Acute Lymphoblastic Leukemia Detection Employing Deep Learning and Transfer Learning Techniques

The death due to cancer increases daily due to late or even no diagnosis. It is necessary for early detection and treatment of such fatal cancers and avoiding carcinogenic agents is an important step for the prevention of cancers. The most frequent kind of cancer, with prostate and lung cancer following closely after. Different types of cancers are detected by collecting blood or urine samples. The majority of the cancers are detected by Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and even X- Ray images. There are different techniques used by the doctors like pattern recognition, seven-point detection, Asymmetry, Border, Color, Diameter (ABCD) method. There are lot many methods used by professionals in this field. However, these methods may not be efficient always. Hence currently there is a huge scope for automation in the medical field for better performance and early diagnosis of fatal diseases to prevent further complexities and also for the betterment of mankind.This paper demonstrates the automation of the categorization of Acute Lymphoblastic Leukaemia. In order to do this, many CNN variations with various activation functions are tested; CNN, Xception, and MobileNetV2 yielded the best accuracy of 100%. These automated diagnostic approaches reduce the risk of cancer in humans because they are more accurate than manual diagnosis methods.