This study aimed to examine fibrosis in histological images of the left ventricle in patients with non-ischemic cardiomyopathy. Additionally, it aimed to create a framework for generating realistic fibrotic structures that could be valuable for studying arrhythmias.
We analyzed the geometrical characteristics of fibrosis in histological images from ten human hearts with non-ischemic cardiomyopathy. This involved labeling connected fibrotic clusters and calculating various geometric properties, including their size, orientation, and solidity. To replicate fibrosis resembling histological samples, we introduced a method derived from the multiple-points simulation technique, an approach commonly employed in geostatistics.
The analysis of fibrosis revealed that the size distribution of fibrotic clusters does not align with uniformly generated fibrosis commonly used in computational studies. Moreover, fibrosis has an average elongation factor of 2.4 and the direction of elongation can vary depending on the location within the heart wall. In the mid-myocardium, fibrosis predominantly aligns with fibers, with the 25th, 50th, and 75th percentile quartiles of the angle between clusters and fibers being -12.6°, 0.8°, and 13.2°, respectively. However, in the sub-epicardium (-36.4°, 15.2°, 50°), and particularly in the sub-endocardium (-44°, -8.5°, 25.8°), elongation is less pronounced. With the developed method, we successfully replicated randomly sampled images from our dataset. The quality of the match was visually assessed and controlled by the parameters of the fibrotic structures. Moreover, we introduced tools for generating fibrotic textures with tailored properties, including density, orientation, and elongation.
In conclusion, our analysis emphasizes the need for image generation methods that accurately capture the properties of fibrotic textures. We introduced tools for generating realistic fibrotic images, which provide new opportunities in studies of arrhythmias.