Rapid Whole-Heart Quantitative Assessment of Myocardial Injury Using an 8-Minute Free-Running Delayed-Phase DCE-CMR

Presenting Author(s)

• 

  Xinheng Zhang, PhD

  Post-doctoral Scientist
  Cedars-Sinai Medical Center
  Redwood City, California, United States

Primary Author(s)

• 

  Xinheng Zhang, PhD

  Post-doctoral Scientist
  Cedars-Sinai Medical Center
  Redwood City, California, United States

Co-Author(s)

• 

  Li-Ting Huang, MD, MSc

  radiologist
  National Cheng Kung University Hospital
  Los Angeles, California, United States
• AM

  Archana V. Malagi, PhD

  Postdoctoral Scientist
  Cedars-Sinai Medical Center
  Los Angeles, California, United States
• XL

  Xinqi Li, MSc

  Ph.D student
  Cedars-Sinai Medical Center
  Berlin, Berlin, Germany
• YH

  Yuheng Huang

  Ph. D. Candidate
  UCLA, United States
• HH

  Hao Ho, PhD

  Assistant Professor
  UCLA, United States
• AK

  Alan C. Kwan, MD, MSc

  Imaging cardiologist
  Cedars-Sinai Medical Center, United States
• 

  Janet Wei, MD

  Assistant Professor
  Cedars Sinai Medical Center
  Los Angeles, California, United States
• 

  Xiaoming Bi, PhD

  Director, Cardiovascular MR Collaborations
  Siemens Medical Solutions USA, Inc.
  Oak Park, California, United States
• 

  Anthony G. Christodoulou

  Associate Professor
  University of California, Los Angeles (UCLA)
  Los Angeles, California, United States
• 

  Debiao Li, PhD

  Professor
  Cedars Sinai Medical Center
  Los Angeles, California, United States
• 

  Rohan Dharmakumar, PhD

  Executive Director
  Indiana University School of Medicine
  Indianapolis, Indiana, United States
• 

  Hsin-Jung (Randy) Yang, PhD

  Associate Professor
  Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA
  Los Angeles, California, United States

Background: CMR is the reference standard for myocardial tissue characterization and functional assessment, essential for managing myocardial injury. Conventional late gadolinium enhancement (LGE) requires a 10–15 min delay, is qualitative only, and limited to focal scar detection. Delayed-phase dynamic contrast-enhanced (dDCE) modeling leverages washout dynamics to reproduce LGE appearance without delay and provide quantitative biomarkers including extracellular–extravascular space (Ve) and microvascular permeability–surface area product (PS)—that reflect remodeling and microvascular injury¹. Current dDCE methods rely on 2D T1 mapping, constrained by breath-holds and limited coverage. To overcome this, we developed a free-breathing, non-ECG-gated 3D low-rank tensor (LRT) dDCE framework. This approach enables motion-resolved whole-heart dDCE mapping in an 8-min free-running scan, allowing rapid quantitative assessment of function, LGE burden, and microvascular integrity. The method was tested in a reperfused MI canine model and validated against conventional 2D breath-hold protocols.

Methods: A free-running 3D IR-prepared Cartesian sequence with randomized Gaussian sampling was implemented at 3T and reconstructed using the multitasking LRT framework^2,3. Images were modeled as a 5-way tensor spanning spatial and temporal dimensions (cardiac, respiratory, T1 recovery, and contrast washout). Reperfused MI was induced in canines (n=5, 1-week post-MI). Each subject underwent two imaging sessions, 48 hours apart to allow contrast washout, with both LRT-dDCE (8 min) and conventional scans (cine, MOLLI, and 2D dDCE). Gadolinium kinetics were fit to a modified two-compartment exchange model(2CXM)⁴ to derive F, Vp, Ve, and PS. Simulated LGE images were generated from dDCE maps and compared with conventional 15-min LGE. A schematic study design is depicted in Fig.1.

Results: Whole-heart LRT-dDCE provided robust, motion-resolved maps of function, LGE, and quantitative parameters (Fig. 2). LVEF and MI volume strongly agreed with conventional methods (LVEF: 32.0±6.2% vs. 32.0±6.5%, R²=0.966; MI volume: 18.1±5.2% vs. 17.6±4.0%, R²=0.969). Ve and PS values correlated with 2D dDCE measurements (P=0.48 and 0.27, respectively) but provided improved whole-heart coverage. In infarcted regions, Ve and PS were significantly elevated compared with remote myocardium (Ve: 57.2±30.6% vs. 4.7±3.0%, P< 0.001; PS: 32.1±19.8 vs. 3.4±4.5 mL/min/100mL, P< 0.001), delineating extracellular expansion, microvascular leakage, and regions of microvascular obstruction.

Conclusion: Free-running 3D LRT-dDCE enables rapid (8 min), whole-heart, motion-robust imaging for simultaneous quantification of function, infarct burden, extracellular–extravascular space, and microvascular permeability. This approach overcomes the limits of breath-hold 2D methods, delivers quantitative biomarkers beyond LGE, and holds strong translational potential for tissue characterization and therapeutic monitoring across a wide spectrum of heart diseases.

Figure 1. Image acquisition and processing pipeline for the proposed method. I. images were acquired post contrast with proposed protocol: 3D Free running continuous IR based sequence in 8 min post-Gd; Protocol 2: conventional 2D single-slice acquisition of post-contrast MOLLI T1 maps. II. for protocol 1, images were reconstructed by LRT algorithm and Gd concentration was computed and put into the modified 2CXM model to derive the corresponding dDCE parameters. III. the EGE images and PS, Ve, Fp and Vp maps were derived, and LGE was simulated based on dDCE maps for further analysis. For Protocol 2, the same pipeline was applied to MOLLI T1 maps for dDCE map validation.
Figure 2. The dynamic curve, LRT-dDCE (PS, Fp, vp, and ve) maps and Gd concentration maps of the representative canine subject were successfully reconstructed. A) the dynamic curves obtained from the 2CXM can differentiate the MI, remote, and blood pool. B), the PS, Fp, Vp, and Ve maps from the LRT-dDCE images showed the corresponding changes related to MI region and MVO. Elevated Ve and PS were demonstrated in the hyperintense region on the LGE image. Fp and Vp were decreased in the hypointense cores on the EGE image. C) LRT-dDCE images showed consistent dynamic Gd concentration images (1-8 min) and simulated late Gd concentration maps can be further simulated (9-15 min).
Figure 3. Visual and quantitative comparison of dDCE parameters from the proposed (8 min) and conventional 2D (30 mins) acquisitions in myocardial infarction (MI) and remote regions. A) LRT-dDCE and conventional 2D dDCE maps (PS, Ve) derived from gadolinium enhancement are compared to corresponding LGE and EGE images. Elevated PS and Ve correlate with MI (hyperintense on LGE, marked by arrowheads). Linear regression of B) LVEF and C) MI volume validates with conventional acquisitions. D) Ve and E) PS show significant differences between MI and remote regions, with no statistical difference between LRT-dDCE and MOLLI-dDCE within respective regions (**P < 0.01, ***P < 0.001).