A Novel Deep-Learning Based Method for Cell Phenotype Classification

Cell phenotype classification represents a crucial task in various medical and biological applications. With fluorescent microscopy emerging as a leading method for biological cell characteristics analysis, cell phenotype classification is often being performed by analysing fluorescence images. This paper proposes a novel approach for performing single-cell classification in a sequence of several preprocessing sub-steps, relevant feature extraction, and a deep learning model for performing the final classification. Due to the lack of relevant publicly available datasets, the algorithm was tested on artificially generated images, obtained by the authors’ own previously developed artificial data generation system. Very promising performance metrics were obtained, testifying to the utility of the proposed method.