Uncategorized

Imaging Trolls | 1K Rads on AI | Streamlined Stenosis


“This isn’t a gray area,”

Northeastern University professor, Gary Young, after finding that patients are far more likely to get referred for inappropriate MRIs when their physician is hospital-employed.



Imaging Wire Sponsors

Arterys | Bayer Radiology | Canon Medical Systems | GE Healthcare
Healthcare Administrative Partners | Hitachi Healthcare Americas
Novarad | Nuance | Riverain Technologies | Siemens Healthineers
United Imaging | Zebra Medical Vision



The Imaging Wire


One Thousand Radiologists on AI

European Radiology just published a massive radiologist survey detailing how they view AI and where they think improvements are needed.

  • Big AI Survey – This study’s large size (1,041 rads) and international reach (54 countries) suggests that it’s very representative, although it’s also likely that AI perceptions have evolved since the April-July 2019 survey period.
  • Positive Changes – The vast majority of respondents (82%) expect “that AI will cause a significant change” to radiology, alter radiologists’ future (85%), and improve radiology (89%).
  • AI’s Role – Most expect AI to evolve into radiologists’ co-pilot, serving as a second reader (78%) and optimizing workflows (77%), while nearly half expect AI to “partially replace” radiologists (47%).
  • AI Barriers – The respondents most commonly listed ethical and legal issues (62%) and a lack of knowledge (57%) as AI implementation barriers. However, rads with an advanced understanding of AI cited generalizability challenges (40% overall) and a lack of labeled images (27% overall) as the primary barriers (they also agreed about the ethics barrier).
  • AI Education – Most surveyed radiologists said they want AI education to be incorporated in residency programs (79%), while the majority of rads plan to learn about AI even if they aren’t required to do so (75%).
  • The Takeaway – Most of these radiologists expect AI to significantly change their specialty (mostly for the better), but they also see a range of challenges that imaging AI has to overcome in order for many of those changes to happen (technical, data, educational, ethical/legal).

Siemens’ PET/CT Game Changer

Whether you’re exploring new research questions or identifying the best clinical approach, a 106 cm axial view could be a game changer for you and your clinical team. See how with Siemens Healthineers’ Biograph Vision Quadra PET/CT.

– This is sponsored content.



Streamlining Stenosis Reporting

A Singapore-based team developed a deep learning model for spinal stenosis detection / classification that’s getting online praise for its progress towards streamlining stenosis MRI reporting.

  • Background – Lumbar spinal stenosis MRI assessments can be repetitive and time consuming, making them a valuable target for deep learning-based efficiencies.
  • The Model – The researchers developed the DL system from 446 patients’ lumbar spine MRIs, using the first CNN for detection/localization and the second CNN for classification.
  • The Study – They tested the system against a 50-exam internal dataset (labeled by 1 MSK specialist as reference standard, tested against 2 subspecialists) and a 100-exam external set.
  • The Internal Results – With the internal dataset, the DL model and subspecialists achieved high agreements for central canal stenosis (k = 0.96 DL model; 0.98 & 0.98 rads), lateral recess stenosis (k = 0.92 DL; 0.92 & 0.95 rads), and neural foraminal stenosis (k = 0.89 DL; 0.94 & 0.95 rads).
  • The External Results – Against the external set, the DL model classified all regions of interest with “almost perfect agreement” (k = 0.95–0.96).
  • The Takeaway – We hear a lot of calls for AI models that help streamline radiologists’ more cumbersome tasks, and tools that extract the features needed to draft radiology reports seem to fit that description.

Einstein Medical’s Case for ClearRead CT

This Riverain Technologies case study details how Einstein Medical Center adopted ClearRead CT enterprise-wide (all 13 CT scanners) and how the solution allowed Einstein radiologists to identify small nodules faster and more reliably.

– This is sponsored content.



