Automatic cell classification in human's peripheral blood images based on morphological image processing

Abstract

A new scheme for automatic analysis and classification of cells in peripheral blood images is presented in this paper. The proposed method can analyze and classify mature red-blood and white-blood cells efficiently. After we identify red-blood and white-blood cells in a blood image captured by a CCD camera attached to a microscope, we extract their features and classify them by a neural network model based on back- propagation learning. While we have fifteen different clusters including the normal one for red-blood cells, there are five different categories for white-blood cells. We also propose a new segmentation algorithm to ex- tract the nucleus and cytoplasm for white-blood cell classification. In addition, we apply the principal component analysis to reduce the dimension of feature vectors efficiently without affecting classification performance. Experimental results demonstrate that the proposed method outperforms the learning vector quantization-3 and the k-nearest neighbor algorithms for blood cell classification.