Optimizing Medical Image Processing Performance through Chip Modeling

This paper presents a novel approach to optimizing the performance of medical image processing through SystemC chip modelling. Our unique contribution lies in the integration of hardware/software co-optimization technique tailored for medical imaging, which is distinct from conventional methods of either using software optimization or hardware acceleration techniques. We focus on three key image processing techniques: a Median filter for noise reduction, a Threshold filter for high-contrast conversion of grayscale or color pixels, and a Sobel filter for edge detection. These techniques, while crucial for identifying features in medical images, often come with a high computational cost. Our method tackles this problem. To demonstrate the effectiveness of our approach, we applied it to a group of images of a patient with cerebral cancer. Our results show that SystemC chip modeling can significantly improve the computational efficiency of medical image processing, with potential applications in various medical imaging technologies. After implementing the proposed hardware model, there was an improvement of nearly 5.5 times compared to the software optimizations, and a 20-fold enhancement over the initial default implementation.