Rough set theory with Jaya optimization for acute lymphoblastic leukemia classification

Early diagnosis of malignant leukemia can enormously help the physicians in choosing the right treatment for the patient. A lot of diagnostic techniques are available to identify leukemia disease, but these techniques are costly. Hence, there is a need for a less time-consuming and cost-effective method for the classification of leukemia blood cells. In this paper, application of graphical user interface technique for the differentiation of acute lymphoblastic leukemia nucleus from healthy lymphocytes in a medical image is described. This method employs backtrack search optimization algorithm for clustering. Five different categories of features are extracted from the segmented nucleus images, i.e., morphological, wavelet, color, texture and statistical features. Feature selection plays a very important role in medical image processing. It reduces the computational time and memory space. The hybrid intelligent framework includes the benefits of the basic models; and in the meantime, it overcomes their limitations. Three different kinds of hybrid supervised feature selection algorithms such as tolerance rough set particle swarm optimization-based quick reduct, tolerance rough set particle swarm optimization-based relative reduct and tolerance rough set firefly-based quick reduct are applied for selecting prominent features. These algorithms incorporate the strengths of evolutionary algorithms. The redundant features are eliminated to generate the reduced set which gives predictive capability equal to that of the original set of features. Jaya algorithm is applied for optimizing the rules generated from classification algorithms. Classification algorithms such as Naive Bayes, linear discriminant analysis, K-nearest neighbor, support vector machine, decision tree and ensemble random undersampling boost are applied on leukemia dataset. Experimental results depict that the above classification algorithms after optimizing with Jaya algorithm improve classification accuracy compared to the results obtained before optimizing with Jaya algorithm.