Enhancing Blood Cell Classification by Applying Big Transfer and (XAI)

Blood cells are considered one of the most significant elements in the human body. The study demonstrates a precise hybrid deep-learning model combined with EfficientNetB6 and Big Transfer (BiT-M-R50x1) for accurate blood cell classification in medical image analysis which is crucial for health diagnosis. The performance of this model surpasses that of existing models, achieving an exceptional precision, recall, and F1 score of 97%, as well as an overall accuracy of 97.16%, respectively by utilizing a 17,092 high-quality blood cell image dataset. The study integrates explainable artificial intelligence (XAI) methods like GradCAM++ to build heatmaps of key regions concerning the model’s classifications to enhance decision-making transparency. The study discusses future directions, including exploring ensemble models, continuous improvement of XAI techniques, and thorough clinical validation while recognizing its limitations. The BiT-EfficientNet model demonstrates significant potential for medical image analysis, offering machine-learning practitioners and healthcare experts interpretability, accuracy, and real world application. Integrating robust performance and transparency using Explainable Artificial Intelligence (XAI) methodologies renders it a helpful instrument for accurate and reliable diagnosis. The progress in utilizing deep learning in the medical field shows excellent potential as researchers investigate ensemble models with enhanced explainable artificial intelligence (XAI) techniques through clinical testing.