As cardiologists increasingly rely on CMR for ventricular function, tissue characterization, and complex diagnostic questions, imaging teams are being asked to deliver more exams without compromising image quality, patient tolerance, or workflow efficiency. The challenge is no longer whether CMR provides value. It is how to make that value easier to access across real-world clinical environments.
Cardiac MRI (CMR) has become one of the most powerful tools we have for understanding the heart. It’s the gold standard for measuring ventricular volumes and function, and it answers questions that other imaging modalities simply can’t. As referrals continue to climb, imaging departments everywhere are feeling the pressure—longer waitlists, tighter schedules, and the constant need to move complex exams along faster.
And let’s be honest: CMR isn’t the easiest exam for patients either. Depending on the protocol, a full exam can run anywhere from 30 minutes to well over an hour. Complex cases can stretch close to two hours. During that time, patients are expected to lie perfectly still, follow precise breathing instructions, and hold their breath—sometimes 40, 50, even 60 times. For many, that’s a lot to ask. Some need extra coaching, some need breaks, some need sedation, and in the worst case scenario some simply can’t finish the exam.
In a 2025 Journal of Cardiovascular Magnetic Resonance study of full free-breathing CMR, Yang et al. noted that repeated breath holds and the recovery periods between them can account for nearly 50% of total scan time, and that patients unable to hold their breath may have difficulty completing the examination or obtaining high-quality images.
So it’s no surprise that the field has been pushing hard toward faster, more robust and patient-friendly CMR exams. Vendors are working to solve this problem by introducing new acquisition techniques like compressed sensing, free-breathing approaches, and single-shot trajectories. Deep learning is beginning to further reshape cardiac imaging as well.
But in real-world clinical environments, access to the latest scanner hardware, software upgrades, or advanced cardiac packages is not always consistent. That creates a practical gap: imaging teams need the benefits of accelerated CMR, but they may not have the newest platform required to access every vendor-native solution.
That’s where SwiftMR comes in—a vendor-neutral deep learning reconstruction platform designed to support CMR exams across any scanner, regardless of hardware or software generation. Put simply, you don’t need the latest scanner platform to benefit from deep learning in cardiac MRI.
A Real-Life Clinical Implementation Example
Performed on a Siemens Healthineers 1.5T system
A typical CMR exam is a mix of pulse sequences—cine, flow, dark blood, tagging, mapping, delayed enhancement—acquired across multiple planes. For this clinical example, let’s zoom in on cine imaging, since it’s the backbone of functional cardiac assessment.
Why Cine Imaging Matters
Cine MRI captures the heart in motion. Data is acquired across multiple cardiac phases and then displayed as a dynamic sequence, allowing clinicians to assess ventricular function, chamber size, myocardial wall motion, and overall cardiac mechanics.
To achieve adequate temporal resolution and signal-to-noise ratio, cine data is commonly acquired over multiple heartbeats per slice. In this site’s standard-of-care protocol, cine imaging was typically acquired over 8–11 heartbeats per slice.
Here’s the challenge: The majority of cine sequences require breath holds.
At 60 bpm, acquiring data over 11 beats, each breath hold is roughly 11 seconds—and with upwards of 50 or more breath holds in a cardiac exam, that burden adds up fast for the patient.
SwiftMR: Reduced Short-Axis Stack Acquisition Time by 69%
Using SwiftMR, the site combined multiple acceleration techniques to acquire cine data in just 4 heartbeats instead of 11, significantly reducing breath-hold time compared with the site’s standard-of-care sequence.
Stacking multiple acceleration methods can introduce trade-offs, including increased noise, lower SNR, and reduced spatial resolution. SwiftMR’s deep learning–powered denoising and super-resolution reconstruction helped address those trade-offs, supporting diagnostic-quality cine images from accelerated acquisitions.
The result was a faster, more efficient acquisition that maintained the functional information clinicians rely on while easing the patient’s breath-hold burden.
The biggest impact was seen when acquiring stacked cine series. Traditionally, this is performed as one slice per breath hold. For example, 10 slices in the stack equals 10 breath holds.
With SwiftMR, the site was able to acquire 2–3 slices in a single breath hold, reducing the required breath holds for a stack series by more than half.
On average, this reduced the short-axis stack acquisition time from:
4:52 → 1:31
Approximately a 69% reduction in short-axis stack acquisition time.
More importantly, it reduces the number and duration of breath holds required from the patient. For CMR, that matters: less breath-hold burden can mean less fatigue, fewer interruptions, improved patient tolerance, and a smoother exam for both the patient and the imaging team.
This is the practical value of applying SwiftMR to established CMR methods: faster acquisition, reduced patient burden, and more efficient cardiac MRI while preserving the functional information clinicians rely on.
As CMR demand continues to grow, improving access does not always have to depend on new hardware. In this implementation, SwiftMR helped accelerate a core component of the cardiac exam on an existing Siemens Healthineers 1.5T system, reducing short-axis stack acquisition time and easing breath-hold burden while preserving the functional information clinicians rely on.
For more clinical details on this implementation or to discuss how SwiftMR can support cardiac MRI workflows at your site, contact Anthony Rodenberg, BA RT(R)(MR), Director, Clinical Programs.