Automated white blood cell diagnostics using federated learning and distributed deep learning

The rise of big data has transformed the way complex problems in fields such as medicine and biology are solved. In the medical field, analyzing White Blood Cells (WBCs) is crucial for diagnosing diseases and evaluating the immune system. While automated tools like cell counters can quickly generate results, manual blood smear analysis remains critical for accuracy and patient monitoring. Unfortunately, this manual process is slow, labour-intensive, and prone to errors, making it challenging to manage large-scale data efficiently. This study combines the strengths of Federated Learning and Big Data to tackle these problems. The authors propose a new approach for classifying WBCs by leveraging Federated Learning (FL) for privacy-preserving, distributed training on large datasets, while utilizing Apache Spark's tools for big data management and processing. Additionally, advanced deep learning models, such as ResNet50, VGG19, and U-Net, enhance WBC classification accuracy by creating five RDDs and training each of the three models on each of the five RDDs. The ResNet50 model achieved the highest accuracy of 94.06% in RDD2 and RDD5, followed by VGG19 with 94.27% in RDD1, and U- Net with 85.99% in RDD4. RDD has its validation accuracy. This study addresses the dual challenges of scalability and privacy. Additionally, distributed data on five Nodes of Resilient Distributed Datasets (RDD) demonstrated that both VGG19 and ResNet50 achieved higher accuracy compared to U-Net, while training each deep- learning model to enhance diagnostic accuracy by integrating Federated Learning with Big Data frameworks and recent deep-learning techniques. This innovative technique highlights the potential of combining these technologies to advance healthcare and biomedical research.