Article Text
PDF
Abstract
Background The GlasgowHeart method is designed to overcome some limitations of currently available feature-tracking methods by incorporating all of the myocardial tissues using an in-house developed intensity-based b-spline deformable registration method. The aim of this pilot study was to ensure that peak circumferential strain (Ecc) estimation is feasible and reproducible with minimal intra- and inter- observer variability.
Methods 20 healthy volunteers underwent 1.5T CMR twice, < 2 days apart. Mid- LV cine sequences, were analysed with the Glasgowheart software (Figure 1). Two observers independently analysed 40 short axis slices for inter-observer variability. One observer reanalysed the 40 short axis slices. Pearson correlation and Bland-Altman analysis were used.
Results 20 participants were used in the analysis (mean age ± SD 49.5 years (17.2) 50% male). Ecc measured on the first set of MRIs by the two observers was highly correlated (R = 0.915, p < 0.001) and in excellent agreement (mean difference = 0.01; 95% LoA: −0.01, 0.02). Repeated image analysis also disclosed a high degree of association in paired measurements of Ecc that was strongly correlated (R = 0.915, p < 0.001) and in excellent agreement (mean = 0.00; 95% LoA: −0.02, 0.01). Ecc measured in the second set of MRIs by 2 observers was well correlated (R = 0.937, p < 0.001) and in excellent agreement (mean = 0.00; 95% LoA: −0.016 and 0.021). The repeated image analysis at follow-up yielded Ecc that was well correlated (R = 0.942, p < 0.001) and in excellent agreement (mean = 0.00; 95% LoA: −0.009 and 0.009). There was no difference between the average global Ecc at different time points (p > 0.05).
Conclusion The GlasgowHeart method is a robust and reproducible method of assessing cine-derived circumferential strain. By tracking a higher proportion of voxels than the currently available feature tracking software, it has clear potential to provide a more accurate assessment of strain.
The Glasgowheart pixel-tracking software. (A) shows border delineation (B) shows automatic segmentation using the inferior right ventricular insertion point. (C) and (D) show graphical output as circumferential and radial strain graphs, respectively