Multi Class Classification of Acute Leukemia using Transfer Learning

Leukemia is a fatal disease which affects blood cells. Detection of this disease using machine learning and deep learning techniques can reduce the burden of doctors and clinicians. The proposed work performs a multi level classification of acute leukemia involving transfer learning technique. An accuracy of 95% and 96% are achieved for support vector machine and logistic regression, respectively, when features extracted by ResNet50 are used for training. 95% test accuracy is obtained for k-Nearest Neighbor classifier (K=5) when features extracted by MobileNet are used for training.