Classification of Acute Leukemia Based on Multilayer Perceptron

Abstract

In general, various artificial neural network have been applied in many areas such as modelling, pattern recognition, signal processing, diagnostic and prognostic. In this paper, artificial neural network are used to detect and classify the white blood cell (WBC) inside the acute leukemia blood samples. There are 25 features have been extracted from segmented WBC, which consist of shape, color and texture based features. Then, it have been fed up as the neural network inputs for the classification process in order to classify the segmented regions into two classes either B or T. The training algorithm for MLP network is LevenbergMarquardt (LM). The MLP network achieves the highest testing accuracy of 96.99% for 4 hidden nodes at state of 5 by using the overall 25 input features. Thus, MLP network trained by using LM algorithm is suitable for acute leukemia cells detection in blood sample.