Fig 1.
Sample images of cardiac tissue.
Heart tissue slices were stained using the Trichrome (Gieson) protocol and scanned with a high resolution scanner. Sample images are cut-outs with a size of 1024 x 768 pixels. (A) Micrograph image section with fibres predominately oriented parallel to the image plane. Details of the green region of interest (ROI) are depicted in Fig 2. (B) Tissue with myocytes oriented rather oblique and perpendicular to the image plane. (C) Region with cells mixed in orientation from parallel to perpendicular to the image plane.
Fig 2.
Zoomed tissue sample of image in Fig 1A (green ROI).
Two nuclei of cardiomyocytes (red circles) and one nucleus of a fibrocyte on the left (blue circle) are shown. The hues of both nuclei types are very similar, whereas the chromatin arrangement and the colour saturation differ markedly.
Fig 3.
Block diagram of the image segmentation steps.
In this figure, the sequence of image processing is visualized. It shows all steps that lead to an image where only segmented nuclei remain.
Fig 4.
Fibre orientation based on the shape of a nucleus.
The left image shows a small detail of a sample image visualizing a nucleus, the image in the middle represents the segmented nucleus as a binary image. The graph on the right depicts the scheme of determination of the angle of orientation. e* denotes the longitudinal axis of the nucleus cutting contour in the xy- image plane. The angle φ represents the in-plane angle of the nucleus with respect to the x-axis. Image size = 20 μm x 20 μm.
Fig 5.
Schematic geometry of nuclei inclinations and intersections.
The blue cylinders represent the lateral view of nuclei of cardiomyocytes with distinct inclination angles. The in-plane intersections of the nuclei (top view) are depicted in red. The long axis of a nucleus is denoted by e, the short axis is denoted by d. The in-plane intersection leads to a long axis denoted by e* and an unchanged short axis d. (A) Parallel (in-plane) nucleus α = 0°. (B) Inclined nucleus. The inclination angle α can be calculated by using a triangle. If α is so small that the cutting plane reaches the end faces of the cylinder, the analytical relation of α, e, e* and d (B) is no longer valid. If a ratio of 4:1 is assumed for the length e and width d for a nucleus, the limit for an accurate calculation of α is about 14°. Calculated angle values below that limit would not be valid. (C) Perpendicular nucleus α = 90°.
Table 1.
Distinct cell components and their ability to give an indication about the in-plane fibre orientation.
Fig 6.
Automated image processing of sample image in Fig 1A.
(A) shows the unprocessed image. (B) Generation of an unsharp mask, multiplied with (A). (C) Noise removal. (D) Colour deconvolution to extract the cyan coloured parts of the image. (E) Application of a Gaussian blur. (F) Adding (E) to (A) generates a mask around the nuclei. (G) Thresholding and removing noise and (H) Gaussian blur extracts the shape of the intersected nuclei from the generated mask.
Fig 7.
Fibre orientations depicted as dashes in the sample images corresponding to Fig 1.
The nuclei were segmented and identified according to the nucleus based image processing procedure NBO (sections 2.4 and 2.5). A macro within Fiji was created which added the orientation dashes automatically to the corresponding nucleus. The orientation of a dash indicates the fibre orientation φ in the image plane and the colour indicates classes of inclination angles α of the fibres. Green represents a flat (0°-30°), yellow a moderate (30°-60°) and red a steep orientation (60°-90°). Three outliers were manually highlighted with a blue circle. (A) One nucleus of a fibrocyte in the tissue was segmented by mistake. (C) A nucleus angle was wrongly determined (yellow dash) and two nuclei were grouped together because they were located directly next to each other (red dash). The macros containing the code of the image segmentation and visualization as well as the input images and the result files were combined in ZIP files for each sample image (S1, S2 and S3 Zips).
Table 2.
Sensitivity and Specificity of the image processing method.
Table 3.
Comparison of several image processing methods capable to calculate the orientation of an image.
Table 4.
Overview of the morphometric investigations for sample images in Fig 7.