Accelerating 3D radial MPnRAGE using a self-supervised deep factor model

Yan Chen; Steven R Kecskemeti; James H Holmes; Curtis A Corum; Nima Yaghoobi; Vincent A Magnotta; Mathews Jacob

doi:10.1002/mrm.30549

Back

Accelerating 3D radial MPnRAGE using a self-supervised deep factor model

Journal article

Open access

Peer reviewed

Accelerating 3D radial MPnRAGE using a self-supervised deep factor model

Yan Chen, Steven R Kecskemeti, James H Holmes, Curtis A Corum, Nima Yaghoobi, Vincent A Magnotta and Mathews Jacob

Magnetic resonance in medicine, Vol.94(3), pp.1191-1201

09/2025

DOI: 10.1002/mrm.30549

PMCID: PMC12202740

PMID: 40457622

Files and links (1)

url

https://doi.org/10.1002/mrm.30549View

Published (Version of record) Open Access

Abstract

To develop a self-supervised and memory-efficient deep learning image reconstruction method for 4D non-Cartesian MRI with high resolution and a large parametric dimension. The deep factor model (DFM) represents a parametric series of 3D multicontrast images using a neural network conditioned by the inversion time using efficient zero-filled reconstructions as input estimates. The model parameters are learned in a single-shot learning (SSL) fashion from the k-space data of each acquisition. A compatible transfer learning (TL) approach using previously acquired data is also developed to reduce reconstruction time. The DFM is compared to subspace methods with different regularization strategies in a series of phantom and in vivo experiments using the MPnRAGE acquisition for multicontrast imaging and quantitative estimation. DFM-SSL improved the image quality and reduced bias and variance in quantitative estimates in both phantom and in vivo studies, outperforming all other tested methods. DFM-TL reduced the inference time while maintaining a performance comparable to DFM-SSL and outperforming subspace methods with multiple regularization techniques. The proposed DFM offers a superior representation of the multicontrast images compared to subspace models, especially in the highly accelerated MPnRAGE setting. The self-supervised training is ideal for methods with both high resolution and a large parametric dimension, where training neural networks can become computationally demanding without a dedicated high-end GPU array.

multi‐contrast MRI

quantitative MRI

single‐shot reconstruction

MPnRAGE

Details

Title: Subtitle: Accelerating 3D radial MPnRAGE using a self-supervised deep factor model
Creators: Yan Chen - University of Virginia
Steven R Kecskemeti - University of Wisconsin–Madison
James H Holmes - University of Iowa, Radiology
Curtis A Corum - University of Iowa, Radiology
Nima Yaghoobi - University of Virginia
Vincent A Magnotta - University of Iowa, Iowa Neuroscience Institute
Mathews Jacob - University of Virginia
Resource Type: Journal article
Publication Details: Magnetic resonance in medicine, Vol.94(3), pp.1191-1201
DOI: 10.1002/mrm.30549
PMID: 40457622
PMCID: PMC12202740
NLM abbreviation: Magn Reson Med
ISSN: 1522-2594
eISSN: 1522-2594
Publisher: WILEY
Grant note: P50HD103556 / NIH HHS R01AG087159 / NIH HHS P50 HD105353 / NICHD NIH HHS R01EB031169 / NIH HHS S10OD025025 / NIH HHS GE Healthcare R01AG067078 / NIH HHS R01HD108868 / NIH HHS R43MH122028 / NIH HHS R01EB019961 / NIH HHS
Language: English
Electronic publication date: 06/02/2025
Date published: 09/2025
Academic Unit: Roy J. Carver Department of Biomedical Engineering; Radiology; Electrical and Computer Engineering; Psychiatry; Iowa Neuroscience Institute
Record Identifier: 9984826428202771

Metrics

3 Record Views