An Optimized GAN Approach for Denoising MRI Brain Tumor Images

Image denoising is a critical pre-processing step in MRI brain imaging that is intended to improve image quality and clarity by reducing noise. The presence of noise in MRI images can be attributed to multiple factors, including the imaging apparatus, patient motion, and transmission inaccuracies. The presence of this noise might hinder the visibility of crucial anatomical features, hence posing challenges for medical practitioners in effectively diagnosing and interpreting the images. An optimized technique has been presented to tackle the aforementioned problem, which combines the Generative Adversarial Network (GAN) with the Spotted Hyena Optimizer (SHO). This technique is known as an optimized GAN (OGAN). The utilization of the Spotted Hyena Optimizer (SHO) in conjunction with GAN-based denoising presents a sophisticated method for enhancing image quality. The SHO optimizes the GAN's parameters, resulting in high-quality denoised images that retain essential characteristics. The effectiveness of this approach has been evaluated using peak signal-to-noise ratio (PSNR), Universal Image Quality Index (UQI), and Structural Similarity Index (SSIM). This methodology exhibits its superior performance compared to existing techniques.