“Also doesn’t mean these algorithms are useless btw”
UMC Utrecht medical statistician and epidemiologist, Maarten van Smeden, on studies showing that AI can’t outperform humans.
Imaging Wire Sponsors
Arterys | Bayer Radiology | Blackford Analysis
Canon Medical Systems | Fujifilm Healthcare Americas
GE Healthcare | Novarad | Nuance
Riverain Technologies | Siemens Healthineers
United Imaging | Zebra Medical Vision
The Imaging Wire
Bad AI Goes Viral
A recent mammography AI study review quickly evolved from a “study” to a “story” after a single tweet from Eric Topol (to his 521k followers), calling mammography AI’s accuracy “very disappointing” and prompting a new flow of online conversations about how far imaging AI is from achieving its promise. However, the bigger “story” here might actually be how much AI research needs to evolve.
The Study Review: A team of UK-based researchers reviewed 12 digital mammography screening AI studies (n = 131,822 women). The studies analyzed DM screening AI’s performance when used as a standalone system (5 studies), as a reader aid (3 studies), or for triage (4 studies).
The AI Assessment: The biggest public takeaway was that 34 of the 36 AI systems (94%) evaluated in three of the studies were less accurate than a single radiologist, and all were less accurate than the consensus of two or more radiologists. They also found that AI modestly improved radiologist accuracy when used as a reader aid and eliminated around half of negative screenings when used for triage (but also missed some cancers).
The AI Research Assessment: Each of the reviewed studies were “of poor methodological quality,” all were retrospective, and most studies had high risks of bias and high applicability concerns. Unsurprisingly, these methodology-focused assessments didn’t get much public attention.
The Two Takeaways: The authors correctly concluded that these 12 poor-quality studies found DM screening AI to be inaccurate, and called for better quality research so we can properly judge DM screening AI’s actual accuracy and most effective use cases (and then improve it). However, the takeaway for many folks was that mammography screening AI is worse than radiologists and shouldn’t replace them, which might be true, but isn’t very scientifically helpful.
UC Davis’ Ultra High Resolution CT Experience
See Dr. Brian Goldner of UC Davis Sacramento detail his experience with Canon’s Ultra High Resolution CT and how it can be applied to cardiothoracic interpretations.
What to Expect from Fujifilm Healthcare Americas
Check out this Imaging Wire Show interview with Fujifilm Healthcare Americas’ Dave Wilson, detailing what we can expect as Hitachi Healthcare becomes part of Fujifilm.
- Canon’s PCCT Acquisition: Canon advanced its photon-counting CT (PCCT) technology portfolio with its acquisition of CZT detector developer, Redlen Technologies. Canon already owned 15% of Redlen, and its additional ~$268m investment will give it full control of the company, while fast-tracking Canon’s PCCT product development strategy. It seems like there’s a PCCT detector race forming, as GE Healthcare similarly acquired its photon-counting CT detector partner, Prismatic Sensors, in late 2020.
- 4D Flow WWS, an Aortic Biomarker: Northwestern researchers found that 4D Flow Cardiac MRI-based wall shear stress (WSS) analysis can help identify patients with higher future risks of aortic degeneration. The researchers retrospectively applied 4D Flow to CMRI exams from 72 patients with bicuspid aortic valves (CMRI exams at baseline & ≥5yrs). The patients who had higher annual ascending aorta growth rates were nearly 4x more likely to have elevated WSS levels in their baseline scans than patients with lower growth rates.
- Optimizing Telemedicine: A Stanford team called for the development of an optimized telemedicine model, given the technology’s rapid adoption and presumed staying power. According to the authors, an optimal telemedicine care model must include: 1) Support for multidisciplinary care team consultations – yes, including radiology; 2) Optimized experiences for both clinicians and patients; 3) Embedded quality and patient-reported outcome metrics.
- ImageLink’s Consumer Pivot: Imaging center chain ImageLink (15 locations in GA, OH, FL) announced a strategic pivot to a consumer-driven business model that caters to patients who seek out their own imaging exams and pay out-of-pocket. ImageLink has big post-pivot ambitions, appointing a new leadership team with deep consumer-focused imaging experience (former Smart Choice MRI execs) and revealing plans to actively acquire more imaging centers.
- Mobile Stroke Units Work: Acute stroke patients who are transported in mobile stroke units (MSUs) receive faster treatment and have better outcomes than patients transported by EMS. That’s from a NEJM study (n = 1,047 eligible patients) that found MSUs’ in-transit stroke diagnoses result in 36 minutes faster tissue plasminogen activator (tPA) treatment times (72 vs. 108 min), higher 90-day Rankin scores (0.72 vs. 0.66, 0-1 range), and lower 90-day mortality rates (8.9% vs. 11.9%).
