Initial experience with synthetic MRI of the knee at 3T: comparison with conventional T₁ weighted imaging and T₂ mapping

Abstract

Objective:

To assess the feasibility and accuracy of synthetic MRI compared to conventional T₁ weighted and multi-echo spin-echo (MESE) sequences for obtaining T₂ values in the knee joint at 3 Tesla.

Methods:

This retrospective study included 19 patients with normal findings in the knee joint who underwent both synthetic MRI and MESE pulse sequences for T₂ quantification. T₂ values of the two sequences at the articular cartilage, bone marrow and muscle were measured. Relative signal intensity (SI) of each structure and relative contrast among structures of the knee were measured quantitatively by T₁ weighted sequences.

Results:

The mean T₂ values for cartilage and muscle were not significantly different between MESE pulse sequences and synthetic MRI. For the bone marrow, the mean T₂ value obtained by MESE sequences (124.3 ± 3.6 ms) was significantly higher than that obtained by synthetic acquisition (73.1 ± 5.3 ms). There were no significant differences in the relative SI of each structure between the methods. The relative contrast of bone marrow to muscle was significantly higher with conventional T₁ weighted images, while that for bone marrow to cartilage was similar for both sequences.

Conclusion:

Synthetic MRI is able to simultaneously acquire conventional images and quantitative maps, and has the potential to reduce the overall examination time. It provides comparable image quality to conventional MRI for the knee joint, with the exception of the bone marrow. With further optimization, it will be possible to take advantage of the image quality of musculoskeletal tissue with synthetic imaging.

Advances in knowledge:

Synthetic MRI produces images of good contrast and is also a time-saving technique. Thus, it may be useful for assessing osteoarthritis in the knee joint in the early stages.

Introduction

MRI provides high soft tissue contrast and can be used to assess the internal derangement of joints. However, MRI scans can take a long time, and unlike CT, the signal intensities (SIs) obtained with conventional MRI are not quantitative. Therefore, the intensities of pathological processes obtained by this technique can neither be used as a comparison for follow-up examinations nor can they be compared to reference normal values.¹

SyMRI is a synthetic MRI method based on a quantitative approach in which a single saturation recovery turbo spin echo sequence is used to estimate absolute physical properties, proton density (PD), longitudinal relaxation rate and transverse relaxation rate, including correction for B1 inhomogeneities.² Rather than predetermining the acquisition parameters such as echo time (TE), repetition time (TR) and inversion time, to maximize tissue contrast, synthetic MRI produces a free range of synthetic weightings based on a single sequence through mathematical inference.^3–5 The quantitative sequences used in synthetic MRI measure inherent tissue properties (T₁, T₂ and PD), and these measurements can be used for individual patient follow-up and comparison between patients.

The clinical use of synthetic MRI has been demonstrated in the brain in various disease processes.^1,2,6–8 However, there have been no reports regarding its application to the musculoskeletal system. In this study, we assessed the image contrast of synthetic MRI compared to corresponding conventionally obtained clinical MRIs of the knee joint. This study was performed to assess the T₂ values in the knee, as measured by synthetic MR sequences compared to 2D fast spin-echo (FSE) multi-echo spin-echo (MESE) sequences. The secondary aim was to evaluate synthetic MRI in a clinical setting by assessing the relative SI and contrast compared to conventional T₁ weighted images.

Methods and Materials

Case selection

This study received Institutional Review Board approval, and the requirement for informed consent was waived. The study population consisted of 218 consecutive patients who had undergone knee MRI from March 2016 to May 2016. The inclusion criteria were normal knee MRI findings in the routine protocol sequences with additional synthetic sequences, comparable conventional T₁ weighted sequences and 2D FSE MESE sequences used for T₂ mapping. Patients with the following abnormal findings were excluded: osteoarthritis with or without meniscal degeneration or tear (n = 103), substantial trauma with fracture or ligament injury (n = 44), inappropriate MR protocol (n = 23), previous operative status (n = 11), septic or inflammatory arthritis (n = 10) and bone or soft tissue tumour (n = 8). Thus, a total of 19 patients were included and evaluated in this retrospective study. The mean age was 39.3 years (range, 17–54 years); 12 patients were males and 7 were females.

MR parameters

All of the MR studies were acquired on a 3.0 T MR unit (Discovery MR750W; GE Healthcare, Waukesha, WI) using a transmit-receive quadrature knee coil (GE Healthcare). Sagittal T₁ weighted, 2D FSE MESE sequences for T₂ mapping and synthetic MR sequences were acquired during clinical MRI in addition to the conventional MRI sequences in the knee at our institution. The imaging parameters of each sequences are summarized in Table 1.

