A Novel Approach for Leukemia Classification using Multi-Neural Networks

Leukemia is a type of cancer that originates in the bone marrow and affects the blood-forming cells. These abnormal cells, typically white blood cells, multiply uncontrollably, hindering the production of normal blood cells. Because of its various genetic and molecular properties, leukemia, a complex and heterogeneous group of blood malignancies, provides major hurdles in correct subtype categorization. Traditional categorization approaches often fail to reflect the complexities of leukemia subgroups. In this paper, this research offer the Multi-Neural Network (MNN), a ground-breaking strategy for addressing these difficulties by exploiting hierarchical information and merging specialized neural networks. The dataset was collected from Kaggle repository. After collecting dataset this research use Non adaptive threshold for Image denoising. After denoising, this research use Adam optimization algorithm for optimization process. After optimization this research use HOG algorithm for feature selection. The proposed MNN architecture is made up of a unique collection of neural networks, each adapted to a certain hierarchical level of leukemia subtypes. An Improved Convolutional Neural Network (CNN), a DenseNet, and an improved VGG19 are among the specialized networks. These networks are painstakingly developed to extract distinguishing information from leukemia cell pictures, enhancing classification accuracy. This research uses the cross-entropy loss function in combination with the Adam optimization approach to improve the performance of our MNN even more. This combination improves the training process, allowing our MNN to discover complex patterns and correlations from the leukemia dataset.