Abstract
It has been suggested that local elastic moduli of soft biological tissues relate closely with their histology. An image analysis method is proposed in which structural components in histological images are classified based on their color tones to obtain area fraction values and the position of the center of gravity of each component. similarity between colors was determined with Mahalanobis'generalized distance in (R, G, B)color space from predetermined color tones. In the present analysis, several characteristic color tones representing various colors in the input histological image(24-bit color)were specified first by an operator. Then, in the region of interest the pixel which represented the smallest distance from a color tone was classified into this color tone. The difference between areas, extracted manually and automatically was less than 6.6%and on comparison these regions were found to be practically equal. Thus the accuracy of this method was deemed sufficient for further analysis.
