Fibrosis quantification and operator expertise in the risk stratification of apical and non-apical hypertrophic cardiomyopathy

Abstract Background Hypertrophic cardiomyopathy (HCM) is a progressive genetic cardiac disease with significant clinical variability. Late gadolinium enhancement (LGE) on cardiac magnetic resonance (CMR) imaging quantifies myocardial fibrosis, with LGE ≥15% of left ventricular mass being a criterion for cardioverter-defibrillator implantation according to 2023 European Society of Cardiology (ESC) Guidelines. Different LGE quantification methods demonstrate variable reproducibility, but the impact of operator expertise on method performance remains unexplored. Purpose To evaluate intra- and inter-observer reproducibility of different LGE quantification methods and assess reproducibility based on operator expertise in HCM patients. Methods One hundred CMR studies of HCM patients (34 apical pattern, 66 non-apical) were analysed between September 2010-2023. An EACVI level III operator quantified LGE using five methods: 2-SD, 5-SD, 6-SD, full-width half-maximum (FWHM), and manual thresholding. Intra-observer variability was assessed by repeat measurements after one month. Inter-observer variability was evaluated comparing measurements between the level III operator and both another level III operator and a level I operator on 50 patients. LGE was quantified in grams and as percentage of LV mass. Reproducibility was assessed using intraclass correlation coefficients (ICC) with 95% confidence intervals. Results LGE assessment demonstrated high reproducibility between two EACVI level III operators for all methods except 2-SD in all HCM patients. Manual thresholding showed best intra-observer reproducibility (ICC 0.97, 95% CI 0.95-0.98). Among automated methods, FWHM (ICC 0.93, 0.87-0.96) and 6-SD (ICC 0.93, 0.88-0.96) yielded highest inter-observer reproducibility between level III operators across both phenotypical patterns. However, inter-observer reproducibility between level III and level I operators was poor across all quantification methods in both apical and non-apical HCM. In apical HCM subgroup analysis, manual thresholding and FWHM methods achieved high agreement amongst experienced operators only. Conclusions LGE quantification techniques are not interchangeable in HCM, with reproducibility dependent on both method and operator expertise across apical and non-apical variants. FWHM and 6-SD methods were most reproducible in both phenotypical patterns but performed poorly when utilised by less-experienced operators. No method ensured sufficient ease of use and reproducibility with inexperienced operators. LGE quantification by experienced operators is warranted in clinical practice for reliable scar quantification and risk stratification in both apical and non-apical HCM.