AI Trough | Sharing POCUS | Effective abMRI

“It is an illusion to assume that radiologists will be able to cover all of these ultrasonography studies in a hospital by themselves around the clock in a timely manner,”

Robert & Thomas Kwee, encouraging radiologists to share POCUS interpretations with their frontline colleagues.

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.
  • Canon Medical Systems – Delivering innovative imaging solutions and services through industry leading partnerships to improve the quality of life for all people.
  • 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.
  • United Imaging – Our mission, Equal Healthcare for All, pushes us beyond conventional boundaries to help clinicians expand modern, digital, intelligent care to more people within their communities.
  • Zebra Medical Vision – Transforming patient care with the power of AI.

The Imaging Wire

A Week in the Disillusionment Trough

AI is coming off a tough publicity week that featured a pair of high-visibility exposés about its data and research challenges, followed by a study detailing one commercial AI tool’s clinical underperformance. Here are some details on AI’s first official week in the “Trough of Disillusionment.”

  • ML Reckoning – IEEE Spectrum detailed healthcare AI research’s “lack of reproducibility and poor quality,” suggesting that AI research has a lot of work to do. The article detailed one study that found healthcare machine learning papers were far less reproducible than other machine learning subfields, and another recent COVID AI paper (already widely covered) that found COVID imaging AI studies were generally flawed, not reproducible, and nowhere near ready for clinical use.
  • Foundational Errors – Two days later, Wired highlighted a recent MIT study that found ImageNet and nine other key AI datasets (many non-medical) are “riddled with” labeling errors. The MIT researchers advised developers who plan to use these datasets to first “clean the errors out,” while the article also called into question how labels are produced (low-paid workers) and even whether labels are the best option for most AI training.
  • UW’s Generalizability Concerns – The University of Wisconsin published a study detailing how Aidoc’s intracranial hemorrhage AI tool performed worse than expected, calling for improved AI generalizability testing standards. Their review of 3,605 non-contrast head CTs found that the Aidoc tool and UW’s neuroradiologists were concordant on 96.9% of exams (3,494 concordant exams w/ 349 ICHs), while Aidoc’s 81.3% positive predictive value and 92.3% sensitivity were both “unexpectedly lower” than the tool’s previous studies. The researchers also weren’t able to identify the cause of this discrepancy.
  • Some Context – People tend to paint AI with a wide brush (all good or all bad), and that’s especially true after last week. However, most imaging AI insiders would agree that academic papers can be sloppy, AI training would benefit from a lot more high-quality labeled images, and that we need better ways to validate generalizability. Most would also agree that we’re in the very early innings of the AI ballgame (lots of evolution coming) and AI is already helping clinicians every day. Stories like these just make better headlines.

