Automated Classification of Blood Cell Cancers Using EfficientNetB3 Model

Blood cell classification plays a crucial role in medical diagnostics, aiding in the detection of various diseases, including leukemia and other hematological disorders. Previous studies have utilized models such as CNN, VGG-16, and ensemble techniques for blood cell classification, achieving varying levels of success. However, these models often struggle to balance accuracy and computational efficiency. This study proposes an advanced deep learning-based approach utilizing the EfficientNet-B3 model, selected for its ability to optimize accuracy with fewer parameters through compound scaling, making it suitable for medical image analysis. Leveraging a dataset containing 12,500 augmented images of blood cells, the model was trained and evaluated to classify six different cell types with high precision. Comparative analysis with recent studies implementing CNN and VGG-16 models demonstrates the superior performance of the proposed model, achieving a remarkable accuracy of 99%, with precision, recall, and F1-score consistently at 0.99 across all classes. The results validate the effectiveness of EfficientNet-B3 against existing methods, highlighting a significant improvement in classification performance. Furthermore, the model's efficiency and reliability suggest its potential for integration into automated diagnostic systems, providing pathologists with an accurate and time-efficient tool for blood cell analysis. This research highlights the impact of deep learning in medical image classification and sets a foundation for further improvements through model optimization and explainability techniques. Future work will explore enhancing model interpretability and integrating real-time diagnostic capabilities for broader clinical applications.