Cell segmentation and cell splitting based on gradient flow tracking in microscopic images

Abstract

We introduce a new approach for segmentation and splitting of cells in different types of microscopy images. Our approach is based on gradient flow tracking followed by local adaptive thresholding to extract nuclei and cells from the background. In comparison to previous flow tracking-based approaches, we introduce a new criterion for the detection of sinks, a new scheme for their combination, and filtering steps for more robust and accurate results. Experiments using different types of image data show that the approach yields good results for single and touching cells of different sizes, shapes, and textures. Based on quantitative results we found that that our approach outperforms previous approaches.