Abstract— Magnetic resonance imaging (MRI) at high spatial resolution is essential for accurate detailed and quantitative physiologic assessment. However, single image super resolution (SISR) techniques with traditional long scan times, reduced spatial masking, and low signal-to-noise ratio (SNR) to acquire MR images with high spatial resolution have emerged as a solution promising high-resolution reproduction from relatively simple MR images. In this work, we introduce a new method called ESRGAN. ESRGAN builds on the GAN framework and improves its design and operation. Key features of ESRGAN include optimized network design, improved adversary loss functions, and sensitivity loss integration for image quality.