University of Southern California

The General Hospital Corporation

Justin P Haldar

Lawrence Leroy Wald

Richard M Leahy

Bo Zhao

Tae Hyung Kim

Methods, systems, devices and apparatuses for generating a high-quality MRI image from under-sampled or corrupted data The image reconstruction system includes a memory. The memory is configured to store multiple samples of biological, physiological, neurological or anatomical data that has missing or corrupted k-space data and a deep learning model or neural network. The image reconstruction system includes a processor coupled to the memory. The processor is configured to obtain the multiple samples. The processor is configured to determine the missing or corrupted k-space data using the multiple samples and the deep learning model or neural network. The processor is configured to reconstruct an MRI image using the determined missing or corrupted k-space data and the multiple samples.

Scan-specific recurrent neural network for image reconstruction

Systems and methods for acquiring magnetic resonance fingerprinting (MRF) imaging data from a subject using a magnetic resonance imaging (MRI) system are provided. The method includes receiving an indication of an MRF imaging process to be performed by the MRI system and receiving a desired design objective for the MRF imaging process and a configuration metric associated with the MRF imaging process. The method further includes using the configuration metric to bound a variance of tissue parameter estimates associated with the MRF imaging process and determine imaging parameters that achieve the desired design objective. The method also includes performing the MRF imaging process using the determined imaging parameters to acquire MRF data using the MRI system.

Systems and methods for designing magnetic resonance fingerprinting imaging parameters

Samples of a Fourier transform of a signal may be received that are distributed in a generally spherically-shaped pattern on a surface of at least one sphere. The samples may be assembled to form a profile function having a domain that is a surface of at least one sphere. Information indicative of at least one property of the signal may be determined by applying a spherical linear transform to the profile function. The spherical linear transform may use more than just an equator of the profile function for each of multiple orientations.

Growing community of inventors

Los Angeles, CA, United States of America

Justin P Haldar