SAM Filter Based Convolution Neural Network Alogrithm for Leukocyte Classification

Abstract

In biomedical field, the analysis of red blood cells (RBC) and white blood cells (WBC) were of vital importance for diagnosing diseases. As for WBC, it can be classified into basophils (B), lymphocytes (L), neutrophils (N), monocytes (M), and eosinophils (E) five components. Based on varieties methods of hyperspectral imaging, a novel white blood cell classification method, which was a new implementation algorithm in the field of medical research, was designed by three main blocks: the realization of spectral angle match algorithm, morphological processing method and basic structure of the convolution neural network system. In the case of basophils, eosinophils, lymphocyte and neutrophils, the classifications accuracies were 95.3%, 93.2%, 90.8%, 92.7% respectively, improved by nearly 10% with respect to the SAM-only cases.