Using the monogenic signal for cell-background classification in bright-field microscope images

Abstract

Some cell detection approaches which deal with bright-field microscope images utilize defocussing to increase the image contrast. The latter is related to the physical light phase through the transport of intensity equation (TIE). Recently, it was shown that it is possible to approximate the solution of the TIE using a modified monogenic signal framework. We show empirically that using the local phase of the previous monogenic signal in place of the defocused image improves the cell-background classification rate. The evaluation was performed on L929 adherent cell line with more than 1000 manually labeled cells. The improvement was 6.8% using a random forest classifier and 10% using a support vector machine classifier with a radial basis function kernel.