Simultaneous acquisition of T1ρ and T2 quantification in knee cartilage – reproducibility and diurnal variation

Abstract

Purpose

To develop a robust sequence that combines T_1ρ and T₂ quantifications and to examine the in-vivo repeatability and diurnal variation of T_1ρ and T₂ quantifications in knee cartilage.

Materials and Methods

Six healthy volunteers were scanned in the morning and afternoon on two days using a combined T_1ρ and T₂ quantification sequence developed in this study. Repeatability of T_1ρ and T₂ quantification was estimated using root-mean-square coefficients-of-variation (RMS-CV). T_1ρ and T₂ values from morning scans were compared to those from afternoon scans using paired t-tests.

Results

The overall RMS-CV of in-vivo T_1ρ and T₂ quantification was 5.3% and 5.2% respectively. The RMS-CV of AM scans was 4.2% and 5.0% while the RMS-CV of PM scans was 6.0% and 6.3% for T_1ρ and T₂ respectively. No significant difference was found between T_1ρ or T₂ values in the morning and in the afternoon.

Conclusions

A sequence that combines T_1ρ and T₂ quantification with scan time less than 10 minutes and is robust to B₀ and B₁ inhomogeneity was developed with excellent repeatability. For a cohort with low-level daily activity, although no significant diurnal variation of cartilage MR relaxation times was observed, the afternoon scans had inferior repeatability compared to morning scans.

Introduction

Cartilage degeneration is a characteristic manifestation in pathologies such as osteoarthritis (OA). It is well established that biochemical changes of macromolecules within the cartilage matrix are among the initiating events of cartilage degeneration, and take place prior to morphologic changes of the tissue. Therefore, efforts have been made during the last decade on developing imaging techniques for cartilage matrix composition. Among the techniques developed, magnetic resonance (MR) T_1ρ and T₂ relaxation times have been proposed as promising imaging markers for early cartilage degeneration in OA (1, 2).

MR T_1ρ and T₂ quantification in cartilage may provide complementary information associated with cartilage degeneration during OA due to their different relaxation mechanisms. In particular, while some studies show T_1ρ is more sensitive than T₂ (which is more dominated by hydration and collagen changes) to proteoglycan loss in cartilage (3, 4), some others suggest both T_1ρ and T₂ are sensitive to changes in proteoglycan contents in OA cartilage (5, 6). The sensitivity of T_1ρ and T₂ to proteoglycan changes in cartilage matrix is still under investigation. However, studies investigating voxel-by-voxel relationship between T_1ρ and T₂ are very limited (7). One difficulty is these two parameters are normally acquired with separate sequences and the correlation study may be sensitive to motion and potential variation caused by different sequence acquisition. It has been documented that T₂ quantifications were significantly different between using spin-echo, fast spin-echo (FSE), spoiled gradient echo (SPGR) and magnetization-prepared sequences (8). The T₂ values are also significantly higher when longer echo train length (ETLs) are used in FSE acquisitions (8). Furthermore, T_1ρ and T₂ quantitative techniques normally need a long acquisition time because multiple T_1ρ- or T₂-weighted images need to be acquired to reconstruct the relaxation time map. Techniques towards shortening the overall acquisition time are desirable. A relatively short acquisition time (normally less than 10 minutes) is also critical to make these techniques feasible for clinical applications.

A couple of previous studies examined the diurnal variation in cartilage morphology (thickness and volume) in the knee joint (9, 10). However, no study has been documented yet on the potential diurnal variation of relaxation times in knee cartilage. The goals of this study were therefore three folds: 1) To develop a robust sequence that combines T_1ρ and T₂ quantification; 2) To examine the in-vivo repeatability of T_1ρ and T₂ quantification; 3) To examine the potential diurnal variation of T_1ρ and T₂ quantification in cartilage in young healthy adults.

Methods

Sequence development

The sequence was developed based on the MAPSS (magnetization-prepared angle-modulated partitioned k-space spoiled gradient echo snapshots) T_1ρ quantification sequence developed previously (11). The T_1ρ preparation pulses contained continuous hard 90_x (tip down pulses) - spin lock pulses - 90°_-x (tip up pulses). Multiple k-space lines (views per segmentation, VPS) are acquired immediately after each magnetization preparation. RF cycling was applied to eliminate the adverse impact of longitudinal relaxation on quantitative accuracy. This RF cycling scheme also yields a transient signal evolution that is independent of the prepared magnetization, and consequently a same variable flip angle train can be applied to provide a flat signal response to eliminate the filtering effect in k-space caused by transient signal evolution after each spin-lock (11). In this work, we added the following features to this sequence:

Improving the sequence robustness to B₁ and B₀ inhomogeneity

In the original MAPSS sequence, the phase of the second half of the spin-lock pulse was shifted 180° from the first half to reduce artifacts caused by B₁ inhomogeneity (12, 13). However, significant banding artifacts are often observed when there is B₀ inhomogeneity due to the off-axis rotation after the tip-down pulses. In this work, composite pulses, as originally proposed by Dixon et al. (14) were used. The tip-down/tip-up 90° hard pulses were followed and preceded by a hard 135° pulse with the opposite phase (Figure 1A). The composite tip down pulses will bring the magnetization near the spin-lock axis, and the composite tip up pulses bring the magnetization back near the longitudinal (z) axis even under the condition of a large off resonance (14).

