Abstract by Andrew Carr
Medical image reconstruction using deep learning
Medical imaging is a thoroughly studied field with significant impact on human well-being; however, conventional imaging systems are expensive and unwieldy due to the physical constraints required to achieve high resolution imaging. We propose a novel reconstruction method for Magnetic Resonance Imaging (MRI) using a Deep Convolutional U-Net to learn manifold features in sensor space. This method, called AwesomeSauceImaging, relaxes physical constraints by using a higher capacity model which can achieve comparable results with smaller, less expensive devices.