Identification and Classification of Leukemia Disease Using Deep Learning Techniques
Aziz Makandar, Arpana Bhandari · 2024
This paper introduces a novel methodology for the identification and categorization of leukemia using microscopic images, using advanced image processing and Machine Learning (ML) technique. To enhance detection accuracy, this study explores automated and deep learning algorithms trained on extensive blood smear image datasets to differentiate between normal and abnormal cells. The proposed approach involves enhancing blood micrographs, which are then analyzed using an active contour method to isolate white blood cell regions. These regions are subsequently examined by three deep learning models: VGG16, InceptionV3 and ResNet50. Evaluation of the C-NMC 2019 dataset with these models, coupled with an SVM classifier, yielded accuracies of 81 %, 95% and 89%, respectively. The methodology includes data preprocessing, image augmentation and feature extraction, followed by classification with an SVM classifier. The InceptionV3 model achieved the highest performance, with a classification accuracy of 95% and an Fl score of 93 %.