Nuclei Extraction Based on Multi-channel Information

To analyze the nuclei distributed in the glomeruli of the renal biopsy images automatically, the problem of missing or wrong extraction of them caused by the complex characteristics of the renal biopsy images must be solved. This paper introduces a dynamic thresholding method using eigenvalue feedback strategy based on multi-channel information. Firstly, the nonlinear thresholding surface adjusted by area eigenvalue feedback strategy is used to R-channel images of RGB color space and C-channel images of CMYK color space respectively to obtain the binary images. Through fusing the two images, the undetermined regions, which consist of the genuine and spurious nucleus, can be marked. And the reality degree is calculated to distinguish which one is genuine. Secondly, an area proportionality eigenvalue is considered in feedback strategy to divide the nuclei clusters into single ones. Then the nuclei extraction can be realized. The experimental results demonstrate the effectiveness of this method.