FPGA implementation of a Convolutional Neural Network for Alzheimer's disease classification

In the field of medicine, the analysis and processing of images, particularly in the context of Alzheimer's disease (currently incurable), plays a critical role. Positron emission tomography (PET), essential in neuroscience research, reveals brain activity through analysing scans. Precise classifications of brain PET data are crucial for accelerating the diagnosis of Alzheimer's disease, allowing for earlier intervention and potentially better prognoses. To unravel the complexities of PET data in Alzheimer's diagnosis, we will deploy advanced CNN-based classification techniques. However, it is important to be aware of the computational and memory requirements of these techniques. The unique combination of high performance and low power consumption in SoC-FPGAs makes them perfect for resource-constrained environments. This opens doors for deploying advanced techniques like CNNs for medical image analysis. This paper proposes an Alzheimer's disease classification system using convolutional neural networks (CNNs) implemented on a Pynq_Z2 board. By evaluating this technique in terms of accuracy and cost through simulations, we were able to demonstrate its efficiency and validate the feasibility of the proposed design approach