Addressing Bone Marrow Classification Hurdles Using EfficientNetB5 Technique

Multi-class classification of bone marrow images faces challenges such as class imbalance, significant intra-class variability, and the need for precise and interpretable diagnostic models. These issues complicate the development of reliable automated clinical diagnostic systems. Accurate classification of bone marrow images is crucial for diagnosing a range of hematological disorders. However, difficulties arise in achieving precise and comprehensible results due to morphological variability and class imbalance. This study examines the application of EfficientNetB5, an advanced convolutional neural network, to address these issues in bone marrow image classification. By implementing customized loss functions, advanced data augmentation, and interpretability methods, our approach achieves an impressive 96% accuracy. These results highlight how EfficientNetB5 can enhance diagnostic reliability and precision in clinical settings. Our approach overcomes challenges by addressing class imbalance with tailored loss functions, increasing dataset diversity with innovative augmentation techniques, and ensuring model transparency with interpretability strategies. Leveraging EfficientNetB5's advanced architecture, we achieved a significant improvement with a 96% accuracy rate in classification.