Automatic nuclei and cytoplasm segmentation of leukocytes with color and texture-based image enhancement

Afaf Tareef, Yang Song, Tom Weidong Cai, Yue Wang, David Dagan Feng, Mei Chen

Abstract

Designing a single automatic and accurate segmentation approach for different classes of white blood cells is a challenging task. This paper presents a fully automated segmentation framework to segment both nuclei and cytoplasm of five major classes of white blood cells in the peripheral blood smears based on color and texture enhancement. Particularly, a new gray-scale transform is generated based on three representative color channels to separate the nuclei from the cytoplasm and background by Poisson distribution based minimum error thresholding. For cytoplasm segmentation, discrete wavelet transform (DWT) and morphological filtering based enhancement procedure is utilized to highlight the cytoplasm and eliminate the small details inside the cells. Finally level set-based refinement and false candidates filtering are applied to obtain the accurate cell segmentation. The proposed approach is evaluated on two sets of peripheral blood smears, and it demonstrates improved segmentation performance when compared with existing methods.
BibTeX
@inproceedings{tareef2016automatic,
  author = {Afaf Tareef and Yang Song and Tom Weidong Cai and Yue Wang and David Dagan Feng and Mei Chen},
  title = {{Automatic nuclei and cytoplasm segmentation of leukocytes with color and texture-based image enhancement}},
  booktitle = {ISBI},
  year = {2016},
  doi = {10.1109/isbi.2016.7493418},
}

Loading related content...

Figure