Automated cell counting

Abstract

Cell counting is a common procedure which uses in many biological and medical experiments. It is found that manual cell counting is time consuming and easily causes error due to cell concentration or cell visibility. Therefore, automated cell counting is needed to be developed for increasing accuracy and reducing time use while performing cell counting procedure. In this project, it is mainly divided into two parts, image processing and cell classification. Illumination and a MATLAB toolbox, CellSegm is used for performing cell segmentation of the surface-stained cells. In the toolbox CellSegm, smoothing, ridge enhancement, watershed segmentation and classification are performed by tuning the parameters in the algorithm to delineate cells. However, the classification in the software is not ideally performed. SVM is used which can help to improve the classification results. By extracting the cell features from the extracted cells and put them in the SVM model, thus perform a better classification of cells. Calculation of the sensitivity and the positive predictive value are used throughout the algorithms in the project. In the result, a relatively high value of sensitivity and a relative high positive predictive value are hoped to be attained to indicate the performance of the automated cell counting algorithm.