Application of GPU Embedded Systems for Medical Image Analysis: Brain Tumour Segmentation

Abstract

GPU-embedded devices for image processing in medicine, this work explores accurate brain tumour segmentation, potentially revolutionising neurology's diagnostic precision and treatment planning. This work intends to improve brain tumour segmentation accuracy and efficiency by using GPU processing capability, opening new avenues for medical imaging technology developments. As medical data grows for research and diagnosis, healthcare practitioners automate techniques for reliable and rapid picture analysis, such as segmentation or restoration. Many current solutions for these jobs rely on Deep Learning techniques, which demand strong hardware and are not suitable for the power consumption issues outlined above. Demand exists for developing cost-effective image analysis systems with enhanced performance. The study proposes an automated brain tumour segmentation approach using a Convolutional Neural Network on a low-cost GPU-integrated platform using Deep Learning. By redefining core operations, successfully verified the strategy of segmenting brain tumours using the BRaTS 2022 dataset, demonstrating that artificial neural networks can be trained and deployed in medical fields with minimal resources.