Histological validation of dark-blood LGE quantification methods in rat myocardial infarction models: A 3.0 T CMR...

Current studies of cardiac magnetic resonance late gadolinium enhancement (CMR-LGE) in rat myocardial infarction (MI) models are mostly based on 7.0 T MRI, which is not widely available. In addition, differences in infarct volume detected by various quantification methods for dark-blood LGE remain unclear. This study is the first to systematically validate different dark-blood LGE quantification methods using 3.0 T MRI in an SD rat MI model. The aim was to assess the reliability, consistency, and agreement with histology, providing a methodological reference for in vivo quantification of myocardial fibrosis in rat MI models and offering experimental and theoretical support for optimizing quantitative analysis strategies in future clinical studies.

Methods: Finally, 20 SPF male SD rats were included (5 in the control group and 15 in the myocardial infarction model group). MI models were created by open-chest ligation of the left anterior descending coronary artery. Four weeks later, 3.0 T CMR dark-blood LGE scans were performed. After heart excision, continuous 4-6 μm paraffin sections were prepared. Masson trichrome staining was used to determine collagen volume fraction as the histological standard. LGE images were analyzed with CVI42 software. Quantitative results were obtained using manual delineation, 2SD threshold, 3SD threshold, 5SD threshold, and full-width half-maximum (FWHM) methods. Consistency with histology was assessed using intraclass correlation coefficients (ICC), concordance correlation coefficients (CCC), Pearson correlation, and Bland-Altman analysis.

Consistency analysis revealed significant differences among the five LGE postprocessing methods compared with the histological gold standard (P < 0.05). The manual method demonstrated excellent intra- and inter-observer consistency (ICC > 0.95) and significantly outperformed the automated methods. Lin's concordance correlation coefficient and Pearson correlation analyses indicated the highest agreement with histology for the manual method (CCC = 0.895, 95 % CI: 0.745-0.958; r = 0.926, P < 0.001), surpassing all automated threshold methods.Correlation analyses confirmed a strong agreement between the manual method and histology, consistent with the CCC results. Among the automated methods, FWHM (CCC = 0.689, 95 % CI: 0.393-0.856) and 5SD (CCC = 0.682, 95 % CI: 0.388-0.850) performed relatively well, whereas 3SD (CCC = 0.617) and 2SD (CCC = 0.474) showed poorer agreements. Bland-Altman analysis supported this trend: the manual method exhibited the smallest systematic bias (mean bias = -1.00) and narrowest 95 % limits of agreement (-5.88 to +3.88). Among the automated methods, 5SD had the smallest bias (mean bias = +0.18), whereas 2SD showed the largest bias (mean bias = +5.96) and the widest limits of agreement.

Using 3.0 T dark-blood LGE imaging, in vivo assessment of myocardial fibrosis in SD rat MI models is feasible. Manual delineation showed the highest agreement with histology and the best quantification accuracy. Among the signal intensity threshold methods, FWHM and 5SD demonstrated relatively high consistency and stability, providing good reference data for the in vivo evaluation of myocardial infarction in rats.