“. . . we will end up reading 10% more studies for the same pay until the insurance companies and hospital systems decide AI + non-physicians signing off is good enough for their customers.”
One Auntminnie.com forum commenter’s take on what happens after CMS cuts radiology reimbursements by 10% next year. Other comments were even more pessimistic.
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.
- 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.
The Imaging Wire
One for All AI
A team of German scientists unveiled a new “one for all” self-learning imaging AI segmentation method that can develop models using a variety of datasets (different modalities, sites, brands, etc.) for a wide range of tasks without requiring a specialist’s knowledge or high-powered computers. Those are some big claims considering the way we develop AI models now, so you deserve some details:
- nnU-Net Intro – The team’s nnU-Net deep learning-based segmentation method automatically configures itself (including: preprocessing, network architecture, training, post-processing) for any task and any imaging dataset, “allowing even people with limited prior expertise to configure self-learning algorithms.”
- nnU-Net Comparison – In international competitions, nnU-Net achieved the top results for 33 out of 53 different segmentation tasks, despite competing against algorithms that were specifically developed for each of the tasks (and developed by specialists). That’s a better record than the 2020 Patriots.
- Up Next – The researchers are making nnU-Net available as an open source tool to support large-scale clinical studies and potentially/eventually clinical use. That’s right, nnU-Net can work with any dataset, it doesn’t require specialist involvement, and it will be available open source.
The Wire
- Orthopod Advantage: Orthopedic surgeons can diagnose surgically relevant rotator cuff tears when performing their own ultrasound scans with equal or greater accuracy than radiologists. That’s from an International Journal of Orthopaedics study (you guessed it) comparing the results of 88 surgeon-performed shoulder ultrasounds against radiologist-read MRIs and ultrasound scans from the same 88 patients, finding that Surgeon-US had significantly greater specificity (98.2 vs. 81.8), positive predictive value (97.6 vs 79.6), and overall accuracy (95 vs. 84) than the radiology-read scans. The study suggests that these results support the use of one-stop shoulder clinics.
- Fired from Home: A veteran NYC radiologist is suing New York City Health for allegedly forcing him out of his Bronx hospital after he asked to work from home during the COVID emergency due to concerns over his autoimmune disease (ulcerative colitis). A day after his WFH request in March the radiologist received a negative six-month review citing his low accuracy rates (93.8%) and asking him to resign or be fired.
- An Interesting RaaS Patent: Teleradiology company USARAD Holdings announced a new U.S. patent for a decentralized method of matching medical imaging facilities with AI-supported remote radiologists, and managing associated financial transactions using blockchain tech. This is the core of what USARAD is calling radiology-as-a-service and it would also play a major role in what its closely-related partner Nanox has been promoting as its “medical screening as a service” (MSaaS) strategy. This all might seem futuristic, but the idea of automating image interpretation using matchmaking workflow and AI diagnostics and then facilitating payments using blockchain might prove to be pretty realistic (in the future…).
- An MRI Mask Warning: The FDA warned patients and providers to make sure they wear face masks without metal parts or coatings during MRI exams after a patient’s face was burned during a neck MRI. Noting that it might be hard to tell whether a mask has metal in it, the FDA encouraged patients to confirm with their MRI technologist whether a mask includes any metal.
- AstraZeneca & Qure.ai Think Globally: AstraZeneca will use Qure.ai’s qXR chest X-ray AI solution to support its early lung cancer detection efforts in emerging regions (Latin America, Asia, Middle East, Africa). AstraZeneca will use qXR to identify patients with suspicious lung abnormalities in countries that have traditionally faced radiologist shortages and then guide them to diagnosis and treatment.
- CCTA Matches SPECT: A new study in JAHA found that coronary CTA (CCTA) matches SPECT myocardial perfusion imaging for guiding the treatment of patients with stable angina. A trial featuring 1,050 patients (518 w/ CCTA, 532 w/ SPECT) over 16 months found that the two patient groups had similar major cardiac event rates (MACE), while patients with negative CCTA results were less likely to have MACE or revascularization than patients with negative SPECT results (1.2% vs. 3.2%).
- Fixing AI Measurements: Dr. Luke Oakden-Rayner shared another awesome blog last week (and associated pre-print study) arguing that just about every AI vs. radiologist study is measuring the wrong things — and systematically underestimating doctors in the process. Instead of comparing AI performance against a group of radiologists’ average sensitivity and specificity (pretending they aren’t inversely correlated) or using ROC scores (relying on clinically unrepresentative radiologist confidence ratings), Dr. Luke suggests using each radiologists’ findings as a point on a summary ROC (SROC) curve. With this “simple off-the-shelf method” we can better understand the performance of a group of radiologists and accurately compare AI models against the “average radiologist.”
