A Novel Automated Blood Cell Counting Method Based on Deconvolution and Convolution and Its Application to Neural Networks

Abstract

Blood cell counting and disease detection are very important in healthcare, biomedical research and pathology. Conventional cell counting techniques have problems of their own due to cost, complexity, skill requirement, and time consumption. Alternative image processing techniques are also challenging for huge computational load. Moreover, the modern image processing is quickly moving toward the neural network and machine learning-based smart systems rather than algorithms. They are, however, heavily resource-hungry and complex,  essentially, games of humongous number crunching and monstrous computing workforce. The amount of data required to train the networks is also very large. Here, three differnt methods have been presented which provide relatively simple alternative to the above challgnes–A) blood cell counting using deconvolution–convolution algorithm,  B) cell counting and disease detection using convolution and finally, C) blood cell counting using neural network aided by deconvolution–convolution method for clustering and classification. They are simpler, robust, faster and less resource-hungry as far as the requirement of the computational power is concerned.