Signal-to-Noise Ratio in 4D Flow CMR: Sequence Determinants and Impact on Kinetic Energy, Energy Loss, and Vorticity – Insights from the TREK Study

Presenting Author(s)

• 

  Marc Lee, MD

  Assistant Professor of Pediatrics
  Nationwide Children's Hospital
  Columbus, Ohio, United States

Co-Author(s)

• RO

  Ryan P. O'Hara, PhD

  Postdoc
  Children's National Hospital
  Washington, District of Columbia, United States
• JK

  Jared D. Klein, MD

  Assistant Professor of Pediatrics
  Joe DiMaggio Children's Hospital
  Fort Lauderdale, Florida, United States
• 

  Richard M. Friesen, MD

  Assistant Professor
  Children's Hospital Colorado
  Aurora, Colorado, United States
• MD

  Matthew Dove, MD

  Pediatric Cardiologist
  Vanderbilt
  Jackson, Tennessee, United States
• 

  Anudeep K. Dodeja, MD

  Associate Director Adult Congenital Heart Disease
  Connecticut Children's
  Hartford, Connecticut, United States
• HT

  Hieu Ta, MD

  Cardiologist
  Cincinnati Children's Hospital Medical Center, United States
• 

  Jonathan Kochav, MD

  Cardiology Fellow
  Columbia University Irving Medical Center
  New York, New York, United States
• 

  Adam B. Christopher, MD

  Assistant professor
  UPMC Children's Hospital of Pittsburgh
  Pittsburgh, Pennsylvania, United States
• 

  Rajesh Krishnamurthy, MD

  Professor
  The Ohio State University
  Columbus, Ohio, United States
• EB

  Elias Balaras, PhD

  Professor
  George Washington University, District of Columbia, United States
• FC

  Francesco Capuano, PhD

  Assistant Professor
  Universitat Politecnica deCatalunya Barcelona Tech
  Barcelona, Catalonia, Spain
• BM

  Brett A Meyers, PhD

  Research Assistant Professor
  Purdue University
  West Lafayette, Indiana, United States

Primary Author(s)

• 

  Yue-Hin Loke, MD

  Associate Professor of Pediatrics
  Children's National Hospital
  Washington DC, District of Columbia, United States

Co-Author(s)

• 

  Yue-Hin Loke, MD

  Associate Professor of Pediatrics
  Children's National Hospital
  Washington DC, District of Columbia, United States

Background: Advanced 4D flow cardiac magnetic resonance (CMR) biomarkers such as kinetic energy (KE), energy loss (EL), and vorticity are increasingly applied to characterize congenital heart disease such as repaired Tetralogy of Fallot (rTOF). However, use in pediatric populations often involves variation in acquisition parameters and image quality, which may impact hemodynamic measurements and confound physiologic interpretation. Understanding both the determinants of signal-to-noise ratio (SNR) and its independent impact on derived flow metrics is critical for harmonization and multicenter adoption.

Methods: We analyzed 276 complete 4D flow CMR datasets across six institutions from the multicenter Tetralogy of Fallot 4D flow Registry of Energetics and Kinetics (TREK) study (rTOF patients and controls, median age 19 years, 46% female). Imaging was performed on 1.5T and 3T Siemens/GE scanners with varying voxel sizes, flip angles, temporal resolutions, and contrast protocols. Sequence metadata were extracted. SNR was measured via ROI analysis on Pixel (Tempus), defined as the ratio of peak aortic signal intensity to standard deviation of air signal intensity. 4D flow biomarker extraction was performed using ITFlow (CardioFlow Design, Tokyo, Japan), yielding mean systolic KE (mJ/mL), mean systolic EL (mW/mL), and peak systolic vorticity (1/s). Multivariable models evaluated (1) sequence determinants of SNR, and (2) independent associations of SNR with flow biomarkers after adjustment for case type and clinical covariates (BSA, rTOF vs control, ventricular size/function).

Results: SNR was strongly influenced by acquisition parameters (Table 1), with significant independent contributions from voxel volume (p=0.026), flip angle (p=0.030), temporal resolution >50 ms (p=0.007), and the use of ferumoxytol agent (p< 0.001). Field strength showed a trend toward higher SNR at 3T vs 1.5T (p=0.052), while vendor (Siemens vs. GE) and echo time were not significant. For the 4D flow biomarkers (Table 2) after adjusting for clinical variables, SNR was independently associated with EL (β=–0.53, p< 0.0001) and vorticity (β=–0.16, p< 0.0001), but not KE (β=–0.03, p=0.46). Lower SNR inflated EL and vorticity values (Figure 1). KE in the aorta was unaffected by SNR, and was significantly lower in rTOF compared with controls (p< 0.0001), with further contributions from RVEDVi and LVEF.

Conclusion: SNR is determined by acquisition parameters, and in turn selectively influences downstream 4D flow biomarkers. EL and vorticity appear to be highly SNR-sensitive, whereas KE remains physiologically robust, demonstrating lower KE in rTOF patients which may have prognostic significance. These findings underscore the need to account for image quality in multicenter 4D flow studies and support harmonization strategies to ensure accurate interpretation of advanced flow metrics. Future work will assess cross-vendor reproducibility to enable reliable clinical translation.

Table 1. Sequence Determinants of SNR.

Acquisition Parameter	β (95% CI)	p-value
Contrast Use
Gadolinium vs None	+5.51 (–2.50, +13.5)	0.18
Ferumoxytol vs None	+18.8 (+8.81, +28.9)	0.0003
Siemens vs GE	+8.51 (–2.82, +19.8)	0.14
3T vs 1.5T	+10.5 (–0.09, +21.1)	0.052
Echo time (per ms)	–8.73 (–23.5, +6.06)	0.25
Flip angle (per °)	+1.13 (+0.11, +2.14)	0.030
Effective temporal resolution > 50 ms	+9.27 (+2.53, +16.0)	0.007
Voxel volume (per mm³)	+0.77 (+0.09, +1.45)	0.026

Table 2. Independent Associations of SNR with Aortic Flow Biomarkers (log-transformed outcomes, adjusted against clinical covariates)

4D Flow Variable	log(SNR) β (95% CI)	p-value	rTOF vs Control β (95% CI), p</span>	p-value	Other significant clinical covariates
Kinetic Energy (KE) (log)	–0.034 (–0.124, +0.056)	0.456	–0.305 (–0.383, –0.227)	< 0.0001	log(RVEDVi) +0.390, p=0.0042; log(LVEF) +0.688, p=0.0077
Vorticity (log)	–0.163 (–0.218, –0.109)	< 0.0001	–0.099 (–0.147, –0.052)	< 0.0001	BSA –0.097, p< 0.0001; log(LVEF) +0.449, p=0.0044
Energy Loss (EL) (log)	–0.527 (–0.656, –0.398)	< 0.0001	–0.098 (–0.210, +0.014)	0.087	BSA –0.149, p=0.0031; LVEF +1.227, p=0.0010

Forest plot of standardized β coefficients (adjusted for clinical covariables).