Artificial Intelligence-Assisted CMR Scanning vs Standard-of-Care: Comparative Analysis of Clinical Benefits From...

Clinical adoption of cardiovascular magnetic resonance (CMR) is hampered by procedural complexity and magnetic resonance imaging access.

The objective of the study was to investigate the benefits of artificial intelligence (AI)-assisted CMR in clinical practice.

Patients at a tertiary-care center underwent CMR via conventional, operator-directed (ODS) or AI-assisted (AIA) scanning over 3 years. Scan time, scan time variation, image quality using a validated 5-point scale, scan failure, and survival were compared between groups, stratified across CMR indications and technologists' experience.

From 2020 to 2024, 6,080 consecutive patients (age 57 ± 17 years, 55% males) underwent 6,545 studies (1,635, 25% AIA, 4,910, 75% ODS) for 5 primary indications. Average scan time and its coefficient of variation were reduced by 23% and 31%, respectively, with AIA compared to ODS (37.9 ± 6.9 vs 49.2 ± 13.0 min; P < 0.001). Scan time reduction by AIA was consistent across indications and range of technologists' experience, and remained significant after adjusting for key demographic and clinical differences (P < 0.0001) or propensity score. Scan quality score was higher in AIA for cine images (4.09 ± 0.8 vs 4.03 ± 0.8; P < 0.001), late gadolinium enhancement (4.08 ± 0.8 vs 3.98 ± 0.7; P < 0.001), and perfusion (4.28 ± 0.6 vs 4.14 ± 0.6; P < 0.001). Scan failure leading to 90-day repeat CMR was lower in AIA compared to ODS (0.6% vs 1.4%, P < 0.001). After its introduction in 2021, AIA use progressively increased to 78% by 2023. At a median follow-up of 29 months, patient survival did not differ between groups (log-rank P = 0.500).

In this nonrandomized comparison, AIA improved scan time, scan time consistency, image quality, and 90-day scan failure across top CMR indications, compared to ODS.