Figure 1.

(A) Pulse diagram of the combined T_1ρ/T₂ quantification based on the modified MAPSS sequence. Either the T_1ρ or the T₂ preparation could be run for any echo, and the T_1ρ- and T₂-weighted images will be acquired in one scan sequence. Combining the composite pulses and RF-cycling can simultaneously correct B₀ magnetic field inhomogeneity effects and B₁ RF inhomogeneity effects. (B) The scheme of the k-space trajectory used in the sequence based on radial-centric view ordering.

Combining T_1ρ and T₂ weighting

A combined T_1ρ and T₂ sequence was designed containing an interchangeable preparation, in which either the T_1p or the T₂ preparation could be run for any echo, and the T_1ρ- and T₂-weighted images will be acquired in one scan sequence. The T₂ preparation contained an MLEV (Malcolm Levitt's composite-pulse decoupling sequence) train of nonselective composite 90°_x180°_y90°_x refocusing pulses (with the pulse duration of 1.624 ms) (15). In the case of the first TSL = 0 and first TE = 0, T_1ρ and T₂-weighted images share the first image.

Applying radial-centric view ordering to minimized eddy current effects

Multiple k-space lines are acquired after each magnetization preparation, traversing in segmented radial-centric view ordering (16), as shown in Figure 1B, to minimize the eddy current effect (as compared to the segmented elliptic-centric order in the previous MAPSS sequence (11)).

The combined T_1ρ/T₂ sequence was validated in phantoms and control subjects by comparing the results to the published MAPSS sequence (11).

Subjects

Six healthy volunteers (age 22-35 years, 3 females) were recruited for this study. The scans were performed using a GE 3T scanner (Signa HDx, GE Healthcare, Milwaukee, WI) and an 8-channel phased array knee coil (Invivo, Gainsville, FL). Each subjects were scanned in the morning (between 8-10am) and in the afternoon (between 5-7pm), and scanned again on a second day after one week using the same protocol, resulting in four scans per subject (namely AM1, PM1, AM2, PM2). Each subject filled out modified International Physical Activity Questionnaires (IPAQ) for their activity levels for the past year (before AM1 scan), 1 week (before AM1 and AM2), and 1 day (before PM1 and PM2). The study was approved by the Committee for Human Research at our institution, and the informed consent was obtained from all subjects.

Imaging Protocols

The in-vivo imaging protocol included sagittal T₂-weighted fat-saturated fast spin-echo (FSE) images (TR/TE = 4300/51 ms, FOV = 14 cm, matrix = 512 × 256 slice thickness = 2.5 mm, gap = 0.5 mm, echo train length (ETL) = 9, bandwidth = 31.25 kHz, NEX = 2), sagittal 3D fat-saturated high-resolution spoiled gradient-echo (SPGR) images (TR/TE = 15/6.7 ms, flip angle = 18, FOV = 14 cm, matrix = 512 × 512, slice thickness = 1 mm, bandwidth = 31.25 kHz, NEX = 1), and the T_1ρ/T₂ quantification sequence (FOV = 14 cm, matrix = 256 × 128, slice thickness = 4 mm, Views Per Segment = 64, time of recovery = 1.2 s, number of slices = 26; For T_1ρ: TSL = 0/10/40/80 ms, spin-lock frequency = 500 Hz; for T₂: preparation TE = 0/13.7/27.3/54.7 ms; total acquisition time = 9 mins 30 sec). Parallel imaging ASSET (Array Spatial Sensitivity Encoding Technique) with acceleration factor (AF) as 2 was used for FSE and SPGR images. ARC (Autocalibrating Reconstruction for Cartesian imaging) parallel imaging with phase direction acceleration factor of 2 was used in T_1ρ/T₂ sequences. In each set of in-vivo scanning, phantoms with different concentrations (1% to 4%) of agarose were scanned with the T_1ρ/T₂ quantification sequences using the same parameters.

