Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2347
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-03T06:01:29.618787Z
status: published
lens: g_space
sector: health
headline: Performance of artificial intelligence versus clinician left-ventricular segmentation of cardiac MRI for mortality prediction: a retrospective, multicentre, prognostic study
dek: Background: Ejection fraction, a key measure of cardiac function, is used in health care for clinical decision making and risk stratification. We hypothesised that a deep-learning artificial intelligence (AI) segmentation of ejection fraction obtained from cardiac MRI (CMR) outperforms expert clinician segmentation for the prediction of all-cause mortality.
Methods: In this retrospective, multicentre, prognostic study, we studied a cohort of 27 382 consecutive patients who underwent cardiac MRI (CMR) in the UK a…
gain_title: In 27,382 cardiac MRI patients, AI-derived ejection fraction fit mortality data better and discriminated marginally better than clinician-derived values (C index 0.669 vs 0.637).
problem_title: (none)
trace_subject: (none)
gain_reading: In 27,382 cardiac MRI patients, AI-derived ejection fraction fit mortality data better and discriminated marginally better than clinician-derived values (C index 0.669 vs 0.637).
gain_evidence: AI ejection fraction showed better model fit (AIC 57 936 vs 58 097, ΔAIC 161 favouring AI) | C index 0·669 vs 0·637
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers applied a deep-learning tool to segment the left ventricle on cardiac MRI scans from patients in London and Pittsburgh, then compared the ejection fraction it produced with expert clinician measurements. Both were tested against all-cause mortality over a median follow-up of 5.6 years, during which 13% of patients died.

The result suggests automated measurement could serve as a first-line step with clinician oversight, since the AI values tracked outcomes at least as well as human ones. The advantage was small after adjusting for age and sex, and the authors say prospective studies are still needed to confirm any practical benefit.
limitation: The study was retrospective, and the authors state that the AI's potential for first-line measurement with clinician oversight should be investigated in prospective studies.
tag: Evidence-backed gain
key_points: The cohort included 27,382 patients, 26,106 from Barts Heart Centre and 1,276 from University of Pittsburgh Medical Centre. | 3,458 patients (13%) died during a median follow-up of 5.6 years. | Adding AI ejection fraction to clinician values significantly improved model fit (χ2=168), while adding clinician values to AI helped only slightly (χ2=7.5). | After adjustment for age and sex, discrimination was 0.781 for AI versus 0.777 for clinicians.
rundown: The study was retrospective and multicentre, using consecutive patients referred for cardiac MRI. UK patients were scanned between October 2014 and December 2019, and US patients between June 2010 and March 2016, with mortality taken from national and local databases.

Cox proportional hazards models compared AI-measured and clinician-measured left-ventricular volume metrics. Fit was assessed using Akaike Information Criterion, discrimination with C indices, and added value using likelihood-ratio tests of nested models.

The authors describe the AI's prognostic edge as marginal and support its use for first-line ejection-fraction measurement only with clinician oversight. Funding came from the National Institute for Health and Care Research and the British Heart Foundation.
sources:
- peer_reviewed | The Lancet Digital Health | https://doi.org/10.1016/j.landig.2026.101042 | 2027-06-16
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

b7a0aa7d816ffe2228c0b6db05e9c7308d2e9ff7790db7977305faf7cd5bcad6

previous

0000000000000000000000000000000000000000000000000000000000000000