کنفرانس بین المللی فناوری اطلاعات و دانش

صفحه اصلی / چهاردهمین کنفرانس بین المللی فناوری اطلاعات و دانش

Improving Deep Neural Network Accelerator for Malaria Diseased Blood Cells using FPGA

نویسندگان :

Hadi Rezaeikarjani¹ Mojtaba Valinataj²

1- دانشگاه صنعتی نوشیروانی بابل 2- دانشگاه صنعتی نوشیروانی بابل

کلمات کلیدی :

Hardware Accelerators،Malaria Disease،FPGA،Disease Detection with Neural Networks،Neural Networks،Medical Diagnosis

چکیده :

The escalating computational demands of deep neural networks across various applications have driven the adoption of hardware accelerators. These specialized hardware devices are tailor-made for specific computational tasks, offering enhanced efficiency compared to conventional computer systems. In medical diagnosis applications, particularly the detection of malaria-infected blood cells, hardware accelerators play a pivotal role. This paper explores the augmentation and acceleration of malaria-infected blood cell detection by leveraging FPGA-based hardware accelerators with deep neural networks. The significance of this research is twofold. Firstly, rapid and precise processing of medical images is imperative in diagnosing malaria. FPGA-based hardware accelerators excel in parallel processing and high efficiency, significantly expediting disease detection, a crucial advantage during outbreaks. Secondly, the intricate architectures and numerous parameters of deep neural networks demand efficient implementation. Hardware accelerators, notably FPGA-based ones, facilitate precise and efficient model execution, enhancing diagnosis accuracy, a paramount factor in disease detection. The study adopts an artificial neural network with a Multilayer Perceptron (MLP) architecture and implements various hardware units, resulting in substantially faster malaria-infected cell detection. The outcomes demonstrate an impressive accuracy increase from 94.76% to 98.27% and a significant reduction in latency from 5.93 nanoseconds to 0.397 nanoseconds in the hardware implementation. Moreover, the output representation has been improved, transitioning from a matrix display to a visually interpretable format with distinct colors, enabling real-time disease detection.