#415 – The Wire

  • Vara’s Interval Solution: A new European Journal of Radiology study showed that Vara’s mammography AI solution could help radiologists prevent interval and advanced cancers. The researchers used Vara to analyze screening mammograms from 2,396 women who later developed interval cancers, finding that the algorithm detected and localized 27.5% of the false negatives (103 of 374) and 12.2% of exams with “minimal signs of cancer” (57 of 468). The researchers used Vara at 99% specificity to limit false positives (above the program’s 97% threshold), suggesting that a lower AI specificity would have caught even more false-negatives.
  • Aduhelm’s Rapid Decline: Aduhelm’s controversial FDA clearance and string of major rejections (culminating with UnitedHealthcare on Monday) finally proved to be too much for Biogen, which eliminated most commercial operations built around the once-promising Alzheimer’s drug and announced its CEO’s resignation. Aduhelm was originally expected to drive a wave of neuroimaging exams (it requires brain PET for diagnosis & ongoing MRIs for monitoring), but it appears that these neuroimaging volumes will have to wait for future treatments.
  • At Home Efficiency: A recent Academic Radiology study shared a solid argument against radiology’s version of ‘return-to-office,’ finding that at-home reporting can improve turnaround times and reduce radiologist burnout. Six months after COVID forced Penn State Hershey Medical Center to shift some staff to home-based teleradiology, they found that remotely-interpreted reports were returned much faster (e.g. emergency CT 28 vs. 35 min; inpatient MRI 80 vs. 312 min) and the home-based rads had higher wellness survey scores.
  • Samsung’s New Mobile X-Ray: Samsung’s Boston Imaging brand announced the FDA clearance and U.S. launch of its new GM85 Fit mobile X-ray system. The GM85 inherits many features from Samsung’s premium AccE GM85 system, while adding improvements to its battery life, driving experience, detector weight, operating noise, and imaging processing engine. 
  • AI’s Unintended Consequences: A new editorial from Chris Wood (founder of Reveal-Dx, Clario, and Confirma) warned of AI adoption’s unintended consequences, urging AI developers and adopters to address potential issues before they take place. Wood reviewed PACS’ early history of unintended consequences and the major efforts to fix them (e.g. Wood’s Clario solution fixed PACS’ worklist issues), while suggesting that the same could happen with imaging AI tools. 
  • POCUS for AAA Follow-Ups: A new Sacramento VA study showed that point-of-care ultrasound could be valuable for abdominal Aortic Aneurysm (AAA) follow-up assessments, in addition to its proven role in AAA diagnosis. The researchers had a non-physician perform AAA follow-up POCUS exams on 80 patients after receiving three hours of training, finding that the average POCUS-based AAA measurements closely matched standard-of-care measurements from trained vascular technologists (4.1cm vs. 4.0cm). 
  • Peak Imaging: A UPenn, Qure.ai, and MinXray-led team just completed a mission across the Tibetan Himalayas to deploy AI-powered mobile X-ray systems at some of the world’s highest altitude hospitals (11k to 14.5k ft), including at Mount Everest Basecamp (18k feet). In addition to giving these remote facilities their first X-ray systems and only X-ray diagnostics, this on-foot expedition created some great social media content.
  • Low-Dose Thickness Protocol: The University of Chicago radiology department developed a new patient thickness-based X-ray protocol that cut radiation dosage in enteric tube placement exams by 80%, without sacrificing diagnostic accuracy. The enteric tube thickness protocol is based on a number of factors (BMI, ventilator use, X-ray signal distance), suggesting that similar calculations might allow lower-dose protocols for other diagnostic tasks with similar image quality requirements.
  • United Imaging Flips the Script: United Imaging shared a comprehensive update on its activities and initiatives, revealing notable account and service team expansions, its mission and values-focused employee training programs, and its decision to “flip the script” on industry events in 2022. United Imaging will continue to leverage its Texas headquarters as its permanent booth, while taking a “tailored approach” to major industry events (including not always by having a booth), and increasing its presence at smaller shows.
  • Screening Inconsistencies: A new JAMA study revealed significant inconsistencies between U.S. states’ breast cancer screening plans and the federal USPSTF recommendations. Only 16 (31%) states’ plans matched the USPSTF recommendations for average-risk women’s mammography screening starting age and frequency, while 26 (51%) states were partially consistent, and 9 state plans (18%) were not consistent with any guideline components. The authors attributed these inconsistencies to a variety of factors, but suggested that specifying age/frequency guidelines for higher-risk women might create more overall consistency.
  • The Saliency Map Effect: A University of Sydney-led study revealed new insights into how AI-based saliency maps impact radiologist diagnostic decisions, and what could make these tools more effective. The researchers had 36 radiologists add annotations to 20 LD-CT exams that used saliency maps to highlight relevant image areas (10 benign & 10 malignant; 3 false positives & 2 false negatives). The radiologists’ annotations varied widely (only 12% agreement), suggesting that AI-based annotations using common radiologist descriptions could be valuable additions. The researchers also found that radiologists weren’t influenced by the AI false positives, but were more likely to miss AI false negatives.

You're signed up!

It's great to have you as a reader. Check your inbox for a welcome email.

-- The Imaging Wire team