Study on Recognition Method of Label-free Red and White Cell Using Fecal Microscopic Image

In clinical studies, fecal microscopy images are rich in human pathology information, among which the type and the quantity of cells are important clues for disease diagnosis of human digestive systems. In this paper, we study an automatic identification method of red and white cells in fecal microscopy images using a 20x magnification system. Firstly, this paper adopts a strategy based on the combination of logic operations for image segmentation. The edge detection is carried out on grayscale image and B channel image of the original color image separately, and the results are fused by or operation. The morphological processing is applied afterwards to remove the imperfections of segmentation, e.g. cell adhesion. And select relevant features to remove other visible components except red and white cells and impurities to obtain red and white cell segmentation images. Then according to the red and white cells in morphology, Fast Fourier transform (FFT) and Canny image edge detection difference, obtained six related features to form a feature vector. Finally, the supported vector machine (SVM) classifier is trained by these features, and the trained SVM classifier is used to identify. The experimental results indicate that the recognition method proposed can achieve an accuracy about 90%, which satisfies the demand from the practice.