Image and Data Analysis

Cartilage of the lateral/medial femoral condyles (LFC/MFC), the lateral/medial tibia (LT/MT) and patella (P) were segmented semi-automatically in the SPGR images using in-house developed software (17). T_1ρ and T₂ maps were reconstructed by fitting the T_1ρ- and T₂-weighted images voxel-by-voxel to the equations below:

$$S\left(\mathit{\text{TSL}}\right)\propto exp(-\mathit{\text{TSL}}/T_{1\rho })$$ [1]

$$S\left(\mathit{\text{TE}}\right)\propto exp(-\mathit{\text{TE}}/T_2)$$ [2]

The images acquired in the combined T_1ρ/T₂ sequence with TSL/TE = 0 were rigidly registered to the high-resolution SPGR images, and the transformation matrix was applied to T_1ρ and T₂ maps using the Visualization ToolKit Computational Imaging Science Group (VTK CISG) Registration Toolkit. The cartilage contours previously generated from SPGR images were overlaid on the registered T_1ρ and T₂ maps, Figure 2. The mean and standard deviation of T1ρ and T2 relaxation times were calculated in each defined compartments.

Figure 2.

The diagram of image processing and quantification of the study. The cartilage was segmented on the high resolution SPGR images (red contours). T_1ρ- weighted images with TSL = 0 were registered to SPGR using VTK registration toolkit. The registration transformation matrix was applied to the T_1ρ or T₂ maps which were reconstructed from T_1ρ- or T₂ -weighted images. Example T_1ρ and T₂ color maps overlaid on SPGR images are also shown.

Statistical Analysis

Repeatability was estimated using root mean square coefficients of variation (RMS-CV). The T_1ρ and T₂ values from morning scans were compared to the values from afternoon scans using a paired t-test and the significance level was 0.05. Diurnal variation was quantified as percentage change from morning to afternoon: (PM-AM)/AM × 100%.

Results

Comparison of the new sequence vs. MAPSS sequence

Figure 3 shows the images of phantoms and in-vivo knees using the original MAPSS sequence and the modified sequence. Significant banding artifact was observed at the edge of the phantom due to B₀ inhomogeneity, which was diminished using the composite tip-down/tip-up pulses (Figure 3A, 3B). The signal curve from the phantom showed non-exponential signal oscillation in the data without composite pulses, which introduced a poor fitting and an underestimate of T_1ρ quantification; while the data using composite pulses had very good fitting of the exponential decay curve (Figure 3C). Figure 2D showed in-vivo T_1ρ -weighted images that were acquired using the original MAPSS sequence in a healthy control and showed significant banding artifact due to B0 inhomogeneity (Figure 3D). This banding artifact was significantly reduced when the composite pulses were applied for the tip-down/tip-up pulses (Figure 3E). Using images free of artifact from both sequences, the RMS-CV of quantified T_1ρ and T₂ in phantoms between the new sequence and the previously published MAPSS sequence was 1.3% for T_1ρ and 0.7% for T₂, respectively. The RMS-CV of quantified T_1ρ and T₂ in-vivo between the new sequence and the MAPSS sequence was 2.6% for T_1ρ and 1.9% for T₂, respectively.

Figure 3.

T1ρ-weighted images of phantoms (time of spin-lock = 40 ms, spin-lock frequency = 500 Hz) and in vivo knees (lower row, time of spin-lock = 40 ms, spin-lock frequency = 500 Hz) with single 90 tip-down/tip-up pulses (A, D) and with composite tip-down/tip-up pulses (B, E) as shown in Figure 1. The banding artifacts (at the edge of the phantom, and indicated with blue arrows in the knee) were significantly reduced in images using composite pulses. The signal curve from the phantom (C) showed non-exponential signal oscillation in the data without composite pulses, which introduced a poor fitting and underestimate of T_1ρ quantification; while the data using composite pulses had very good fitting of the exponential decay curve.

Repeatability

Phantom Results

The RMS-CV of intra-day repeatability in phantoms was 1.0% for both T_1ρ and T₂ quantification. The RMS-CV of inter-day repeatability in phantoms was 2.2% for T_1ρ and 2.0% for and T₂, respectively.

In-vivo Repeatability

The overall RMS-CV of the scan-rescan quantification of T_1ρ and T₂ was 5.3% and 5.2% respectively. When the data was separate for AM scans and PM scans, the RMS-CV of all AM scans was 4.2% and 5.0% for T_1ρ and T₂ respectively, while the RMS-CV of all PM scans was 6.0% and 6.3% for T_1ρ and T₂ respectively. The RMS-CV of each compartment ranged 2.2% - 5.9% for T_1ρ AM, 4.8-8.8% for T_1ρ PM, 4.0% - 6.2% for T₂ AM and 4.2% - 8.5% for T₂ PM, respectively, Figure 4.

Figure 4.

In-vivo T_1ρ (A) and T₂ (B) quantifications of each defined compartments in each session, and the inter-day repeatability of T_1ρ (C) and T₂ (D) quantifications. LFC: lateral femoral condyle; LT: lateral tibia; MFC: medial femoral condyle; MT: medial tibia; P: patella.

