- Pediatric Imaging Disparities: A large JAMA study (38 hospitals, 27 states, 12M ED visits) showed that US children’s hospitals that serve more Hispanic and Black patients have greater imaging utilization discrepancies. Black patients were consistently less likely to receive diagnostic imaging than White patients at every hospital, while there was a significant correlation between a hospital’s proportion of minority patients and the imaging utilization difference between White and Black patients (correlation coefficient, −0.37).
- An AI Education Lesson: A team from the University of British Columbia shared valuable insights into how best to teach future doctors about AI. After training 350 med students in a 5-week workshop, the authors identified a series of educational challenges (knowledge disparities, curricular depth vs. breadth, knowledge retention) and successes (addressing AI concerns, open-access resources, and multidisciplinary collaboration) that they encountered. Based on their experience, they recommend that future AI educators: (1) standardize their curricular structure, (2) create AI case studies, (3) use experiential learning, and (4) involve a diverse group of trainees.
- Pros & Cons of WBCT: A new Emergency Radiology study highlighted the pros and cons of implementing whole-body CT (WBCT) for lower-extremity trauma patients in the ED. The authors evaluated patient throughput during the six months before and after Indianapolis’ University Methodist Hospital adopted WBCT (n=58 & 58), finding that the protocol decreased the time patients spent in the ED (416 to 340 min) and reduced X-ray images per patient (2.8 to 0.8), but significantly increased overall imaging costs ($23.66 to $95.53) and radiation exposure (4.03 to 7.61 mRem).
- Aidoc in Barcelona: Aidoc announced a new AI partnership with Hospital Clínic de Barcelona (HCB), revealing plans to deploy three AI detection modules (ICH, PE, incidental PE) that will be used to improve HCB’s patient care and train its radiology residents. HCB joins 12 other European academic medical facilities using Aidoc AI, in addition to a growing number of North American institutions.
- Pandemic Cancer Screenings: A survey of 480k US adults led by the American Cancer Society found that preventative cancer screenings dropped by as much as 80% during the first few months of the pandemic (vs. 2018), with full-year 2020 screenings declining 6% for breast cancer, 11% for cervical cancer, and 16% for colorectal cancer.
- GE Adds Q-IT’s Helix Suite: GE Healthcare led off SIIM 2022 with the addition of Q-IT’s Helix Radiology Performance Suit workload management solution to the GE PACS portfolio. Developed by Q-IT (a subsidiary of large rad practice Quantum Imaging) the Helix Suite uses predictive analytics to prioritize and assign exams on the worklist based on real-time assessments of a practice’s available radiologists and their skill sets. Quantum Imaging found that Helix can boost radiologists’ reading capacity by up to 20%, with similar improvements to super STAT and STAT/ED turnaround times (16.7% & 20.5%).
- Shields Cyberattack: Major New England imaging center and ASC company, Shields Health Care Group (30 locations), disclosed that a cyberattack exposed 2M of its patients’ information. Shields confirmed that the individuals’ full names, social security numbers, dates of birth, home addresses, provider information, diagnoses, billing information, medical record number, patient IDs, and more were accessible and/or removed from its systems. Although health system ransomware attacks justifiably get all the publicity, at least seven US practices and imaging centers have disclosed security incidents since the start of 2021.
- COPD AI 5yr Predictions: Korea-based researchers developed a chest X-ray-based deep learning algorithm that predicted COPD patients’ 5-year survival rates more accurately than conventional methods. When tested using three external CXR datasets (n=394, 416, 317), the algorithm predicted 5-year outcomes more accurately than “forced expiratory volume in 1 second” measurements (FEV1) with two of three cohorts (AI vs. FEV1 AUCs: 0.73 vs. 0.63; 0.67 vs. 0.60; 0.76 vs. 0.77).
- Mirada’s Growth Round: Mirada Medical completed a $17M growth round (total funding now $26M) after private equity firm Apposite Capital doubled down on its original 2019 investment. Mirada develops AI autocontouring software for personalized radiation oncology treatment planning, and the fresh funding should help it expand its partnerships, product lineup, and customer base.
- YOLO BC Predictions: Aiming to reduce mammography screening false-negatives, University of Louisville researchers developed a You-Only-Look-Once (YOLO)-based AI model that can predict which patients with “normal” prior mammograms would have cancerous screenings roughly one-year later. The authors used two image-to-image translation techniques (Pix2Pix and CycleGANs) to create synthetic mammograms from 413 patients’ “normal” priors and current cancer-positive exams. The YOLO-based model was able to detect and classify different types of abnormalities with 88%-95% accuracy in current exams, and 36%-50% accuracy in prior exams.
- GE & NCCS AI Alliance: GE Healthcare announced a research collaboration with National Cancer Centre Singapore (NCCS) focused on developing new AI-powered data analytics and clinical workflows solutions that will be used to create more personalized cancer treatment methods. The collaboration will combine GE’s imaging and monitoring expertise with NCCS’s expertise in R&D, testing, and consultation.
