Image Reconstruction (including machine learning)

1356447 - Highly-Accelerated Real-Time Myocardial Tagging for Exercise CMR at 3T

Wednesday, January 25, 2023

Location: E-POSTER

Presenting Author(s)

MM

Manuel A. Morales, PhD

Post-doc Research Fellow
Beth Israel Deaconess Medical Center
Boston, Massachusetts, United States

Primary Author(s)

MM

Manuel A. Morales, PhD

Post-doc Research Fellow
Beth Israel Deaconess Medical Center
Boston, Massachusetts, United States

Co-Author(s)

SY

Siyeop Yoon, PhD

Post-doc Research Fellow
Beth Israel Deaconess Medical Center
Boston, Massachusetts, United States
Ahmed S. Fahmy, PhD

Instructor in Medicine
Harvard Medical School
Boston, Massachusetts, United States
JR

Jennifer Rodriguez

Clinical Trials Specialist
Beth Israel Deaconess Medical Center
Boston, Massachusetts, United States
DH

Daniel A. Herzka, PhD

Associate Professor
Case Western Reserve University, United States
Warren J. Manning, MD

Professor of Medicine and Radiology
Harvard Medical School
Boston, Massachusetts, United States
RN

Reza Nezafat, PhD

Professor
Harvard Medical School
Boston, Massachusetts, United States

Disclosure(s):

Warren J. Manning, MD: No relevant disclosure to display

Reza Nezafat, PhD: No financial relationships to disclose

Background: Exercise cardiovascular magnetic resonance (Ex-CMR) is a stress imaging test with promising applications in cardiovascular disease (CVD). Ex-CMR myocardial tagging could provide insights into myocardial deformation post-exercise. However, current ECG-segmented gradient-echo myocardial tagging sequences are not robust for this task. The goal of our study is to develop a real-time myocardial tagging sequence that can be used for quantification of beat-to-beat variation of myocardial deformation after exercise.

Methods: We implemented a highly-accelerated SSFP-based real-time tagging sequence with higher SNR than gradient-echo, thus more robust to high acceleration rates at 3T (Fig 1A). Highly accelerated images were reconstructed using compressed-sensing (CS) and deep learning (DL). Imaging parameters were: temporal resolution = 28.8 ms, resolution = 2.0 × 2.0 mm2, FA = 30º, slice thickness = 8 mm, TE = 1.1 ms, TR = 2.6 ms. Tag lines were perpendicular to phase-encoding (PE) to preserve harmonics. Fat-saturation was used to spoil fat off-resonance, and 5 ramp-up pulses were used to minimize imaging delay. We used vendor-provided CS reconstruction per beat to create images from 6-fold under-sampled k-space. Resulting blurred images (2-fold reduced PE spatial resolution) were input to DL recon using resolution enhancement generative adversarial inline network, i.e., REGAIN (Fig 1B) to deblur low-resolution CS-reconstructed images. We recently developed REGAIN for cine imaging using breath-hold ECG-gated segmented cine. REGAIN was readily used to reconstruct tagging data without modifications. 12-fold accelerated k-space were reconstructed in-scanner in real-time. Images were available upon completion of scan.

Performance was initially evaluated by recruiting 17 healthy subjects (7 male, 30 ± 8 years) and 26 CVD patients (19 male, 54 ± 16 years). Real-time tagging images were collected in the 2-CH and short-axis views at basal, mid-ventricular, and apical levels. A subset of subjects (13 healthy, 4 patients) participated in an Ex-CMR.

Results: Real-time tagging (Fig. 2) using spatial resolution of 2×2 mm2 and temporal resolution of 28.8 ms was successful. We observed similar tag sharpness pre- and post-exercise in 2-CH (1.5 mm-1, 1.5 mm-1) and short-axis views (1.6 mm-1, 1.5 mm-1)(Fig. 3A). Measurements of aMW-GCS made in 1st rest, 2nd rest and post-exercise images in healthy subjects were 18.7%, 19.1%, 23.2%. Measurements in patients were 17.6%, 18.6% and 23%. Measurements from 1st and 2nd rest images showed ICC=0.71 (Fig. 3C).

Conclusion: We developed highly accelerated SSFP real-time tagging at 3T, reconstructed with CS and REGAIN, with 28.8 ms temporal resolution for evaluation of beat-to-beat myocardial deformation post-exercise.