Synopsis

Keywords: Quantitative Imaging, Quantitative Imaging, T1 mapping

An innovative cylindrical phantom is used to mimic a dynamic acquisition with temperature-induced changes in T₁. Single reference (SR) variable flip angle (VFA) T₁ maps are acquired at several rotational positions of this phantom, and are compared to corresponding T₁ values acquired with the dual-angle variable flip angle and inversion recovery T₁ mapping methods. The SR-VFA method of T₁ mapping is shown to produce T₁ accuracy within 4-10% of both inversion recovery and dual-angle VFA T₁ measurements.

Introduction

The rise of MR-guided focused ultrasound ablative surgeries has led to an increased need for fast, accurate temperature mapping in all tissues, including in tissues such as fat and bone, whose temperature change cannot be measured with standard proton resonance frequency shift (PRFS) thermometry. Quantitative T₁-based thermometry has emerged as a potential solution to this problem. Recently, a technique called the single-reference (SR) variable flip angle (VFA) method¹ of T₁ mapping was developed as a way to produce one T₁ map per dynamic image, thereby potentially matching the temporal resolution of PRFS thermometry. However, this method has not been rigorously compared to the clinically accepted VFA method, nor to the gold standard T₁ mapping method, inversion recovery. This work aims to validate the SR-VFA T₁ mapping method against the gold standard inversion recovery method using an innovative phantom to mimic the dynamic processes of temperature-induced T₁ change.

Methods

An innovative cylindrical phantom was developed with three concentric rings, each of which was filled with 0.75% agar solution with varying MnCl₂ and CuSO₄ concentrations^2,3 so as to produce T₁ values varying smoothly in the range of ~200 to 2500 ms (see Figure 1), and T₂ values of ~20 ms to 250 ms. The phantom was centered on a rotating stand, and single-slice turbo-spin-echo inversion recovery measurements were taken at inversion times of 25, 75, 150, 400, 900, 1500, and 3500 ms. Following these measurements, 3D variable flip angle (VFA) stack-of-stars measurements (TR = 20 ms, TE = 1.24 ms) were taken with two scans at flip angles of 7° and 25°, with the center of the slab at the same location as the 2D inversion recovery slice. 3D B₁⁺ mapping with the AFI method⁴ was performed with a flip angle of 60° and TR₂/TR₁ = 150 ms / 30 ms. Following these initial “baseline” measurements, the phantom was manually rotated by 30° from 0° to 360°, and the VFA measurements were repeated after each rotation.
Images were reconstructed offline via in-house MATLAB (Mathworks, Natick, MA) code. Inversion recovery T₁ measurements were created using the reduced dimension nonlinear least squares fit method⁵. Other T₁ maps were reconstructed with the VFA and SR-VFA methods, with B₁⁺ correction from the B₁⁺ map acquired at 0° rotation. For the single reference method, the high flip angle scans at each incremental rotation of the phantom served as the “dynamic” images, whereas the two scans at 0° rotation served as the “baseline” images.
Following T₁ map creation, a 3x3 median filter was applied, and 120 evenly spaced 3-voxel-radius circular regions of interest (ROIs) were selected circularly along each of the three concentric rings for a central slice of the SR-VFA T₁ map at each of the 13 rotational points. These were registered with corresponding ROIs in the inversion recovery T₁ maps and VFA T₁ maps for comparison. Figure 2 gives a visual description of these methods.

Results

Figure 3 gives a Bland Altman plot of the difference in T₁ values between SR-VFA and the inversion recovery method as a function of the mean T₁ value. The SR-VFA T₁ values from small ROIs were within ±4-10% of the corresponding inversion recovery T₁ values. Similar results were obtained with the comparison between the SR-VFA and VFA T₁ mapping methods, as shown in Figure 4. Notably, in both method comparisons, the outer ring and inner ring exhibited the largest and smallest offset respectively in both method comparisons. The largest ring and smallest ring also correspond with the largest (200-1400 ms) and smallest (300-900 ms) “dynamic” range of T₁s from one end of the rotated phantom to another.

Discussion

This work shows that the single-reference variable flip angle (SR-VFA) method of T₁ mapping provides comparable accuracy to the VFA method. For this to happen, a sufficiently low TE must be attained such that TE << T₂*.¹ Otherwise, some estimate of dynamic T₂* changes must be attained, such as through multi-echo acquisition.⁶ Notably, as discussed by Malmberg et al. in 2022, the SR-VFA method tends to be more accurate for smaller changes in T₁ relative to baseline⁶. Thus, we would expect that the SR-VFA method’s accuracy would be better for the inner ring than the outer ring due to the smaller dynamic T₁ range. This is confirmed experimentally with Figures 3 and 4. Further, errors in B₁⁺ mapping, especially in the outside ring where B₁⁺ uniformity would be the worst, could account for some of the errors in T₁ .
Future work will evaluate a true dynamic comparison of the SR-VFA method to the gold standard via a rotating platform powered by an MRI-compatible servo motor,⁷ and also seek to perform T₁ validation in-vivo.

Conclusion

The single reference variable flip angle T₁ mapping method is shown to be a valid, faster substitute for the VFA method in low TE acquisitions in phantoms. This work is a key step towards clinical use of faster T₁ mapping in combined PRF/T₁ thermometry for focused ultrasound ablative surgeries, as well as dynamic contrast-enhanced imaging.

Acknowledgements

This research was made possible thanks to the Mark H. Huntsman endowed chair, and NIH grants R01EB028316, R37CA224141, and R03EB029204.