Classification of Leukemia using Fine Tuned VGG16

Abstract

Leukemia is a hematological disorder which affects the ability of the body to resist against diseases and infection. Early detection of the disease can play a vital role in the treatment of a patient. Computer aided detection system based on machine learning and deep learning algorithms can reduce the burden of doctors and the mortality rate due to leukemia. Transfer learning technique is frequently used in biomedical field due to unavailability of huge and well annotated dataset. The proposed work applies transfer learning to classify leukemia using 1358 microscopic images of blood smears. Pre-trained VGG16 is fine tuned on the leukemic dataset to classify an image as acute leukemia instance, chronic leukemia instance or a healthy instance with an accuracy of 93.01%.