A Deep Learning Approach Bone Marrow Cancer Cell Multiclass Classification using Microscopic Images

Abstract

White blood cells play a key role in identifying a person’s illness. One type of cancer that can affect the white blood cells (WBC) in the bone marrow is leukemia and myeloma (cancer of plasma cells). White blood cell identification, counting, and segmentation are crucial steps in the effective study of a few malignant tumors. The goal of this study is to create an automatic classification method for plasma cell cancer and the two main types of leukemia cancer (ALL, AML). The model on the bone marrow images is prepared using the Multiple Myeloma (MM) with the parameterized convolutional neural network and also contrasts with CNN framework (InceptionV3, ResNet50, and Vgg16) to achieve accurate classification results. The best model was chosen based on the smallest loss for the validation data. The deep learning models were developed by monitoring the training and validation loss every epoch. Utilizing this reduces classification time, condenses the information image element, and speeds up meeting times with more precise weight limits. While manual microscopy is time-consuming, this research based on hyper-tuned convolutional neural networks will be promising to attain higher accuracy more quickly.