|
AI for Chest X-Ray Varies, AI Risk Prediction, and Workplace Bullying
May 21, 2026
|
|
|
|
|
Together with
|
|
|
|
“The next 1-2 years of clinical AI won’t be won by the models with the highest AUCs alone. They’ll be won by the systems that consistently produce … the case the physician would not have seen on their own.”
|
|
Elad Walach, CEO of Aidoc.
|
|
|
New imaging IT tools are becoming available to help radiologists work more efficiently and manage growing imaging volumes. In this edition of The Imaging Wire Show, we talked to Brian Cavanaugh and Chris Barnett of Altamont Software about how their solutions can help by connecting disparate IT applications into a single modern platform. Enjoy the show – and the long weekend! See you next Thursday. – Brian Casey, Managing Editor
|
|
|
|
Not all AI is created equal when it comes to analyzing chest X-rays. A new study in Radiology found wide variation in performance for seven commercially available chest X-ray algorithms to detect lung cancer.
X-ray is by far the most widely used imaging modality. Radiography is often the first imaging exam a patient receives, and it frequently serves as a gateway to other more advanced imaging modalities.
- But radiography also has well-known shortcomings (which is why advanced imaging is needed for follow-up). Could AI help unlock X-ray’s value and make it more useful?
That’s what a host of AI algorithm developers are banking on, but the wide variety of solutions can create confusion for clinicians.
- So U.K. researchers decided to hold an AI bake-off, comparing commercially available algorithms from seven developers for detecting lung cancer on chest X-rays.
The competing companies included Annalise/Harrison.ai, Gleamer, Infervision, Milvue, Oxipit, Qure.ai, and Rayscape. Researchers anonymized performance results from the different products.
In all, chest radiographs from a dataset of 5.2k patients with a real-world lung cancer prevalence rate were included, with researchers finding…
- Significant variance in algorithm performance by each of the major accuracy measures: sensitivity (21%-78%), specificity (59%-98%), and positive predictive value (1.5%-28%).
- All the algorithms increased the number of false positives, and with significant variation. One model generated only 10 more false positives than radiologists, while another produced – wait for it – over 2k.
- If used to triage patients for follow-up CT exams, one model would generate $1.6k in additional costs while another would produce $327k.
What accounts for the variation? An underlying factor is most likely differences in the datasets used for model training.
- In any event, the study underscores the need for more head-to-head comparisons to determine the strengths and weaknesses of individual AI algorithms.
The Takeaway
This week’s study on how AI performance varies between commercially available algorithms initially seems disturbing and might suggest a need for stronger regulatory oversight. But AI’s diversity could be its strength in a future where every patient case is analyzed by multiple different algorithms, each with its own advantages. This could ultimately produce a more complete picture of the patient than any one algorithm on its own.
|
|
Can You Meet Conditional MRI Requirements?
How confident are you that your MR practice can meet the MR Conditional requirements? Philips has developed ScanWise Implant to simplify scanning patients with MR Conditional implants. Learn more on this page.
|
|
- New X-Ray Use Guidelines: X-ray is an inexpensive technology and delivers useful images at a relatively low radiation dose. But that doesn’t mean it should be used all the time. With the goal of reducing inappropriate utilization, a group of specialty medical societies published new guidelines in Radiology covering best practices for X-ray utilization, especially for low-value applications across 12 anatomical sites. The guidelines focus specifically on non-radiology specialists who may rely on radiography for clinical decision-making, such as for trauma applications.
- AI Breast Cancer Risk Prediction: Swedish researchers developed a new algorithm for predicting long-term risk of breast cancer from AI analysis of mammograms, and compared it to other risk techniques in a new paper in Science Translational Medicine. In testing with 8.7k patients, the algorithm performed better at predicting the proportion of invasive breast cancers over 10 years than the image-based Mirai algorithm and clinical models like Tyrer-Cuzick (AUC = 0.72 vs. 0.65-0.66). The study highlights the rapid progress being made in image-based breast cancer risk prediction.
- Workplace Bullying and Harassment: Radiographers/radiologic technologists are frequent targets of bullying and harassment. A new meta-analysis in Radiography documents the problem, surveying 10 studies worldwide in which 1.4k respondents reported their experiences. Bullying rates ranged from close to 100% in African countries to 46% in Taiwan. Patients were most frequently the perpetrators, followed by caregivers and family members. Long patient wait times were often reported as organizational factors that contributed to confrontations – many of which were never reported.
