Phenotypic patterns in feline heart failure: A natural model for understanding variable disease severity in humans

Figures

• 

  Figure 1. Fibrosis quantification (A) Fibrosis pixel classifier. Representative histological images of fibrotic myocardial tissue stained with Masson’s trichrome (left: blue = fibrosis, red = cardiomyocytes). The corresponding pixel classification outputs (right) display fibrosis (red), cardiomyocytes (yellow), and background. Blood vessel lumens and erythrocytes are classified as background (arrow). (B) Examples of fibrosis quantification (red) in myocardial tissue (yellow) sections. In some samples, excessive endocardial fibrosis was observed. For the quantification of myocardial fibrosis, fibrotic areas at the inner and outer borders were filtered out. The brown line delineates the segmentation borders used to define myocardial fibrosis. (C) Boxplots display mean total fibrosis percentage, mean myocardial fibrosis percentage, and mean non-myocardial fibrosis percentage per outcome group (congestive heart failure [HF], HF with arterial thromboembolism [ATE], and no reported cardiac cause of death), with each dot representing an individual cat (HF: n = 7, ATE: n = 9, no cardiac death: n = 18). Pairwise comparisons were performed using unpaired t tests (mean total fibrosis percentage) or…

• 

  Figure 2. Vessel segmentation (A) The user provides an initial bounding box as input, after which the segment anything model (SAM) automatically segments the vessel structures. (B) Sample area selection. Vessels were segmented in the left ventricle and septum as the region of interest for analysis (dotted red outline). (C) Exemplary wall-to-lumen ratios (WLRs). Histological images of segmented vessels illustrate variations in wall thickness and lumen size across different cats. Calculated WLR and outer diameters (OD) are shown for each example. (D) WLR calculation. Schematic representation of the formula: WLR = (OD − lumen diameter)/lumen diameter. (E) Boxplots show WLR measurements categorized by vessel size group (small: <100 μm, n = 573 vessels; medium: 100–250 μm, n = 198 vessels; large: >250 μm, n = 95 vessels from 35 cats), from 37 cats total, stratified by outcome group (congestive heart failure [HF], HF with arterial thromboembolism [ATE], and no reported cardiac cause of death). Each dot represents one vessel. Overall group differences were assessed using Kruskal-Wallis tests; pairwise comparisons for the large vessel group were performed using a log-Gaussian linear mixed model…

• 

  Figure 3. Nucleus segmentation and morphometric analysis (A) Area selection. Six longitudinal and six transversal areas from HE-stained sections were sampled from the left ventricle and septum for nucleus segmentation and analysis. (B) Representative images of nuclei detected in transversal (left, dark gray border) and longitudinal (right, red border) myocardial sections, highlighting differences in nuclear morphology depending on orientation. Scale bars: 50 μm. (C) Eccentricity calculation: defined as the ratio between the minor and major axes of an ellipse fitted to the nucleus, measuring deviation from circularity. (D) Solidity calculation: defined as the ratio of nucleus area to its convex hull area, indicating the degree of nuclear irregularity. (E and F) Nucleus characteristics grouped by outcome group (congestive heart failure (HF), HF with arterial thromboembolism (ATE), and no reported cardiac cause of death), showing area (μm2), solidity, and eccentricity in longitudinally cut (E) and transversally cut (F) nuclei. Each dot represents one cat. Pairwise comparisons were performed using estimated marginal means from log-linear mixed models with cat as a random effect, with p values…

• 

  Figure 4. Correlation matrix of phenotype quantifications This heatmap displays pairwise correlation coefficients between histological measurements and individual characteristics from up to 37 cats, where not all measurements were available for all individuals (see Table 1). Pearson or Spearman correlation was used depending on the normality of each variable (see Table S4). Positive correlations are shown in red and negative correlations in blue, with color intensity reflecting correlation strength. Numbers indicate correlation coefficients. Correlations with p < 0.05 are indicated in bold.

• 

  Figure 5. Transcriptome-wide enrichment of biological pathways associated with histological and nuclear phenotype parameters Dot plot showing the top three significantly enriched Gene Ontology Biological Process terms per phenotype parameter (adjusted p < 0.05, Benjamini-Hochberg correction) across 12 cats with available transcriptomic data. Dot color represents the normalized enrichment score (NES), indicating the direction and strength of the association; dot size reflects -log10 of the adjusted p value. Parameters include age, nuclear morphology (area, eccentricity, and solidity) measured in longitudinal and transversal myocardial sections, myocardial fibrosis percentage, vascular wall-to-lumen ratio, and number of small, hematoxylin-dense nuclei. See also Table S5.