Introduction

Synovitis, inflammation of the synovial membrane surrounding synovial fluid, is a common finding in osteoarthritis. Synovitis is typically assessed using T1-weighted contrast-enhanced MRI (DCE), but it is not commonly included in clinical imaging due to cost, imaging time, and potential toxicity. Non-contrast diffusion-weighted imaging using a modified double echo steady state (DESS) sequence has been proposed as an alternative for the assessment of synovitis^1,2. This method uses a weighting parameter (β) to null joint fluid through a linear combination of the two DESS echoes. This study aims to determine the ideal β value to null joint fluid and enhance visualization of the synovium using diffusion-weighted DESS.

Methods

We scanned both knees of ten osteoarthritic patients on a 3T MRI scanner using two modified quantitative DESS (qDESS) sequences. The first sequence contained a high diffusion gradient (DESS High) of 20 cycles/pixel (slice encode direction) while the second contained a low diffusion gradient (DESS Low) of 2 cycles/pixel (slice encode direction). Scan parameters for both scans are included in Table 1. Six β values were then chosen in order to get a wide range of values; Table 2 gives these values, and the methods for determining them are described below. The first β was determined by the Extended Phase Graph (EPG) signal model^3,4, using the following values based on previous estimates of synovial fluid²: T1 = 36­­20 ms; T2 = 767 ms; ADC = 2.6 μm2/ms2. The second β was labeled fluid ROI mean suppression (FRMS), and was determined by creating an ROI in the joint fluid and choosing a β that would null the average value of the fluid to zero according to the following equation: $$${\beta}=\frac{average\: ROI \: signal \:echo\: 1}{average\: ROI\: signal\: echo\: 2}$$$. The third was an average of values found using the EPG and FRMS models (EPG/FRMS Average). The fourth, fifth, and sixth values were chosen by multiplying the β from the EPG model by 0.75, 0.50, and 0.25, respectively, to get a wider range of values. The first and second qDESS echoes were then recombined in MATLAB to make a hybrid image for each β value using linear combination according to the following equation, also described in Figure 1: $$${Echo\: 1}- {\beta}\cdot{Echo\: 2}$$$.
Hybrid images were randomized, blinded and then read by 2 radiologists, one who evaluated all 10 subjects and one who evaluated a subset of 3 subjects. Radiologists scored each hybrid image set on a scale from 1 to 5 [Figure 2]; 1 indicating synovial thickening was not distinguishable from joint fluid, 3 indicating synovial thickening was somewhat distinguishable from joint fluid, and 5 indicated synovial thickening was distinguishable from joint fluid.

Results

Figure 3 shows representative hybrid images created for one subject and a reference DCE image. The β values given by the EPG model were 2.8 for DESS High and 1.15 for DESS Low. The mean DESS Low FRMS β value was very similar to the EPG model and showed low variance, at 1.17 ± 0.04. In contrast, the mean DESS High FRMS β value was far from the EPG model and showed higher variance, at 4.14 ± 1.26. Figure 2 shows the mean and standard deviations of radiologist image scores for DESS High and DESS Low hybrid images. DESS Low hybrid images created using EPG, FRMS and EPG/FRMS Average β values showed similar scores (p=1.0) and were significantly higher than the other three β values (p<0.01). For DESS High, hybrid images created using FRMS and EPG/FRMS Average β values showed statistically higher ratings than EPG * 0.75, EPG * 0.50, and EPG * 0.25 (p<0.01) but not EPG (p=1.0).

Discussion

Non-contrast qDESS hybrid images created with β values using the EPG model, FRMS, and EPG/FRMS Average were scored favorably by radiologists. This demonstrates the potential for utilizing these techniques for evaluation of synovitis without contrast in routine knee scans. Ideally, the EPG model would be used, as it does not require manual segmentation of a fluid ROI to create hybrid images. Results from the DESS Low scans indicate that the EPG model can be used, as it scored statistically equivalent to FRMS and EPG/FRMS Average, and is also very close in value to the FRMS β as shown in the results. Results from the DESS High scans are less clear, as FRMS and EPG/FRMS Average scored higher than EPG but their use would be less automated or would require new methodology for robust application. We are currently working to compare the diagnostic accuracy and confidence of these images against acquired clinical-standard T1-weighted contrast-enhanced images for assessment of synovitis.

Conclusion

Hybrid images using a weighted difference between echos of the DESS sequence showed favorable contrast for evaluation of synovitis. Our results found that the EPG model provided an optimal β value for DESS Low hybrid images while the FRMS method was best for DESS High hybrid images. Both methods showed potential for non-contrast evaluation of synovitis for routine knee scanning.

Acknowledgements

This work was funded by grant support from GE Healthcare and NIH grants R01-EB002524-14, NIH K24-AR062068-07, NIH R01-AR063643-05, and NIH R00-EB022634