Table 1.

MR parameters

Imaging parameter	Synthetic	Conventional T1	MESE
Acquisition plane	Sagittal	Sagittal	Sagittal
Field of view (cm)	16	16	16
Matrix	320 × 224	320 × 224	320 × 224
Section thickness (mm)	4.0	4.0	4.0
Slices	24	24	24
Interslice gap (mm)	0.4	0.4	0.4
Flip angle	120	90	90
TR (ms)	4000a	640–800	800
TE (ms)	22, 90a	7	7.2, 14.4, 21.6, 28.8, 36, 43.2, 50.4, 57.6
Echo train length	12	3	1
Bandwidth (Hz/pixel)	81.37	162.77	244.14
Number of excitations	1	2	1
Parallel factor	2	1	1
Acquisition time	5 min 36 s	1 min 40 s	9 min 3 s

MESE, multi-echo spin-echo; TE, echo time; TR, repetition time.

^aAcquisition parameters. Synthetic images were generated using TR and TE matching the conventional or traditional sequences.

Synthetic MRI was performed using MAGiC, which is a customized version of SyntheticMR’s SyMRI software. The MAGiC sequences are 2D FSE multi dynamic, multi-echo sequences that use an interleaved slice-selective 120° saturation and multi-echo acquisition, and images are obtained with different combinations of TE and saturation delay time. Each acquisition led to eight complex images per section with different combinations of four saturation delays and two TEs. The acquisition time of the synthetic MR sequences was 5 min, 36 s. Sagittal conventional T₁ weighted and MESE sequences for T₂ values were acquired with section thickness and in-plane resolution matching those of the synthetic MR sequences. Imaging parameters (TR, TE) were also selected to provide visual image contrast similar to the synthetic MR images. T₂ quantification sequences were obtained using a sagittal MESE acquisition with eight TEs (7.2, 14.4, 21.6, 28.8, 36, 43.2, 50.4 and 57.6 ms). The T₂ map from MESE was obtained using dedicated software (Discovery 750w, T₂ Map; GE Healthcare). Synthetic T₁ weighted images and T₂ maps were created from raw quantification data with SyMRI software in the GE 3T scanner console. Figure 1 shows examples of a conventional T₁ weighted image, a T₂ map and synthetic MR images.

Figure 1.

Example of conventional and synthetic images of the knee. (a) Conventional T₁ weighted image (TR/TE = 650/7). (b) T₂ map from the MESE sequences (scale 25–75 ms). Standard conventional images are shown in the top row (a, b), and the corresponding synthetic images for the same patients are shown in the bottom row (c, d). Regions of interest were placed within the superior lateral trochlear cartilage (not shown), medial femoral condyle in the first full slice from the intercondylar notch for bone (1) and gastrocnemius muscle medial head (2). MESE, multi-echo spin-echo; TE, echo time; TR, repetition time.

Image analysis

Quantitative assessment was performed using a picture archiving and communication system and Advantage Workstation (GE Healthcare). In all of the subjects, the regions of interest (ROIs) were measured for cartilage, bone and muscle.^9,10 ROIs on a sagittal plane were positioned on the superior lateral trochlear cartilage, medial femoral condyle in the first full slice from the intercondylar notch for bone and gastrocnemius muscle medial head, as done in previous studies.^9,11 The circular ROI for cartilage measurement was 3 mm in diameter, and that for the bone marrow and muscle was 9 mm in diameter.¹² ROIs for each patient were identical in size and placed in identical positions on matching sections. Each measurement was performed by two radiologists with 14 and 9 years of experience, respectively. To secure reproducibility, tiny dots were marked and recorded at the point of the previously measured area by the first reader. The second reader measured the values with ROIs around the dots.¹³ We measured the mean T₂ values at each ROI on the T₂ map generated from both MESE and synthetic MR sequences, and compared the differences in T₂ values between the two sequences. The means and standard deviation of the SI within each ROI on conventional and synthetic T₁ weighted MR images were recorded. The relative SI and relative contrast were used for direct comparison of image quality between the conventional and synthetic T₁ weighted images. The relative SI of each structure was calculated as SI/SD, and the relative contrast of structure A (a) to structure B (b) was calculated as (SI_a–SI_b) / (SD_a² + SD_b²)1/2 ^14–16

