Colorectal cancer image classification using image pre-processing and multilayer Perceptron

Manual screening of colorectal biopsy tissue under microscope to conform the presence of cancerous cell is difficult and time consuming. The criteria in diagnosing colorectal cancer cell are gland shape and nucleus size. In this paper, we proposed a method of automatic image pre-processing to extract important feature of colorectal tissue images. Images captured under microscope may vary in color brightness due to different staining concentration and the size of biopsy tissue. In this paper we proposed a method using HSV color to remove element outside the area of nucleus. In order to extract the gland shape, we proposed a gland tracking boundary and segmentation. By using the result of gland tracking, nucleus size that forms the glands are measured. Multilayer Perceptron is being used to detect the shape of glands. By combining result of gland shape and nucleus size, we perform the image classification. The result shows that classification achieves 94% accuracy by using the proposed methods.