StyleUNet: An Enhanced Style Transfer for Brain MRI Images using StyleGAN with U-Net

This research aims to improve the quality of brain MRI images through the combined application of StyleGAN and U-Net (StyleUNet), a powerful tool in the field of deep learning. StyleUNet, known for its capability to manipulate and improve images, is employed to transform brain MRI scans, making them clearer and more visually appealing. By utilizing StyleGAN and U-Net generators, the work seeks to refine the details and characteristics of brain MRI images, thereby facilitating better interpretation and analysis by medical professionals and researchers. This process involves a sophisticated interplay of algorithms and image-processing techniques to achieve optimal results. The research comprised four experimental trials using datasets containing T1-styled and T2-styled images. The model exhibited a style consistency loss of 4.9 and 3.34 throughout the training phase, notably lower than alternative models. The generator loss (T1-2.3, T2-1.5) and discriminator losses (T1-1.7, T2-0.91) indicate a harmonious operational performance. The model demonstrated superior efficacy in generating images of the same style, achieving peak SSIM scores of 0.93 and 0.95 and PSNR values of 21.23 and 20.67, respectively. These outcomes highlight the robustness and effectiveness of the proposed methodology in image generation tasks. This work is a significant step forward in leveraging deep learning methods for medical imaging applications, highlighting the importance of innovative approaches in healthcare technology.