Statistical analysis

To assess differences between measurements of T₂ values, we generated Bland–Altman plots for plotting the difference between measurements (y-axis) and mean of measurements (x-axis). The horizontal dashed lines reflect the mean difference between measurements (2 × SD of difference in measurements). The interobserver agreement of T₂ values between MESE and synthetic MR sequences was quantified using the intraclass correlation coefficient. The r values were classified as follows: 1.0, perfect agreement; 0.81–0.99, almost perfect agreement; 0.61–0.80, substantial agreement; 0.41–0.60, moderate agreement; 0.21–0.40, fair agreement; and ≤ 0.20, slight agreement.¹⁷ Wilcoxon’s signed-rank test was used to determine if there were statistically significant differences in relative SI and relative contrast obtained with conventional and synthetic T₁ weighted images. Statistical analyses were performed using the SPSS (ver. 22; IBM corp., Armonk, NY) and MedCalc (ver. 16; MedCalc, Ostend, Belgium) software packages. In all of the analyses, p < 0.05 was taken to indicate statistical significance.

Results

For cartilage and muscle, the mean T₂ values in the MESE sequences were slightly lower than those in the synthetic MR sequences. However, the differences were not statistically significant (cartilage, 36.7 ± 3.4 ms and 37.2 ± 3.5 ms, p = 0.686; muscle, 34.2 ± 2.7 ms and 34.3 ± 2.0 ms, p = 0.863). The mean T₂ value of bone marrow was significantly higher for MESE (124.3 ± 3.6 ms, p < 0.001) than for synthetic MR sequences (73.1 ± 5.3 ms) (Figure 2). Bland–Altman plots indicated good agreement between the sequences, as demonstrated by the small differences between T₂ values of cartilage and muscle. The mean differences in T₂ values were -0.5 ± 3.5 ms at the cartilage and 0 ± 4.7 at the muscle (Figure 3). The interobserver agreements for each parameter are summarized in Table 2. The T₂ values for cartilage, bone marrow and muscle exhibited substantial or almost perfect agreement.

Figure 2.

Means and standard deviations of T₂ values for multi-echo spin-echo and synthetic MR sequences.

Figure 3.

Bland–Altman plots of T₂ values of cartilage (a) bone marrow (b) and muscle (c). The y- and x-axes indicate the difference and average between multi-echo spin-echo and synthetic MR sequences, respectively. The blue lines show the means of differences, while the 95% confidence intervals are denoted by the pairs of dotted red lines.

Table 2.

Intraclass correlation coefficient for agreement between MESE and synthetic MR sequences for T₂ values

		MESE	Synthetic
Cartilage	Spearman correlation (ρ)	0.740 (0.326–0.900)	0.731 (0.303–0.897)
p-value	0.003	0.004
Bone marrow	Spearman correlation (ρ)	0.763 (0.386–0.909)	0.762 (0.393–0.908)
p-value	< 0.001	0.002
Muscle	Spearman correlation (ρ)	0.814 (0.516–0.928)	0.655 (0.121–0.866)
p-value	< 0.001	0.016

MESE, multi-echo spin-echo.

Numbers in parentheses are 95% confidence intervals.

The mean values for the relative SI of each structure and relative contrast of bone marrow to cartilage and bone marrow to muscle are shown in Table 3. There were no significant differences among the structures, although the mean relative SI for cartilage and muscle were slightly higher for the conventional T₁ weighted sequences, except in the bone marrow. The relative contrast of bone marrow to cartilage in the synthetic sequences was similar to that for the conventional T₁ weighted sequences. The conventional T₁ weighted sequences showed significantly higher relative contrast of bone marrow to muscle compared to the synthetic MR sequences (p = 0.011).

Table 3.

Comparison of the relative SI of each structure and relative contrast between conventional T₁ weighted and synthetic MR images

	Conventional T1	Synthetic T1	p-value
Relative SI
Cartilage	17.2 ± 7.5	15.2 ± 6.1	0.370
Bone marrow	12.1 ± 3.8	13.8 ± 3.1	0.075
Muscle	13.3 ± 4.9	12.0 ± 3.8	0.418
Relative contrast
Bone marrow to cartilage	7.3 ± 2.9	6.7 ± 2.0	0.624
Bone marrow to muscle	7.5 ± 2.0	6.1 ± 1.6	0.011

SI, signal intensity.

Discussion

The MAGiC sequence is very fast; the data are acquired in a single scan, the image contrast can be adjusted after scanning by manipulating TR, TE and inversion time, and quantitative maps provide absolute values of the physical properties of patients. The majority of the literature regarding synthetic MRI and its clinical application has focused on the pathology of the brain. In our initial experience with this technique in the knee joint, we showed that synthetic MRI in the knee yields similar T₁ weighted contrast and T₂ values as conventionally acquired images, suggesting that it may be useful for evaluating the internal derangement of the knee joint.

