“because healthcare is hard . . .”
Business Insider on why Walmart is taking its time with its health clinic rollout.
Imaging Wire Sponsors
- Arterys – Reinventing imaging so you can practice better and faster.
- Bayer Radiology – Providing a portfolio of radiology products, solutions, and services that enable radiologists to get the clear answers they need.
- GE Healthcare – Enabling clinicians to make faster, more informed decisions through intelligent devices, data analytics, applications and services.
- Healthcare Administrative Partners – Empowering radiology groups through expert revenue cycle management, clinical analytics, practice support, and specialized coding.
- Hitachi Healthcare Americas – Delivering best in class medical imaging technologies and value-based reporting.
- Novarad – Transformational imaging technologies that empower hospitals and clinicians to deliver clinical, operational and fiscal excellence.
- Nuance – AI and cloud-powered technology solutions to help radiologists stay focused, move quickly, and work smarter.
- Riverain Technologies – Offering artificial intelligence tools dedicated to the early, efficient detection of lung disease.
- Siemens Healthineers – Shaping the digital transformation of imaging to improve patient care.
- Zebra Medical Vision – Transforming patient care with the power of AI.
The Imaging Wire
AI Advice Influence
AI-based diagnostic advice isn’t always correct and it isn’t always trusted, but only one of those factors influences physicians’ diagnostic accuracy. That’s from a new NPJ Digital Medicine study and here are the details:
- The Study – The researchers provided radiologists (n = 138) and internal/emergency medicine physicians (IM/EM, n = 127) with eight chest X-ray cases and corresponding diagnostic advice that they were told either came from human colleagues or an AI model. The physicians were asked to evaluate advice quality and make final diagnoses.
- The Catch – All of the advice actually came from humans (not AI) and some advice was inaccurate.
- Human Bias – As you might expect, the radiologists rated the ‘AI-based’ advice as lower quality than the ‘human-based’ advice, while the IM/EM physicians did not.
- Confirmation Bias – As you might also expect, the radiologist and IM/EM physicians’ diagnostic accuracy was “significantly worse” when they were provided inaccurate advice (rads -40.10%, IM/EMs -37.53%), regardless of whether they thought the advice was ‘AI-based’ or ‘human-based.’
- The Takeaway – In addition to confirming the importance of using accurate advice, this study shows that AI-based advice does indeed influence clinical decision making (other studies have said it doesn’t) and that physicians struggle to overrule inaccurate advice even if it came from a source that they view as lower quality (radiologists w/ AI).
- The Applications – These findings might reveal an opportunity for clinical teams and AI developers to create ways to avoid confirmation bias (e.g. only providing AI-based advice when requested, providing advice confidence scores). They also reveal a need to build radiologists’ trust in AI, as systems like this are adopted (e.g. additional training, more hands-on experience).
- Reconsidering Walmart Clinic: Walmart might be scaling back its plans to open 4,000 standalone health clinics (some w/ in-house imaging) by 2029 due to leadership changes, shifting priorities (COVID & eCommerce), and because the retail giant is still experimenting with its health clinic strategy. Although Walmart has only opened about 20 of the 4,000 clinics so far, this has been a major strategic initiative (w/ a $3b budget) and has the potential to change how/where many Americans received their healthcare.
- Multiple Nodule Malignancy Predictor: A China-based research team developed a machine learning tool that uses chest CT and socioeconomic data to accurately predict malignancy risk of patients with multiple pulmonary nodules. The PKU-M tool (trained w/ 520 patients’ CT nodule characteristics & socioeconomic data) predicted cancer within a 220-patient test set with a 0.89 AUC, outperforming four other logistic regression-based models (0.68 to 0.81). Then with a separate 78-patient set, PKU-M predicted cancer with an AUC of 0.87, beating three thoracic surgeons (0.73 – 0.79), a radiologist (0.75), and a previously developed AI tool (0.76).
- icometrix Adds Ischemic Stroke: icometrix expanded its icobrain portfolio with the U.S. and European launch of its new icobrain cva solution (now FDA & CE cleared), which analyzes acute ischemic stroke patients’ CT perfusion scans and provides physicians with tissue perfusion status to support treatment decisions. The icobrain portfolio also includes AI-based solutions for the detection/assessment of epilepsy (MRI), MS (MRI), dementia (MRI), and TBI (CT).
- CTC AI for Polyp Differentiation: A German research team developed a deep learning model capable of differentiating benign and premalignant colorectal polyps using CT colonography radiomics data, suggesting that this method could eventually be used as a CTC second reader. The algorithm (trained on: 107 polyps, 169 segmentations, 63 patients; tested on: 77 polyps, 118 segmentations, 59 patients) differentiated benign and premalignant polyps with a 0.91 AUC, 82% sensitivity, and 85% specificity.
- Probo Goes to France: Probo Medical continued its European expansion with its acquisition of French pre-owned imaging equipment, service, and parts company, IMAX Medical. This is another notable expansion milestone for Probo, which first expanded beyond ultrasound and outside of the U.S. through a pair of February 2020 acquisitions and just completed its second UK equipment/parts/service company acquisition last month.
- More COVID Vaccine False Positives: We’ve read/written plenty about the COVID-19 vaccines’ potential to mimic signs of breast cancer in mammograms, but a new case series published in Radiology reveals vaccine side effects can also be confused for axillary lymphadenopathy (change in lymph node shape/consistency) across “almost all modalities.” The article detailed multiple exams that initially revealed signs of metastasis (leading to biopsy and/or additional imaging) and were later found to be due to the vaccine.
