T2 Mapping from Super-Resolution-Reconstructed Clinical Fast Spin Echo Magnetic Resonance Acquisitions

Hélène Lajous, Tom Hilbert, Christopher W. Roy, Sébastien Tourbier, Priscille de Dumast, Thomas Yu, Jean-Philippe Thiran, Jean-Baptiste Ledoux, Davide Piccini, Patric Hagmann, Reto Meuli, Tobias Kober, Matthias Stuber, Ruud B. van Heeswijk, Meritxell Bach CuadraMICCAI 2020 · DOI

Abstract

Quantitative MRI (qMRI), particularly T2 mapping, offers significant potential for assessing fetal brain maturation and identifying neurological abnormalities in utero. However, traditional mapping techniques are prohibitively slow for moving subjects. This paper presents a novel pipeline that leverages super-resolution (SR) reconstruction of standard clinical fast spin echo (HASTE) sequences acquired at multiple echo times. By fusing low-resolution 2D slices into a high-resolution 3D volume before quantitative fitting, the method generates accurate T2 maps in under 9 minutes. Validation on a NIST phantom demonstrates high repeatability and accuracy, with relative errors under 11% compared to gold-standard sequences.

Introduction

A major challenge in fetal medicine is obtaining high-quality quantitative brain maps from patients who cannot remain stationary. While T2 mapping reveals critical information about brain microstructure, "gold standard" sequences often require hours of scan time, making them incompatible with unpredictable fetal movement. Clinical practice currently relies on Fast Spin Echo (HASTE) sequences, which are rapid but provide only low-resolution 2D slices insufficient for precise quantitative measurements. This study addresses the gap between fast clinical scans and slow quantitative precision.

Methods

The proposed pipeline consists of three primary stages: variable acquisition, super-resolution reconstruction, and quantitative fitting.

Overview of the MRI acquisition and 3D super-resolution reconstruction pipeline.
  1. Variable Acquisition: Standard 2D HASTE images are acquired at multiple Echo Times (TE) across three orthogonal planes (axial, coronal, and sagittal).
  2. Super-Resolution (SR) Reconstruction: A Total-Variation (TV) SR algorithm is employed to fuse these low-resolution 2D slices into a single, isotropic high-resolution 3D volume for each TE.
  3. Quantitative Fitting: A mono-exponential decay model is fitted to every voxel in the reconstructed volumes to calculate specific T2 relaxation times.

The study was validated using the NIST/ISMRM System Standard Model 130 phantom, focusing on T2 values (180ms to 430ms) representative of fetal and neonatal brain tissue.

Results and Key Insights

The SR-based approach significantly improves upon standard HASTE mapping. A key finding is that the SR process acts as a filter, reducing Gibbs ringing artifacts by combining data from multiple orientations.

Comparison of T2 maps and standard deviation maps for SE, HASTE, and SR methods.
  • Efficiency and Accuracy: The method produces a full 3D volume in under 9 minutes, a dramatic reduction from the 7+ hours required for gold-standard single-echo spin-echo (SE) sequences. The relative error for relevant T2 values was maintained below 11%.
  • Optimal Echo Selection: Researchers identified that using 6 echo times provides the best balance between acquisition speed and mapping accuracy.
  • Mitigating Overestimation: Standard HASTE sequences often overestimate T2 due to stimulated echoes. The integration of SR reconstruction brings measurements significantly closer to the gold standard.
  • Robustness: The results were highly repeatable (CV < 8%) and remained stable even in the presence of simulated noise.

Discussion and Future Work

While the phantom validation demonstrates the technical feasibility and accuracy of the SR-qMRI pipeline, the study's primary limitation is its reliance on a static model. Moving forward, the researchers plan to evaluate the method in vivo. Although SR techniques are inherently designed to mitigate motion artifacts, the specific interaction between fetal movement and variable echo timing requires further investigation. Initial pilot studies on adult brains are planned to refine the technique before clinical fetal trials.

Summary updated

BibTeX
@inproceedings{lajous2020t,
  author = {Hélène Lajous and Tom Hilbert and Christopher W. Roy and Sébastien Tourbier and Priscille de Dumast and Thomas Yu and Jean-Philippe Thiran and Jean-Baptiste Ledoux and Davide Piccini and Patric Hagmann and Reto Meuli and Tobias Kober and Matthias Stuber and Ruud B. van Heeswijk and Meritxell Bach Cuadra},
  title = {{T2 Mapping from Super-Resolution-Reconstructed Clinical Fast Spin Echo Magnetic Resonance Acquisitions}},
  booktitle = {International Conference on Medical Image Computing and Computer-Assisted Intervention},
  year = {2020},
  doi = {10.1007/978-3-030-59713-9\_12},
}
Figure