A 5G network based conceptual framework for real-time malaria parasite detection from thick and thin blood smear slides using modified YOLOv5 model.

Objective: This paper aims to address the need for real-time malaria disease detection that integrates a faster prediction model with a robust underlying network. The study first proposes a 5G network-based healthcare system and then develops an automated malaria detection model capable of providing an accurate diagnosis, particularly in areas with limited diagnostic resources. Methods: The proposed system leverages a deep learning-based YOLOv5x algorithm to detect malaria parasites in thick and thin blood smear samples. The YOLOv5x network architecture was modified by introducing two squeeze-and-excitation network (SENet) layers just before the Upsample layers. The system is designed to operate over 5G networks efficiently, enabling remote and smart healthcare solutions. Results: The modified YOLOv5x model demonstrated improved accuracy and precision in detecting malaria parasites on microscopic slides. The inclusion of SENet layers optimized the network's performance, making it suitable for real-time disease detection over a 5G network. Conclusion: Our model exemplifies how a generic one-stage object detection algorithm, such as YOLOv5x, can be repurposed to detect objects as small as malaria parasites from microscopic visuals in a cost-effective manner over the 5G network. By integrating the computational efficiency of deep learning with the connectivity of 5G networks, this system can significantly enhance remote diagnostic capabilities and contribute to smart healthcare solutions.