FPGA BASED BRAIN TUMOR CLASSIFICATION WITH CLIENT-SERVER ARCHITECTURE

Abstract

The accurate and timely classification of brain tumors is essential for effective treatment planning and patient care. This project explores the use of FPGA (Field-Programmable Gate Array) technology to develop a high-performance brain tumor classification system with a client-server architecture for efficient processing. Various deep learning models, including different Convolutional Neural Network (CNN) architectures, were trained on a dataset of brain tumor images to identify the most effective approach. The ResNet architecture, selected for its performance, was then optimized and deployed on an FPGA using its Deep Learning Processing Unit (DPU) for real-time inference. The system's clientserver framework allows MRI scans to be sent from the client to the server, where the FPGA processes and returns the classification results. This setup achieves faster inference times compared to traditional CPU and GPU-based systems while maintaining high accuracy. The project demonstrates that the FPGA-based approach significantly reduces processing time without compromising accuracy, highlighting the potential of FPGA technology in enhancing the speed and efficiency of medical image analysis, and offering a practical solution for real-time diagnostic applications.