Validation of sub-segmental visual scoring for the quantification of ischemic and nonischemic myocardial fibrosis using late gadolinium enhancement MRI

Purpose: To determine the accuracy and reproducibility of late gadolinium enhancement (LGE) MRI scar quantification using visual sub-segmental analysis (VSSA) versus signal threshold-based analysis in ischemic and nonischemic cardiomyopathy. Materials and Methods: One-hundred sixty-one patients with abnormal LGE imaging underwent VSSA and signal threshold-based analysis. VSSA was performed using a 68 sub-segmental model. Signal threshold-based analysis was performed using cutoffs of ≥2, ≥3, and ≥5 standard deviations (SD) above the mean signal of normal myocardium. Comparison of VSSA and signal threshold-based analysis was performed by linear regression and Bland Altman plots. Results: Seventy (44%) patients had ischemic scar, 76 (47%) had nonischemic scar, and 15 (9%) had a combined pattern. Correlation coefficients for VSSA versus signal threshold-based analysis at ≥2, ≥3, and ≥5SD thresholds were r = 0.63, r = 0.79, r = 0.81 (P < 0.001) for all patients, r = 0.74, r = 0.81, r = 0.81 (P < 0.001) in those with ischemic scar, and r = 0.46, r = 0.69, r = 0.72 (P < 0.001) in those with nonischemic scar. Bland Altman analysis revealed no significant bias in total scar volume among all patients (-4.3 ± 7.9%), those with ischemic scar (-4.8 ± 7.8%), or those with nonischemic scar (-2.6 ± 7.6%). Intra-observer and inter-observer variability of the VSSA technique was excellent with a mean difference in total percent scar of 0.3% (-8.3-8.9%) and -0.4% (-9.5-8.5%), respectively. Conclusion: A VSSA-based model of myocardial scar quantification is accurate and reproducible in ischemic and nonischemic cardiomyopathy.