Deep Learning-Based Analysis of Blood Smear Images for Detection of Acute Lymphoblastic Leukemia
Nitla Gokulkrishnan, Tushar Nayak, Niranjana Sampathila · 2023
Leukemia, a type of cancer affecting the blood and bone marrow, involves the abnormal production of leukocytes and can impact the immune system. While more prevalent among children, it can also affect adults. Early detection plays a critical role in effective treatment and patient recovery. In this paper, we have used an open source four-class Acute Lymphoblastic Leukemia (ALL) dataset that has been segmented using color thresholding. Subsequently, these images have then been trained on pre-trained Convolutional Neural Networks (CNNs): ResNet-50 and ResNet-101, with hyperparameter tuning to classify between benign and three stages of malignant ALL lymphoblast cells. The results demonstrate that our proposed method achieved accuracies exceeding 98% in detecting ALL, indicating the potential of deep learning-based classifiers in aiding hematologists accurately detect ALL and improving patient outcomes.