Lymphoma sub-type classification using DenseNet 169-based Transfer Learning Architecture

Lymphoma is a serious disease with significant potential complications, including heart disease, lung disease, and increased susceptibility to infections due to immune system suppression. This study aims to classify lymphoma cells using a digital pathology dataset comprising three types: Chronic Lymphocytic Leukemia (CLL), Follicular Lymphoma (FL), and Mantle Cell Lymphoma (MCL), with 113, 139, and 122 images, respectively. Image augmentation techniques were applied to address the limited dataset, resizing the images to 40% from the original 1388 x 1040 pixels. The classification was performed using a deep learning approach with transfer learning, specifically employing the DenseNet169 model from TensorFlow. The DenseNet169 model, initialized with ImageNet weights and optimized with specific parameters, achieved an accuracy of 99.1%, surpassing previous research by 0.4%. The study demonstrates that image augmentation, combined with transfer learning, effectively enhances classification accuracy in the context of limited datasets.