Deep Kernel Method for Dynamic MRI Reconstruction

We introduce a deep kernel model for the recovery of real-time dynamic MRI from highly undersampled measurements. The proposed scheme uses the cascade of two deep convolutional neural networks (CNN) for the kernel representation of images. Unlike the supervised CNN approaches for image reconstructions that require extensive fully-sampled training data for learning the network, the parameters of the two CNNs in the proposed method are learned from the undersampled measurements directly in this work. The main benefits of the proposed scheme are (a) the elimination of the empirical choice of the feature map and kernel function in the kernel method, and (b) the unsupervised nature of the proposed framework.