The effect of thresholding techniques on 3D quantitative scar metrics in patients with implantable cardioverter defibrillator

LGE images are reconstructed using either magnitude (MAG) or phase-sensitive inversion recovery (PSIR), each producing different signal intensities due to distinct handling of longitudinal magnetization. These differences in signal intensity can influence LGE quantification, which typically relies on full width at half maximum (FWHM) or n-standard deviation (n-SD) thresholding to predict adverse cardiac events.

This study compared the impact of FWHM and n-SD on MAG and PSIR-derived scar characteristics and sudden cardiac death (SCD) risk stratification.

Ischemic cardiomyopathy (ICM) patients undergoing LGE imaging prior to ICD implantation were retrospectively identified. MAG and PSIR images were post-processed using ADAS LV (ADAS LV Medical, Barcelona, Spain) with cut-off values to differentiate scarred from healthy myocardium set at 40-60% of maximum signal intensity for FWHM and 2-5 standard deviations above the mean for n-SD. Scar metric comparisons between the specified acquisition and thresholding techniques were made. Additional interface metrics were obtained using customized scripts.

Among the 90 patients included, 80 (88.9%) had PSIR images. MAG analysis showed significant differences between FWHM and n-SD in core scar (11.15±7.75g vs. 23.59±14.96g, p<0.001), border zone (BZ) (12.57±6.85g vs. 16.43±8.15g, p<0.001), conduction corridor (corridor) mass (4.16±3.80g vs. 2.65±2.74g, p=0.024), corridor length (186.08 ± 160.31mm vs. 132.16 ± 121.86mm, p=0.038) and total corridor protectedness (89.36±95.87mm vs. 53.64±58.27mm, p=0.027). Additionally, the interface metrics between healthy myocardium and scar (84.30±45.56cm2 vs. 105.13±54.58cm2, p<0.001) and the interface between BZ and scar (77.62±43.78cm2 vs. 91.15±49.27cm2, p=0.001) were significantly different between FWHM and n-SD.

For PSIR reconstruction, core scar, BZ, corridor mass and length, total corridor number and protectedness did not differ significantly between FWHM and n-SD (p= 0.58, 0.91, 0.18, 0.51, 0.29 and 0.53, respectively). Interestingly, the interface between healthy myocardium and scar (112.47±60.50cm2 vs. 101.38±54.37cm2, p=0.004), as well as BZ and scar (97.74±51.59cm2 vs. 90.06±50.57cm2, p=0.014) were different for FWHM and n-SD using PSIR reconstructed images.

The choice of thresholding technique significantly impacts quantification of myocardial scar characteristics in MAG images, with FWHM and n-SD yielding markedly different measurements. This variability is not present in PSIR images, where both thresholding methods produce consistent results. These findings highlight the importance of considering the reconstruction technique and thresholding method when analysing LGE-images, to ensure accurate assessment and informed clinical decision-making.Figure 1  Figure 2.