Elephant herd optimization with neural networks: A case study on acute Lymphoblastic Leukemia diagnosis

Abstract

There are several types of cancer; each is classified by the type of cells that are affected. Leukemia is a kind of cancer that caused by excessive production of leukocytes that replaces normal blood cells. According to the growth speed overproduction of leukemic cells, they can be classified into four major types. This work focuses only on Acute Lymphoblastic Leukemia (ALL), which is also called childhood leukemia. The main goal of this work is to classify the Acute lymphoblastic leukemia cells normal or affected. The proposed approach starts by identifying and segmenting each blood cell then extracting features and finally, classifying them by a hybrid neural network. In this paper, the feed-forward neural network is trained by the Elephant Herd Optimization (EHO) algorithm which updates the weights and the biases of the network. The objective function is the reduction of the misclassification rate. ALL-IDB2 dataset is used in this work. It contains 260 microscopic images. EHO achieves acceptable results as it outperforms other classification methods as well as it overcomes neural networks that are optimized by the other optimization algorithms regarding diagnosing ALL.