- 18M Women Behind on Cancer Screening: Nearly 18M women in the U.S. are behind schedule on at least one cancer screening test. A new study in JAMA Network Open reviewed self-reported screening adherence for breast, cervical, and colorectal cancer in 68.1k women. In all, 50% of women were up to date on all three tests, 29% with two tests, 13% with just one, and 8% with none. Women due for all three screenings were more likely to be uninsured and to use medical care less often.
- Imaging Use Varies in Children: Children from lower socioeconomic backgrounds and racial minorities were less likely to get imaging for emergency care. In a new study in JAMA Network Open, researchers analyzed 857k emergency department visits in the U.S., finding that children with public insurance were less likely to receive imaging for asthma, head trauma, and abdominal trauma (AOR = 0.85, 0.77, and 0.59). Similar discrepancies occurred for non-Hispanic Black and Hispanic children, even in hospitals specifically set up to handle pediatric cases.
- Blood Tests Help Target CT Lung Screening: What’s the best way to find people who would benefit from CT lung cancer screening but who don’t meet standard screening criteria? Researchers in JAMA describe their clinical study of the INTEGRAL-Risk protein-based blood test for estimating lung cancer risk. In 1.7k people with smoking histories, the test identified 85% of lung cancer cases occurring within one year, more than either USPSTF 2021 and PLCOm2012 models (63% and 70%). INTEGRAL-Risk could be used to refine eligibility for CT screening.
- United Imaging’s ISMRM 2026 Highlights: United Imaging Healthcare highlighted a range of new MRI technologies at the just-concluded ISMRM 2026 meeting in South Africa. The company showcased its uMR Jupiter 5T scanner’s ability to perform metabolic imaging and detect early-stage morphological changes, while on the uMR Ultra 3T scanner, United showed how its uAIFI platform enables dynamic whole-body imaging. And imaging with lower carbon emissions will be possible with a new gradient power amplifier architecture under development (FDA clearance pending).
- AI Reads Cardiac MRI Scans: An open-source AI algorithm for cardiac MRI performed better than general-purpose AI models in a new study in Nature Communications. Researchers from Carnegie Mellon University and Cleveland Clinic developed CMR-CLIP to connect dynamic cardiac MR images with the impressions sections of radiology reports and trained it with 13k patient studies. In testing, they found that the algorithm outperformed general AI models by over 35% in some cases, and had accuracy as high as 99% for some heart conditions. CMR-CLIP can be downloaded on GitHub.
- Philips Updates Angio Dose Reduction: Philips this week introduced SmartIQ, a new coronary imaging technology for its Azurion angiography system. SmartIQ can help interventional physicians manage the tradeoff between radiation dose reduction and image quality. It includes an ultra-low-dose protocol for coronary interventions that reduces dose by 50% compared to ClarityIQ, the company’s previous technology. SmartIQ is being tested in the RADIQAL trial to compare dose reduction versus ClarityIQ while maintaining performance.
- AI Detects Fake Medical Images: A recent study in Radiology found that many radiologists couldn’t tell AI-generated deepfake medical images from real ones. But maybe AI can help. A company called AI or Not ran the images from the Radiology study through its AI content detection algorithm, which correctly identified 92% of the authentic images as real and missed none of the AI-generated images, achieving 100% accuracy in detecting synthetic content. The company believes its tool could be used for revealing insurance fraud, analyzing legal evidence, and other applications.
- DeepHealth Gets Regulatory Nods: DeepHealth is reporting regulatory authorizations in the U.S. and Europe for new AI solutions. The company received FDA 510(k) and CE mark authorizations for Prostate AI, an AI algorithm for MRI that’s part of the Prostate Suite package DeepHealth debuted at RSNA 2025. Other approvals include the CE mark for Brain Health and Brain Age in the company’s Neuro Suite package (both already have 510(k) clearance), and the CE mark for LumbarMR, an algorithm for assessing low back pain on MRI.
