Automatic Detection of Leukemia through Convolutional Neural Network

Abstract

Leukemia is a fatal cancer disease that develops in blood-forming tissue by the excessive development of white blood cells (WBCs) in the human body. However, a bone marrow test is recommended by the pathologist to diagnose leukemia and further types of leukemia. Leukemia has two classes i.e., acute and chronic leukemia. Therefore, early leukemias detection enables preventative actions to be taken to avoid any harm to human life. In addition, several manual and automatic methods have been proposed, however, they possess some drawbacks and are inefficient for the precise detection of leukemia. This research proposes a deep learning-based framework for precise and automatic leukemia identification using microscopic images. The proposed framework comprises four stages which are pre-processing, data augmentation, segmentation, and the classification of leukemia. Moreover, pre-processing is utilized to clean the dataset images and eliminate the noise. Following that, data augmentation approaches have been employed to increase the number of images, and remove the class imbalance, and overfitting problems. The modified Convolutional Neural Network (CNN) based model is employed to segment the leukemia images. A well-known pre-trained AlexNet architecture has been used for classification. Besides that, a publicly available dataset Acute Lymphoblastic Leukemia Image DataBase (ALL-IDB) has been utilized to train and test the proposed model. The proposed model yielded 98.05% accuracy, a specificity of 97.59%, 100% of recall, and a 99.06% of F1-score. The experimentation results demonstrate that this model is effective and reliable for leukemia identification using the ALL-IDB dataset and suitable for deployment in clinical applications.