Introduction

Magnetization transfer (MT) ratio (MTR) of peripheral nerves correlates with disability in patients with Charcot-Marie-Tooth Type 1A (CMT1A)¹, which is an inherited dysmyelinating neuropathy. Although MTR reports on myelin content², it is semi-quantitative, lacks pathological specificity, and is sensitive to scanner hardware and acquisition parameters^3–5. Quantitative MT (qMT) methods provide tissue-specific parameters^2,6–9 (e.g., macromolecular pool-size-ratio, PSR) that more specifically report on myelin; however, these methods require long scan times that limit clinical translation. More recently, MT saturation (MTsat)^10,11, which isolates MT from T₁ relaxation effects, has been proposed to improve pathological specificity while balancing scan time constraints. As such, this study systematically compared peripheral nerve MTR and MTsat values in terms of their relationships with myelin-specific PSRs, scan-rescan repeatability, and ability to differentiate heathy from dysmyelinated nerves, with the goal of establishing biomarkers of disease progression for future longitudinal studies in CMT1A.

Methods

The thigh of 13 controls (21-43 y.o., 5 female) and five CMT1A patients (19-60 y.o., 3 female) was imaged on a 3.0-T MRI Philips scanner. MT data were acquired in each subject with a MT saturation pulse (20-ms sinc-Gauss at 1, 2, 4, 8, and 16 kHz off-resonance and 350°, 600°, and 800° effective flip angles), segmented EPI readout (5 lines/shot), and water-selective excitation pulse (1331, 6°). Additional parameters included: TR/TE=52-55/12.5 ms, SENSE factor=1.5, resolution=1.0x1.0x6.0mm³, and field-of-view=192x192x96mm³. A series of T₁-weighted scans (6-30°, TR/TE=25/12.5 ms) were also acquired using the same sequence without an MT pulse. In addition, B₀ was measured using a dual-gradient echo method¹², and B₁ map was estimated using the Bloch-Siegert method¹³.

All data were non-rigidly co-registered using advanced normalization tools (ANTs)¹⁴. MTR was calculated from the reference data (no MT pulse) and a single MT-weighted image (1 kHz off-resonance at each MT pulse angle), and then was corrected for B₁ inhomogeneities (MTRc)¹⁵. MTsat was calculated using the same images with each of the five T₁-weighted images individually^10,16 to determine the best subset of MT/T₁-weighted acquisition parameters for MTsat. PSR was estimated by fitting all MT-weighted data to a two-pool MT model¹⁷ along with independent estimates of T₁ (from all five T₁-weighted scans), B₁, and B₀. The sciatic nerve was then manually segmented, and the mean slice-wise MT metrics were calculated for comparison.

Statistical Analysis

Linear relationships between PSR and MTR/MTsat (as a function of T₁-weighted image flip angle and MT pulse angle) were evaluated using Pearson correlation in all subjects and ranked according to significant differences in Fisher-transformed correlation coefficient z-tests. The signal to noise ratio (SNR) was calculated for each estimated metric as the ratio of mean ROI values to the standard deviation across slices. The MT metrics were ranked based on SNR with ties assigned by the Wilcoxon signed-rank test. To evaluate scan-rescan repeatability, the relative scan-rescan limits of agreement were estimated based on two scans acquired in five control subjects (5-166 days between scans)^18,19. Finally, clinical applicability of the metrics was assessed by evaluating significant differences between CMT1A patients and controls using the Wilcoxon rank-sum test. The effect size was assessed according to Cohen’s D calculated as the mean difference between groups divided by the standard deviation, which was assumed to differ across groups. Statistical significance was defined at the p<0.05 level.

Results

Figure 1 shows representative parameter maps. Note the reduced sensitivity of MTsat to B₁ variations (see posterior regions) relative to MTR, especially at lower MT pulse angles. All MTR and MTsat metrics significantly correlated with PSR (MTR without B₁ map correction was weakest: r=0.71, p=0.001; MTsat when the T₁ flip angle was 30° was the strongest: r=0.94, p<0.001, Figure 2). A saturation effect leading to non-linearities between MTsat and PSR was observed at higher values; however, this was reduced for larger T₁ flip angles. The relationship between PSR and MTR/MTsat was largely consistent across MT pulse angles/offset frequencies. SNR was highest for the MTR methods and lowest for MTsat with T₁ flip angle 6° (Figure 3). SNR significantly increased for MTR at higher powers, while no difference was observed across MT powers for MTsat. Additionally, B₁ correction schemes had no significant effect on correlation to PSR or SNR for MTsat (data not shown). In terms of repeatability, all MT metrics showed acceptable levels of consistency across scans (Table 1). Finally, all MT metrics were significantly different between controls and CMT1A patients (Figure 3).

