Pixelwise gradient gan model (PGGM) for image fusion and gadolinium-free contrast-enhanced MRI (GFCE-MRI) synthesis

Abstract

This thesis explores the transformative potential of integrating Artificial Intelligence (AI), particularly Generative Adversarial Networks (GANs), into Magnetic Resonance Imaging (MRI) to enhance diagnostic imaging for nasopharyngeal cancer (NPC). Amid growing concerns over the safety and accessibility of traditional contrast agents, this research introduces an innovative Pixelwise Gradient GAN Model (PGGM) designed for Virtual Contrast Enhancement (VCE) and image fusion. By leveraging AI, PGGM aims to simulate the effects of gadolinium-based contrast agents, enhancing MRI images without their associated health risks. This approach not only promises to mitigate the adverse effects associated with conventional contrast agents but also to improve the accessibility and quality of MRI imaging, particularly for renal-impaired patients. Image fusion can also be performed using PGGM. By combining the desired features from different image modalities, fused images capturing the desired features of the input images will be produced by PGGMs. The thesis details the development and application of PGGM, highlighting its dual utility in enhancing T1-weighted and T2-weighted MRI images. The model architecture incorporates advanced AI techniques, including LSGANs and NSGANs, to address and optimize the contrast and texture representation of MRI scans. Through meticulous hyperparameter tuning and the application of various normalization methods, PGGM demonstrates superior performance in producing high-quality, interpretable MRI images for NPC. Quantitative and qualitative analyses validate PGGM's effectiveness against existing models, showcasing its ability to produce images with enhanced contrast and texture similar to traditional contrast-enhanced scans. The findings suggest that AI-driven techniques, as exemplified by PGGM, could revolutionize MRI by offering safer, more accessible, and accurate diagnostic tools. This research underscores the potential of AI in medical imaging, offering a beacon of hope for enhancing diagnostic accuracy and patient safety. The integration of AI into MRI through PGGM and similar models represents a pivotal shift in diagnostic imaging, paving the way for safer, more efficient, and accessible imaging techniques for NPC and potentially other conditions.