Image enhancement systems and methods with variable number of excitation (NEX) acquisitions accelerated using compressed sensing for magnetic resonance (MR) imaging are provided. The MR system comprises a body coil adapted to emit electromagnetic waves onto the anatomy of interest and receive the signals emitted from the anatomy of interest. The system comprises variable number of excitations (NEX) based compressed sensing by acquisition of different points in k-space using the body coil. A first neural network comprising an image enhancement module is provided to reconstruct the body coil images that provides a high-quality image. The high-quality image is stored in an image database. A processor is configured to connect the image database to a second neural network that is a deep learning network trained to assess the image quality and provide feedback to the processor and the first neural network.