<![CDATA[Cardiac magnetic resonance imaging (CMR) is a vital tool for noninvasively assessing heart shape and function, offering exceptional spatial and temporal resolution alongside superior soft tissue contrast. However, CMR images often suffer from noise and artifacts due to cardiac and respiratory motion or patient movement impacting diagnostic accuracy. While real-time noise suppression can mitigate these issues, it comes at a high computational and financial cost. This paper introduces a method that includes a complete way to clean up medical images by using a new Denoising Generative Adversarial Network (D-GAN). The D-GAN architecture incorporates a recursive residual group-based generator and a discriminator inspired by PatchGAN.The recursive residual group-based generator and the Selective Kernel Feature Fusion (SKFF) mechanism are part of a new D-GAN architecture that makes denoising work better. A PatchGAN-based discriminator designed to improve adversarial training dynamics and texture modeling for medical images. These innovations offer improved feature refinement and texture modeling, enhancing the denoising of cardiac MRI images. allows the model to get a doubling context of local and global, informational, and hierarchical developed features located in the generator. Our technique outperforms other methods in terms of PSNR and SSIM. With scores of 0.837, 0.911, and 0.971 for noise levels of 0.3, 0.2, and 0.1, and PSNR scores of 29.48 dB, 32.58 dB, and 37.85 dB, the results show that the D-GAN method is better than other methods.]]>
