Time efficient whole-brain coverage with MR Fingerprinting using slice-interleaved echo-planar-imaging

Benedikt Rieger, Mehmet Akçakaya, José C. Pariente, Sara Llufriú, Eloy Martínez‐Heras, Sebastian Weingärtner, Lothar R. SchadScientific 2018 · DOI

Abstract

Magnetic resonance fingerprinting (MRF) is a promising method for fast simultaneous quantification of multiple tissue parameters. The objective of this study is to improve the coverage of MRF based on echo-planar imaging (MRF-EPI) by using a slice-interleaved acquisition scheme. For this, the MRF-EPI is modified to acquire several slices in a randomized interleaved manner, increasing the effective repetition time of the spoiled gradient echo readout acquisition in each slice. Per-slice matching of the signal-trace to a precomputed dictionary allows the generation of T<sub>1</sub> and T<sub>2</sub> maps with integrated B<sub>1</sub><sup>+</sup> correction. Subsequent compensation for the coil sensitivity profile and normalization to the cerebrospinal fluid additionally allows for quantitative proton density (PD) mapping. Numerical simulations are performed to optimize the number of interleaved slices. Quantification accuracy is validated in phantom scans and feasibility is demonstrated in-vivo. Numerical simulations suggest the acquisition of four slices as a trade-off between quantification precision and scan-time. Phantom results indicate good agreement with reference measurements (Difference T<sub>1</sub>: -2.4 ± 1.1%, T<sub>2</sub>: -0.5 ± 2.5%, PD: -0.5 ± 7.2%). In-vivo whole-brain coverage of T<sub>1</sub>, T<sub>2</sub>* and PD with 32 slices was acquired within 3:36 minutes, resulting in parameter maps of high visual quality and comparable performance with single-slice MRF-EPI at 4-fold scan-time reduction.

Introduction

Quantitative MRI (qMRI) is often considered the "holy grail" of medical imaging, as it provides objective physical values—such as relaxation times—rather than qualitative, weighted images. However, the clinical adoption of qMRI has been hampered by its slow acquisition speeds. Magnetic Resonance Fingerprinting (MRF) was introduced to address this by acquiring unique signal evolutions ("fingerprints") and matching them to precomputed dictionaries.

While MRF is inherently fast, many high-speed implementations like MRF-EPI have been limited to single-slice coverage. Scaling these methods to the whole brain often requires significant "dead time" to allow for longitudinal magnetization recovery between slices, which negates the speed benefits. This study proposes a slice-interleaved approach that utilizes this idle time to achieve rapid volumetric coverage without sacrificing precision.

Methods

The researchers modified the existing MRF-EPI sequence to incorporate a slice-interleaved acquisition scheme. The core innovation involves acquiring a group of slices (optimized to four) in rapid succession following a single global inversion pulse.

Schematic of the slice-interleaved MRF-EPI sequence (A) and the randomized patterns for TR (B), TE (C), and flip angle (D).

To ensure that the signal traces remain distinct and easily matchable to the dictionary, the acquisition order of slices within these groups is pseudo-randomized. This randomization ensures that the inversion recovery curve is sampled effectively across all slices. The method also integrates quantitative Proton Density (PD) mapping by compensating for coil sensitivity and normalizing the signal to the cerebrospinal fluid (CSF).

Validation was performed through:

  1. Numerical Simulations: To identify the optimal number of interleaved slices.
  2. Phantom Scans: Comparing results against gold-standard reference measurements (IR-TSE for T1 and GRE for T2*).
  3. In-vivo Human Scans: Testing feasibility in healthy volunteers and patients with Multiple Sclerosis (MS).

Results and Discussion

The slice-interleaved approach achieved a four-fold reduction in scan time compared to traditional single-slice MRF-EPI. The system successfully acquired 32 slices for whole-brain coverage in just 3:36 minutes.

Phantom trials confirmed high accuracy, with minimal deviations from reference values (T1: -2.4%, T2: -0.5%, PD: -0.5%). In-vivo results demonstrated high visual quality in the resulting parameter maps. In patients with MS, the method clearly delineated lesions in T1, T2, and PD maps, showing significantly higher values in damaged tissue compared to healthy white matter.

Simulations revealed that interleaving four slices provides the best balance; while more slices could further accelerate the scan, it would eventually lead to a marginal loss of precision by sampling the T1 recovery curve too sparsely.

Conclusion and Future Work

By turning "dead time" into data acquisition time, slice-interleaved MRF-EPI enables efficient whole-brain quantitative mapping. While the current implementation maps T2*, future iterations could incorporate refocusing pulses for pure T2 mapping. Additionally, further optimization of flip angle and echo time (TE) patterns specifically for interleaved sequences may boost accuracy even further. This technique represents a significant step toward making rapid, whole-brain quantitative MRI a routine clinical tool.

Summary updated

Abstract

Magnetic resonance fingerprinting (MRF) is a promising method for fast simultaneous quantification of multiple tissue parameters. The objective of this study is to improve the coverage of MRF based on echo-planar imaging (MRF-EPI) by using a slice-interleaved acquisition scheme. For this, the MRF-EPI is modified to acquire several slices in a randomized interleaved manner, increasing the effective repetition time of the spoiled gradient echo readout acquisition in each slice. Per-slice matching of the signal-trace to a precomputed dictionary allows the generation of T<sub>1</sub> and T<sub>2</sub>* maps with integrated B<sub>1</sub><sup>+</sup> correction. Subsequent compensation for the coil sensitivity profile and normalization to the cerebrospinal fluid additionally allows for quantitative proton density (PD) mapping. Numerical simulations are performed to optimize the number of interleaved slices. Quantification accuracy is validated in phantom scans and feasibility is demonstrated in-vivo. Numerical simulations suggest the acquisition of four slices as a trade-off between quantification precision and scan-time. Phantom results indicate good agreement with reference measurements (Difference T<sub>1</sub>: -2.4 ± 1.1%, T<sub>2</sub>*: -0.5 ± 2.5%, PD: -0.5 ± 7.2%). In-vivo whole-brain coverage of T<sub>1</sub>, T<sub>2</sub>* and PD with 32 slices was acquired within 3:36 minutes, resulting in parameter maps of high visual quality and comparable performance with single-slice MRF-EPI at 4-fold scan-time reduction.
BibTeX
@inproceedings{rieger2018time,
  author = {Benedikt Rieger and Mehmet Akçakaya and José C. Pariente and Sara Llufriú and Eloy Martínez‐Heras and Sebastian Weingärtner and Lothar R. Schad},
  title = {{Time efficient whole-brain coverage with MR Fingerprinting using slice-interleaved echo-planar-imaging}},
  booktitle = {Scientific},
  year = {2018},
  doi = {10.1038/s41598-018-24920-z},
}
Figure