AI ejection-fraction measurements better predicted death risk in heart-scan study

The researchers compared AI-generated ejection-fraction measurements with measurements made by experienced clinicians using cardiac MRI scans.[1] Ejection fraction measures how effectively the heart pumps blood.[1] Doctors use it to inform decisions about heart-failure treatments, implanted devices, procedures to improve blood flow, surgical risk and chemotherapy monitoring.[3] The study was published online on November 2, 2026, and is described in the paper as a retrospective, multicentre, prognostic study.[2]

The analysis covered MRI scans from 27,382 patients in the United Kingdom and United States who were being investigated for known or suspected heart conditions.[1] Patients were followed for an average of more than five years, during which 3,458 deaths occurred.[1] AI-derived measurements were slightly but significantly better than clinician measurements at predicting which patients faced increased risk of death.[1] The difference remained significant after the researchers accounted for factors including age and sex.[1] The team described the work as the largest study of its kind.[1]

The AI tool analyzed a patient’s scan in approximately two to three seconds.[1] In earlier work, the research team found that expert clinicians typically took around 13 minutes to analyze a scan.[1] The researchers reported an AI error rate of around 1% at the patient level.[3] They said clinicians should verify AI results and interpret scans in their broader clinical context.[1] The findings support using AI as a first-line tool for cardiac MRI analysis, with expert review when needed; the AI was not intended to replace clinicians.[3]

The reported improvement in identifying higher-risk patients was small but meaningful, according to the researchers.[3] The team plans to combine AI measurements from heart scans with other tests, particularly electrocardiograms, to detect heart-muscle problems earlier.[3] The researchers said this could allow treatment to begin before lasting damage occurs.[3] The study received funding from an NIHR Clinical Lectureship, the NIHR Biomedical Research Centre at UCLH and a British Heart Foundation Accelerator Award.[3]