Skip to main content
2025 SCMR Conference Main banner
Final Program


Schedule at a Glance
Back

Rapid Fire Abstracts

Rapid Fire 03- Rapid MRI

AI-CINE Increases Speed, Decreases Breath holds, and Enhances Quantification (RF_TH_252)

Thursday, January 30, 2025
3:30 PM – 3:40 PM East Coast USA Time
Location: Blue Room PreFunction
Background: Cardiovascular disease patients presenting with mild dyspnea, tachypnea, arrhythmias and restlessness challenge the technologist to shorten study duration and minimize breath holds. Tremendous developmental efforts using highly accelerated MR sequences coupled with advanced reconstruction algorithms have been tried. In this study, Artificial intelligence (AI)-assisted accelerated compression sensing (CS) CINE technology (AI-CINE) integrates CS, parallel imaging and half Fourier analysis into a single technique that markedly decreases image acquisition time relative to conventional CINE (C-CINE). This study compared quantitative left (LV) and right ventricles (RV) structure-functional metrics at 1.5T.

Methods: Cardio-oncology patients and healthy volunteers (n=89) received C-CINE and AI-CINE acquisitions in the same session.  Image acquisition time and quality were assessed. Biventricular structure, volumes and function [end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), ejection fraction (EF), muscle mass, and wall thickness] were analyzed with an automated in-line image analysis package. C-CINE and AI-CINE data were compared by Bland-Altman analysis with calculation of intra-class coefficient (ICC).

Results: Eighty-nine participants (58.5 years±16.8, 42 male, 47 female) were evaluated. AI-CINE acquisition time (37s) was 84% faster than C-CINE (238s) with preserved anatomic structural clarity and slightly lower image contrast versus C-CINE.  Bland Altman plots and ICC revealed excellent agreement between C-CINE and AI-CINE (ICC, 0.73≤r≤0.98). Figure 1  LV measurements showed very good to excellent reliability (ICCabs=0.89–0.98). The RV ICCs were in the “good” range (ICCabs = 0.73-0.89), which may relate to the derivation of RV metrics from stacked LV short-axis rather than RV optimized slices. Table 1. Among subjects plagued with claustrophobia, dyspnea, arrhythmias, and restlessness, AI-CINE images were notably acquired faster and more easily than C-CINE, benefiting the patients and technologist. Figure 2 presents C-CINE and AI-CINE in a 76-year-old female patient diagnosed with cardiac amyloidosis. These 2-, 3-, and 4-chamber views exhibit outstanding agreement between C-CINE and AI-CINE showing a mildly dilated LV and small to moderate pericardial effusion (indicated by arrows) highlighting the  detection of pathological features with both sequences. In eight cases significant disparity between AI-CINE and C-CINE was noted and related to inferior quality of C-CINE images resulting in less reliable metrics. AI-CINE produced reliable images for qualitative and quantitative analysis in 97/100 subjects while C-CINE did so in 89/100 subjects.

Conclusion: In cardio-oncology patients and healthy volunteers, regardless of LV and RV size and function, AI-CINE accelerated cardiac image acquisition, preserved anatomical detail and diminished patient-related motion artifact, particularly for challenging clinical cases.