Pediatric Heart Disease

1338246 - Adaptive k-space sampling enables multi-shot real-time imaging in pediatric patients

Friday, January 27, 2023

8:00 AM – 8:10 AM

Location: Silver Pearl 2

Presenting Author(s)

NC

Nuri Chung, BSc

Graduate Researcher
University of California, San Diego, California, United States

Co-Author(s)

AR

Ana E. Rodríguez-Soto, PhD

Project Scientist
University of California, San Diego, United States
Sanjeet Hegde, MD, PhD

Director of Research, Heart Institute/ Medical Director of Cardiac MRI
UC San Diego/Rady Children's Hospital
San Diego, California, United States
Eleanor L. Schuchardt, MD

Pediatric Cardiologist, Assistant Clinical Professor
UC San Diego/Rady Children's Hospital
San Diego, California, United States
Brent M. Gordon, MD

Pediatric Cardiologist
University of California, San Diego, California, United States
Ileen F. Cronin, MSc, RN, BN

Nurse Practitioner
University of California, San Diego
Washington, District of Columbia, United States

Primary Author(s)

Francisco Contijoch, PhD

Assistant Professor
University of California, San Diego
La Jolla, California, United States

Disclosure(s):

Sanjeet Hegde, MD, PhD: No financial relationships to disclose

Brent M. Gordon, MD: No relevant disclosure to display

Background: Adaptive radial k-space sampling (ARKS) has enabled the possibility of multi-shot, real-time imaging by identifying similar periods of the cardiac cycle and determining the optimal sampling angle, all within a single bSSFP TR1. The benefit of this technology in children is unknown, especially those with pediatric heart disease,as they can have different heart rates/rhythms than adults. We evaluated whether ARKS can robustly analyze pediatric electrocardiograms (ECGs) recorded while patients are in a MRI scanner and quantify the improvement in data sampling and sampling uniformity over single-shot or predetermined multi-shot approaches.

Methods: Retrospective analysis was performed on ECGs recorded from 37 pediatric patients (median age: 12 years, interquartile range: 8 to 15 years old, 46% female) during ECG-gated, free breathing 4DFlow scans with IRB approved waiver of informed consent. 200 seconds of ECG data was utilized to simulate 2D ARKS cine imaging with a bSSFP TR = 2.6 ms and 10 MRI samples combined from each shot. Using MATLAB, ARKS evaluated a 10-RR interval search window with a template signal that matched the median duration of the RR interval of that patient. The position of each MRI sample in the cardiac cycle (as a percentage) was used as a measure of physiologic agreement and determined using QRS triggers detected by the MRI scanner. The impact of increasing cross-correlation thresholds on number of beats accepted, cardiac cycle percentage error, and angular sampling was evaluated.

Results:
The median heart rate (HR) was 84.8 bpm (IQR: 69.8 to 96.9 bpm). Median RR variance was 5.5% (IQR: 2.0 to 8.8%) and decreased with HR (rho = -0.53, p < 0.01).

Using a 90% strength of correlation cutoff, 6 or more heartbeats were identified by ARKS for multi-shot image reconstruction during 72.9% of the scan durations. Maximum error in terms of percent cardiac cycle discrepancy (%CCD) was < 10% for 96.9% of the scan duration. ARKS distributed radial views in k-space more evenly than predefined golden angle sampling. The standard error of the angular gap between adjacent radial views decreased from 0.24° to 0.17°.

A higher strength of correlation cutoff decreased the number of beats found. With a 95% cutoff, 50.8% of the scan duration had 6 or more beats identified for multi-shot reconstruction. This led to 97.5% of the scan duration combining beats with %CCD < 10%, but led to lower radial sampling uniformity (ARKS angular gap standard error 0.31°)

Conclusion: In clinical pediatric patients with ECGs recorded while in the MRI environment, ARKS robustly identified prior beats to synthesize for image reconstruction and improved sampling uniformity over multi-shot golden angle imaging. As expected, we observed RR variations which decreased at higher HRs. Further, the strength of correlation cutoff affects how many beats are used for multi-shot imaging and should be optimized based on the application.