Application of AI-assisted compressed sensing cine sequence in the assessment of cardiac function in children with Kawasaki disease.

Abstract

Background: Cardiac magnetic resonance (CMR) imaging is an important tool for assessing the cardiac structure and function in Kawasaki disease. To perform a prospective analysis comparing the newly developed temporal artificial intelligence (AI)-assisted compressed sensing cine (ACS-cine) and conventional cine (Conv-cine) in the assessment of cardiac function in Kawasaki disease patients.

Methods: This study included 45 patients (35 males), who were scanned using both ACS-cine and Conv-cine sequences. Left and right ventricular ejection fractions (LVEF and RVEF), left and right ventricular end-systolic volumes (LVESV and RVESV), and left and right ventricular end-diastolic volumes (LVEDV and RVEDV) were measured. The image quality and intraclass correlation coefficient (ICC) of the two methods were also assessed.

Results: The mean acquisition time of ACS-cine was significantly shorter than that of Conv-cine (17.18±2.45 vs. 251.12±38.01 seconds, P<0.05). In terms of cardiac function measurements, the median LVEF measured by Conv-cine was higher than that of ACS-cine (63.0% vs. 61.0%, P<0.05), while median LVESV was lower in Conv-cine than in ACS-cine (26.5 vs. 27.9 mL/m², P<0.05). The image quality score of ACS-cine was slightly higher than that of Conv-cine (4.41±0.50 vs. 3.87±0.65, P<0.001). The inter-observer consistency was almost perfect for both ACS-cine (ICC: 0.882-0.978) and Conv-cine (ICC: 0.952-0.991).

Conclusions: ACS-cine technology demonstrates significant time efficiency advantages in CMR with satisfactory image quality, although there are statistically significant differences in the measurement of certain cardiac function parameters (LVEF and LVESV) compared to Conv-cine.