METHOD AND DEVICE FOR DYNAMIC MAGNETIC RESONANCE IMAGE RECONSTRUCTION WITH ADAPTIVE PARAMETER LEARNING

摘要

Provided are a method and device for dynamic magnetic resonance image reconstruction with adaptive parameter learning; the method improves the regularization term in a CS-MRI model, and comprises using DCT in the spatial domain and TV in the time domain to de-redundant a dynamic magnetic resonance image, and using a convolutional neural network to adaptively learn a large number of parameters in CS-MRI and establish a magnetic resonance image reconstruction model; reconstructing a sample image by means of the established magnetic resonance reconstruction model to obtain a reconstructed image; calculating the difference between a fully sampled image and the reconstructed image; according to the difference, using a back-propagation algorithm in the network to update the parameters in the model, comprising DCT, TV filter operator, and regularization parameters. The method can efficiently reconstruct a highly undersampled image to obtain an image having high reconstruction accuracy and reconstruction speed, thus the time of magnetic resonance scanning can be effectively reduced without losing spatial resolution.