Introduction

Gadolinium(Gd) retention in body after administration of Gadolinium based contrast agents(GBCAs) in clinical routinely contrast enhanced MRI attracts attentions of clinicians and scientists, and bone is reported major target of Gd retention(1,2). Delayed release of Gd from bone would lead to potential chronic toxicity to body(3). Thus, assessing deposited Gd in bone would benefit clinical decision and management of this risk factor in body. To date, noninvasive assessing Gd deposition in bone is still lacking(4). In this work, we aimed to investigate the feasibility of assessing Gd deposition in rabbit cortical bone by using MRI based T1 mapping.

Method

Animals: This study was approved by local Ethics Review Board. Twenty-eight male adult rabbits(3-3.5 Kg) were involved and randomly allocated into: control group, linear GBCAs group, macrocyclic GBCAs group and high-dose macrocyclic GBCAs group(n=7 for each group), and daily administrated by 0.9ml/kg bodyweight saline, 0.3 mmol/kg bodyweight Magnevist, 0.3mmol/kg bodyweight Gadovist and 0.9 mmol/kg bodyweight Gadovist, respectively, for 5 consecutive days per week over period of 4-weeks. Rabbits were then allowed for 4-week rest and sacrificed for left-tibia collection(Figure 1).
Sample preparation: Harvested tibias were cleared of external muscle and soft tissues, and bone marrows were removed by a scalpel. Tibias were cut into segments with approximate length of 4 cm, and then froze at -20℃ in an freezer for short-term storage. Samples were allowed to thaw in phosphate buffered saline solution at 4℃ for 24 hours prior MR imaging. Each bone sample was immersed in Fomblin in a 5 ml centrifuge tube during MR imaging to minimize susceptibility effects at bone-air interface.
MR imaging: MRI data were acquired with volumetric transceiver coil on a 7T animal MR scanner(PharmaScan, Bruker BioSpin, Ettlingen, Germany) whose equipped gradient system have a maximum gradient strength of 380 mT/m and a maximum slew rate of 3420 mT/m/ms. Non-selective rectangular pulse with maximum available power of 329 W and duration of 10 us was employed for signal excitation. Central-out radial trajectory based 3D UTE sequence, whose data sampling started at k-space center and acquired along gradient ramp via non-linear sampling, with a nominal TE of 11 us was implemented for signal readout. 3D UTE with variable repetition time(TR) was implemented for T1 quantification according to VTR-UTE scheme. Details of MR parameters were: TR = 4, 8, 16, 32, 64, 100, 200, 400 ms, TE = 11 us, nominal flip angle = 35 degree, field of view = 40Х40Х60 mm3, matrix = 128Х128Х128, under-sample factor = 1.5, number of projections = 34242, band-width = 500kHz,. For B1 field correction, 3D UTE based AFI which interleaved acquired signals at two TRs, i.e. TR1 = 20 ms and TR2 = 100 ms, was implemented with same excitation pulse and geometry parameters as VTR-UTE.
Data analysis: Flow chart of MR images analysis is shown in Figure 2. 3D VTR-UTE images and 3D AFI-UTE images were reconstructed from acquired k-space data by non-uniform fast Fourier transform(5). Bone T1 maps were generated according to articles by Ma et.al(6).
First, longitudinal magnetization mapping function $$$f_{z\left ( \alpha ,\tau ,T_{2} \right )}$$$ was estimated in 3D AFI-UTE according to:
$$f_{z\left ( \alpha ,\tau ,T_{2} \right )}\approx \frac{rn-1}{n-r}$$
Where $$$r$$$ is ratio of steady state signal acquired in $$$TR_{2}$$$ and $$$TR_{1}$$$, $$$n$$$ is the ratio of $$$TR_{2}$$$ and $$$TR_{1}$$$.
Second, because of same RF pulse was implemented in both VTR-UTE and AFI-UTE, estimated longitudinal magnetization mapping function therefore can be substituted to signal expressions of VTR-UTE, which is same as signals of steady-state spoiled gradient echo(SPGR) sequence:
$$S=M_{0}f_{xy}\left ( \alpha ,\tau ,T_{2} \right )\frac{1-E}{1-Ef_{z}\left ( \alpha ,\tau ,T_{2} \right )}$$
Where $$$E=e^{-\frac{TR}{T_{1}}}, f_{xy}\left ( \alpha ,\tau ,T_{2} \right )$$$ denotes the transverse magnetization mapping function. Combining TR-independent parameters $$$M_{0}$$$ and $$$f_{xy}\left ( \alpha ,\tau ,T_{2} \right )$$$ into single unknown parameters, T1 can be determined through non-linear least square (Levenberg-Marquardt(7,8)) optimization by in-house developed scripts in Matlab(Matlab 2014, MathWorks, Natick, MA). Fitting quality was indicated by the coefficient of determination $$$R^{2}$$$.
Regions of interest(ROIs) was first delineated by region-growing algorithm to semi-automatic generate mask over interested bone, and pixels with fitting quality $$$R^{2}<0.98$$$ was neglected. Histogram of T1 values was generated according to generated mask, and mean of center 80% T1 in the histogram was considered as T1 values for the interested bone at current slice. Same operations was repeated for other center slices and total 30 slices were analyzed. Mean T1 value over these 30 slices was considered as the T1 value for the interested bone.

Result

Figure 3 shows representative images at different TR, and quantified T1 map and R2 map. Figure 4 plots bone T1 values of all specimens. Significant lower bone T1 value was observed in macrocyclic group(323±22,P<0.05), high-dose macrocyclic group(305±28ms,P<0.01) and linear GBCAs group(290±22ms,P<0.001) than that in control group(349±19ms). In addition, T1 values in linear GBCAs group was observed significant higher than that in macrocyclic group(P<0.05).

Discussion & Conclusion

Preliminary results demonstrated bone T1 values in GBCAs exposure group is significant lower than that in control group, suggested T1 values might be a potential biomarker for evaluating Gadolinium deposition. Further research is required to investigate the correlation between quantitative Gd assessment by inductively coupled plasma mass spectrometry and T1 values.