- GE Security Warning: Medical cybersecurity firm CyberMDX reported that a range of GE Healthcare medical imaging devices (XR, CT, MRI, US, mammography, IR, PET/CT) could be hacked, exposing patient data or allowing hackers to disrupt device operation. CyberMDX attributed the vulnerability to the devices’ use of hardcoded default passwords (they have to be changed by a GE engineer), which hackers could use if they gained access to the same network as the devices (e.g. an employee opens an email with malware). GE has already addressed these issues in newer models and is supporting older models through its service teams.
- Same Day Conflict: A Northwell Health survey revealed that patients and referring physicians have very different imaging results communication preferences. Although 82% of surveyed patients (n = 368) prefer that radiologists communicate their imaging results on the same day as their exams and 66.4% experience anxiety waiting for their results, 87.6% of referring physicians (n = 168) don’t even want radiologists to directly share results with their patients. The study called for a major patient-centric shift in how / when results are communicated.
- Zebra’s X-ray Bone Measurement FDA: Zebra Medical Vision announced the FDA approval of its new orthopedic surgery planning AI solution that extracts bone measurements from X-ray scans with the same quality as a CT scan (its 7th FDA-cleared product, first in orthopedics). The new solution enables the use of standard X-rays for surgery planning, which Zebra suggests could transform orthopedic surgery by allowing surgeons to perform implant size measurements in offices or regions where CT scans and MRIs are not available/affordable.
- MPFS Cuts Breakdown: With the 2021 evaluation and management (E/M) MPFS changes now official, Healthcare Administrative Partners detailed how bad the radiology reimbursement cuts will actually be next year. Big picture: the conversion factor will fall by 10.2% to $32.41, bringing notable reductions for diagnostic radiology (-10%), IR (-8%), nuclear medicine (-8%), and radiation oncology/therapy (-5%), and cutting most diagnostic reimbursements by 10% to 11% (although CT Thorax will be cut by 17%). HAP analyzed these cuts from a lot of angles, so check out their analysis and/or their upcoming webinar to learn how you’ll be impacted.
- Bayer Radimetrics v3.0: Bayer Radiology announced the launch of Radimetrics v3.0, adding a range of workflow and compliance improvements to its enterprise dose management application. Radimetrics v3.0 supports many of the latest accreditation standards and brings upgrades to the system’s UI, visualization/analytics, data tracking, and cybersecurity features (among other upgrades).
- Taking MEG to Market: The University of Nottingham and Magnetic Shields Ltd (MSL) just created a new public/private spin-out company called Cerca Magnetics Limited to launch their “world’s most advanced brain scanner.” Cerca Magnetics’ wearable magnetoencephalography (MEG) scanner is designed to allow people to move around while scanning is taking place, providing “an unprecedented window on brain function” for people with severe neurological illnesses (e.g. epilepsy) including younger patients who struggle to remain still. This is the second time this scanner has made it onto The Imaging Wire and it seems like they’ve made solid progress over the last year.
- Aidoc & Imbio’s PE Partnership: Aidoc and Imbio will co-develop the first end-to-end AI solution for pulmonary embolism (PE), combining Aidoc’s PE detection AI tool to prioritize suspected pulmonary embolism and Imbio’s quantitative imaging analysis technology to automatically calculate RV/LV ratio. The solution is currently only available for investigational use.
The Resource Wire
– This is sponsored content.
- Check out NVIDIA’s Blog post about AI Startups, like Arterys, and their contributions to the fight against COVID. Arterys tells the epic story of its Marketplace launch, and details the various AI models developed to detect COVID in CT images and X-rays.
- To honor Breast Cancer Awareness Month, Siemens Healthineers hosted a special webinar on Breast MRI featuring Dr. Susan Weinstein. Register to access the recording and watch it at your own convenience.
- Learn how rapid breast cancer diagnosis clinics are reducing diagnoses-to-consultation turnarounds “from weeks to hours,” and how they are helping patients and clinicians in this new GE Healthcare Insight.
- In this Bayer Radiology video, East Texas Medical Center Radiology Director, Bill Tobin, details how they used Bayer’s MEDRAD Stellant Smart Injector and contrast dose management to reduce contrast volumes and repeat scans.
- Wondering what to look for in your next CT? This Hitachi blog goes beyond slice count, detailing the top three features that your next CT has to have. Here’s a hint: they will help you care for your >30% patients who are overweight.
- CPT updates, E/M services changes, CDS, and MIPS are just some of the topics covered in Healthcare Administrative Partners’ upcoming 2021 MPFS Updates & Radiology Reimbursement Impact webinar. Register today!
- It says a lot when a solution works so well for a radiology department that they decide to perform a study to quantify its benefits. In this Imaging Wire Q&A, University Hospital of Zurich’s Thomas Frauenfelder discusses his experience and study on Riverain Technologies ClearRead CT.
- In its latest Q&A, Nuance sat down with Nines CEO David Stavens, PhD, and medical director Dr. Jean-Paul Dym, MD, to discuss their experiences with PowerScribe One and their perspectives as a Nuance innovation partner.
