Early detection of meniscal degeneration is essential for developing and targeting treatment strategies for osteoarthritis [1]. Quantitative magnetization transfer (QMT) may be particularly useful for this purpose because it probes macromolecules in tissues, which degenerate in osteoarthritis. To validate QMT metrics, comparisons to histology and biochemical markers must be carried out. Either whole cadaver knee specimens or menisci obtained from total knee replacement surgeries are most often used. The latter are easy to procure and are comparatively inexpensive, but tissue-air susceptibility artefacts are an issue; as such, these specimens are often scanned in perflubron (a fluid without hydrogen) to avoid compromising the dynamic range in the image. The aim of this work was therefore to compare QMT parameters of menisci in situ (whole cadaver specimen) and ex situ (menisci only, in perflubron).

Meniscus MT data were acquired for three cadaver knee specimens at 3T (GE MR 750, GE Healthcare, Waukesha, WI) using a 16-channel receive-only coil in situ and a single channel surface coil ex situ with a SPGR sequence (FA=10°, TR=48 ms, 0.625x0.625 in-plane resolution, 3 and 2 mm slice thickness in situ and ex situ) and a MT sensitizing pre-pulse (Fermi pulse, duration 8 ms). MT pulse flip angles of 300, 450° and offset frequencies of 1.5, 3.5, 7.8, 17.8 and 40.5 kHz were used for 10 total samples in the Z-spectrum. The required observed $$$T_1$$$ ($$$T_{1obs}$$$) and $$$B_1$$$ variations were mapped with the DESPOT1 [2] and dual angle methods [3], respectively. $$$T_2$$$ data was acquired [4] and mapped [5]. The menisci from each cadaver knee were excised and submerged in perflubron. Medial and lateral menisci were segmented manually, slice-by-slice, for analysis purposes.

For QMT mapping, custom software (MATLAB® 2015b, the MathWorks, Natick, MA) was written based on the rectangular pulse model of off-resonance saturation [6] modified for the Fermi pulse shape. A voxel-wise Levenberg-Marquardt fit was used to find the best QMT parameters. Overfitting of the QMT parameters occurred when using 5 free QMT model parameters with 10 measurement points, so two strategies were used to obtain robust estimation. First, the number of free parameters was reduced to 4 by fixing the $$$T_{1f}/T_{2f}$$$ ratio to 55 and 80 for in situ and ex situ specimens respectively, based a priori knowledge of $$$T_{1obs}$$$ and $$$T_2$$$ measurements. Second, in the fitting process, a regularization term $$$\lambda || k_f ||_2 $$$ (L2-norm of the exchange rate, $$$k_f$$$) was added to penalize large, physically unrealistic values of $$$k_f$$$. The regularization parameter $$$\lambda$$$ was set to 0.001 to balance the bias introduced by regularization and the variability of $$$k_f$$$. The final QMT parameters calculated were: the restricted pool fraction $$$f$$$, exchange rate $$$k_f$$$, $$$T_1$$$ relaxation time of the free pool ($$$T_{1f}$$$) and $$$T_2$$$ relaxation time of the restricted ($$$T_{2r}$$$) pool. Differences in QMT parameters between the in situ and ex situ cases and lateral and medial menisci were determined using Friedman’s two-way non-parametric test.

Mean $$$f$$$, $$$T_{1obs}$$$ and $$$T_{1f}$$$ varied significantly between the in situ and ex situ case, while mean $$$k_f$$$, $$$T_{2r}$$$ and $$$T_2$$$ were relatively constant (Figure 1). Representative QMT parameter maps are presented in Figure 2. Values of $$$f$$$ were systematically lower (p<0.01) in the ex situ samples than in situ samples, and consistently lower in the medial menisci (p=0.03). $$$T_{1obs}$$$ and $$$T_{1f}$$$ in the menisci increased in all specimens after excision (p<0.01 for both). $$$T_{2r}$$$ showed no obvious variation after extraction or between lateral and medial samples. A small trend towards lower values of $$$k_f$$$ after excision was noted, but the high variability of $$$k_f$$$ may limit the interpretation of this change.We developed an imaging and post-processing protocol for QMT parameter mapping in the meniscus and showed that some QMT parameters vary between the in situ and ex situ cases. Estimates of $$$f$$$, $$$T_{1f}$$$ and $$$T_{2r}$$$ were robust, but our confidence in estimates of $$$k_f$$$ is slightly lower due to the regularization and the fixed $$$T_{1f}/T_{2f}$$$ ratio. $$$T_2$$$ values (between 9-16 ms) were similar to previously published $$$T_2$$$ values in the meniscus [7-9] (average of 11 ms). In the future a greater number of Z-spectrum points may need to be collected or the particular points collected optimized. Variations in QMT parameters may be due to the diffusion of perflubron into the ex situ samples; further studies of perflubron dynamics are necessary to understand the observed differences. Overall, QMT may be a more sensitive measure of meniscal degeneration because it directly probes the changes in the macromolecular pool ($$$f$$$, $$$k_f$$$, $$$T_{2r}$$$) that occur in osteoarthritis.