An HARDI compressed sensing super-resolution reconstruction method based on deep dictionary learning

Abstract

The invention relates to an HARDI compressed sensing super-resolution reconstruction method based on deep dictionary learning. The method comprises the steps of preprocessing an acquired high-angle diffusion image to obtain training data, establishing a deep network model which is used for dictionary learning and comprises a plurality of layers of dictionaries, training the constructed deep network model through the training data; extracting orthogonal vectors in sequence by adopting orthogonal triangular decomposition to serve as an initial dictionary; the method comprises the following stepsof: solving a last layer of learning dictionary; adding a sparsity constraint item to carry out sparse representation, collecting data of which the density is far lower than that of original data astest data, obtaining a sparse representation coefficient based on the test data, and finally generating a reconstructed three-dimensional diffusion magnetic resonance image related to a human body through a direction distribution function obtained through radial integration. According to the method, the sampling data amount required for reconstructing the diffusion magnetic resonance image with the same resolution is smaller. And the data sampling speed is higher. And the nerve fiber reconstruction capability is better.

Description

technical field [0001] The invention relates to the fields of compressed sensing and medical imaging, in particular to a high-angle-resolution diffusion imaging compressed sensing super-resolution reconstruction method based on deep dictionary learning. Background technique [0002] For HARDI imaging, the existing compressive sensing technology mainly performs compressive sensing reconstruction of data based on the spatial resolution or angular resolution of the data. The brain is sampled by reducing the number of spatial or angular upsampling, and then using these small sampled data to obtain high-resolution images through dictionary reconstruction. Joint compressive sampling methods for both spatial and angular aspects have recently emerged. At the same time, the number of samples in space and angle is reduced, which further reduces the number of data samples required to reconstruct the image. Most of the dictionaries used are a priori dictionaries, which are dictionarie...

AI Technical Summary

Problems solved by technology

The existing dictionary learning method for joint compressed sensing is mainly designed for HARDI image denoising, and the learning algorithm is mainly evolved according to the conventional classic dictionary learning algorithm, and the expression ability of the obtained dictionary is relatively weak

Images