White Blood Cell Segmentation Based on Visual Attention Mechanism and Model Fitting

Abstract

White blood cell segmentation is a crucial step in developing a computer-aided automatic cell analysis system. To improve the precision of the leukocyte segmentation, this paper presents a white blood cell segmentation algorithm based on visual attention mechanism and model-fitting. The proposed method first employs a color space volume based on visual attention mechanism and an adaptive threshold method to segment the nucleus. Then, the edge region of the image is removed and the initial white blood cell region at the center is obtained. After that, the edge detection is performed to extract the whole leukocyte. The cytoplasm of the leukocyte is obtained by subtracting the nucleus from the entire leukocyte. Finally, the model-fitting method is used to solve the problem of leukocyte adhesion. Experimental results on an image dataset containing 300 leukocyte images show that the proposed method performs well over the state-of-the-art methods.