engleski

Estimation of bacterial culture growth region using image segmentation in Matlab

In this work an algorithm for identification of homogeneous bacterial colonies’ growth region in agar Petri dishes is presented using image segmentation implemented in Matlab®. The goal is to achieve recognition of regions where homogeneous bacterial colonies were developed as well as an estimation of their radius. Isolation of bacterial cultures’ region from the dish was achieved using two segmentation methods based on histogram: Top-down curve fitting – a novel algorithm based on curve fitting ; Gaussian mixture models with EM – based on probabilistic theory. Homogeneous bacterial colonies were developed from a single type of bacteria which exponentially multiplied forming a colony. The area of growth can be approximated to a circular formation, although the true form can vary. Separation of circular-like shaped colonies from non-circular ones was achieved using a normalized shape descriptor using shape identification techniques. In order to compare different aspects of the segmentation process presented in this work a graphical user interface (GUI) was developed to help visualize the obtained results, as well as observe and compare the effect of different parameters used. At the end results of the proposed segmentation process are shown using images collected from the Department of microbiology at the Institute of Public Health in Zagreb County as well as artificially generated images (created to test different aspects of the algorithm) and images downloaded from the internet.

image segmentation; homogeneous bacterial colonies; Gaussian mixture models with expectation maximization; curve fitting

nije evidentirano