- POCUS Developer Raises $12M: Point-of-care ultrasound developer Vortex Imaging raised $12M in a funding round to support further development and regulatory clearance of what the company calls computational ultrasound. Vortex’s technology relies on powerful GPU-based cloud image processing to enable the acquisition of 3D volumes at the point of care. The company plans to initially focus on urology and nephrology and then expand into other clinical areas.
- Visage Lands $50M Contract: Visage Imaging scored another large contract, signing a $50M agreement to install its enterprise image management system at Beth Israel Lahey Health of Boston. Beth Israel operates 14 hospitals in eastern Massachusetts and southern New Hampshire, and Visage will install its cloud-based Visage 7 Enterprise Imaging Platform under a transactional licensing model, with go-live targeted at Q1 2027.
- Harrison Adds Sycai to Open Platform: Harrison.ai has added AI solutions from Sycai Medical to its Open Platform, the AI orchestration network the company launched last year as an alternative to proprietary platform offerings. Sycai’s algorithm automates detection and follow-up of pancreatic cysts on abdominal CT scans. With the addition of Sycai, Open Platform now includes 23 clinical AI solutions across five modalities.
- HOPPR Debuts Breast AI Foundation Model: Foundation model company HOPPR launched a new vision-language model designed to help AI developers create new algorithms for 2D mammography. The company’s EB 2D Mammo Narrative Model was trained on 200k mammography images and generates narrative language that describes characteristics of breast images. AI developers can use the model as a foundation for breast imaging applications that can be configured and validated for specific environments.
- Sectra Scores Toronto Health System: Sectra inked a contract to install its cloud-based medical image management software at Unity Health Toronto, where its Sectra One Cloud solution will be used to consolidate radiology and breast imaging from three hospital sites. The health system performs over 500k imaging exams annually.
- AZmed Secures FDA Clearance: AZmed landed the third FDA 510(k) clearance for its AZtrauma algorithm for analyzing X-rays to detect musculoskeletal issues. The new clearance covers the detection of joint effusions and dislocations and builds on AZtrauma’s previous clearances for fracture detection. AZtrauma is available as part of the company’s Rayvolve AI suite.
- Mach7 Signs Teleradiology Firm: Mach7 Technologies signed an agreement to provide its eUnity Viewer software to teleradiology firm American Radiologist Network in a contract valued at $1.2M. The five-year agreement is structured as a subscription license based on a minimum of 675k studies a year, and volumes are projected to eventually reach over 1M. Mach7 noted the contract’s accelerated installation time, with first use expected within 45 days.
|
|
Connect Imaging Across Every Care Setting
Join Intelerad at SIIM 2026 to learn how you can connect across every care setting, eliminate disks, and accelerate diagnosis with a network clinicians trust. Book a demo today or drop by booth #305.
|
|
A Radiology Question Bank with Analytics
Medality’s new DXIT/Core Radiology SmartBank powered by TrueLearn includes over 800 practice questions aligned to the Diagnostic Radiology In-Training and ABR Qualifying (Core) Exam Domain Blueprints. Assess your readiness for test day now.
|
|
Elevating Breast Cancer Detection
Breast Suite from DeepHealth is a new package of AI-powered solutions delivering increased breast cancer detection rates, risk stratification tools, and viewing and reporting workflow acceleration. Find out how it can benefit your practice today.
|
|
- Next-Generation 1.5T MRI: Echelon Synergy from Fujifilm Healthcare Americas is a powerful and affordable next-generation 1.5T MRI system featuring Synergy DLR deep-learning reconstruction, fast exam times, and patient-friendly design. Discover how it can help you achieve faster workflow and improved image quality.
- Visit Quibim at ASCO 2026: Stop by Quibim’s booth at ASCO 2026 to learn how they are redefining the future of oncology through their innovative AI-based diagnostic and predictive imaging biomarkers. Book a meeting today or drop by at booth #27149.
- Radiology Automation Simplified: CARPL is an enterprise-grade radiology AI validation and deployment platform with 250+ AI applications across 85+ AI vendors that empowers healthcare providers to access, assess, and securely integrate imaging AI in their practice. Book a demo today.
- Smarter Reporting and Smoother Workflows: Join Sectra executives on May 27 at 12 pm ET to learn about how smarter reporting and smoother workflows can enable next-generation radiology. Reserve your seat today.
- Radiology Case Report: A man in his 40s presented with a known metastasis within his abdomen. Learn how contrast-enhanced MRI helped to diagnose the extent of his disease.
