Mastering Multi-Class Bone Marrow Classification with EfficientNetB5: Challenges and Innovations

Classifying bone marrow pictures into different categories is challenging due to issues like uneven numbers of examples in each category, differences within the same category, and the need for very accurate and easy-to-understand diagnostic models. These challenges make it hard to create reliable automated systems for medical diagnosis. To find different blood diseases, it’s important to correctly identify images of bone marrow. There are some challenges in multi-class classification, like the need for clear and correct results, differences in shapes, and uneven numbers of classes. This work looks at using a modern type of computer model called EfficientNetB5 to help sort images of bone marrow. Our method reaches a high accuracy of 96% by using special loss functions, new ways to increase data variety, and tools to understand the results. Using the strong design of EfficientNetB5, our method reaches an impressive 96% accuracy in classifying items, which is a big improvement.