MRI acquisition time is critical in the musculoskeletal area. Compared to other anatomical structures, the PD-weighted sequences, which has excellent signal distinction among fluid, hyaline cartilage and fibrocartilage are essential for the assessment of joints. Furthermore, an additional imaging plane that has an oblique orientation for evaluating structures such as ligaments may add time to the examination and limit workflow. Several authors^18–22 have reported that a 3D sequence capable of reformatting in various planes is a promising method for imaging the musculoskeletal system and would allow faster isotropic acquisition of musculoskeletal MR images. Previous studies^23,24 have shown that the mDixon technique rapidly generates multiple images in a single acquisition, and the use of mDixon images as substitutes for T₂ weighted images and fat-suppressed T₂ weighted images would reduce the total examination time. In addition, Andreisek et al²⁵ reported that image generation with the synthetic TE technique was a potentially viable alternative to standard T₂ weighted images obtained at different TEs for evaluation of meniscus and articular cartilage in the knee joint. Where previous studies have focused mostly on reducing the time of sequences associated with T₂ contrast, synthetic MR sequences provide synthesized images including T₂ and PD-weighted sequences as well as T₁ weighted sequences, and also generate quantitative measurements such as T₁ and T₂ relaxation maps.

T₂ mapping sequences have been commonly used to assess the articular cartilage of the knee joint.^26–28 The T₂ value for articular cartilage reflects the water content, collagen content and collagen fibre orientation in the extracellular matrix, with longer T₂ values thought to represent cartilage degeneration.^27,29–32 In this study, the accuracy and reproducibility of T₂ values using the synthetic sequences were evaluated in knee joints. We found that the T₂ values of cartilage and muscle in the knee joint, calculated from the synthetic MR sequences, were highly consistent with those calculated by the MESE sequences. Although the value of bone marrow was significantly different for the synthetic sequences, the accuracy and reproducibility of T₂ relaxation times of the knee joint, including cartilage, muscle and bone marrow were substantial or almost perfect. The differences in relative SI and contrast between conventional T₁ weighted and synthetic MR sequences showed similar results. In contrast to cartilage and muscle, relative SI of bone marrow and relative contrast of bone marrow to muscle were significantly different for the synthetic MR sequence. The significant differences in relative SI and relative contrast in the bone marrow may be mainly attributable to the lack of optimization for musculoskeletal tissues in the synthetic sequences, and/or the inherent limitations of the MAGiC sequences (e.g. the limited relaxation points). In a clinical setting, MAGiC sequences allow one to choose the field of view, matrix size, slice thickness, slice gap and acceleration factor. The echo train length can be chosen in a limited range of 10–16. The TR has a minimum value of 4000 ms. The TEs are chosen automatically (approximately 20 and 95 ms). Therefore, the SI of the musculoskeletal tissue, such as bone marrow, subcutaneous fat and meniscus, which have very long or short T₂ values, can be measured as unexpected values for the inherent properties of musculoskeletal tissue. The results could also be due to differences in the acceleration factor for parallel imaging. Future technical efforts should be directed towards optimization for musculoskeletal application.

There were several limitations in this study. First, our study included a small number of patients with only normal findings in knee MRI. Additional studies are necessary to compare the diagnostic abilities of synthetic and conventional sequences for the detection of various pathological conditions of internal derangement of the knee. Second, we did not evaluate synthetic sequences from T₂ weighted and PD-weighted images. Third, in this study, ROI measurements in musculoskeletal tissues showing low signal intensities were not performed because the current MAGiC sequence has not been optimized for tissues with low T₂ values such as the meniscus or ligaments. Fourth, based on our results indicating differences in the SI of bone marrow and relative contrast of bone marrow to muscle in the synthetic MR sequences, there is a potential limitation in the assessment or comparison of subchondral bone marrow oedema associated with osteochondral lesions between initial and follow-up studies.

In summary, synthetic MRI is a promising new MRI technique that obtain multi-contrast images simultaneously, making it ideal for evaluating the knee joint. Although evaluation of musculoskeletal tissues with high T₂ values, such as bone marrow, is limited, this method would be very useful for quantitatively studying and diagnosing diseases of the musculoskeletal system after further optimization. To the best of our knowledge, this is the first application of synthetic MRI to the knee joint.