- Leveraging AI-Powered Discovery for Image Exchange: Southwest Medical Imaging, a premier physician-owned radiology practice in the southwestern U.S., partnered with Medicom to streamline their workflow. Discover how they utilize Medicom’s AI-powered Smart Search, which leverages an LLM to automatically detect and surface imaging data, to eliminate manual searching and accelerate patient care.
- Unlock Next-Level Diagnostic Possibilities: Photonova Spectra from GE HealthCare is designed to realize the full potential of photon-counting CT in oncology, cardiology, neurology, and more. Learn more about the difference its Deep Silicon technology makes on this page.
- Enterprise Imaging Without the Lock-In: Mach7 helps healthcare organizations unify imaging access, modernize workflows, and maintain control of their imaging data. See how.
- A Breakthrough in Imaging Data Standardization: Enlitic’s Ensight 2.2 is a breakthrough in imaging data standardization that gives health systems a clearer, more detailed understanding of imaging data, accelerating the path from implementation to impact. Find out what it can do for you today.
- AI-Assisted Radiology Reporting: Move beyond legacy radiology reporting with AI-assisted reporting from Mosaic Technologies. See how Mosaic Reporting can help your organization move faster today while building for tomorrow.
- Visit Visage at SIIM 2026: At this year’s SIIM 2026, Visage Imaging will demonstrate its Visage 7 solution operating across the entire Apple ecosystem, including on Apple Silicon-powered workstations with multiple Studio Display XDRs. Book a priority demo today or drop by booth #404-408.
- Intelligent Imaging in Radiography: As a technologist, you face tight schedules, complex exams, and the need for consistent quality. Check out this article from Siemens Healthineers to learn more about their intelligent imaging solutions and see what experienced colleagues have to say about using them.
- Reporting Shouldn’t Add Friction: Clicks, delays, and workarounds add up over time. Modern reporting should protect focus, reduce cognitive load, and help radiologists move faster. See how Rad AI Reporting streamlines your workflow. Book a demo.
- Fewer Biopsies, Better Accuracy: AI is converging with TI-RADS and BI-RADS in ways that go beyond automation. Read this article from Kailo Medical to learn how structured reporting is reducing unnecessary biopsies, improving consistency, and reclaiming clinical time.
- Shifting the Stage in Lung Cancer Screening: Watch this video from Riverain Technologies to learn how their ClearRead CT solution for lung cancer screening can drive enrollment, earlier detection, and seamless management of incidental findings.
- 8 Ways Merge Supports Enterprise Imaging Providers: Merge enterprise imaging solutions deliver measurable value to imaging providers through continued innovation, thoughtful design, and flexible deployment. Request a demo today to see them in action.
- See Your One-Year Return in 30 Seconds: What’s the return on investment for echo AI? Check out this online calculator from Us2.ai to find out how much capacity AI adds, the additional echo scans that capacity buys you, and the net new revenue.
- New Tools for Detecting Cardiovascular Plaque: New imaging tools are advancing the detection of cardiovascular plaque – a key risk factor for heart attack. In this video, ClearCardio’s John Osborne, MD, PhD, discusses these innovations, including United Imaging’s uCT ATLAS CT scanner and their Software Upgrades for Life program.
- A Best in KLAS Hat Trick: AGFA HealthCare was named Best in KLAS in three enterprise imaging segments this year: PACS under 300k studies, universal viewer, and vendor-neutral archive. Find out what makes customers keep coming back on this page.
- AI-Powered Population Health: DeepHealth is assembling radiology’s largest portfolio of AI-enabled radiology solutions for population health. Learn more about their focus and their recent acquisition of Gleamer in this video interview.
- Designed for Your Emergency Challenges and Beyond: The Rembra CT scanner from Philips offers speed, access, and precision for excellent image quality at low dose, with fast, adaptable workflow for everything from routine exams to critical care. Discover its advantages today.
- Bring Your Radiology AI into Your Clinical Workflows: CARPL enables healthcare providers and researchers to develop, test, and deploy their own AI models within existing clinical infrastructure. From seamless data ingestion and de-identification to model training, packaging, and live deployment, CARPL provides an end-to-end environment tailored for radiology.
|
|
|
|
|
