The aim of our study was to study the suitability of HPF and Segment software in analyzing rotation and torsion of the left ventricle in different study groups. Analysis was successful in all cases and usage of both software was robust. Absolute end-systolic rotation and torsion values differed significantly between the two software, Segment showed systematically larger values compared to HPF. Software-specific reference values are required when comparing absolute values of rotation and torsion between different software. Correlations between the rotation and torsion values obtained with HPF and Segment were statistically significant, most significantly in the apical plane.
Based on the results of both software, absolute torsion of the left-ventricle does not differ significantly between HCM patients, mutation carriers, and healthy controls. Rotation of the apical level of the left ventricle is increased in mutation carriers. On the contrary, basal rotation is slightly lower with this group (Fig. 2). As torsion is calculated as the angle difference between apical and basal rotations, the differences in rotation are not seen in the torsion curves (Fig. 3). This could indicate that HCM mutation without actual hypertrophy could already affect the motion of the left ventricle. However, the sample size is too small to draw definitive conclusions.
The effects of HCM on the motion of the left ventricle have been studied before but as He et al. summarized, the effects of location and extent of hypertrophy on left ventricular function have not attracted much attention, and further studies should be conducted to deepen the understanding of disease mechanics (19). Previous studies show that peak torsion in HCM patients can be slightly higher when compared to control groups (20). However, in the case of apical hypertrophic cardiomyopathy, apical rotation can be markedly decreased while basal rotation is preserved (21). In three of the HCM patients in our study, end-systolic apical rotation was negative, leading to an impaired torsion value. As HCM is a heterogeneous disease with diverse effects on the motion of the heart, more work is needed to understand the mechanics of the HCM heart.
Previous studies indicate end-systolic torsion in healthy subjects, between basal and apical levels of the left ventricle was 7.7º ± 1.4º (Rüssel at al. (22)). Our result with Segment was similar (8.0° ± 2.6°). However, the result of the HPF was significantly lower (3.8° ± 1.3°). The definition for torsion used by Rüssel at al. uses the mean radius of apical and basal levels divided by the mean distance between the two planes to normalize the torsion between different sized hearts. The definition used by HPF for torsion in our work is absolute and does not take in to account the size of the heart.
Gabor filter-based motion tracking has been previously shown to result in accurate segmentation of the tagging lines (23). Gabor filtering achieves optimal locating in both spatial and Fourier domain making it more suitable for tag line motion analysis compared to pure HARP (24). The performance of the HPF algorithm has been assessed by the developers using a computational model as well as clinical data. They concluded that with a sufficient image quality, HPF tracks motion correctly within sub-pixel accuracy (10). Similarly, Segment strain tagging module has been validated by the developers, and the underlying non-rigid elastic registration based motion analysis algorithm has been show to yield clinically reproducible results (14, 25).
Apical tagging imaging plane should be chosen carefully, as apical rotation is dependent on the imaging plane. The generally lower torsion values of HPF in our study could be due to the apical plane being close to the mid-level of the left ventricle. Increased end-systolic apical rotation and end-systolic torsion values were detected in subjects with longer distance between the tagging planes. Our study is limited to the rotation and torsion analysis of systole only, and the sample sizes are small to draw definitive conclusions on the acquired rotation and torsion values.