Enhancing Two Dimensional Magnetic Resonance Image Using Generative Adversarial Network

Abstract

Magnetic Resonance Imaging is often used in medical imaging techniques. The particular magnetic resonance imaging needs to be clear and sharp for precise and effective medical diagnosis. The image quality can be severely harmed by a slight movement in the muscle or the intended area. It is difficult to obtain high-quality scans due to hardware limitations and health risks associated with magnetic resonance imaging radiation. The existing research has shown that the generative adversarial network approach with deep neural networks gives impressive results compared to traditional approaches such as bicubic interpolation. In the proposed methodology, generative adversarial networks is used to improve the resolution and quality of the magnetic resonance imaging. The proposed architecture converts the low-resolution image input to high-resolution image output. Two different neural networks are used in the generative adversarial network i. e., the discriminator and the generator. These two architecture compete against one another to enhance the final output. The high-resolution results are generated by a generator, and the generator's performance is improved by a discriminator.