Detection of Leukemia using Machine Learning

Abstract

Leukemia is a type of cancer which damages the blood-forming tissues that produces rbc's, wbc's and blood platelets, which is present inside bone marrow. Leukemia is the cancer that mainly occurs within blood forming tissues in the human body. It may slowly affect all the organs inside the body without any external indication leading to several other disorders. The diagnosis of leukemia in early stages of the disease is difficult as the current technologies are more time cons uming. Studies suggest that India ranks 3rd highest in reported cases of Blood Cancer, after USA and China. If a person is already diagnosed with any one among other types of cancer, then the chances of developing leukemia is nearly 30%. The symptoms of leukemia are tiredness, flu, fatigue and weight loss which makes it very difficult to predict manually by clinical methods. Since it is malignant and fatal it is neces s ary that it mus t be predicted earlier in order to undergo sufficient treatment and thereby saving millions of lives. As compared to all other types cancers, leukemia cases have been increasing dramatically in the entire world. Since the symptoms of leukemia are being very common, people fail to detect it in the early stage. The main objective of our project is to diagnose leukemia in the primary stage by analysing peripheral blood smear images and thus treatment could be given in specific and successful way. Thus our project focusses on using the best machine learning algorithms for detecting leukemia in a very early stage. This res earch study collects peripheral blood smear images (PBS), which are microscopic images of the blood samples. The images collected are preprocessed and then segmentation is performed where the images are segmented based on pixels and enlarged portion of affected area of the blood s ample image is us ed for feature extraction process. Finally classification is done using CNN in order to detect leukemia.

Topics & Concepts