Discussion and Conclusion

This study found that all MT metrics could differentiate healthy from dysmyelinated nerves. MTR was found to have higher precision than the other MT metrics and the narrowest limits of agreement but demonstrated the weakest relationship with PSR. In contrast, MTsat had the strongest correlation with PSR when T₁ flip angle was 18-30° and achieved acceptable levels of precision and repeatability. In addition, MTsat was found to have reduced B₁ sensitivity relative to MTR, although some sensitivity was observed at larger MT pulse angles. As a result, we are actively investigating more robust B₁ correction strategies for MTsat. Future work will also include acquisition in a larger, longitudinal patient cohort to evaluate relationships between each MT metrics and disability and disease progression.

Acknowledgements

This work was supported in part by an NIH/NCATS R21 TR003312 and Philips Healthcare.

References

1. Dortch RD, Dethrage LM, Gore JC, et al. Proximal nerve magnetization transfer MRI relates to disability in Charcot-Marie-Tooth diseases. Neurology 2014;83(17):1545–1553.
2. Samsonov A, Alexander AL, Mossahebi P, et al. Quantitative MR imaging of two-pool magnetization transfer model parameters in myelin mutant shaking pup. Neuroimage 2012;62(3):1390–1398.
3. Berry I, Barker GJ, Barkhof F, et al. A multicenter measurement of magnetization transfer ratio in normal white matter. J. Magn. Reson. Imaging 1999;9(3):441–446.
4. Henkelman RM, Huang X, Xiang Q-S, et al. Quantitative interpretation of magnetization transfer. Magn. Reson. Med. 1993;29(6):759–766.
5. Stanisz GJ, Odrobina EE, Pun J, et al. T1, T2 relaxation and magnetization transfer in tissue at 3T. Magn. Reson. Med. 2005;54(3):507–512.
6. Romero IO, Sinha U. Magnetization transfer saturation imaging of human calf muscle: Reproducibility and sensitivity to regional and sex differences. J. Magn. Reson. Imaging 2019;50(4):1227–1237.
7. Underhill HR, Rostomily RC, Mikheev AM, et al. Fast bound pool fraction imaging of the in vivo rat brain: Association with myelin content and validation in the C6 glioma model. Neuroimage 2011;54(3):2052–2065.
8. Gochberg DF, Kennan RP, Gore JC. Quantitative studies of magnetization transfer by selective excitation and T1 recovery. Magn. Reson. Med. 1997;38(2):224–231.
9. Schmierer K, Tozer DJ, Scaravilli F, et al. Quantitative magnetization transfer imaging in postmortem multiple sclerosis brain. J. Magn. Reson. Imaging 2007;26(1):41–51.
10. Helms G, Dathe H, Kallenberg K, Dechent P. High-resolution maps of magnetization transfer with inherent correction for RF inhomogeneity and T1 relaxation obtained from 3D FLASH MRI. Magn. Reson. Med. 2008;60(6):1396–1407.
11. Lema A, Bishop C, Malik O, et al. A Comparison of Magnetization Transfer Methods to Assess Brain and Cervical Cord Microstructure in Multiple Sclerosis. J. Neuroimaging 2017;27(2):221–226.
12. Skinner TE, Glover GH. An extended two-point dixon algorithm for calculating separate water, fat, and B0 images. Magn. Reson. Med. 1997;37(4):628–630.
13. Sacolick LI, Wiesinger F, Hancu I, Vogel MW. B1 mapping by Bloch-Siegert shift. Magn. Reson. Med. 2010;63(5):1315–1322.
14. Vercauteren T, Pennec X, Perchant A, Ayache N. Diffeomorphic demons: Efficient non-parametric image registration. Neuroimage 2009;45(1, Supplement 1):S61–S72.
15. Ropele S, Filippi M, Valsasina P, et al. Assessment and correction of B1-induced errors in magnetization transfer ratio measurements. Magn. Reson. Med. 2005;53(1):134–140.
16. Karakuzu A, Boudreau M, Duval T, et al. qMRLab: Quantitative MRI analysis, under one umbrella. J. Open Source Softw. 2020;5(53):2343.
17. Underhill HR, Yuan C, Yarnykh VL. Direct quantitative comparison between cross-relaxation imaging and diffusion tensor imaging of the human brain at 3.0 T. Neuroimage 2009;47(4):1568–1578.
18. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat. Methods Med. Res. 1999;8(2):135–160.
19. Bland JM, Altman DG. Agreement Between Methods of Measurement with Multiple Observations Per Individual. J. Biopharm. Stat. 2007;17(4):571–582.