Study of Machine Learning Algorithms on Early Detection of Leukemia

Abstract

Leukemia is a type of cancer that impacts the blood cells and the bone marrow it can be classified into acute and chronic. Early detection is crucial for reducing mortality rates. Acute leukemia progresses rapidly, causing bleeding issues, infections, and anemia due to premature of white blood cells accumulating blood vessels in bone marrow. Chronic leukemia advances slower, leading to an aggregation of abnormal cells. Detecting leukemia more effectively and accurately, automated and machine learning algorithms are being developed. Training algorithms on extensive blood smear images datasets allow these methods to differentiate normal cells from abnormal ones. With faster and more standardized detection of leukemia, medical pathologist can make more informed decisions. While machine learning algorithms may improve detection, skilled healthcare professionals remain essential for interpreting results and providing optimal patient care. The proposed model give a demonstration of a convolutional neural network (CNN) and TensorFlow framework, this method predicts leukemia cells from healthy blood samples this technique has gained popularity as a valuable tool for diagnosing leukemia as well as treating its accuracy of 92.62%. The second approach is that to classify large dataset images of the malignant cells from the normal cells, we employ a VGG19, a ResNet50, and a ResNet101 neural network, as well as batch normalization of the images achieved better accuracy and F1-score.