16th Edition | Two Clearances are Better Than One

Campbell Arnold, Satvik Tripathi
Jun 3
6 min read

“Given that up to 90% of stroke risk and 50% of dementia risk is modifiable, it is imperative that our healthcare networks develop infrastructure to proactively monitor brain health.”

— Johnson et al., AJNR 2025

Welcome to Radiology Access! your biweekly newsletter on the people, research, and technology transforming global imaging access.

In this issue, we cover:

Hyperfine Nets Back-to-Back FDA Clearances for Software and Next-Gen Device
AJNR Showcases the Power and Potential of Low-Field MRI
Frozen No More: FDA’s Gen AI, Elsa, Aims to Speedup The Agency’s Workflow
AI Informatics Take Center Stage at SIIM 2025 in Portland

If you want to stay up-to-date with the latest in Radiology and AI, then don't forget to subscribe!

Hyperfine Nets Back-to-Back FDA Clearances for Software and Next-Gen Device

And what these clearances mean for current and prospective users.

In a significant milestone for low-field imaging, Hyperfine has received two back-to-back FDA clearances—one in mid-May for its Optive AI reconstruction software, and another yesterday for its next generation Swoop portable MRI system. Together, these advances represent a major leap in image quality and clinical potential for portable MRI.

The Optive AI software uses deep learning to enhance resolution, contrast, and noise suppression—delivering what the company calls the “most substantial image quality improvement ever for the Swoop system.” I’ve personally witnessed Hyperfine’s impressive image quality improvements over the past six years (left: T1-weighted image of me circa 2020, right: recent FLAIR acquired at ASNR in May). Hyperfine will start to roll out Optive AI to scanners in Q3 2025.

The newly cleared next-generation Swoop system builds on these software advances with some significant hardware upgrades:

Higher gradient strength for improved DWI and signal-to-noise performance
Redesigned head coil that boosts signal, especially in the posterior fossa
More open architecture, improving access and comfort for larger, pediatric, and anxious patients.
Redesigned RF shielding to reduce noise artifacts
New symmetric yoke design to improve image uniformity

These changes also reflect a broader shift in Hyperfine’s go-to-market strategy: the company recently announced plans to target neurologists’ offices for in-office imaging—potentially opening a new frontier for portable MRI in outpatient care.

Looking ahead, the Hyperfine team sees these developments as a defining moment. “We believe the technological advances with the new Swoop system [deliver] a transformative leap in image quality, providing the level of clarity, uniformity, and anatomical detail needed to help bring portable MRI into mainstream clinical practice,” said Rafael O’Halloran, Hyperfine’s Vice President of Technology.

Bottom line: Hyperfine receives dual FDA clearances for new deep learning-powered software and next-gen Swoop scanner—with a pivotal increase in image quality.

AJNR Showcases the Power and Potential of Low-Field MRI

What does Hyperfine’s recent clearance mean for the future of low-field?

Just a few weeks before Hyperfine’s new FDA clearance, the American Journal of Neuroradiology (AJNR) released a special collection on low-field MRI, spotlighting the transformative potential of portable and accessible imaging. Sponsored by Hyperfine, this issue brings together cutting-edge research that pushes the boundaries of what low-field MRI can do—especially in underserved and resource-limited settings.

As AJNR Editor-in-Chief Dr. Max Wintermark and Hyperfine CMO Eddie Knopp noted, “These studies reflect the diverse capabilities of low-field MRI in addressing critical clinical needs.” From resource-constrained hospitals to the patient bedside ICUs, low-field MRI is expanding neuroimaging access worldwide.

Here’s a summary of what’s inside the collection:

Two studies evaluating portable MRI for assessing ventricular size showed promising utility in acute care settings, particularly for hydrocephalus patients and pediatric populations.
A team explored low-field MRI for diagnosing optic neuritis at the point of care, offering a new way to identify multiple sclerosis related disorders in settings where traditional MRI is unavailable.
Researchers developed an approach to predicting white matter hyperintensities using portable MRI, opening up the possibility for brain health screening in primary care.
A study confirmed that cerebral malaria brain volume scoring methods are valid at low field—an important finding for pediatric neurologic care in low-resource regions.
In Malawi, a low-field scanner was deployed in a resource-constrained hospital, demonstrating how imaging infrastructure can be reimagined for global health.

