MRI tagging is a valuable method for evaluating regional heart function. Currently, there are a number of different techniques for analyzing the tagged images, which are based on different analysis algorithms. The purpose of this study is to compare the harmonic-phase (HARP) and sine-wave modeling (SinMod) tagging analysis techniques for evaluating myocardial strain and torsion in healthy controls (HC) and patients with type-1 diabetes (T1DM).

HARP isolates the harmonic signal peak in the k-space of the tagged image and constructs a harmonic-phase (HARP) image, as illustrated in Figure 1. The basic underlying principle of HARP tracking is that harmonic phase is an imposed material property.¹ Therefore, by tracking the harmonic phase vector of each pixel over time, one can track the position and, by extension, the displacement of each material point over time. The SinMod technique is a frequency-based method for extracting myocardial motion from the tagged images based on sinusoidal approximation.² In SinMod, the intensity distribution in the neighborhood of each pixel is modeled as a summation of sinusoidal wavefronts, which are locally attached to the moving myocardial tissue. For each sinusoidal wave, the displacement perpendicular to the wavefront is estimated for each pixel, resulting in a map showing this displacement component. The flowchart of the SinMod algorithm is shown in Figure 2.

Thirteen T1DM patients and eight matched HC (Table 1) underwent MRI exam that included cine; tagged; and mitral flow images, which were analyzed to measure ventricular mass and ejection fraction (EF); myocardial strain and ventricular torsion (= (basal rotation - apical rotation) / base-to-apex distance); and mitral early-to-atrial flow rate (E/A; as a measure of diastolic dysfunction), respectively. The tagged images were analyzed by two experts using the Diagnosoft HARP and inTag SinMod software packages (Figure 3). The tagging analysis was conducted on short-axis slices at the base, mid-ventricular, and apical levels to measure circumferential strain and apical-to-base torsion. Correlation analysis was conducted to evaluate the relationship between HARP and SinMod measurements and to study inter-observer agreements. Student’s t-test was conducted to evaluate the significance of the measurements’ differences between HARP and SinMod and between HC and T1DM (P<0.001 was considered significant).

As shown in Table 2, all SinMod measurements were significantly (P<0.001) larger than those by HARP. Nevertheless, there existed consistency in the measurements by each technique, as seen by the good correlation between the HARP and SinMod measurements in both controls and patients, except for apical strain (in both patients and controls) and mid-ventricular strain in patients. Further, the two techniques resulted in close P values (show on Table 2) between the measurements in both patients and controls, except for apical strain and torsion. All the measurement differences were insignificant between the patients and controls. Overall inter-observer correlations were acceptable, except for a few outliers in the data. The inter-observer agreements were similar in the patients and controls. For all studied subjects, the inter-observer correlation coeficients for HARP/SinMod were 0.71/0.79, 0.75/0.82, 0.69/0.45, 0.72/0.74, and 0.88/0.79 for basal strain, mid-ventricular strain, apical strain, global strains, and apical-to-basal torsion, respectively. All correlations between inter-observer measurements were significant (P<0.001), except for SinMod apical strain where P=0.04.

The results showed measurements’ exaggeration by SinMod compared to HARP. However, there existed consistency in the measurements by each technique, as described in the Results section. Most differences occurred in apical measurements, which can be attributed to the small anatomical size (i.e. fewer taglines across the heart wall) and lower image quality in apical images. The insignificant differences between the patients and HC are not unexpected in T1DM due to the patients’ young age and nature of the disease, e.g. compared to type-2 diabetes. The inter-observer agreement was better in SinMod than in HARP for both torsion and strain (except at apex). Generally, the inter-observer correlation was better for global than segmental measurements, and for basal and mid-ventricular than apical strain.

In a previous study,³ the SinMod technique has been compared to HARP for assessing mid-ventricular strain, where the results showed good agreement for measuring global strain, although the agreement on the segmental level was substantially lower. Alternatively, our study investigated the techniques’ performance for measuring strain at different levels (basal, mid-ventricular, apical) as well as apical-to-basal torsion, which provides more detailed analysis of the HARP and SinMod differences.

In conclusion, care should be taken not to mix measurements from different tagging analysis techniques, especially in longitudinal or multi-center studies, as the measurement differences could be related to the implemented techniques, not to real differences between studied subjects.