Advanced Hybrid Deep Learning Model for Precise Multiclass Classification of Bone Marrow Cancer Cells

Bone marrow cancer is when rogue blood cells overgrow in the bone marrow by disrupting regular blood cell production. Accurate classification of these cancers is crucial for effective treatment planning and patient management. Leukemia and myeloma (plasma cell cancer), one types of malignancy that can damage the white blood cells (WBC) within the bone marrow. White blood cell identification, counting, and segmentation are crucial steps in effectively studying a few malignant tumors. In this study, an automated classification method has been proposed for plasma cell cancer which are Multiple Myeloma (MM), Acute Lymphocytic Leukemia (ALL), and Acute Myeloid Leukemia (AML). This bone marrow model image is pre-processed and trained with the parameterized hybrid convolutional neural network and also compared with the CNN framework (InceptionV3, ResNet50, and Vgg16) to achieve accurate classification results. The optimal model was selected by identifying the one with the lowest loss for the validation data. Achieving a high accuracy rate of 99.58% was made possible through the development of hybrid model algorithms, which were carefully crafted by monitoring training loss and validation loss to identify the optimal value. This process of monitoring the training and validation of a deep learning model can help identify the optimal accuracy and loss values. This proposed model can reduce classification time, condense image information, and speed up processing times with more precise weight limits.