A Pioneering and Streamlined Pixel Based Approach of Hyperspectral Image Classification for White Blood Cell Using Convolution Neural Network

Abstract

In medical field its necessary to do cell classification as its purpose helps in diagnose,treatment and control of major diseases especially the white blood cells classification, it helps in serving the purpose.Rather than visually inspecting microscopic images more accurate and fast diagnosing methods are available  which includes automation segmenting for classifying microscopic white blood cells samples, in this paper segmentation is done using proposed methods which provide accurate result by integrating some digital image processing algorithms.In the process it involves to pre-process the data which consist of microscopic blood smear sample and then the pre-processed image is classified using the algorithm.In the proposed model it is better than traditional approach because it includes the information like spatial and spectral data which is combined with hyper-spectral imagery and microscopic imaging.This paper is proposed for blood-cell classification framework which is coalition of modulated Gabor operator and Convolution nueral network kernels is also called as MGCNN.In this method for each convolution layer dot product of gabor operator and initial CNN kernel is taken which results in transforming the CNN kernels into recurrence area to cause it to learn features,the highlights learned by tweaked gabor as for various frequencies and directions empowers it to be more point by point and more classifiable. The proposed method performance turn out to be better than the traditional convolution neural networks and SVM approaches especially in cases of small training data-set the proposed methods is still more efficient.