The Wire

  • RapidAI Trolled: RapidAI’s week took a negative turn after a Florida practice’s Informatics & AI Director told all of Twitter that after one week using the stroke AI tool, all involved physicians think the “half baked software” is “absolute junk” and “nowhere near ready for clinical use.” Plenty of AI implementations have rocky starts, but this might be the most public imaging product complaint to hit social media. (Editors note: this Twitter post has since been removed).
  • The Problem with Dysplasia Screening: A new Clinical Imaging study revealed that ultrasound might not be the best way to screen for developmental hip dysplasia. The study (n = 2.9m U.S. infants, 6,806 w/ dysplasia) found that hip ultrasound screening rates increased from 0.4% in 2007 to 2.2% in 2017, while hip dysplasia prevalence remained at 1.7 per 1k infants.
  • RP’s AAA Follow-Ups: Radiology Partners just detailed an initiative that significantly increased how often its radiology reports for asymptomatic abdominal aortic aneurysms included “best practice” follow-up recommendations. The initiative (defining best practice, creating structured reporting macro, training, ongoing measuring & reporting) drove a significant increase in follow-up recommendations for AAAs measuring 2.6-5.4cm (2.1% before to 58% after).
  • SOMATOM X.ceed: Siemens Healthineers unveiled its new SOMATOM X.ceed CT scanner (single-source, 82cm gantry, 50cm spectral FoV), highlighted by its new myExam Companion (guides/automates rad tech operations) and myNeedle Companion (supports CT-guided needle procedure planning & insertion) solutions. The SOMATOM X.ceed CT is expected to gain FDA and CE Mark approval later in 2021.
  • FDG PET/CT Depression Screening: FDG PET/CT scans can detect major depression in cancer patients, potentially creating another use for cancer therapy monitoring exams. That’s from a University of Arkansas study that reviewed whole-body FDG PET/CT images from 134 multiple myeloma patients (38 w/ major depression), finding that depressed patients’ brains had significantly lower metabolic activity.
  • Occult Imaging: A new JAMA study detailed a deep learning system that effectively identifies occult (invisible) scaphoid fractures in X-rays. The DCNN’s first model distinguished scaphoid radiographs with/without fractures (0.955 AUROC) and the second model re-examined the negative cases (w/ an 0.810 AUROC), combining to identify 90% of the occult fractures.
  • DiA & SonoScape Partner: DiA Imaging Analysis and SonoScape Medical announced a partnership that combines SonoScape’s PoC ultrasound systems with DiA’s AI-based cardiac and abdominal ultrasound analysis tools. SonoScape joins DiA’s growing list of ultrasound OEM partners, which also includes Philips, GE Healthcare, Konica Minolta, and Terason.
  • A Case for Smart Scheduling: A German hospital’s CTPA data revealed wide exam variations at different times of the day, illustrating how time series analysis could improve radiology department scheduling. The CTPA scans revealed that: 1) Emergency CTPAs were lowest between 6am-2pm and peaked at 4:23pm; 2) In-patient CTPAs were highest between 6am-2pm and peaked at 1:54pm; 3) CTPA processing times were shortest between 10pm and 6am.
  • UltraSight’s Series B: UltraSight (formerly On-Sight Medical) just wrapped up a $13m series B round to fund its cardiac ultrasound AI guidance platform. The startup aims to bring echocardiography to point-of-care settings that traditionally don’t have enough trained sonographers (e.g. smaller hospitals, clinics, ambulances, and remote areas).
  • Subchondral Bone Length: BU researchers developed a deep learning bone shape measurement system, called subchondral bone length (SBL; analyzes MRIs for cartilage loss & bone flattening), that could become a new biomarker for knee osteoarthritis severity. In an initial test, SBL effectively predicted knee OA patients’ joint space narrowing, pain, disability, and replacement surgeries.
  • Shimadzu’s RADspeed Pro Style Edition: Shimadzu Medical Systems USA officially launched its RADspeed Pro style edition X-ray, highlighted by its new POWER GLIDE Technology, which reduces technologists’ physical exertion when moving the X-ray tube.
  • Ultrasound’s Flexor Tendon Advantage: Ultrasound’s speed, economics, and accuracy could make it “indispensable” for evaluating flexor tendon hand injuries. That came from a UMB study (n = 35 with hand/wrist trauma, 50 injured tendons) where ultrasound diagnosed complete or partial flexor injuries with “high accuracy,” including 100% accuracy with full-thickness tears and tenosynovitis.
  • Advancing Radiogenomics: A UCLA team developed a new radiogenomics deep learning system that revealed new associations between lung cancer patients’ gene expression, histology, and CT imaging features. Scientists have been studying radiogenomics for quite a while, but this new system applies a “gene masking” technique to produce a wider range of associations and link each specific gene expression to imaging features.

Novarad Simplifies Ditching the Disk

Do your patients text more than they use CDs? Find out how Novarad’s CryptoChart simplifies image access, combining secure QR codes and text and email communications to help providers and patients ditch the disk.

– This is sponsored content.



The Resource Wire

– This is sponsored content.

  • Radiology is among the top five specialties reporting significant burnout. Join The Nuance Challenge to discover how to optimize workflow to improve provider satisfaction and accelerate care delivery.
  • See why Children’s National Medical Center calls Arterys’ Cardio AI 4D flow “an important diagnostic tool that enhances diagnostic sensitivity and has the potential to improve surgical planning and patient outcomes” in this JACC study.
  • Know how your practice measures up? In its latest post, Healthcare Administrative Partners details the key benchmarking quality metrics and how they can help radiology practices improve.
  • This new paper in Entrepreneur details the impact that AI can have on population health programs, touching on Zebra-Med’s own efforts to expand population health AI.
  • Live Site Planning Sessions are one of United Imaging’s favorite collaboration moments. Their team comes on-site to take as-built measurements of the Medical Imaging suite and then creates a preliminary drawing for live discussion, giving customers and their design team a chance to adjust equipment placement, make changes to the floor plan, and review equipment specifications in a live virtual environment, saving time in the preliminary drawing process.
  • 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.
  • Learn how Hitachi Healthcare’s Sonticus system alerts physicians when critical findings thresholds are exceeded, helping them address urgent issues and avoid diagnostic errors.

You're signed up!

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

-- The Imaging Wire team