#424 – The Wire

  • BrainKey Adds Genetics: BrainKey added genetics to its unique AI-based brain health analysis platform, joining its existing brain MRI and demographic analysis capabilities. The BrainKey Platform produces reports that help individuals understand their brain health (w/ longitudinal imaging) and the factors influencing their future brain longevity (w/ genetics), potentially allowing early detection and personalized treatments. Although BrainKey is quite unique, its platform leverages a number of imaging-related applications that seem to be gaining momentum, including multimodal analytics, patient-driven preventative imaging, and personalized treatments.
  • CEM Reduces Biopsies: Italy-based researchers found that recombined contrast-enhanced mammography (rCEM) exams could significantly reduce biopsy rates following screening recalls compared to standard assessments (SA = additional views, DBT, or ultrasound). The researchers analyzed post-recall assessments from 207 women with 225 suspicious findings, revealing that rCEM would have referred 16.4% fewer women for biopsy than SA (94 vs. 131 biopsies). rCEM achieved 93.8% sensitivity and 65.9% specificity, while each of its five false negatives were ductal carcinoma in situ (earlier-stage, less-invasive).
  • FDA Can’t Regulate AI: A new opinion piece in PLOS Digital Health makes the case that AI’s pace of innovation is simply too quick for the FDA to regulate alone, and that a “distributed approach” is needed for proper regulation. The Harvard researchers argue that health systems should self-regulate by developing standards and training protocols that ensure AI tools are safe at the local level, while the FDA should only step in to regulate AI that’s intended for nationwide use or that has a high potential to negatively impact patient health.
  • NPPs & Head CTs: Emergency departments’ noncontrast head CT volumes have skyrocketed in recent years, but the growing presence of ED nurse practitioners and physician assistants (NPs/PAs) isn’t to blame. That’s from a new University of Colorado Anschutz study (n = 6,624) that found no significant difference between the rate of pathology-positive head CTs ordered by physicians and NPs/PAs (4.6% vs 3.7%). They did however identify patient and situational factors associated with higher (patient age, daytime exams, indications of malignancy) and lower CT positivity rates (trauma, vertigo/dizziness).
  • Veye Lung Validation: A new PLOS One study highlighted Aidence’s Veye Lung Nodules solution’s ability to improve radiologists’ diagnostic performance. The researchers had two radiologists evaluate 470 pulmonary nodules from 337 chest CTs, with and without Veye Lung Nodules, finding that the solution significantly improved lung nodule detection (80.3% w/ AI vs. 71.9% unaided) and “minimally increased” their false positive rates (0.16 vs. 0.11). In fact, this study was used to support Aidence’s CE Mark approval.
  • Motilent’s New Funding: Gut-imaging AI startup Motilent secured $1.5M from the UK’s NIHR government agency to support the development and roll-out of its MRI post-processing software GIQuant, which tracks intestinal movement. Motilent will use the new funding to further develop GiQuant, deploy the solution across 10 additional UK hospitals, and study GIQuant’s ability to support the management of Crohn’s Disease.
  • CMR for Heart Transplant Monitoring: A team of Australian researchers found that cardiac MRI multiparametric mapping can effectively monitor patients for heart transplant rejection, suggesting that it could become an alternative to invasive endomyocardial biopsy (EMB). Analysis of 401 CMR and 354 EMB patients over a two-year follow-up period (40 w/ heart transplants) found that CMR accurately identified patients who would have rejected transplants (AUC: 0.92; NPV: 99%), while allowing 94% of patients to avoid EMBs (6% of CMR patients later received EMB).
  • Parents’ AI Perceptions: A Lurie Children’s Hospital survey (n = 1,620) found that most parents are open to emergency clinicians using AI tools to manage children with respiratory illnesses. The majority of parents were comfortable with the use of AI to assess the need for antibiotics (77.6%) or bloodwork (77.7%), and to interpret radiographs (77.5%). Black parents and parents between 18-25 years reported greater discomfort with AI than White parents (odds ratio: 1.67) and parents over age 46 (odds ratio: 2.48).
  • GE Reaching Full Contrast Production: GE Healthcare announced that its Shanghai facility’s production of iodinated contrast media should be close to 100% this week, suggesting that the end of the contrast shortage is in sight. Although many providers will still have to wait for contrast shipments to make their way west, GE will work with customers “to help them plan several weeks ahead as supply progressively recovers.”
  • Crowdsourced AI: Historically it’s been difficult to train deep learning models to identify rare injuries, but a recent study published in Skeletal Radiology showed that CNNs trained with images sourced from the internet can successfully identify rare joint dislocations. Researchers trained CNNs using 100 online radiographs of four joints (shoulder, elbow, hip, and hip replacements). When tested against external datasets with 100 corresponding scans, the internet-sourced CNNs achieved high AUCs (0.88-0.99).
  • SR Adds RCL: Strategic Radiology (SR) added Radiology Consultants of Lynchburg (11 radiologists, 5 hospitals, 1 imaging center, 45yrs in practice) to its consortium of independent radiology practices. The partnership is SR’s second in Virginia and increases its membership to 32 practices and 1,500+ radiologists.

You're signed up!

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

-- The Imaging Wire team