How MedStar Health built radiologist confidence and scaled SwiftMR® across its enterprise imaging network

Presented by MedStar Health at the RSNA 2025 annual meeting

MedStar Health is the largest healthcare system in the Maryland–Washington, D.C. region

Why MedStar explored AI reconstruction

One solution had to satisfy two very different teams

From skepticism to confidence: earning radiologist trust in AI

SwiftMR reduced scan times in both care settings

From clinical validation to enterprise-wide deployment

Scale consistent image quality across your enterprise

What radiologists needed

What enterprise IT and operations needed

Appoint radiologist champions

Structured communication and follow-up

Collaborative on-site reviews

Test the challenging cases

Enterprise-grade security and reliability

13 imaging locations using SwiftMR across MedStar

38–39% total scan-time savings on inpatient and outpatient scanners

60-day radiologist-led clinical evaluation before go-live

Former Chief, Division of Neuroradiology

MedStar Georgetown University Hospital

Building radiologist confidence in AI reconstruction

Neuroradiologist specializing in head, neck, and skull base imaging. Dr. Jay led the radiologist-side clinical evaluation of SwiftMR and has held leadership roles with the ACR Institute of Radiology Pathology and the Radiology Leadership Institute.

Scaling AI across an enterprise: operational and IT excellence

Technical lead for MedStar's 10 hospitals and 90+ imaging locations and a charter member of MedStar's AI Center of Excellence. Carl led the enterprise rollout across radiology, IT, and operations.

MedStar Health operates 10 hospitals and more than 90 imaging locations across Maryland and the Washington, D.C. area, spanning both inpatient and outpatient settings. Its enterprise imaging team supports radiology, cardiology, orthopedics, and other major service lines.

As the network grew through the acquisition of community imaging centers, MedStar inherited a diverse MRI fleet spanning multiple vendors, field strengths, and equipment ages. The challenge: deliver the same diagnostic quality to every patient, regardless of which scanner they happened to be imaged on.

Two problems pushed MedStar to evaluate AI reconstruction. Both had to be solved without disrupting the radiologists and technologists who relied on the existing fleet every day.

The goal: achieve enterprise-wide consistency in image quality in a cost-effective manner, enabling scalability across diverse scanner types.

MedStar evaluated AI reconstruction against a clear set of requirements. A solution had to win over the radiologists reading the images and the IT and operations teams running the network. SwiftMR met every criterion.

Unlike AI tools that add overlays or prioritization cues, SwiftMR changes the image acquisition process itself. That makes clinical collaboration essential during testing, validation, and rollout. Radiologists had to trust the images before anything else could scale.

Assigning a lead radiologist per body part created clear ownership, ensuring consistent feedback and sign-offs throughout the evaluation.

Regular touchpoints with radiologists closed the feedback loop quickly, so concerns were addressed before they slowed adoption.

The AIRS team visited every two to three weeks to make protocol adjustments based directly on radiologist feedback.

Reviewing tough cases, such as lesion studies, during the evaluation phase reinforced diagnostic reliability and built confidence.

Inpatient and outpatient scanners of the same model both achieved 38–39% total time savings, with distinct benefits for each setting.

On the inpatient floor, where exams run longer and schedules are unpredictable, SwiftMR delivered consistent scan-time reductions across every protocol. The value here was schedule adherence and predictable workflows under constant interruption.

At the outpatient center, the same scanner model achieved comparable savings, which translated into shorter appointment slots and higher daily throughput without adding scanners or staff.

Total time savings reflect a full protocol set on Siemens Aera 1.5T scanners. Results vary by scanner model, field strength, pulse sequence, and clinical protocol.

With radiologist confidence established and consistent image quality confirmed across scanners, the next phase was scaling SwiftMR across the enterprise. That meant satisfying IT and operations without disrupting the people doing the scanning and reading.

A single cloud deployment, centrally managed by MedStar IT, supports every MRI site across the network. Technologists send images exactly as they always have , IT auto-routes them based on predefined logic, and everything arrives in PACS, so there is no change for radiologists.

One deployment supports all MRI sites across the enterprise, with no workflow change for technologists or radiologists.

Patient data never leaves the hospital network, on fully HIPAA-compliant, US-based cloud infrastructure.

All data in the cloud automatically deletes every 24 hours.

The vendor never trains its models using customer data.

If a network interruption occurs, scanning continues on the standard protocol, and images can be re-submitted to SwiftMR anytime.

SwiftMR's deep learning technology reconstructs MRI scans, delivering sharper images in a fraction of the time, on the scanners you already own. Health systems like MedStar have used it to standardize diagnostic quality, increase throughput, and scale across inpatient and outpatient sites with no workflow disruption. Talk to AIRS Medical about bringing SwiftMR to your network.

SwiftMR® is FDA 510(k) cleared as an AI-powered MRI image enhancement solution intended for noise reduction and increased image sharpness on non-contrast enhanced MRI images in DICOM format. Supported body parts, pulse sequences, field strengths, patient populations, and MRI manufacturers may vary by country.

Content adapted from a presentation by MedStar Health at the RSNA 2025 annual meeting. Scan-time data reflects Siemens Aera 1.5T inpatient and outpatient scanners. Individual outcomes may vary. Results vary by scanner model, field strength, pulse sequence, and clinical protocol; individual practices should not expect identical results. For investigational purposes only for scan-time reduction over 50%.

Quotes and data published with MedStar Health’s permission. This page is intended for healthcare professionals.

© 2026 AIRS Medical Inc. SwiftMR® is a registered trademark of AIRS Medical Inc.

“Reviewing challenging cases during the evaluation period reinforced diagnostic reliability and helped transform early skepticism into confidence.”

“SwiftMR's impact differs across settings. For inpatients, it stabilizes the workflow and ensures consistent image quality under unpredictable conditions.”