Nuclear Boundary and Pore Detection in SEM Images

Abstract

Scanning Electron Microscope (SEM) images are commonly used for cancer research in order to investigate and simulate mechanisms in specific target cells. High throughput methods generate a number of images that makes automatic computer analysis inevitable. The intracellular space on such images however, might be highly complex and hence an accurate detection of the nucleus boundary is challenging. A novel combination of known algorithms can address this problem. The circular Hough Transform will be used to select candidate objects of cell nuclei at first. The accurate detection of boundary structures with pores is furthermore conducted by edge detection and morphological filtering. This paper also presents a short analysis of the impact of the use of different edge detectors on the accuracy of resulting detection. Nuclei membrane with clearly visible pores can be detected with high accuracy in complex SEM images by the presented method.