Bridging Modalities with VarVit-GAN: A Generative Adversarial Network for Multi-Modal Brain MRI Translation
Kaliprasad Pani, Indu Chawla · 2023
In the realm of medical imaging, capturing diverse aspects of brain structures is crucial for accurate diagnoses of brain tumors. Magnetic Resonance Imaging (MRI) has emerged as the most trusted imaging modality due to its superior soft tissue contrast and non-invasiveness. However, the complexity of brain pathologies necessitates different modalities like T1, T2, T1CE and FLAIR for a comprehensive understanding. However, this process poses challenges, such as the need for contrast agent injection and extended scanning times, which can entail potential risks, leading to patient discomfort and resource inefficiencies. To address these issues, we propose an approach named VarVit-GAN, harnessing the power of Generative Adversarial Networks (GANs). By leveraging the capabilities of both Autoencoders and Vision Transformer techniques, the proposed model enables the generation of high-quality T1CE images from the T1 modality. Through comprehensive experiments on a brain MRI dataset, VarVit-GAN demonstrates its effectiveness. Rigorous evaluations using metrics such as Structural Similarity Index (SSIM) and Peak Signal-to-Noise Ratio (PSNR) validate its superior performance compared to traditional methods. In summary, VarVit-GAN offers a promising avenue for multi-modal brain image translation, benefiting medical practitioners and researchers in improved diagnostics and patient care.