Application Of Hough Transform In Segmentation Of Microscopic Cell Image

Cytologic diagnosis is a basic diagnosis method of clinical medicine. It is significant for the early stage diagnosis and qualitative diagnosis of cancers. Pleural effusions is a common clinical symptom of cancers and some other diseases, so the cytologic diagnosis of pleural effusions is a main means of disease diagnosis. At the present time, clinical cytologic diagnosis of pleural effusions lies on the pathologists. With the application of digital image processing technology and pattern recognition technology in the domain of medicine, it is possible to found a pleural effusions computer-aided diagnosis system. Thus the manual experiential qualitative diagnosis is going to be superseded by the more scientific computer-aided quantitative diagnosis which can reduce the pathologists'work intensity, break the subjective experience restriction and increase the diagnosis speed and accuracy. Automatic segmentation of the pleural effusions microscopic cell image, as a key step of the pleural effusions computer-aided diagnosis system, directly affects the analysis and diagnosis in the next approach. Although pleural effusions image contains many kinds of cells, generally, it contains plenty of normal lymphocytes and mesothelial cells. So this paper aims at segmenting these normal lymphocytes and mesothelial cells from the pleural effusions image to simplify the image background and decrease the invalid processing and analysis, thus the subsequent work will focus on the segmentation, analysis and diagnosis of abnormal cells. Under the pleural effusions image's complex background, the normal lymphocytes and mesothelial cells have regular shape, which could be roughly described as circular or elliptic, at the same time these cells'size have a certain physiological range. This paper makes a simple review and survey of the main methods of microscopic cell image segmentation. According to the features of pleural effusions image, the paper adopt a robust Hough transform to segment the normal circular and elliptic cells based on the detection of circles and ellipses. The study of Hough transform theory shows that all kinds of modified Hough transform for circles and ellipses detection are not fit for the pleural effusions image's segmentation very well. Especially these algorithms can't make full use of the features of the image. So according to these features of pleural effusions microscopic cell image, this paper presents an algorithm for the detection of circles and ellipses with modified randomized Hough transform. The algorithm consists of two steps: first, locate the centers of ellipses(including circle, as the special case of ellipse) using the ellipse's geometrical properties and apriori knowledges; second, calculate ellipse's three parameters besides the center coordinates using multiple three-points random samples in restricted area. Experimental results show that this algorithm is capable of detecting both circles and ellipses with excellent efficiency and accuracy. It also performances well when it was applied in the segmentation of pleural effusions microscopic cell image. It is capable of segmenting the normal lymphocytes and mesothelial cells from pleural effusions image with complicated background rapidly and correctly.