Diurnal Variation

No significant difference was found in T_1ρ or T₂ between the values obtained in the morning and in the afternoon in this cohort (P > 0.05 for both T1ρ and T2). T_1ρ and T₂ measured in AMs were significantly correlated with the measures in PMs (R = 0.81 and R = 0.85, P < 0.01). The change percentage from AM scans to PM scans ranged from -3.4% to 1.8% in the defined five compartments, with a large variation among each individual subjects (as indicated with large SD of the percentage change, Table 1).

Table 1. Variation of T_1ρ and T₂ from morning to afternoon: (AM-PM)/AM×%.

	LFC	LT	MFC	MT	Patella
T1ρ	1.8%±5.6%	-3.4%±8.7%	1.2%±7.8%	-0.5%±14.8%	-0.3%±6.3%
T2	0.6%±7.4%	-1.2%±8.6%	1.0%±6.2%	-0.6%±8.1%	-0.1%±6.2%

L/MFC: lateral/medial femoral condyle; L/MT: lateral/medial tibia

Discussion

A 3D sequence that can simultaneously quantify T_1ρ and T₂ relaxation times was developed. Concatenation of the two measurements in a single scan will allow direct (pixel-by-pixel) comparison between T_1ρ and T₂ maps by removing potential quantification variations caused by different acquisition sequences (spin-echo vs gradient-echo sequences for example) and minimizing chances of motion between these two quantifications. The developed sequence combines T_1ρ and T₂ quantification and the current study saves scan time compared to separate T_1ρ and T₂ quantification sequences by sharing the TSL = 0 and TE = 0 image. Currently the total scan time is less than 10 minutes to obtain both quantifications, which will greatly facilitate potential clinical applications of these advanced MR techniques. Further, combining T_1ρ and T₂ acquisition in one sequence will increase the parameter dimension and can potentially help with further acceleration using advanced techniques, which we will explore in the future.

In this work, the composite pulses originally proposed by Dixon et al. were used for tip-down and tip-up pulses during T_1ρ and T₂ preparation. Chen et al. showed earlier that combining the composite pulses and RF-cycling can simultaneously correct B₀ magnetic field inhomogeneity effects and B₁ RF inhomogeneity effects (image artifact and T_2ρ contamination) (18). Images from phantoms and in-vivo knees from this study confirmed that the modified sequence is robust for potential B₀ and B₁ inhomogeneity.

Our study showed that the developed sequence provided excellent in-vivo repeatability for T_1ρ and T₂ quantification. The in-vivo inter-day repeatability is comparable to the previously published intra-day repeatability (5). However, the repeatability of the PM T_1ρ and T₂ were inferior to the AM scans, which may potentially be due to the variation introduced by different activity levels during the day. Therefore, for large cohort and longitudinal studies, it may be optimal to scan subjects early in the morning to minimize potential variation in relaxation times caused by differences in activity levels and loadings during the day. An alternative strategy to minimize the loading effect is to equilibrate the loading condition right before the scan, by for example sitting in the wheel chair for more than 30 minutes.

No significant diurnal variation of cartilage MR relaxation times was found in this cohort. Previous study in young healthy volunteers (n=6 and 10 respectively) reported significant decrease cartilage thickness from mornings to evenings, especially in medial tibial and femoral condyle (9, 10). The subjects included in the study performed by Waterton et al. spent the day predominantly in the laboratory, mostly standing at a bench, which introduced quite different loading condition compared to subjects in our study. Our study cohort is composed with subjects who spent most of their day sitting (the hours for sitting ranged from 6-10 hours on the scan days based on the IPAQ questionnaires). This low level of daily activity may introduce very mild loading to the knee joint, therefore introduce minimal variation of the MR relaxation times.

Previous studies of vertebral discs demonstrated significant decrease in T₂ from mornings to evenings (19, 20) (n = 6 and 18 respectively), which can be explained by that the vertebral discs normally experience continuous loading during the day. Interestingly, Karakida et al. showed that this significant decrease of disc T₂ was only observed in healthy normal discs before the age of 35 years, but not in normal discs after 35 years and in degenerated discs (20). This result suggested that the response of disc matrix to loading is affected by aging and degeneration. Future studies for diurnal variation of knee cartilage shall include subjects who spend more time standing or in other loading-inducing activities, and explore the potential relationship between degeneration and diurnal variation.

In conclusion, a sequence that combines T_1ρ and T₂ quantification with a total scan time less than 10 minutes and is robust to B₀ and B₁ inhomogeneity was developed with excellent in-vivo repeatability. Such sequences have the great potential to be applied in future clinical trials and studies. For a cohort with relatively low level of daily activity, although no significant diurnal variation of cartilage MR relaxation times was observed, the scans in the afternoon had inferior repeatability compared to those in the morning.