AI measurements could sharpen a central cardiac assessment
Artificial intelligence measurements of heart function from MRI scans were more strongly associated with patient survival than measurements made by clinicians in a large study spanning the United Kingdom and United States.
The research, led by City St George’s, University of London and University College London and published in The Lancet Digital Health, examined ejection fraction: a widely used measure of how effectively the heart pumps blood. The finding does not suggest that clinicians should be removed from the process. It does indicate that AI-derived measurements could become a useful first-line way to assess an especially consequential cardiac metric.
Ejection fraction is used throughout cardiology. It helps clinicians guide treatment decisions for people with heart disease, including decisions about specialist heart-failure treatments, implanted devices and procedures intended to improve blood flow to the heart. It can also inform assessments of surgical risk and monitoring of heart function during chemotherapy.
Study compared AI and experienced clinicians
Researchers analyzed cardiac MRI scans from 27,382 patients who were being investigated for known or suspected heart conditions. The patients were in the UK and the US and were followed for an average of more than five years. During that period, 3,458 deaths occurred.
The team compared ejection-fraction measurements produced by an AI tool with measurements made by experienced clinicians reviewing the scans. AI-derived ejection fraction was slightly, but significantly, better at identifying patients at increased risk of death. The association remained significant after the researchers accounted for factors including age and sex.
That distinction matters because ejection fraction is often treated as a gatekeeper measurement. A value can influence whether a patient is considered for particular therapies or more intensive intervention. If the measurement itself more closely tracks longer-term outcomes, it could give care teams a stronger starting point for evaluating risk.
Speed is part of the potential benefit
The AI tool analyzed a scan in roughly two to three seconds per patient. The article contrasts that with previous work by the same team finding that expert clinician analysis typically takes around 13 minutes.
That time difference should not be read as a simple contest between an algorithm and a clinician. Cardiac imaging involves far more than producing a single number, and treatment decisions depend on a patient’s history, symptoms and other clinical information. But a rapid automated measurement could reduce a repetitive part of the workflow and make it easier to process large imaging volumes consistently.
The study’s scale is also notable. Rather than evaluating a narrow set of cases, the researchers used tens of thousands of scans from patients undergoing investigation for established or suspected heart problems. The survival follow-up gives the comparison clinical weight beyond whether two measurement methods merely agree with each other.
What the result does—and does not—show
The reported advantage was slight, not a claim that AI can independently diagnose heart disease or determine treatment. The result is specifically about the relationship between AI- and clinician-derived ejection-fraction measurements and subsequent survival in this study population.
Still, the work points to a practical use for AI in imaging: producing a fast, outcome-relevant measurement that clinicians can incorporate into their assessment. In settings where specialists face growing imaging workloads, a tool that can generate a measurement in seconds may be particularly valuable if it performs reliably across patients and clinical sites.
- The analysis included 27,382 cardiac MRI patients in the UK and US.
- Patients were followed for more than five years on average.
- There were 3,458 deaths during follow-up.
- AI measurements took approximately two to three seconds per scan.
The next questions concern implementation. Healthcare systems will need to determine how such tools fit into established imaging workflows, how results are reviewed and how performance holds up in the patient populations they serve. The study supplies evidence that the AI measurement is meaningfully associated with survival; translating that evidence into routine care will require clinical judgment and careful validation.
For now, the research strengthens the case that AI can contribute not only speed but potentially useful prognostic information in cardiac MRI. Ejection fraction has long been one of cardiology’s most important numbers. This study suggests the way it is measured may itself affect how well that number reflects patient outcomes.
This article is based on reporting by Medical Xpress. Read the original article.
Originally published on medicalxpress.com
