Acute Lymphoblastic Leukemia Detection by Tuned Convolutional Neural Network

Convolutional neural networks (CNNs) represent a relatively new type of neural networks but it has been already widely used for different image classification problems and have shown promising results. Medicine represents a science that has a high demand for image processing and classification methods. In this paper, we used CNN for classifying microscopic white blood cell images. The problem with CNN is that it contains a large number of hyperparameters that need to be tuned and it represents a hard optimization problem. In this paper, we proposed a bare bone fireworks algorithm for tuning a subset of CNN hyperparameters. The proposed method was tested on a standard benchmark dataset for acute lymphoblastic leukemia detection. The proposed method was compared with CNN without hyperparameter tuning and optimized SVM method from literature. The proposed optimized CNN method achieved higher accuracy compared to the mentioned two methods from the literature.