Using deep DenseNet with cyclical learning rate to classify leukocytes for leukemia identification

Background The examination, counting, and classification of white blood cells (WBCs), also known as leukocytes, are essential processes in the diagnosis many disorders, including leukemia, a kind cancer characterized by uncontrolled proliferation carcinogenic leukocytes marrow bone. Blood smears can be chemically or microscopically studied to better understand hematological diseases disorders. Detecting, identifying, categorizing cell types for disease therapy planning. A theoretical practical issue. However, methods based on deep learning (DL) have greatly helped classification. Materials Methods Images microscopic smear were collected from GitHub, public source that uses MIT license. An end-to-end computer-aided (CAD) system has been created implemented part this study. introduced comprises image preprocessing enhancement, segmentation, feature extraction selection, WBC By combining DenseNet-161 cyclical rate (CLR), we contribute an approach speeds up hyperparameter optimization. We offer one-cycle technique rapidly optimize all hyperparameters DL models boost training performance. Results dataset split into two sets: approximately 80% data (9,966 images) set 20% (2,487 validation set. 623, 620, 624 eosinophil, lymphocyte, monocyte, neutrophil images, whereas 2,497, 2,483, 2,487, 2,499, respectively. suggested method 100% accuracy images 99.8% testing Conclusion Using combination recently developed pretrained convolutional neural network (CNN), DenseNet, one fit cycle policy, study describes WBCs leukemia detection. proposed is more accurate compared state art.