Acute Lymphoblastic Leukemia Detection Using Deep Learning

Acute Lymphoblastic Leukemia (ALL) is a rapidly growing leukemia of the blood and bone marrow, especially in the pediatric and young adult population. Early and precise diagnosis is essential for the increase in survival rates, however, the traditional approach is highly manpower-intensive and demands domain level expertise. This study leverages deep learning to develop an automated model for detecting ALL from blood smear and bone marrow images, aiming to streamline diagnosis and improve accessibility to advanced diagnostic tools. We developed a convolutional neural network (CNN)-based model for distinguishing between normal cells and leukemic cells with excellent accuracy, sensitivity, and specificity. This model encompasses improvement of diagnostic sensitivity and specificity, efficiency of processing, and aiding of clinicians for loading and timely decisions. Through training on a heterogeneous cell image set, the model can learn and identify discriminative morphological features specific to ALL to enable fast and accurate detection. Our findings illustrate that the model identifies whether the cell is an ALL or a normal cell.