To replace the regular ‘Dual-echo in the steady-state’ (DESS) sequence by an improved, multi-echo version capable of simultaneously generating DESS-like images and T₂ maps, for knee osteoarthritis (KOA) studies. The T₂ maps can be obtained without any increase in scan time.

DESS images [1] provide excellent image contrast for knee MRI and are used extensively for software quantification of cartilage loss [2]. DESS is a modified version of steady-state sequences such as FISP or GRASS, whereby one more signal type, the so-called PSIF (for ‘inverted FISP’) is sampled as well. Both the FISP and the weaker (but more heavily T₂-weighted) PSIF signals are reconstructed and typically summed in magnitude to produce the final displayed image. The resulting DESS (i.e., FISP+PSIF) image provides excellent contrast for cartilage segmentation purposes.

Studies have also shown that measurements of T₂ can be used as a biomarker for the early stages of KOA. T₂ quantification using DESS images have been proposed [3]; however, with previous methods errors in T₁ and uncertainties on the flip angle, often unavoidable with imperfect slice/slab profiles, might corrupt T₂ values. In contrast, the present approach avoids these confounding factors by focusing on the signal evolution of each pathway signal within the TR period. Based on data from a multi-TE (mTE) DESS sequence and associated processing, maps of T₂ and of T₂^* are generated as a bonus, in addition to DESS images, at no increase in scan time.

In DESS acquisitions, the FISP and PSIF signals follow (R₂+R₂′) and (R₂-R₂′) relaxation, respectively [4], where T₂=1/R₂ and T₂^*=1/(R₂+R₂′). An mTE DESS sequence was developed to resolve the (R₂+R₂′) and (R₂-R₂′) signal evolution over the TR period, and a non-linear least-square method was used to extract R₂ and R₂′.

Ten healthy volunteers (5 male) were recruited and informed consent was obtained. All in vivo experiments were performed on a 3.0 T scanner (Siemens Trio) with an 8-channel knee coil (Invivo Corporation). The imaging sessions included: 1) Regular 3D DESS with TR=16.3 ms, TE_FISP/TE_PSIF=5.3/10.8 ms, flip angle=25°, FOV=140x140x123 mm, matrix size=384x308x176, partial Fourier=0.75, acquisition time = 11m02s. 2) Accelerated 3D DESS with R=2.56, acquisition time = 5m45s. 3) Two repeats of our proposed mTE DESS, before and after subject repositioning, with TE_FISP=3.95/8.63/11.33ms, TE_PSIF=4.94/7.63/12.32 ms, R=1.44, acquisition time = 10m14s. Hard pulses were employed, frequency-encoding was in the superior-inferior direction to limit the FOV extent, and a 1-2-1 composite RF pulse provided water-only excitation [2] (Fig. 1). Variable density sampling in both phase encoding directions provided acceleration. Synthetic DESS images were generated using the fitted parameters of mTE DESS images, and cartilage volume was measured for all image sets using semi-automatic software that was designed to segment cartilage in focused regions of the medial compartment femur [5].

One selected slice of regular and synthetic DESS images is shown in Fig. 2. Comparable image contrast was achieved in regular (Fig. 2a), regular accelerated (Fig. 2b) and synthetic (Fig. 2c) DESS images. Yellow contours in Fig. 2a-c show how one given extent of cartilage was identified and segmented in all images [5]. Figure 2d-e shows T₂ overlays, where higher T₂ values in some pixels may be indicative of increased water content of the cartilage. Bland-Altman plots are presented in Fig. 3 to compare cartilage volume measurements, and to compare mean cartilage T₂ values before and after subject repositioning.

As seen from Fig. 3, good agreement in cartilage volume was obtained between the proposed mTE DESS sequence and both non-accelerated and accelerated versions of regular DESS. This result suggests that no loss in diagnostic quality occurred when replacing regular DESS by our mTE version. The advantage of using our version comes from the T₂ maps that are also generated. As seen in Fig. 3, good repeatability was achieved on T₂ measurements in a scan-rescan repeatability test with subject repositioning.

A bi-directional readout (i.e., non-flyback) was employed here, possibly creating discontinuities and chemical shifts between images acquired on odd vs. even echoes. Although chemical shifts are small at high bandwidths such as used here, spatial resolution is also high and even small shifts might lead to blurring. Advanced image co-registration might help reduce the small mismatch between odd- and even-echo images, and is considered as future work.

In conclusion, multi-TE DESS acquisitions were proposed, with a non-linear least-square fitting to quantify signal evolution within TR. Synthetic DESS images provided cartilage volumes comparable to those of regular DESS, and the additional T₂ maps could potentially be used to help detect early cartilage changes in future longitudinal studies of knee OA.