2.2 Acute leukemia classification
This subtopic presents. Many studies have been devoted to the development of accurate methods for automatically detecting different types of leukemia. Huang et al. reported several methods for the recognition and classification of leukemia [3]. They focused on the seminal work of Golub et al. who presented the first microarray-based and bioinformatics-oriented approaches for identifying and classifying tumor types [4,5]. In that study they used a signal-to-noise statistic to select a small set of genes, before developed aschemebasedon microarray gene expression analysis to distinguish ALL from AML, and they reported recognition rates of 94.1%. Other research studies were inspired by Golub et al. and they used the same ALL/AML data sets presented in [3]. These studies applied models, such as multilayer perceptron networks, support vector machines, and the k-nearest neighbor method, where the accuracy ranged from 58% to 97%.