DeepLeukNet—A CNN based microscopy adaptation model for acute lymphoblastic leukemia classification

Acute Lymphoblastic Leukemia is one of the fatal types of disease which causes a high mortality rate among children and adults. Traditional diagnosing of this disease is achieved through analyzing the microscopic images of white blood cells by a clinical pathologist. However, this procedure relies on manual observation and often provides inaccurate results. This research proposes an automated system for diagnosing Acute Lymphoblastic Leukemia disease using a convolutional neural network technique. For this purpose, simulation work has been performed over the Acute Lymphoblastic Leukemia-IDB 1 and Leukemia-lDB 2 datasets. However, data augmentation techniques have been employed to generate images to handle the overfitting problem in the model. Qualitative analysis has been performed by visualizing the intermediate layer activation, ConvNet filters and heatmap layers, and a comparative study has been performed with existing methods to validate the efficiency of our proposed model. However, the results showed that our proposed model attained 99.61% accuracy in Acute Lymphoblastic Leukemia diagnosis. The high accuracy reveals that it provides a more effective way to detect Acute Lymphoblastic Leukemia disease than existing works reported in the same area.