Reproducibility Assessment of Biventricular Strain Derived from Long-Axis Feature Tracking in Travelling Volunteers – A Study in the Berlin Research Network for Cardiovascular Magnetic Resonance

Presenting Author(s)

• JG

  Jan Gröschel, MD

  MD
  Charite
  Berlin, Berlin, Germany

Primary Author(s)

• 

  Clemens Ammann, MD

  Physician
  Charité – Universitätsmedizin Berlin
  Berlin, Berlin, Germany

Co-Author(s)

• RT

  Ralf Felix Trauzeddel, MD

  Researcher- Medical Doctor
  Charité - Universitätsmedizin Berlin
  Berlin, Berlin, Germany
• MM

  Maximilian Müller, MD

  Doctoral candidate
  Charité - Universitätsmedizin Berlin
  Berlin, Berlin, Germany
• LG

  Leonhard Grassow

  Medical student
  Charité – Universitätsmedizin Berlin, Germany
• 

  Thomas C. R Hadler, PhD

  Postgraduate
  Charité - Universitätsmedizin Berlin
  Berlin, Berlin, Germany
• DV

  Darian Viezzer, PhD

  Project Manager
  Helios IT Service GmbH, Berlin, Germany, Germany
• 

  Richard Hickstein, MD

  Physician and Researcher
  Charité – Universitätsmedizin Berlin
  Berlin, Berlin, Germany
• LZ

  Leonora Zange, MD

  Resident
  Charité – Universitätsmedizin Berlin
  Berlin, Berlin, Germany
• 

  Edyta Blaszczyk, MD

  Cardiologist
  Charité - Universitätsmedizin Berlin
  Berlin, Berlin, Germany
• ED

  Elias Daud, MD

  Physician
  Charité – Universitätsmedizin Berlin, Germany
• 

  Jeanette Schulz-Menger, MD

  Head Working Group Cardiac MRI
  Charité/ University Medicine Berlin and Helios
  Berlin, Berlin, Germany
• JG

  Jan Gröschel, MD

  MD
  Charite
  Berlin, Berlin, Germany

Background: Assessment of left and right ventricular (LV, RV) strain can help identify early changes in myocardial structure and function (1). With the advancement of post-processing methods, the analysis of feature tracking (FT)-based strain has become feasible using routinely acquired cine images in cardiovascular magnetic resonance (CMR) (2). Global longitudinal strain (GLS) measures systolic ventricular shortening and is a sensitive predictor of adverse clinical outcomes (3). This study aims to analyse the multicentre reproducibility of LV and RV GLS assessments in a cohort of travelling volunteers in the Berlin research network for cardiovascular magnetic resonance (BER-CMR) (4).

Methods: Eighteen travelling volunteers were prospectively scanned at four sites with same-vendor (Siemens Healthineers, Forchheim, Germany) scanners at 3.0 T (sites I, II, III) and 1.5 T (site IV). Cine imaging in three long-axis views was performed using a segmented balanced steady-state free precession sequence with 30 cardiac phases except site II with 25 phases. At one site a scan-rescan was performed to assess reproducibility and establish intervals for tolerable mean deviations based on 1.96 standard deviations (5). Post-processing with dedicated software (cvi42, Circle Cardiovascular Imaging, Calgary, Canada) was carried out to derive GLS from manually quality controlled auto-segmentations. The significance (α = 0.05) of mean differences between sites was assessed using analysis of variance (ANOVA) and t-tests for post-hoc pairwise comparisons.

Results: Post-processing was carried out successfully in a core lab setting. Pairwise comparisons revealed significant differences in LV GLS between sites I and II (p < 0.001) and sites II and IV (p = 0.013), as well as in RV GLS between sites I and IV (p = 0.027) (Figure 1). RV GLS values were significantly higher at 3.0 T (p = 0.024), whereas field strength had no significant impact on LV GLS (p = 0.153) (Figure 2). Conversely, the use of 25 cardiac phases at site II was associated with significantly lower LV GLS values (p < 0.001), while RV GLS remained unaffected (p = 0.825) (Figure 2). Equivalence of mean results was demonstrated for all parameters except LV GLS between sites I and II, and sites II and IV (Figure 3).

Conclusion: When applying FT-based strain in a multicentre study, careful consideration should be given to the temporal resolution for LV longitudinal strain and to magnetic field strength for RV longitudinal strain.

Figure 1: Boxplots for left and right ventricular global longitudinal strain. Boxplots represent median (solid line inside the box), interquartile range (IQR, box) and Q1 - 1.5*IQR or Q3 + 1.5*IQR (whiskers) for A) left and B) right ventricular global longitudinal strain at each site (site I=3T, site II=3T, site III=3T, site IV=1.5T). p-values: one-way ANOVA (overall); t-test (pairwise; for significant comparisons).
Figure 2: Boxplots for left and right ventricular global longitudinal strain by imaging phases and field strength. Boxplots represent median (solid line inside the box), interquartile range (IQR, box) and Q1-1.5*IQR or Q3+1.5*IQR (whiskers). p-values: independent t-test.
Figure 3: Equivalence testing for left and right ventricular global longitudinal strain across all sites. The underlying tolerance intervals (green areas) were derived from scan-rescan comparisons at site I (±1.96 standard deviations of the difference). Error bars represent the 95% confidence interval for the mean bias from inter-site comparisons. Equivalence is assessed if the error bar lies completely within its corresponding tolerance interval (true for all comparisons except site I vs. II and site II vs. IV for left ventricular global longitudinal strain).