- Intelerad’s UK Acquisition: Intelerad continued its acquisition spree, acquiring UK-based enterprise imaging company, Insignia Medical Systems. That’s Intelerad’s fifth acquisition in the last year, following Heart Imaging Technologies (cardiac viewing/reporting), LUMEDX (CVIS), Digisonics (cardiovascular & OB/GYN PACS), and Radius (cloud tech). However, this seems to be more than just another technology/portfolio add-on for Intelerad, as the acquisition also establishes Intelerad in the United Kingdom.
- Same-Day Impact: Reading screening mammograms on the same day they are performed improves time-to-diagnosis and reduces racial/ethnic disparities. That’s the takeaway from MGH’s immediate-read program that launched during COVID’s May 2020 peak. During June-October 2020, 60.7% of patients with abnormal screenings received same-day diagnostic imaging, up from just 14.8% during June-October 2019 (n = 7,235 & 8,222 screenings, 359 & 521 abnormal). The program also eliminated racial disparities in same-day diagnostic imaging follow-ups, which were previously “significant” for Black and Hispanic women.
- SOMATOM X.ceed’s FDA: Siemens Healthineers’ new SOMATOM X.ceed single-source CT scanner is now FDA-approved. In addition to its premium specs (large 82cm gantry, 50cm spectral FoV, fast 0.25s rotation speed), the SOMATOM X.ceed is highlighted by its new myExam Companion (guides/automates technologist operations) and myNeedle Companion (supports CT-guided needle procedure planning/insertion) solutions.
- Less is More: Now that the ONC’s ban on healthcare information blocking is live, a new JAMA Oncology viewpoint examined the downside of sharing detailed medical reports with patients. While complete transparency gives patients a clear view of their situation, the authors found that the blunt analytical tone of these notes often leads to avoidable patient confusion and anxiety.
- SR Adds Radiology Associates, Corpus Christi: Strategic Radiology added Corpus Christi, Texas’ Radiology Associates LLP (18 radiologists, 5 imaging centers, 80yrs in practice) to its consortium of independent radiology practices. SR now has 31 member practices (2 in Texas) with over 1,300 radiologists.
- COVID’s Rule of 7: Spanish researchers detailed how lung ultrasound (LUS) and patient/clinical factors can be combined to accurately triage/prioritize suspected COVID patients, which might be particularly helpful in low-resource settings. Using data from 228 patients, the researchers’ new “Rule of 7” predicted mortality among patients who were over 70 years-old, had LUS scores over 7, and C-Reactive Protein levels (CRP) over 70 g/L with a 0.813 AUC. Overall, the patients who died and survived had very different average ages (81.8yrs vs. 57.9yrs), lung scores (13.9 vs. 9.1), and CRPs (122.8 vs. 43.3).
- contextflow’s Series A: Austrian AI startup contextflow just closed an $8m Series A round that it will use to fund its commercialization in the US and Europe, support FDA approvals, and expand into new modalities and organs. Contextflow takes a different approach than many other AI players (it’s not modality/disease-specific, not triage-focused, and it is explainable). Its “SEARCH” 3D system detects disease patterns in various scans to aid radiologists’ clinical decision making.
Improving Efficiency with Arterys Lung AI
See how Arterys Lung AI allowed radiologists to reduce their thoracic CT interpretation times by between 35% and 55%, while increasing their nodules reported per study.
The Resource Wire
- More efficiency and accuracy – less burnout and IT overhead. Those are the key results from adopting cloud speech technology detailed by Nuance in this infographic.
- As part of United Imaging’s Equal Healthcare For All mission, they embed their most innovative technology in all MI systems. This forward innovation approach means that even their most financially accessible scanner can do a total body PET exam in five minutes, without compromising imaging quality.
- Learn how Novarad’s new CryptoChart solution gives radiology departments a simple way to ditch the disk, and the headaches and costs that come with them.
- Check out this Imaging Wire Show interview with Blackford Analysis founder and CEO, Ben Panter, detailing how to solve AI’s assessment and deployment problem, AI’s downstream value, and what it will take for AI to have its greatest impact.
- See why AI-powered cardiovascular screening’s detection and prevention benefits make it the new frontier of population health.
- This Riverain Technologies case study details how the University of Colorado Hospital enhanced its chest X-ray workflow with ClearRead Bone Suppress.
- See how GE Healthcare’s AI-ready PACS will help radiology unleash the promise of AI.
- Learn how Salem Regional Medical Center improved its radiology workflows and cut service and syringe expenses after adopting Bayer’s MEDRAD Stellant FLEX system.
- Dynamic imaging on Dual Source CT can help you provide more confident, precise, and accurate answers. Download Siemens Healthineers’ clinical case collection to see powerful examples where functional imaging with DSCT is making a difference.