This special issue reinforces what many in the global radiology community already believe: low-field MRI is not just a workaround—it’s a new frontier. Notably, these studies were conducted using the earlier-generation device. With Hyperfine’s newly cleared system and AI-powered reconstruction now entering the market, the next wave of research will likely unlock even more capabilities.

Bottom line: The AJNR special collection highlights low-field emergence as a powerful tool for expanding neuroimaging access in both global health and acute care settings.

Frozen No More: FDA’s Gen AI, Elsa, Aims to Speedup The Agency’s Workflow

And what the agency wide rollout means for radiology AI developers.

Yesterday, the FDA decided to let it go with an ahead of schedule launch of Elsa, a generative AI tool designed to streamline internal workflows across the agency.

“This rollout is ahead of schedule and under budget,” said FDA Commissioner Dr. Marty Makary, highlighting the agency’s push to modernize its operations.

The tool aims to support FDA staff—including scientific reviewers and field investigators—with tasks like summarizing adverse events, reviewing clinical protocols, and identifying high-priority inspection targets.

Built within Amazon Web Services’ secure GovCloud, Elsa runs entirely within FDA firewalls. Importantly, the agency has stated that Elsa does not train on any proprietary data submitted by regulated industries, ensuring sensitive research and clinical trial data remain protected.

What it means for radiology AI developers:While Elsa won’t be directly evaluating medical products, it may allow the FDA to put more eyes on submissions. Hopefully this won’t lead to “submission bloat” aimed at appeasing AI review tools or cause the kind of automation bias we’ve seen in radiology.

More broadly, Elsa also signals a more AI-literate, tech-enabled FDA—a shift that could improve review efficiency and lead to a better institutional understanding of AI-based tools. For radiology AI developers, that’s a positive sign: a modernized FDA may be better positioned to evaluate and keep up with the next wave of radiology innovations.

Bottom line: The FDA’s launch of Elsa signals a more AI-literate and efficient agency—streamlining reviews to handle the rapid pace of radiology AI innovation.

AI Informatics Take Center Stage at SIIM 2025 in Portland

And how radiology IT foresees generative and agentic AI clinical deployment.

The SIIM25 Annual Meeting in Portland brought together a vibrant community of radiologists, informaticians, and developers, to explore the evolving role of AI and data science in medical imaging workflows. From dynamic keynotes to hands-on demos in the AI Playground, attendees gained insight into the cutting-edge tools shaping the future of healthcare. Key focuses of this year’s conference included:

Applications in generative and agentic AI
An emphasis on multi-specialty enterprise imaging
Ethical frameworks for clinical deployment

Capturing the spirit of the meeting, Dr. Tessa Cook (Past Chair) remarked, “The stress of the work and the preparation sometimes overshadows the excitement, but once you reach the meeting and can’t walk 10 feet without running into old friends or meeting enthusiastic early career and first-time attendees, it feels like exactly where you belong.”

A major takeaway from the conference was the call for stronger validation—especially for generative AI tools. In an opening session on bridging medical AI and policy, Rohini Kosoglu reminded attendees that while AI is advancing rapidly, “politics and policies still listen at human speed. So, please use your voice before the rules are written for you.”

SIIM 2025 reinforced the imaging informatics community’s shared commitment to advancing imaging responsibly—pushing innovation forward without cutting corners. Thank you to Satvik Tripathi for editorial input on this article!

Bottom line: SIIM 2025 showcased a dynamic and growing community focused on practically integrating AI and data science into medical imaging workflows.

Feedback

We’re eager to hear your thoughts as we continue to refine and improve RadAccess. Is there an article you expected to see but didn’t? Have suggestions for making the newsletter even better? Let us know! Reach out via email, LinkedIn, or X—we’d love to hear from you.

References

Disclaimer: There are no paid sponsors of this content. The opinions expressed are solely those of the newsletter authors, and do not necessarily reflect those of referenced works or companies.