Variations in threshold values for border zone and dense scar produce significant changes in scar parameters obtained by ADAS-3D.

Background: ADAS-3D software elaborates cardiac magnetic resonance (CMR) images to obtain a quantitative evaluation of dense scar and border zone (BZ), including BZ channels, which can be useful for ventricular tachycardia ablation and risk stratification. However, most prior reports with ADAS-3D used flexible thresholds (60% ± 5% and 40% ± 5% of maximum pixel signal intensity) to define dense scar and BZ. The impact of such variations of the threshold values on the measurements obtained with ADAS-3D is unknown.

Objective: This study aimed to quantify the degree of change in ADAS-3D measurements when different thresholds for dense scar and BZ are employed.

Methods: A single-center retrospective observational cohort study including 87 consecutive patients with previous myocardial infarction who underwent CMR was conducted. ADAS-3D software semiautomatically processed CMR sequences. We compared the scar measurements obtained with the 9 possible combinations of thresholds (55%/60%/65% and 35%/40%/45% of maximum pixel signal intensity).

Results: The overall comparison between thresholds showed highly significant differences (P < .001) in all scar parameters. Not a single patient maintained the same number of BZ channels with all the thresholds settings. A percentage difference of up to 200% in BZ channel numbers and channel mass was observed in all 36 comparisons. An absolute difference of up to 10 channels was also recorded. Of note, the highest median channel mass (obtained with the thresholds 35-65) was 59-fold higher compared with the lowest one (obtained with the 45-55 cutoffs).

Conclusion: Variations in threshold values result in statistically significant and high-magnitude changes in the quantification of scar parameters by ADAS-3D.