Overview

This repository contains two pulse sequence developments using LIBRE water-excitation combined with 3D radial phyllotaxis k-space trajectories, implemented using MATLAB Pulseq.

$T_{1}$ LIBRE 3D Radial Sequence

• Folder: /t1libre_3d_radial
• Description: T₁-weighted LIBRE water-excitation with 3D radial phyllotaxis trajectory.
• Run with:

run('dev/t1libre_3d_radial/run.m')

$T_{2}$* LIBRE 3D Radial Sequence

• File: /t2prep_0716.m (under progress)
• Description: $T_{2}$*-weighted LIBRE water-excitation with 3D radial phyllotaxis trajectory.

Archive of Previous Development

Earlier versions and experimental scripts are available in the archive/ folder: /archive/. You can browse this directory for older implementations and research iterations.

---

Usage Instructions

1. Setup MATLAB Path
   • Ensure you cloned pulseq, and add the package into your path.
2. Inspect or Modify Parameters /t1libre_3d_radial/configs/libre_default.json Each sequence script includes configurable parameters such as:
   • RF pulse design (flip angle, sub-pulse duration, phase)
   • FOV, resolution, sampling trajectory
   • Timing (TE, TR) and spoiling controls, etc Feel free to adjust them as needed.

---

Successful Results (ISMRM 2026)

This section summarizes the positive outcomes reported in the abstract.

Figure 1: Sequence equivalence and trajectory implementation

• Pulseq and Siemens IDEA LIBRE timing diagrams are matched in design intent.
• Frequency-offset sign convention differences were handled in Pulseq to achieve equivalent water-selective excitation.
• A 3D pole-to-pole trajectory was implemented to reduce trajectory-related artifacts.

Figure 2: Phantom agreement (Site 1, IDEA vs Pulseq)

• SNR/CNR trends were closely matched across the full PDFF range (0-100% fat fraction).
• Strong linear agreement between IDEA and Pulseq for phantom SNR/CNR (r > 0.99, p < 0.01).
• Pixel-wise differences in vial maps were minimal.
• High volume-wise similarity (SSIM = 0.96).

Figure 3: In-vivo reproducibility across implementations and sites

• Comparable eye image quality and fat suppression were observed for:
  • Vendor IDEA-LIBRE (Site 1),
  • Pulseq-LIBRE (Site 1),
  • Pulseq-LIBRE (Site 2).
• Key orbital structures remained consistently visible (globe, rectus muscles, optic nerve, orbital fat), supporting stable water excitation.

Figure 4: Quantitative agreement in vivo (SNR/CNR)

• Intra-site (Site 1, IDEA vs Pulseq):
  • SNR correlation: r = 0.96-0.99, p < 0.01.
  • CNR correlation: r = 0.76-0.98, p < 0.025.
• Inter-site (Pulseq Site 1 vs Site 2):
  • SNR correlation: r = 0.83-0.99, p < 0.02.
  • CNR correlation: r = 0.61-0.99, p < 0.04.
• Within-site agreement was tighter than cross-site agreement, as expected.

Table 5: SSIM consistency

• Phantom and in-vivo SSIM indicate strong implementation consistency.
• Reported highlights include:
  • Up to SSIM = 0.96 in phantom comparison.
  • SSIM > 0.9 for in-vivo, same-site implementation comparisons.
  • Lower but acceptable inter-site SSIM due to expected scanner/site variability.

Results

Figure A1. Sequence implementation and trajectory (abstract figure set).

Figure A1. Phantom characterization panel.

Figure A2. Phantom agreement / difference panel.

Figure A2. In-vivo cross-implementation / cross-site comparison.

Figure A3. In-vivo reproducibility across implementations and sites.

Figure A4. Quantitative agreement panel (Intra-site SNR/CNR).

Figure A4. Quantitative agreement panel (Inter-site SNR/CNR).

Figure A5. SSIM comparison table panel.

---

Possible Issues and Notes

Key Practical Notes

• Shimming quality is critical for eye-region imaging. Please make sure the shim box does not include too much air region.
• FOV center affects image quality:
  • Centering FOV at the brain gave better images than centering at the eye middle in reported tests.
  • If FOV is centered near the nasal cavity, results may become sensitive to head positioning (with vs without foam support).
• Gradient signs in Pulseq need flipping to align acquisition orientation with IDEA data.
• LIBRE pulse flip angle reminder:
  • If total FA = 12 deg in the paper, each of the 2 sub-pulses should be set as 6 deg.
• Trajectory using PTP can reduce artifacts caused by scanner-specific quirks compared to the original 3D radial phyllotaix.

CHUV 2025-06-24: GRE vs LIBRE (whole-brain views) with bad shimming.

Figure 1. Transverse comparison panel.

Figure 2. Transverse difference map.

Figure 3. Sagittal-left comparison panel.

Figure 4. Sagittal-left difference map.

CHUV 2025-06-24: Around-eye focused views with bad shimming.

Figure 5. Around-eye transverse comparison panel.

Figure 6. Around-eye transverse difference map.

Figure 7. Around-eye sagittal-left comparison panel.

Figure 8. Around-eye sagittal-left difference map.

FOV and trajectory notes

Figure 9. FOV location comparison (head center vs the center of the eye ROI).

Figure 10. FOV and shimming around the center of eye ROI, with the same shimming, signal is inconsistent across two adjacent sequences.

Figure 13. Original vs PTP trajectory comparison. Pole-to-pole sequence is more robust from field inhomogineity.

Figure 14. Head-positioning sensitivity example (center of eye ROI FOV; without foam vs with foam).