Ex-vivo validation of nine algorithms for quantifying infarcts with late gadolinium enhancement cardiovascular magnetic resonance

Fig. 1. Example on ROIs used. The cyan color ROI defines the myocardium of interest in which the scar region (yellow) is delineated using the FWHM method. The n-SD method uses both the cyan ROI and the remote region ROI (orange) for infarct delineation. ROI region of interest, FWHM full-width at half-maximum

Fig. 2. Comparison between high-resolution ex-vivo T1w CMR vs. TTC in a subset of the animals (n = 14) where TTC imaging was available. Filled circles indicate d0 and open circles indicate d7. TTC triphenyltetrazolium chloride, CMR cardiovascular magnetic resonance, T1w T1-weighted

Fig. 3. Agreement between IR in-vivo quantification methods and ex-vivo T1w high-resolution infarct measurements. Left panels show agreement between both measurements including identity line (dashed) and regression line in blue. Right panels show modified Bland–Altman blot plot including bias (solid line) and 95% limits of agreement in dashed lines. Bias in text gives bias ± SD. IR inversion recovery, T1w T1-weighted, SD standard devia

Fig. 4. Agreement between PSIR in-vivo quantification methods and ex-vivo T1w high-resolution infarct measurement. Left panels show agreement between both measurements including identity line (dashed) and regression line in blue. Right panels show modified Bland–Altman blot plot including bias (solid line) and 95% limits of agreement in dashed lines. Bias in text gives bias ± SD. PSIR phase sensitive inversion recovery, T1w T1-weighted

Fig. 5. Comparison of bias and SD between infarct quantification methods. The left column shows data for inversion recovery and the right column shows data for phase sensitive inversion recovery. Top row highlights time point after infarction (d0 vs. d3/7). Middle row highlights infarct morphology (solid vs. patchy). The bottom row highlights scanner vendor. There is no clear trend that any method performs differently on different types of infarctions EWA expectation maximization Weighted A priori information, FWHM full width at half maximum, SD standard deviation, FACT feature analysis and combined thresholding

Fig. 6. Comparison of bias and variability (SD) between infarct quantification methods. Data presented for different methods and infarct morphology (solid vs. patchy), and imaging time point (d0, vs. d3/7). EWA expectation maximization Weighted A priori information, SD standard deviation, FACT feature analysis and combined thresholding

Fig. 7. Comparison of solid and patchy infarctions. Top row shows a subject with solid infarction and the bottom row shows a patchy infarction. Left column shows in-vivo images where the endocardium is indicated in red and the epicardium in green. Middle column shows high-resolution T1w images used as reference standard and right column shows corresponding TTC images. Note that the patchy infarct shows a more insular infarction pattern, with areas of lower pixel intensity in both T1w and in-vivo LGE images. TTC triphenyltetrazolium chloride, T1w T1-weighted, LGE late gadolinium enhancement