Zhenzhou Wang
Chinese Academy of Sciences Shenyang Institute of Automation, China
Title: A New Approach for automatic and robust segmentation and quantification of cells and nanoparticles
Biography
Biography: Zhenzhou Wang
Abstract
With the rapid development of microscopy and nanoscale imaging technology, the requirement for automatic and robust segmentation and quantification of cells or nanoparticles increases greatly. Because of the vast variety of the cell or nanoparticle images, most existing methods are only capable of segmenting some specific type of cells or nanoparticles. In this paper, we propose a more versatile and generalized method that is capable of segmenting and quantifying a variety of cells and nanoparticles automatically and robustly. It consists of five parts: (1) automatic gradient image formation;(2) automatic threshold selection; (3) manual calibration of the threshold selection method for each the specific type of cell or nanoparticle images; (4) manual determination of the segmentation cases for each specific type of cell or nanoparticle images; (5) automatic quantification by iterative morphological erosion. After the parameter, N is calibrated and the segmentation case is determined manually for each specific type of cell or nanoparticle images with one or several typical images, only parts (1), (2) and (5) are needed for the rest of processing and they are fully automatic. The proposed approach is tested with different types of cell and nanoparticle images. Experimental results verified its effectiveness and generality. The qunatitative results show that the proposed approach also achieves significantly better accuracy compared to state of the art methods.