Figure 1.
CSH processing on an infarcted mouse heart.
A) An infarcted mouse heart section cut 8 µm thick. Muscle tissue is scarlet red, while collagen fibers appear blue, and necrotic regions are purple-black. Insets show enlarged areas of muscle, collagen and necrotic region. B) The same mouse heart, post-processing by CSH. The areas that CSH determined as collagen are blue, and the areas that CSH determined as muscle are red. The background is yellow. C) A plot of the pixels from the original heart image mapped to HSV space. The gray arrows indicate the direction from which this 3-D graph will be displayed in the following 2-D images. D) A plot of the pixels from the original image in the Hue-Saturation plane. The borders collagen and the muscle rectangular thresholds are visible at Hue = {200, 300, 385}. E) A plot of the pixels from the original image in the Hue-Value plane. F) A plot of the pixels from the original image in the Value-Saturation plane. This graph most clearly shows the different shapes of the collagen peak (blue) and the muscle peak (red).
Figure 2.
Analytic performance across diverse section thicknesses.
A) A section of infarcted mouse heart cut to 4 µm and stained with Masson's trichrome. Descending from the original image, we see the RGB binary image, the HSV binary image, a density map of the pixels mapped to the RGB color space, and a density map of the pixels mapped to the HSV color space. B) A section of infarcted mouse heart cut to 6 µm and stained with Masson's trichrome. C) A section of infarcted mouse heart cut to 8 µm and stained with Masson's trichrome. D) For each of four experimental hearts and each of the three section thicknesses, the area identified as muscle is plotted next to the area identified as collagen using the RGB method. Because each heart has a different size infarction, these results for each heart are normalized as a percentage of the measured area in the 6 µm sample. As the section thickness increases, RGB analysis decreases the perceived collagen area, despite analyzing adjacent sections of heart. E) For each of four experimental hearts and each of the three section thicknesses, the area identified as muscle is plotted next to the area identified as collagen using the HSV method. Because each heart has a different size infarction, these results for each heart are normalized as a percentage of the measured area in the 6 µm sample. There is no discernible change in perceived muscle or collagen area as the section thickness increases when using the HSV method.
Figure 3.
Additional thresholds are possible with CSH.
An infarcted mouse heart, analyzed by CSH for collagen, muscle, and necrotic tissue. A) A post-processed image of an infarcted mouse heart using the HSV method that includes a binary for necrotic tissue. B) 2-D plots of the original pixels mapped to the RGB color space, but with the same binary applied as in Figure 3A. C) 2-D plots of the original pixels mapped to the HSV color space.
Figure 4.
CSH is consistent between individuals.
Apoe−/− mouse innominate arteries stained with MAC3 antibody for detection of macrophages. RGB1(threshold1) was optimized for Cross Section 1, and overestimates the positive area when applied to Cross Section 2. RGB2(threshold2) was optimized for Cross Section 2, and underestimates the positive area when applied to Cross Section 1. CSH was able to effectively use a single HSV threshold on both cross sections. In the overlays between the HSV and RGB1 and RGB2, yellow area shows where there is agreement between the HSV method and the RGB method. Green area in the overlays may indicate false positive area reported by the RGB method, while red area in the overlays may represent false negative area reported by the RGB method.
Figure 5.
CSH is a powerful tool for a variety of stains.
Common histological stains, displaying the fidelity of CSH. (Top panels): A mouse aorta stained with alkaline phosphatase (ALP, red) for detection of early calcification, with Gill's hematoxylin as counterstaining (purple), which depicts advanced calcification. ALP stain is scarlet red (denoted “A” in the top left panel), while hematoxylin is a shade of purple (denoted “H” in the top left panel). Visually, the hematoxylin interferes with the ALP, making it difficult to see where the ALP stain begins and ends. We analyzed the section for ALP-positive area using both CSH and an RGB-based method. (Bottom panels): A mouse liver stained with picrosirius red staining visualized using polarized light microscopy for detection of fibrosis. We analyzed the section using both CSH and an RGB-based method. The RGB method was unable to register the brightest parts of the stain as positive (gray), and falsely interpreted stain artifacts as positive areas (green in both “Merge” images).