The Wire

  • Sharing POCUS: A new European Journal of Radiology editorial called for radiologists to get over their desire to interpret all POCUS exams, arguing that radiologists aren’t equipped to handle rising POCUS volumes or their quick turnaround requirements. The authors instead encouraged radiologists to collaborate with their frontline colleagues to define who should interpret which exams (based on: patient need, expertise, timelines, available rads, available POCUS-trained physicians, etc.). Even though POCUS is already widely used by non-rads, most previous statements from the radiology side of the POCUS turf debate have been far less collaborative.
  • More abMRI Evidence: A new George Washington University study found that abbreviated breast MRI (abMRI) detects cancer comparably to full protocol MRI (fpMRI) and radiologists can interpret the quick 10-minute scans much faster. In the retrospective study (n = 334 abMRI & fpMRI exams, 286 women, 5 cancers) radiologists spotted all five cancers with both protocols, abMRI and fpMRI had similar recall rates (24.6% vs. 24.3%), and abMRI had far shorter average interpretation times (60.7 seconds vs. 99.4 seconds).
  • Nanox’s First FDA: Nanox hit a key milestone last week with the FDA 510(k) clearance of its single-source Nanox.ARC system. Although Nanox currently plans to commercialize its forthcoming multi-source Nanox.ARC system (not this single-source version), the added credibility and regulatory momentum from this clearance are positives for the often-criticized “imaging disruptor.”
  • Endothelial Innovation: Berkeley Lab announced a new endothelial health assessment technique (predicts heart attack & stroke) that they say is more effective, lower-cost, and easier to operate than current ultrasound-based flow-mediated dilation exams. The technique uses a linear actuator to create an artificial pulse in patients’ radial artery and an ultrasonic Doppler stethoscope to detect the pulse 10-20cm upstream, producing a pulse transit time (PTT) measurement. Berkeley Lab will license the new technology to medtech companies for use in at-home or wearable endothelial health devices.
  • Integrating the UK’s COVID Database: Philips and the Cheshire and Merseyside NHS launched the first data integration hub for the UK’s National COVID-19 Chest Imaging Database. With the integration, the Cheshire and Merseyside NHS will provide the National Database with 15 years of imaging data that scientists can use to better understand COVID-19 and validate COVID AI tools.
  • Mammography Lag: A new National Cancer Institute study (n = 573k mammograms, Jan 2019 – July 2020) revealed that screening and diagnostic mammography volumes approached pre-pandemic levels by July 2020 (89.7% & 101.6% of July 2019 levels), but this rebound lagged among Asian and Hispanic women (51.3% & 72.7% of July 2019 levels). The shutdown and uneven rebound also created a January-July 2020 screening and diagnostic mammogram deficit (66.2% & 79.9% vs. Jan-July 2019 totals) that at least partially still remains today.
  • Siemens Prostate MR Works: A study from Siemens Healthineers and international researchers showed that Siemens’ Prostate MR solution can improve radiologists’ accuracy and efficiency when identifying suspicious lesions in prostate MRI exams. The study used 100 prostate MRI cases (with & without prostate cancer) and had seven radiologists read each case with and without help from the Prostate MR system. The radiologists’ average accuracy detecting clinically significant cases (PI-RADS ≥4) increased from 0.84 to 0.88 when using Prostate MR, while interreader concordance (0.22 to 0.36) and median reading times (103 to 81 seconds) also improved.
  • MILabs On the Block: MILabs BV’s venture capital parent company, Thuja Capital, is reportedly looking to sell the Dutch imaging manufacturer for as much as €150m, potentially to “medical device makers in Asia.” MILabs is best known for its preclinical/research imaging systems (micro-PET, micro-SPECT, micro-CT, and Optical Tomography), but it also produces clinical SPECT systems.
  • AI ROI, More than Codes: Healthcare Administrative Partners just published an AI economics overview, detailing potential AI payment scenarios and ways that AI can contribute to radiology practices until reimbursements become more of a reality. Medicare currently pays for AI use in two specific cases (diagnosis of diabetic retinopathy via MPFS, and diagnosis of LVO strokes via IPPS) although the current fee-for-service environment suggests that AI might have a bigger role in value-based systems (e.g. MQPP’s Alternative Payment Models). However, HAP emphasized that AI can still deliver ROI beyond reimbursements, including solutions that ensure more patients come back for their follow-up scans and tools that streamline the clinical decision support process.
  • The Case for Twice-Weekly COVID CXRs: Repeat chest X-ray exams are useful for monitoring hospitalized COVID-19 patients, but CXRs are only necessary every three or four days. That’s from new research out of Pakistan that analyzed all scans from 112 hospitalized COVID patients, finding that the patients’ CXR symptoms were most severe 10-13 days after symptoms began, followed by CXR improvements at roughly 15 days. Based on this timeline, the researchers suggest that there is no need to perform daily CXRs or ongoing CT scans in most patient scenarios.
  • The Flexible X-Ray Detector: Scientists from Singapore and China are developing a flexible 3D X-ray detector that can wrap around objects, potentially improving 3D image resolution when scanning curved objects. The new detector owes its flexibility to its use of luminescent nanoparticles (doesn’t require bulky circuits like current panels) and it could lead to new applications for point-of-care X-ray, compression-free mammography, and image-guided therapy.
  • CMRI’s Screening Potential: Cardiac MRI doesn’t currently have a role in cardiovascular disease screening, but its list of CVD imaging strengths (evaluating cardiac structure and function; tissue characterization) might eventually earn it a screening role for patients with higher cardiometabolic risk. That’s from a new paper in Radiology: Cardiothoracic Imaging that reviewed the shortcomings of current cardiac modalities and suggested that CMRI may fill some of those gaps with the right future clinical validation efforts.
  • COVID’s Cancer Impact: One year into COVID-related diagnostic and treatment delays, an ASTRO survey (n = 117) revealed that 66% of radiation oncology practice leaders are treating patients with more advanced cancers than before the pandemic, while 73% are treating patients who missed screenings and 66% have patients who experienced treatment interruptions.

The Resource Wire

– This is sponsored content.

  • GE Healthcare just launched the 2021 X-Ray Technologist Awards, recognizing technologists who went above and beyond during the COVID-19 pandemic. If any of your RT colleagues stepped-up in the areas of Patient Experience, Clinical Excellence, and Workflow Improvement over the last year, nominate them today.
  • 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%.
  • Researchers at Saint Joseph Hospital (Paris, France) used Arterys’ Cardio AI solution in this study detailing how 4D flow is becoming essential for assessing cardiovascular disease.
  • Zebra-Med’s Bone Health Solution helps radiologists automatically identify patients at risk for osteoporosis, allowing HMOs to enroll more patients in bone health programs.
  • In this Novarad video, interventional oncologist Gary M. Onik, MD shares how Novarad’s AR surgical navigation system, OpenSight, helps his team accurately access and treat tumors.
  • Every decision United Imaging makes, every innovation they choose to pursue, comes down to their mission: Equal Healthcare For All. Hear more about their mission.
  • This Bayer Radiology case study details how Einstein Healthcare Network reduced its syringe costs, enhanced its syringe loading, and improved its contrast documentation when it upgraded to the MEDRAD Stellant FLEX CT Injection System.
  • In this quick video, Sharp Memorial Hospital interventional radiologist, Jim Lyon, MD, describes the image quality and dose advantages of Canon Medical Systems’ Alphenix Sky+ system.

You might also like

You might also like..

Select All

You're signed up!

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

-- The Imaging Wire team

You're all set!