Analysis of Multispectral Images of Excised Colon Tissue Samples Based on Genetic Algorithms

Abstract

We present in this paper a method to estimate four significant biological parameters of colon tissue. The interaction of light with colon tissue is modeled by two layers parameterized by biological parameters, which describe optical properties of the colon. This model is reversed using an optimization framework based on genetic algorithms. From a multispectral image of colon, we compute biological parameters of the colon, this noninvasive optical biopsy might lead to better diagnosis of cancer. We present in this paper experimental results analyzing multispectral images of excised colon tissue samples. We analyze the following three categories of colonic tissue: healthy tissue, with a polyp and with cancerous cells.