- Brainlab’s FDAs: Brainlab announced the FDA approval of its Loop-X Mobile Imaging Robot (already has CE Mark), calling it the first fully robotic mobile intraoperative imaging device on the market. The Loop-X features independently moving imaging source and detector panels (supports patient positioning and non-isocentric imaging) and integrates with digital surgery systems, including Brainlab’s.
- AB MRI Screening Effective Over Time: New research out of South Korea revealed that abbreviated breast MRI screenings maintain diagnostic value over multiple years, adding more evidence supporting the fast/efficient protocol’s mainstream potential. The researchers performed annual AB-MRI screenings on 1,975 women for 3 years (97% mid-to-high risk), finding that many key diagnostic factors improved from the first to the third year including CDR (6.9 to 10.7 per 1k), sensitivity (75% to 80%), and specificity (93.5% to 94.1%), while overall outcomes were consistent throughout the three years. AB-MRI caught 29 of 38 cancers, and the 9 cancers that it missed were invasive, but node negative and caught by other modalities at an early stage.
- MedX JV: Medica Group (UK/Ireland-based telerad provider) and Integral Diagnostics (Australia/NZ-based imaging services provider) just launched their new MedX teleradiology joint venture. MedX will initially staff dual-qualified radiologists (both UK/IR & AU/NZ) to perform teleradiology services for both companies, with future plans to offer teleradiology services to other firms and regions.
- A Call for BiEN Training: Radiology education curriculums need to cover biomedical engineering basics or else future radiologists won’t be prepared for, or be able to shape, the specialty’s future. That’s from a new Diagnostic and Interventional Imaging editorial that detailed some of the latest medtech breakthroughs that will require radiologist involvement (artificial tissue engineering, 3D printing-based care, stem cell transplants) but aren’t addressed in today’s rad training.
- How to VNAI: A new paper out of Utrecht University Medical Center suggests that the wide gap between AI activity/interest and actual clinical adoption is largely because most institutions don’t have a dedicated vendor-neutral AI deployment infrastructure (VNAI). The good news is, the authors just spent the last four years building UMC Utrecht’s VNAI and they provided a (very detailed) guide that other hospitals could use for creating and deploying their own AI infrastructure.
- VI-RADS for BCa Muscle Invasion: A new study in European Radiology found that VI-RADS accurately predicts bladder cancer (BCa) muscle invasion, while highly recommending the structured reporting system. In this prospective multicenter study, 331 patients with suspected/untreated BCa received bladder mp-MRI exams and four radiologists evaluated the scans using VI-RADS. The study identified VI-RADS 4-5 as the appropriate cutoffs for predicting BCa muscle invasion after patients received their first transurethral resection bladder tumor procedure (TURBT; 84.1% sensitivity, 92.3% specificity) and VI-RADS 2-5 after a second TURBT procedure (89.9% sensitivity, 90.1% specificity).
- Brain MRI Merger: Brain MRI software company SpinTech acquired MRI software peer/partner Magnetic Resonance Innovations (MR Innovations) bringing aboard its team of imaging experts, patents, and global collaborators. The merging Michigan-based companies will rebrand as SpinTech MRI, leveraging their combined capabilities for future product development.
- Developing World Breast Exam Alternative: In developing countries without sufficient mammography access, performing clinical breast exams (without imaging) every two years still helps catch cancers. That’s from a 20-year Mumbai, India-based study that screened 75k women (4 rounds of clinical exams every two years) and provided 76k women with just cancer awareness materials (control group), finding that the screening group had earlier breast cancer diagnoses (55.18yrs. vs. 56.5yrs), far fewer stage III-IV cancers (37% vs. 47%) and 30% lower breast cancer mortality rates among >50yr participants.
The Resource Wire
– This is sponsored content.
- This Riverain Technologies case study details how Duke University Medical Center integrated ClearRead CT into its chest CT workflows, reducing read times by 26% and improving nodule detection by 29%.
- Novarad’s COVID-19 AI Diagnostic Assistant analyzes chest CT scans in seconds, helping physicians quickly identify COVID-19 patients and get them into care. The best news – it’s available to clinicians worldwide free of charge.
- Learn how the University of Oxford increased its vertebral compression fracture detection rate from 50% to 90% using Zebra Medical Vision’s bone health solution.
- Check out the four major shifts driving the era of intelligent CT imaging in a post-COVID world.
- Learn how Salem Regional Medical Center improved its radiology workflows and cut service and syringe costs after adopting Bayer’s MEDRAD Stellant FLEX system.
- Check out how Nuance’s mPower Clinical Analytics solution helped Summa Health improve its incidental lung nodule follow-up rates by nearly 8x.
- Documentation is critical in order to be properly reimbursed for yttrium-90 (y-90) radioembolization procedures. Learn how to document these procedures for maximum reimbursement in this blog article from Healthcare Administrative Partners.
- Learn how Hitachi’s VidiStar PACS system evolved to meet the needs of pediatric cardiologists in this Imaging Wire Q&A.
- Did you know Arterys developed the first ever FDA-cleared algorithm leveraging cloud computing and artificial intelligence? Check out this comprehensive database of AI/ML-based medical technologies approved by the FDA.
- In this GE Healthcare post, University of Wisconsin’s Tim Szczykutowicz, PhD, DABR outlines how hospitals are making their CT workflows more efficient by using technology to automate and simplify tasks.