Adapted clustering method for generic analysis of histological fibrosis staining as an open source tool

Figures

• 

  Figure 1. Description of the animal experiments, image analysis and software design.

• 

  Figure 2. Quantification tool with sample selection panel and one sample for each tissue selected.

• 

  Figure 3. Schematic representation of the workflow and analysis process within the quantification tool.

• 

  Figure 4. Workflow showing k-means classification with subclasses. (A) Standard k-means clustering results (a) compared to single step supervised k-means (b) with each of the three classes (1–3). Arrows depict partially falsely classified tissue, in a.1 healthy tissue is shown as background (yellow arrows), where background was falsely classified as healthy tissue in a.2. (B) Supervised single step k-means with subclasses (= multiple samples for each tissue type) showing improved combined results on all tissue types (a) background pixels, (b) healthy tissue, and (c) fibrotic tissue.

• 

  Figure 5. The results obtained by using the FibroSoft software strongly correlate with the results of Western blot analysis of collagen 1 in TAC model of heart failure. (A) Reduced cardiac function following TAC surgery in mice, measured by echocardiography. (B) Statistical quantification of left ventricular mass in mice subjected to TAC. (C, D) Analysis of collagen 1a1 expression in myocardium of mice by Western blot. Tubulin served as a loading control, proving equal loading of protein samples. Values were normalized to sham operated mice. (E) Representative images showing Masson’s Trichrome staining of cardiac paraffin sections in TAC and sham operated mice. Blue areas represent extracellular matrix deposition, while red stained tissue accounts for heathy heart muscle. (F) Statistical quantification of percentage fibrosis area in Masson’s Trichrome staining images, quantified by FibroSoft. (G) Scatter plot showing in each sample relative protein level against the surface percentage (H) Illustrative images showing Sirius red staining of cardiac sections in sham and TAC mice. (I) Quantification of percentage fibrosis area in histological sections subjected to Sirius red staining. (J)…

• 

  Figure 6. Reference chart and results of accuracy analysis. (A) Defined reference chart which combines randomly selected samples per tissue type and then arranged for accuracy evaluation: (a) healthy tissue, (b) fibrotic tissue samples, (c) background samples. (B) Selected sample color points for reference tissue associated subclasses for our k-means algorithm. (C) Resulting image masks of FibroSoft cluster analysis using supervised k-means with subclasses obtained from a run on the defined reference image; (a) identified as healthy tissue, (b) classified fibrotic tissue, and (c) background.