The composite relaxation metric R2-R1ρ, a potential MR imaging biomarker for detecting early cartilage degeneration, was shown as an incomplete orientation-dependent R2 at 3T. Both R2 and R1ρ mappings of femoral cartilage were performed on two subjects, and the constructed composite metrics were compared to the anisotropic R2 values that were extracted from a single 3D T2W dataset using a newly developed method. The preliminary results demonstrated that the orientation-dependent information derived from two completely different methods was comparable, implying that the diagnostically most relevant information in knee or other joint cartilage could be easily and efficiently obtained.
Introduction
A composite metric R2−R1ρ was recently proposed as a potential MR imaging biomarker for detecting early cartilage degeneration1. However, this new metric necessitates lengthy data acquisitions and subsequent involved image post-processing. In this work, we show that the proposed metric basically measured an incomplete anisotropic R2, and compare the measured R2−R1ρ to anisotropic R2 of femoral cartilage in two subjects, where the latter was derived using an efficient and a simple method that was recently developed2.Theory and Method
Free water proton longitudinal relaxation rate in the rotating frame R1ρ reduces to transverse relaxation rate R2 when molecular rotational correlation time τc is in the range of picoseconds (ps)3. In ordered tissues like cartilage, some water molecules are restricted to preferential orientations, resulting in an increased τc to a range of micro- to milliseconds (μs-ms)4. In this slow rotation scenario, R1ρ could be truncated to R2/(1+4ω21τ2c) by neglecting insignificant contributions from high-frequency dipolar interactions, where ω1 is the spin-lock RF amplitude.
Non-zero averaging dipolar interaction was revealed as the dominant relaxation mechanism in cartilage at 3T5. Given fast exchange between free and bound water, R2 could be ascribed to an isotropic (Ri2) and an anisotropic (R(θ)a2=Ra2∗((1−3cos2θ)/2)2) contributions6. Likewise, R1ρ can be characterized with Ri2 and R(θ)a2/(1+4ω21τ2c) , assuming that Ri2 and R(θ)a2 are, respectively, induced by ps and µs-ms time-scale dynamics. Hence, R2−R1ρ can be written as R(θ)a2(4ω21τ2c/(1+4ω21τ2c)) or R(θ)a2 when ω1τc>>1.
R2 mapping was acquired using an interleaved multi-slice multi-echo TSE sequence using TE of n*6.1 ms, n=1 to 8. The reconstructed voxel size was 0.24*0.24*3.00 mm3. TR was 2.5 s and total scan time was 9 minutes. R1ρ mapping was collected using a T1-enhanced 3D TFE sequence with a segmented k-space scheme. A spin-lock RF strength (ω1/2π) was 500 Hz with duration of 0, 10, 20, 30 and 40 ms, respectively. Scan time was 11 minutes with TR/TE of 12.0/6.1 ms, turbo factor of 64, number of TFE shots of 83, shot interval of 2 s.
R2 and R1ρ maps were created by pixel-wise curve-fittings to an exponential decay model with two parameters, and R(θ)a2 map was extracted from a single 3D T2W dataset (TE=48.8 ms)2. Public and in-house software was used for image co-registration and whole cartilage manual or automatic angular-radial segmentation2. Left or right knee on two subjects was collected in the sagittal plane using a Philips 3T MR scanner. All image visualization and data analysis were performed using a software developed in IDL 8.5 (Exelis Visual Information Solutions, Boulder, CO).
Results and Discussions
A schematic drawing is shown in Figure 1a to denote femoral cartilage spatial orientations and angular-radial ROI segmentations. Figure 1b presents a representative medial side T2W image from left knee with ROI segmentations. R2 and R1ρ values from these segmented ROIs in the deep and superficial zones are plotted in Fig. 2a, highlighting the reduced and less orientation-dependent R1ρ. As predicted, R2−R1ρ was predominantly smaller than R(θ)a2 as shown in Fig. 2b that resulted from the restricted spin-lock RF strength (500 Hz) used in clinical scans. Previous investigations demonstrated that cartilage R1ρ anisotropy could be completely removed when using ω1/2π > 2000 Hz7,8.
Original R2 (a, e) and R1ρ (b, f), along with composite R2−R1ρ (c, g) and derived R(θ)a2 (d, h) parametric maps from both the deep (a-d) and superficial (e-h) zones are shown in Figure 3 and Figure 4 for left and right knee, respectively. The white arrows in Figure 3 indicate the spatial location for the image shown in Fig, 1b. Compared with R2 maps, R1ρ had comparable values in both zones that were largely determined by the orientation-independent contributions.
An isotropic chemical exchange (Rex) effect hardly contributes to R2 and R1ρ at 3T5, consistent with the simulated result of 0.056 [1/s] (see ref. fig.5 caption)9. This conclusion was further supported by a recent study showing that R2 and R1ρ had very similar values for human normal knees measured at 3T and 7T10.
Consistent with the theoretical predication, the orientation-dependent information derived from R2−R1ρ and R(θ)a2 was almost indistinguishable especially in the deep zone for two subjects, suggesting that the diagnostically most relevant information in cartilage8 could be easily and efficiently obtained with a minimum cost to clinical MR scan times2.
Conclusion
A composite metric R2−R1ρ was shown as an incomplete anisotropic R(θ)a2, and preliminary comparable results supported the theoretical treatments on both R2 and R1ρ. The potential to significantly reduce clinical MR scan times in deriving diagnostically most relevant information in cartilage warrants further evaluations and validations for the newly developed method in large clinical studies.