one system trained across the whole body gives developers a stronger foundation. It ships through Forward Deployed Services, which tunes the model on customer data, designs labeling workflows and wires it into existing systems.
What it is not
HOPPR is explicit that the release is a software component, not a medical device. It carries no FDA clearance, is not meant for diagnostic use, and is offered in beta for evaluation.
The name is a deliberate nod to history. The letters “FS” belong to Florence Ada Stoney, born in 1870 and dead by 1932, generally treated as Britain’s pioneering female radiologist. Women got their own X-ray suite from her in 1902 at the Elizabeth Garrett Anderson hospital. Fulham Military then handed her a unit to run while the 1914-18 war was under way.
