Introduction

Inhomogeneous magnetization transfer(ihMT) aims to quantitatively image myelin but is known to exhibit nonzero ihMT signal in substances other than myelin¹. However, if myelin has the longest dipolar relaxation time(T_1D), we can improve myelin specificity by using T_1D filtering to filter out the short T_1D components². To measure T_1D of myelin in ex-vivo rat spinal cord, we combined ihMT with CPMG to measure the ihMT signal of myelin water(MW) and intra-/extra-cellular water(IEW), and modelled the signal by integrating the bi-component T_1D model³ with the four-pool model^4–6.

Methods

The MR data was collected using a solenoid coil on a 7T Bruker Biospec system(Bruker Biospin, Ettlingen, Germany).

MR experiments: We imaged nine formalin-fixed rat spinal cords at 38 ± 1°C, with an ihMT prepulse module combined with a CPMG readout. Three healthy rats were sacrificed, and six rats were injured with a dorsal column transection at the C5 region. The injured rats were sacrificed 3-weeks post-injury(n=3) and 8-weeks post-injury(n=3), and imaged 5mm cranial to the injury site. We performed five scans with the saturation schemes: single negative offset(S_­-­­­­), single positive offset(S₊­), two alternating-frequency dual offset(S_+- , S_-+­), and no saturation(S₀). Scan parameters were Δf = +9655Hz and -10345Hz, B_1,rms = 5.8μT, number of Hann-shaped pulses = 1440, pulse width = 1ms, interpulse spacing = 0.3ms, TR/TE = 2750/6.753ms, 48 echoes, echo spacing = TE, FOV = 1x1cm² , matrix = 64x128, and slice thickness = 4mm, and 2 averages. We manipulated our T_1D filter by varying τ_switch at 11 different switch times: 0.8ms, 1.6ms, 2.4ms, 4ms, 4.8ms, 6.4ms, 8ms, 9.6ms, 12ms, 16ms, and 19.2ms(Fig. 1). An extra CPMG scan with identical scan parameters using 6 averages and no presaturation was collected at room temperature. A diffusion tensor imaging(DTI) scan was acquired at 38 ± 1°C using a multi-slice spin-echo sequence. The DTI scan parameters were TR = 1050ms, TE = 21.337ms, matrix = 64x128, FOV = 1x1cm², slice thickness = 1mm, and 4 averages.

Data analysis: We used a Non-Local Means algorithm⁷ to denoise our raw images before performing non-negative least squares(NNLS)⁸, with 32 echoes to improve fitting, on the CPMG scan taken at room temperature. The resulting T₂ distribution was used to obtain the T₂ modal value and area, under each peak. The no saturation ihMT-CPMG scan was then fitted with a bi-exponential with 48 echoes, using the previous peak area and T₂ modes as initial guesses to obtain the new amplitudes and T₂ times for MW and IEW. To fit the pre-saturated ihMT-CPMG data(S₊, S_-, S_+-, S_-+), we fixed the T₂ times in the bi-exponential model with the fitted T₂ values from S₀, and used the fitted S₀ amplitudes as initial guesses to find the new amplitudes. The ihMT ratio(ihMTR = (S₊ + S_- - S_+- – S_-+)/(2*S₀)) was calculated for MW and IEW by replacing S with the fitted MW and IEW amplitudes, and the conventional ihMTR was calculated from the 1^st echo images. These three ihMTR(τ_switch) decay curves were fitted using a four-pool model including dipolar order reservoirs(Fig. 2). Note, some model parameters were taken from previous MT measurements⁹. For each pixel in our ROI, we measured the T_1D(myelin), T_1D(non-myelin), fraction of the semisolid pool related to myelin(fD), and the cross-relaxation time between MW and IEW(T_cr). Myelin water fraction(MWF) was calculated by taking the ratio of the area under the MW peak(T₂<25ms) to the total signal.

Results and Discussion

The ihMTR-MW/-IEW/-1^st echo maps at each τ_switch were simultaneously fitted using our four-pool model with dipolar order reservoirs to obtain the parameter maps seen in Fig. 3. To visualize how well our model fitted the ihMTR(τ_switch) decay curves we fitted the mean ihMTRs from the ventral white matter (WM) and fasciculus gracilis(FG) ROIs, which appears to overestimate ihMTR-IEW at short τ_switch times(Fig. 4).

Fig. 5 compares our fitted parameters between the ventral WM and FG across the cord groups, and we see the fitted parameters for the ventral WM agrees for all groups. The most significantly different parameter was T_1D(myelin) in the FG of the 3 weeks post-injury cords compared to the controls, which is when myelin debris is still present. Note, we expect the myelin debris to be cleared 8 weeks post-injury. This lower T_1D(myelin) in the 3 weeks post-injury cord suggests a lower dipolar coupling strength from the myelin debris versus functional myelin.

Note, our MWF and fractional anistrotropy(FA) maps agreed with previous results of this injury model¹⁰, except for the no significant difference between the MWF in the 8 weeks post-injury versus healthy cord. We believe this is due to the partial volume effect since our images were acquired at a lower resolution.

Conclusion

Our study combined ihMT with CPMG to measure the ihMT signal in MW and IEW, and the overall ihMTR, which were fitted with our expanded four pool model to measure T_1D. We could not fully model the ihMTR-IEW at short τ_switch times, indicating our T_1D(non-myelin) may not be accurate, and the model needs improvement. We saw a significant drop in T_1D(myelin) in the 3 weeks post-injury cords compared to healthy controls, suggesting T_1D(myelin) could potentially distinguish between functional myelin and myelin debris, although more studies are warranted to confirm this.

Acknowledgements

No acknowledgement found.