Bridging the gap between structural and metabolic neuroimaging via MRI-to-PET synthesis: A tri-attention enhanced GAN approach

Abstract

Magnetic resonance imaging (MRI) and positron emission tomography (PET) are two essential neuroimaging modalities that provide complementary structural and metabolic information about the brain, thereby enhancing diagnostic precision for brain disorders such as Alzheimer’s disease (AD). To address the limitations of missing modality data, we propose a novel 3D GAN-based framework for MRI-to-PET neuroimage synthesis, incorporating a Tri-Attention module to integrate spatial, channel, and frequency attention across multiple scales. The proposed method enables the generation of complementary metabolism information by synthesizing PET scans, effectively bridging the modality gap. The effectiveness of the proposed method is evaluated on a subset of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. Experimental results demonstrate the superiority of our approach, achieving significant improvements in image quality metrics (SSIM: 0.882, PSNR: 26.508) and clinical metrics (SUVR), outperforming state-of-the-art methods. These findings underscore the potential of our framework to bridge the gap between structural and metabolic information, offering a promising tool for cross-modality neuroimage synthesis and clinical applications.