Segmentation of overlapping leucocyte images with phase detection and spiral interpolation

Abstract

Leucocyte segmentation is one of the most crucial functionalities for an automatic leucocyte recognition system. In this paper, an algorithm is proposed to segment the leucocytes from the overlapping cell images. It consists of two main steps. The first step involves generation of a combined image based on the saturation and green channels (CIBSGC) by means of the different distribution characteristics of the leucocyte nucleus. A weight coefficient is used to adjust the CIBSGC for extracting the nucleus and estimating the location of the leucocyte. Second, a method of phase detection and spiral interpolation identifies the overlapping regions of cells and determines the leucocyte edge curve. The performance is evaluated by three parameters: sensitivity, positive predictive value and pixel number error. Experimental results validate that the proposed algorithm can successfully segment the overlapping leucocyte with the satisfactory performance for two cell image datasets under different recording conditions.

A color and shape based algorithm for segmentation of white blood cells in peripheral blood and bone marrow images

“…When red blood cells or other noisy fragments are also found in the results, it is possible to refine these results by applying morphological operations and/or evolving active contours on the segmented cell boundaries . After eliminating such fragments, one can also reconstruct the boundaries of white blood cells by spiral interpolation or using color differences between overlapping cells .…”