AI spots osteoporosis on chest x-rays; FDA warns maker over unreviewed AI change; guidelines set MR elastography threshold

A CT scanner seen from the front at a slight angle, an empty patient table running through its ring gantry, the inner rim of the bore glowing amber.

An AI model detected osteoporosis on routine chest x-rays with an area under the curve of 0.88 to 0.93 in three external cohorts, an npj Digital Medicine study reported. The Food and Drug Administration (FDA) posted a warning letter over an unreviewed machine learning update to CT software, guidelines in Radiology set a 19% threshold for true change on liver MR elastography and AIRS Medical won clearance for MRI body-composition AI.

  • AI Oct 9, 2026
    Researchers developed AI-OsRM using 46,478 paired chest radiograph and dual-energy x-ray absorptiometry (DXA) examinations from 26,757 patients, according to a study in npj Digital Medicine. For osteoporosis detection, the area under the receiver operating characteristic curve was 0.93 and 0.92 in two health checkup cohorts of 24,132 and 122,535 people, and 0.88 in a separate hospital cohort of 6,391. In simulations, AI prescreening could reduce DXA use by approximately 50% while detecting 98% to 99% of osteoporosis cases, the authors wrote. The results are accuracy and simulation data, not patient outcomes.
  • FDA Oct 6, 2026
    The Food and Drug Administration (FDA) on Oct. 6 posted a Sept. 1 warning letter saying version 3.0.0 of O.N. Diagnostics' VirtuOst VFA software, which uses spine-containing CT scans, is adulterated and misbranded. According to the letter, a new machine learning algorithm was introduced to locate points on the vertebrae and the device moved from a desktop to a web-based architecture without a new 510(k) submission. "Each of these changes could significantly affect the safety or effectiveness of the device by introducing new software and cybersecurity risks," the FDA wrote. It also cited corrective action and medical device reporting deficiencies.
  • Guidelines Oct 6, 2026
    Technical guidelines for liver MR elastography developed by the Quantitative Imaging Biomarkers Alliance, whose work is now continued under the Quantitative Medical Imaging Coalition, were published online Oct. 6 in Radiology as a special report. They define acquisition protocols and reconstruction requirements for measuring stiffness in chronic liver disease. A measured change in liver stiffness of at least 19% represents a true biologic change with 95% confidence, the authors wrote. The threshold is derived from a within-subject coefficient of variation of approximately 7%.
  • AI Oct 6, 2026
    AIRS Medical said on Oct. 6 that the Food and Drug Administration (FDA) granted 510(k) clearance for SwiftSight Body Composition, which quantifies muscle and adipose tissue from MR images after a scan of less than five minutes. SimonMed Imaging will be the first customer to deploy it, the company said. The FDA's 510(k) database lists a decision date of Oct. 1.
  • Research Oct 6, 2026
    In a randomized trial of 63 wrists from 39 participants, published online Oct. 6 in Radiology, researchers compared ultrasound-guided hydrodissection of the median nerve with normal saline, perineural corticosteroid injection and hydrodissection with saline plus corticosteroid. At 12 weeks, median nerve cross-sectional area fell 43% and 46% in the two hydrodissection groups versus 11% with steroid alone (P = .0001). With steroid alone, symptoms recurred, and adding corticosteroid to saline provided no additional benefit, the authors wrote. The procedure took 15 minutes on average, according to the Radiological Society of North America.

The weekly innovation roundup for radiology, published every Sunday for the week before; every item links to its sources. Latest radiology news · All specialties