Cedars-Sinai Medical Center

University of Pittsburgh

Anthony Christodoulou

Debiao Li

Yibin Xie

Zhaoyang Fan

Christopher Nguyen

Jaime Shaw

Zhengwei Zhou

Zixin Deng

Nan Wang

Sen Ma

Yijen Lin Wu

Pei Han

Yuhua Chen

A method for performing magnetic resonance imaging on a subject comprises obtaining undersampled imaging data, extracting one or more temporal basis functions from the imaging data, extracting one or more preliminary spatial weighting functions from the imaging data, inputting the one or more preliminary spatial weighting functions into a neural network to produce one or more final spatial weighting functions, and multiplying the one or more final spatial weighting functions by the one or more temporal basis functions to generate an image sequence. Each of the temporal basis functions corresponds to at least one time-varying dimension of the subject. Each of the preliminary spatial weighting functions corresponds to a spatially-varying dimension of the subject. Each of the final spatial weighting functions is an artifact-free estimation of the one of the one or more preliminary spatial weighting functions.

Systems and methods of deep learning for large-scale dynamic magnetic resonance image reconstruction

A method of analyzing a multidimensional image tensor containing a plurality of images comprises: performing imaging scans of a subject imaging data; generating the multidimensional image tensor from the imaging data; determining a spatial basis tensor containing basis images based on the multidimensional image tensor; determining a temporal basis tensor containing basis functions for a temporal dimension based on the multidimensional image tensor; determining a core tensor that relates the spatial basis tensor to the temporal basis tensor; pre-multiplying the core tensor and the temporal basis tensor to produce a modified temporal basis tensor; storing the spatial basis tensor and the modified temporal basis tensor; and generating an image by multiplying at least (i) at least a portion of the spatial basis tensor and (ii) at least a portion of the modified temporal basis tensor.

Storage, display, and analysis of factored multidimensional images

A method for performing real-time magnetic resonance (MR) imaging on a subject is disclosed. A prep pulse sequence is applied to the subject to obtain a high-quality special subspace, and a direct linear mapping from k-space training data to subspace coordinates. A live pulse sequence is then applied to the subject. During the live pulse sequence, real-time images are constructed using a fast matrix multiplication procedure on a single instance of the k-space training readout (e.g., a single k-space line or trajectory), which can be acquired at a high temporal rate.

Systems and methods of on-the-fly generation of 3D dynamic images using a pre-learned spatial subspace

A magnetic resonance imaging method is includes collecting spatially encoded data from a subject using an MRI system, and directly extracting a number of temporal basis functions using at least a first portion of the spatially encoded data. The method further includes, after directly extracting the number of temporal basis functions, iteratively calculating a number of coefficient images using the number of temporal basis functions and at least a second portion of the spatially encoded data. Finally, the method includes generating a 4D image based on the number of temporal basis functions and the number of coefficient images.

System and method for integrated time-resolved 4D functional and anatomical MRI

A method of performing multidimensional magnetic resonance imaging on a subject comprises collecting imaging data for a region of interest of the subject, the imaging data related to one or more spatially-varying parameters of the subject within the region of interest; collecting auxiliary data for the region of interest in the subject, the auxiliary data related to one or more time-varying parameters of the subject within the region of interest; linking the imaging data and the auxiliary data; and constructing an image tensor with one or more temporal dimensions based on at least a portion of the linked imaging data and at least a portion of the linked auxiliary data.

Low-rank tensor imaging for quantitative MRI

A new low rank tensor (LRT) imaging strategy/methodology, specifically for quantitative cardiovascular magnetic resonance (CMR) multitasking, includes performing a low-rank tensor image model exploiting image correlation along multiple physiological and physical time dimensions, a non-ECG data acquisition strategy featuring minimal gaps in acquisition and frequent collection of auxiliary subspace training data, and a factored tensor reconstruction approach which enforces the LRT model.

Growing community of inventors

Los Angeles, CA, United States of America

Anthony Christodoulou