- Adult AI for Pediatric Patients: A new study out of South Korea found that Lunit Insight CXR accurately interpreted pediatric chest X-rays, even though Insight CXR is only developed/approved for adults. The researchers analyzed 2,273 exams using Insight CXR and compared the results to original radiologist interpretations, achieving promising sensitivity, specificity, and accuracy (67.2%, 91.1%, 87.5%) and even better results when they excluded patients ≤2 years old or with cardiomegaly (86.4%, 97.9%, 96.9%). The authors suggest that more adult-to-pediatric AI validations might eventually help address the shortage of pediatric AI solutions, although some AI leaders warn against this approach.
- Cerner and the VA’s Lost Orders: A Cerner EHR glitch at the Spokane VA hospital reportedly led to roughly 11k “lost” referral orders between October 2020 and June 2021, over three-quarters of which were for radiology exams. The lost orders harmed at least 148 vets, including dozens classified as “moderate” or “major” harm. Cerner and the VA are both absorbing their share of criticism, as Cerner reportedly knew about the glitch but initially didn’t fix it or notify the VA, and the VA expanded the software to other facilities after it knew orders were being lost.
- Photon-Counting CT’s MSK Performance: A recent Mayo Clinic feasibility study highlighted photon-counting CT’s strong performance with shoulder and pelvis imaging. The researchers performed PCCT and EID CTs on 32 patients and had them evaluated by two musculoskeletal radiologists, finding that PCCT achieved “significantly better” anatomic-structure visualization (p < 0.001) at far lower radiation doses (-31% for shoulder exams, -47% for hip exams).
- Habersham Ditches the Disk: PocketHealth added Georgia’s Habersham Medical Center to the growing list of hospitals using its image sharing platform. The deal puts Habersham on a path towards ending CD-based image sharing, giving its patients the ability to access and share their medical imaging from any device, and allowing Habersham’s clinicians to view and import all imaging using their existing infrastructure.
- Inpatient Rebounding, Outpatient & Home Growing: Vizient’s Sg2 subsidiary predicts that inpatient volumes will quickly recover to pre-pandemic levels and then grow only 2% over the next decade, while chronic conditions cause an 8% rise in inpatient days over the same period. Growth will be far stronger outside of hospitals, with outpatient volumes projected to grow 16% and home visits slated to increase by 19%.
- Lumbar MRI AI: A Radiology Journal study suggests that AI-based efficiency improvements could be on the way for “repetitive and time consuming” lumbar spinal stenosis MRI interpretations. Researchers had eight experienced radiologists interpret lumbar spine MRIs with and without deep learning support (n = 444 MRIs, 25 patients), finding that DL support reduced their average interpretation times (from 124-274 to 47-71 seconds) and improved their interobserver agreement rates (k values: 0.71 vs. 0.39).
- Novarad’s VisAR FDA: Novarad announced the FDA 501k clearance of its VisAR augmented reality surgical navigation system for precision guided intraoperative spine surgery. VisAR transforms CT data into a 3-dimensional hologram that’s viewed through a Microsoft HoloLens visor, allowing surgeons to view images superimposed onto the patient without looking away at a separate monitor. Although known for its PACS, RIS, and imaging sharing, Novarad launched its original OpenSight AR system in 2018.
- Lung-RADS Follow-Ups Are Too Early: A new AJR study suggests that Lung-RADS’ recommended 3-month follow-up period after a positive LD-CT screening result isn’t long enough to properly evaluate malignant nodule growth. The researchers analyzed LD-CTs from 76 patients who received positive Lung-RADS evaluations and were later diagnosed with lung cancer, finding that the frequency of nodule growth at 3 months was “very low yield” (5% w/ linear measurements, 7% w/ volumetric measurements), and the median malignant nodule time-to-growth was 11 and 13 months.
- Stanford & CARPL.ai’s AI Affiliation: After three years of collaboration, CARPL.ai and Stanford formalized their AI relationship, making CARPL.ai a Stanford AIMI Industry Affiliate. The CARPL.ai platform uniquely supports AI development, testing, and distribution, potentially supporting Stanford’s AI activities from the research lab to the clinic.
- Radiology on TikTok: More than 12% of the global population uses TikTok, but does the platform deserve a role in radiology education? Johns Hopkins researchers tried to answer that question, making their own radiology-focused account that amassed 230+ followers, 600+ likes, and 8,200 video views in just 60 days. Although the authors advised potential radiology influencers to “tread carefully” and not violate their institution’s social policies, they were bullish about TikTok’s ability to help radiology educators reach a far larger audience.
- DiA & Intelerad Partner: DiA Imaging Analysis announced a partnership with Intelerad Medical Systems, combining its LVivo Seamless AI-based cardiac ultrasound analysis system with Intelerad’s IntelePACS solution. The alliance bolsters Intelerad’s cardiac ultrasound capabilities while further expanding DiA’s list of PACS and ultrasound partners (ScImage, Change Healthcare, GE, Philips, Konica, Terason, Sonoscape, Circle CVI, Watson Health).
