Oral Abstract 3: Analysis of regional left ventricular strain in cardiac amyloidosis using 3D cine DENSE CMR

Presenting Author(s)

• GH

  George E. Harpur, MEng

  Research Assistant
  Auckland Bioengineering Institute, New Zealand

Primary Author(s)

• GH

  George E. Harpur, MEng

  Research Assistant
  Auckland Bioengineering Institute, New Zealand

Co-Author(s)

• 

  Vicky Y. Wang, PhD

  Research Fellow
  The University of Auckland, New Zealand
• HH

  Holly Hogan-Murphy, BEng

  Research Assistant
  Auckland Bioengineering Institute, New Zealand
• SW

  Sheng-Ya Wang, BEng

  Graduate Student
  Auckland Bioengineering Institute, New Zealand
• DZ

  Debbie Zhao, PhD

  Research Fellow
  Auckland Bioengineering Institute, New Zealand
• TB

  Thiranja P. Babarenda Gamage, PhD

  Research Fellow
  Auckland Bioengineering Institute, New Zealand
• BL

  Boris S. Lowe, MB

  Cardiologist
  Green Lane Cardiovascular Service, New Zealand
• NE

  Nicola C. Edwards, PhD

  Cardiologist
  Waitematā Cardiology, New Zealand
• ML

  Malcolm E. Legget, MD

  Cardiologist
  The Heart Group, New Zealand
• RD

  Robert N. Doughty, MD

  Cardiologist
  The Heart Group, New Zealand
• TS

  Timothy M. Sutton, MB

  Cardiologist
  The Heart Group, New Zealand
• 

  Alistair A. Young, PhD

  Professor
  King's College London, United Kingdom
• 

  Martyn P. Nash, PhD

  Professor of Biomedical Engineering
  Auckland Bioengineering Institute, New Zealand

Background: Cardiac amyloidosis (CA) is characterised by the deposition of insoluble misfolded fibrillar protein in the myocardium. Cardiac magnetic resonance (CMR) myocardial T1 values are known to be elevated in cardiac light-chain (AL) [1] and transthyretin (ATTR) amyloidosis [2]. CMR derived left ventricular (LV) global strains can also serve as useful diagnostic indicators, but regional strain patterns and their relationships to tissue structure in CA are uncertain. 3D cine displacement encoding with stimulated echoes (DENSE) CMR provides voxel-based myocardial displacements [3], from which cardiac strains can be derived to characterise regional mechanics. This study used 3D cine DENSE CMR for non-invasive assessment of LV function in both patients with CA and control participants.


Methods: Twelve participants (6 healthy controls, 6 CA - 5 ATTR and 1 AL) were prospectively imaged with a standard cine and 3D cine DENSE CMR protocol. Non-contrast T1 mapping was performed in all but four of the control participants. Four mid-ventricular short-axis (SAX) 3D cine DENSE images were used to reconstruct systolic kinematic fields using a finite element modelling framework (Fig 1). This included 5 steps: 1) segmentation and slice correction; 2) phase unwrapping using DENSEanalysis [4,5] to generate an Eulerian displacement field at end-systole (ES); 3) fitting a cylindrical template mesh to parameterise ES geometry; 4) free-form deformation (FFD) fitting of the ES model to the Eulerian displacement data to obtain systolic deformation using end-diastole (ED) as the reference; and 5) using the kinematic model to interpolate regional strains - circumferential (E_CC), longitudinal (E_LL) and circumferential-longitudinal shear (E_CL) - at multiple sample points within six mid-ventricular regions of the LV AHA model (regions 7-12). The regional strain distributions were analysed, and compared with T1 times to investigate the relationship between structural and functional remodelling.


Results: Regional analysis showed significant depression of E_CC in all regions in patients with CA compared to the control group (Fig 2a), with median values that were substantially lower (43% - 51%, Tab 1) in septal and inferior regions (8, 9 and 10). In contrast, E_LL was homogeneously depressed in the CA group (Fig 2b), consistent with the longer T1 times (Fig 2d). Patients with CA gave regionally uniform T1 times compared to the regional differences we expect with controls [6]. E_CL showed no consistent differences between groups (Fig 2c). T1 times exhibited moderate positive correlations with E_LL (Pearson correlation coefficient, r = 0.68, Fig 2f) and E_CC (r = 0.63, Fig 2e), but not with E_CL (r = -0.06, Fig 2g).


Conclusion: Patients with CA exhibited regional kinematic differences with heterogeneous depression in E_CC compared with controls. Non-invasive assessment of 3D strain warrants further study given the potential role in the differential diagnosis of CA from LV hypertrophy phenocopies, which could be used to inform earlier and more appropriate treatment for patients with CA.

Acknowledgements: this research was funded by the Health Research Council of New Zealand (programme grants 17/608 and 23/527).


Figure 1. Finite element modelling framework used to reconstruct systolic mid-ventricular kinematic fields. This included 5 steps: 1) segmentation and slice correction; 2) phase unwrapping using DENSEanalysis [4,5] to generate an Eulerian displacement field at end-systole (ES); 3) fitting a cylindrical template mesh to parameterise ES geometry; 4) free-form deformation (FFD) fitting of the ES model to the Eulerian displacement data to obtain systolic deformation using end-diastole (ED) as the reference; and 5) using the kinematic model to interpolate regional strains.

Figure 2. Mean regional strains and T1 times. Box plots (top row) show mean regional values of (a) ECC, (b) ELL, (c) ECL and (d) T1 for the six mid-LV AHA regions (labelled in the inset). Asterisks indicate statistically significant differences (p < 0.05, Tab 1). Scatter plots (bottom row) show correlation of mean regional (e) ECC, (f) ELL and (g) ECL with T1 for all regions across all cases, along with the Pearson correlation coefficients (r values). Note that strains were available for all six control and all six cardiac amyloidosis (CA) participants, whereas T1 times were not available for four of the control cases.