HOPPR is launching a beta version of its FS MSK Whole-Body Narrative Model, a vision-language foundation model that generates draft findings text for musculoskeletal (MSK) radiographs using up to three views per study.
The model was trained on about 674,000 outpatient MSK studies covering the spine, thorax and chest wall, upper extremity, and pelvis and lower extremity, according to the vendor. It focuses on describing abnormalities rather than normal anatomy. HOPPR said it evaluated the model against 26 MSK finding categories, including fracture, degenerative joint and spine changes, alignment, and hardware.
The model is designed as a software component for developers building radiology reporting and workflow applications, and HOPPR’s Forward Deployed Services team will customize and integrate it for each organization’s data and workflows, the company said. It is not cleared by the U.S. Food and Drug Administration (FDA) and is not intended for diagnostic use, the firm said.
