Quantitative susceptibility and T1 ρ mapping of knee articular cartilage at 3T

Abstract

T1 $\rho$ and Quantitative Susceptibility Mapping (QSM) are evolving as substrates for quantifying the progressive nature of knee osteoarthritis.

Objective: To evaluate the effects of spin lock time combinations on depth-dependent T1 $\rho$ estimation, in adjunct to QSM, and characterize the degree of shared variance in QSM and T1 $\rho$ for the quantitative measurement of articular cartilage.

Design: Twenty healthy participants (10 ​M/10F, 22.2 ​± ​3.4 years) underwent bilateral knee MRI using T1 $\rho$ MAPPS sequences with varying TSLs ([0-120] ms), along with a 3D spoiled gradient echo for QSM. Five total TSL combinations were used for T1 $\rho$ computation, and direct depth-based comparison. Depth-wide variance was assessed in comparison to QSM as a basis to assess for depth-specific variation in T1 $\rho$ computations across healthy cartilage.

Results: Longer T1 $\rho$ relaxation times were observed for TSL combinations with higher spin lock times. Depth-specific differences were documented for both QSM and T1 $\rho$ , with most change found at ∼60% depth of the cartilage, relative to the surface. Direct squared linear correlation revealed that most T1 $\rho$ TSL combinations can explain over 30% of the variability in QSM, suggesting inherent shared sensitivity to cartilage microstructure.

Conclusions: T1 $\rho$ mapping is subjective to the spin lock time combinations used for computation of relaxation times. When paired with QSM, both similarities and differences in signal sensitivity may be complementary to capture depth-wide changes in articular cartilage.

Keywords: Arthritis; Articular cartilage; Magnetic resonance imaging; Microstructural integrity; QSM; T1ρ.

Fig. 1

Sampled voxelwise QSM and T1$\mathit{\rho }$computation output within the knee articular cartilage. (A) Voxel-by-voxel Quantitative Susceptibility Mapping (QSM) of the articular cartilage was derived using the phase and magnitude images, for each knee. Prior to voxelwise QSM computation (2), the magnitude image for each knee was co-aligned with an atlas using a series of linear and non-linear registrations to augment manual segmentation (1) of the bony (black; fibula, femur, patella, tibia) and cartilaginous structures (femoral, red; lateral tibial condyle, green; medial tibial condyle, cyan blue; patellofemoral, yellow). (B–E) Voxel-by-voxel T1$\rho$ mapping was done in alignment with the magnitude image to allow for co-localized sampling of region-of-interest measurements post pre-processing. A sample matching sagittal slice for each spin lock time (TSL) combination (i.e., T1$\rho$-short (B), T1$\rho$-extended (C), T1$\rho$-extended plus (D), T1$\rho$-extended minus (E) and T1$\rho$-chalian-like (F)), is shown on the left-hand side, as well as the resultant voxelwise T1$\rho$ maps on the right-hand side.

Fig. 2

Resultant mono-exponential regression for$\mathit{T}\mathbf{1}\mathit{\rho }$relaxation time (ms) computation based on the combination of spin lock times selected. (A-E) show the averaged mean articular cartilage signal within the knee per echoes (y) plotted against their respective spin lock times (x; ms) for all 40 acquired datasets (20 participants with bilateral knee dataset acquired sequentially), for each spin lock time combination studied; (A) V-SHORT (spin locks ​= ​0, 10, 20, 30, 40, 50 ​ms), (B) V-EXTENDED (spin locks ​= ​0, 10, 30, 60, 90, 120 ​ms), (C) V-EXTENDED(PLUS) (spin locks ​= ​0, 10, 20, 30, 40, 50, 60, 90, 120 ​ms), (D) V-EXTENDED(MINUS) (spin locks ​= ​0, 10, 30, 60, 90 ​ms), and (E) V-CHALIAN-LIKE (spin locks ​= ​0, 10, 40, 90 ​ms). A patch faded over the mean curve represents one standard deviation for each group averages. The individual mono-exponential regressions for each dataset are also plotted in the background. The group averaged whole cartilage $T1\rho$ and signal to noise quotient is listed under each plot.

Fig. 3

Three-dimensional anatomical landmarking for tibiofemoral cartilage parcellation and definition of weight-bearing region-of-interest. Combined anatomical landmarks described in Moran et al., 2022 for the systematic mapping of knee magnetic resonance imaging (A-E) were combined in three dimension to create a three-dimensional anatomical labelling system for articular cartilage of the knee (F). These include coronal (A) and sagittal (B-E) labels from the Whole-Organ MRI Scoring mapping system, as well as the International Cartilage Repair Society method for mapping cartilaginous lesions. The anatomical labels associated with weight-bearing areas of the distal femur (F∗) were combined to mask in weight-bearing articular cartilage overlying the femoral condyles (G, blue). This was combined with the whole tibial cartilage (G, red) to create the resultant three-dimensional tibiofemoral weight bearing region of interest for layer-based analysis. A ​= ​anterior; C ​= ​central; L ​= ​lateral; LSs ​= ​lateral subspine; LT ​= ​lateral trochlea; M ​= ​medial; MSs ​= ​medial subspine; MT ​= ​medial trochlea; N ​= ​notch; P ​= ​posterior; T ​= ​trochlear.

Fig. 4

Depth-based subject-specific sign-distance segmentation of the articular cartilage. Circles were fit along the sagittal slices of segmented articular cartilage (femur sampled here) using least-square fits (A-B) to establish the three-dimensional centroid (B∗) and most anterior-proximal point (B+), which became the minimum radius of the overall three-dimensional cylinder covering whole cartilaginous structure. From there, linear rays were extended in a circular fashion with increments of 1° (green) to establish equal angular bins (C). For each angular bin, the contained voxels were tagged based on the signed distance function between it and the cortical edge of the bone (D), to approximate the depth within cartilage.

Fig. 5

Depth-based T1$\mathit{\rho }$and QSM measurements within the healthy knee cartilage and associated statistical analysis. (A) The mean T1$\rho$ for each depth bin is plotted for each TSL combination along with its respective group standard deviation. All curves transition from higher distance (ie, superficial cartilage) to low (ie, deep cartilage). Each curve is shown individually below, with its matching labelled inflection point (dotted line). (B) The statistical matrices showing the results from the follow up t_-test comparing the resulting T1$\rho$ across TSL combinations for each bin distance (number), respectively. Yellow represents statistical significance. (C) Mean (with standard deviation) QSM signal across depth bins of the articular cartilage. Inflection point landmarked (dotted line).

Fig. 6

Direct linear correlation between T1$\mathit{\rho }$and QSM. The individual T1$\rho$ (x) and susceptibility (y) for each subject and bin distance are plotted, along with a linear best fit line for which the squared correlation coefficient is labelled.