T1W MRI to T2W MRI Image Synthesis Using SSIM-CycleGAN

T1-weighted (T1W) and T2-weighted (T2W) sequences are common in MRI imaging, yet both are challenging to acquire efficiently. To optimize time and scanner availability, cross-modality image synthesis has emerged as a promising solution. The synthesis of T2-weighted MRI images from T1-weighted MRI images is essential in clinical practice, as it enables the extraction of valuable diagnostic information without requiring additional imaging sessions. Recently, CycleGAN networks have been widely utilized for synthesizing images from one modality to another due to their ability to be trained in an unsupervised manner. However, traditional CycleGANs often struggle to accurately capture structural and luminance information. Our proposed SSIM-CycleGAN integrates SSIM loss into the CycleGAN framework, resulting in a substantial improvement in the quality of the generated T2-weighted MRI images from T1-weighted MRI images. By focusing on structural similarity, this approach enhances both perceptual quality and clinical relevance. We conducted a detailed comparison of our method against existing techniques, and our proposed approach demonstrated significant improvements both quantitatively and qualitatively.