Image recognition of juvenile colonies of pathogenic microorganisms in the culture based microbiological method implemented in bioMEMS device for express species identification

Abstract

In this paper an image recognition method for microbiological analysis of species is described. This method is developed for use in bioMEMS hybrid laboratory-on-a-chip for total microbiological analysis and is aimed at recognition of colonies of microorganisms for express species identification and sorting. The described approach is based on the Histograms of Oriented Gradients (HOG) method, developed earlier. A principle of this technique is the computation of intensity gradients directions in local areas of image and determination of objects affinity using special classifier. The method is widely used for people identification, where it proved to be very efficient. In microbiology it appeared to be not as successful, because images of single microorganisms and their colonies are too similar. Here an advanced version of HOG is presented, which enabled a number of microbiological species to be identified. The method comprises the following stages: 1) Initial image acquisition; 2) Search and retrieval of the colony; 3) Simplification of the image, transition to local gradients of the grey; 4) Processing with threshold noise filter; 5) Fourier-transform of local gradients of the grey; 6) Formation of classifier by teaching the expert system by comparison with reference images. The described method is applied for recognizing colonies at earlier stages of growth (juvenile colonies) incorporating about 1000 cells. A sample of investigated microorganisms comprised 18 species from 8 different genera.