Article Text


130 Comparison of harmonic phase imaging with local sine wave modelling for the assessment of circumferential myocardial strain using tagged cardiovascular magnetic resonance images
  1. A N Borg1,
  2. C A Miller1,
  3. C D Steadman2,
  4. G P McCann2,
  5. M Schmitt1
  1. 1University Hospital of South Manchester, Manchester
  2. 2NIHR Leicester Cardiovascular Biomedical Research Unit, Leicester, UK


Introduction Assessment of myocardial strain promises to become an important quantitative tool in early diagnosis of cardiac disease and treatment monitoring. Advances in image processing software have facilitated rapid and clinically feasible analysis of strain from tagged cardiac magnetic resonance (CMR) images. Harmonic Phase Analysis (HARP) or Local Sine Wave Modelling (SinMod) can be used for automated derivation of strain. We obtained tagged CMR images to compare measurements of left ventricular (LV) circumferential strain obtained using a HARP with a SinMod method.

Methods Ten normal controls, 10 hypertrophic and 10 dilated cardiomyopathy patients (mean age 46.6±14.8 years) were included. Spatial modulation of magnetisation using short-axis LV slices at mid-ventricular level, with a temporal resolution of 30–50 mS, were obtained using a 1.5 Tesla scanner (Siemens Avanto) with a 32-channel coil. Global and segmental transmural peak circumferential strains (εcc) were measured using HARP (Diagnosoft, USA, version 2.7) and SinMod (InTag, University of Lyons, France, version 3.6.1). Prior to running the algorithm, both methods involve manual tracing of the endocardial and epicardial borders, and localisation of right ventricle-to-septum insertion points, in one frame. Agreement between HARP and SinMod was assessed by Spearman's correlation co-efficient R and Bland Altman methods. Repeated measurements were carried out on 10 randomly selected scans to assess reproducibility.

Results There was a high level of agreement between HARP and SinMod for global εcc (HARP—SinMod mean difference: −0.12%, 95% limits of agreement: −5.69% to 5.45%, R=0.83, p<0.001) (Abstract 130 figure 1). Agreement was much lower for segmental εcc, ranging from very poor in lateral segments to modest in inferoseptal segments (Abstract 130 table 1). Analysis time using SinMod was significantly shorter than for HARP (84±42 vs 201±120 S, p=0.02). Inter- and intra-observer reproducibility were extremely high for SinMod measurements of global εcc (inter-observer R=0.99, repeatability co-efficient (RC) 2.14; intra-observer R=0.99, RC 1.49). Reproducibility of global εcc measurements by HARP was somewhat lower, but still high (inter-observer R=0.89, RC 4.80; intra-observer R=0.98, RC 2.73). There was much greater variability in segmental εcc measurements using both methods, particularly with HARP (Abstract 130 figure 2).

Abstract 130 Table 1
Abstract 130 Figure 2

Inter- and intra observer variability for HARP local sine wave modelling: repeatability co-efficients.

Conclusions HARP and SinMod methods show a high level of agreement for assessment of global mid-ventricular transmural circumferential strain, with good reproducibility for both individual methods. Agreement is much lower for segmental measurements; poor reproducibility for segmental measurements using both techniques probably reflect user variability in identification of right ventricular septal insertion points and contour tracing.

  • Strain
  • myocardial tagging
  • left ventricular function

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.