Multiclass Classification of Heterogeneous Blood Cells Using Deep Learning and contourlet Transform

Abstract

Abstract The classification of human blood cells is very important in the diagnosis of inflammation, infection and blood disorders such as leukemia. Diagnosis of blood malignancies requires identification and classification of white blood cells in peripheral blood smear. The pathologist spends a lot of time analyzing blood cells to determine the minor differences between blood subsets. Due to the high similarity of blood types, human error is sometimes possible. Manual procedures for diagnosing blood diseases are time-consuming, subjective and prone to human error. Therefore, we need machine analysis of microscopic images. Usually, methods based on image processing contain limitations. On the other hand, with the increase in computational processing power in computer-based clinical diagnosis systems, it has enabled the use of machine learning methods. In this article, we will use the combination of deep learning; Gabor filter and wavelet transform to provide a high accuracy blood cells classification model while extracting features from macroscopic images. The basis of the current research is the classification of blood smear images using the combination of contourlet transform, recurrent neural network and optimization method. Feature extraction is based on the combination of wavelet transform and recurrent neural network and feature selection is based on the African vulture optimization method. Finally, an innovative classifier based on clustering is presented to classify different blood cells. Based on the results obtained on the set of Jiangxi Tecom images, the proposed design has achieved an acceptable accuracy and has been able to increase the precision, recall and F-Measure.