Analysis of multispectral images of the colon to reveal histological changes characteristic of cancer

Colon cancer alters the macroarchitecture of the colon tissue. Common changes include the formation of new blood vessels (angiogenesis) and the distortion of the tissue collagen matrix. Such changes affect the colon colouration. This paper presents the principles of a novel image analysis method capable of extracting parameters depicting histological quantities of the colon from its multispectral images. The method is based on a computational, physics-based model of light interaction with tissue. The colon structure is represented by three layers: mucosa, submucosa and muscle layer, and parametrised by the concentration of blood haemoglobins; the size and density of collagen fibres; and the thickness of the layers. Using the entire histologically plausible ranges for these parameters, a cross-reference is created computationally between the histological quantities and the spectra which can be obtained at each point of a multispectral image set. In a pilot study the method was applied to two excised colon samples comprising normal tissue and adenocarcinomas. The histological parameters extracted from the cancerous regions showed the changes characteristic of the colon cancers, namely the increase in the blood volume fraction, decrease in the collagen density and increase in the thickness of the mucosal layer.