Ultrashort Echo Time Cones Double Echo Steady State (UTE-Cones-DESS) for Rapid Morphological Imaging of Short T₂ Tissues

Abstract

Purpose:

In this study we aimed to develop a new technique, ultrashort echo time Cones double echo steady state (UTE-Cone-DESS), for highly efficient morphological imaging of musculoskeletal tissues with short T₂s. We also proposed a novel, single-point Dixon (spDixon)-based approach for fat suppression.

Methods:

The UTE-Cones-DESS sequence was implemented on a 3T MR system. It employs a short radiofrequency (RF) pulse followed by a pair of balanced spiral-out and spiral-in readout gradients separated by an unbalanced spoiling gradient in-between. The readout gradients are applied immediately before or after the RF pulses to achieve a UTE image (S₊) and a spin/stimulated echo image (S₋). Weighted echo subtraction between S₊ and S₋ was performed to achieve high contrast specific to short T₂ tissues, and spDixon was applied to suppress fat by utilizing the intrinsic complex signal of S₊ and S₋. Six healthy volunteers and five patients with osteoarthritis were recruited for whole-knee imaging. Additionally, two healthy volunteers were recruited for lower leg imaging.

Results:

The UTE-Cones-DESS sequence allows fast volumetric imaging of musculoskeletal tissues with excellent image contrast for the osteochondral junction, tendons, menisci, and ligaments in the knee joint as well as cortical bone and aponeurosis in the lower leg within five minutes. spDixon yields efficient fat suppression in both S₊ and S₋ images without requiring any additional acquisitions or preparation pulses.

Conclusion:

The rapid UTE-Cones-DESS sequence can be used for high contrast morphological imaging of short T₂ tissues, providing a new tool to assess their association with musculoskeletal disorders.

INTRODUCTION

Double echo steady state (DESS) imaging stems from a family of conventional steady state free precession (SSFP) sequences that utilize the steady-state of transverse magnetization that is achieved by repeated radiofrequency (RF) pulses (1). In DESS, an SSFP signal is separated into two components—free induction decay (FID)-like signal (S₊) and echo-like signal (S₋)—by utilizing an extended and unbalanced readout gradient between two consecutive RF pulses (2). DESS has been widely used for clinical applications in musculoskeletal (MSK) (3–10) and neuroimaging (11–16) due to its flexible tissue contrast involving T₁, T₂, and diffusion weighting. In DESS imaging, fat signal commonly appears with high signal intensity due to its high T₂ over T₁ ratio, a signal which typically needs to be suppressed when attempting to image non-lipid tissues with specificity. For this purpose, a chemical shift-based water excitation technique using a spectral-spatial RF pulse has been widely used (5,6,8,9,13,14).

In the literature, ultrashort echo time (UTE) imaging has been actively investigated and successfully applied in many cases as a method for imaging tissues with short T₂s or T₂*s (17). UTE imaging significantly reduces the time delay between RF excitation and data readout so that the FID signal from the short T₂ component is captured before it decays to near zero. In UTE imaging, a rewinding gradient is typically removed and replaced instead with a center-out radial or spiral data readout performed immediately after RF excitation (18–20). A 3D Cones trajectory has also been proposed for fast volumetric UTE imaging (21). The UTE Cones (UTE-Cones) sequence utilizes twisted trajectories (i.e., 3D spiral arms) to efficiently encode the 3D k-space for data readout (22,23). Moreover, various data analysis and signal preparation schemes including T₁/T₂* mapping (24–26), quantitative susceptibility mapping (27–30), magnetization transfer (31,32), adiabatic inversion recovery (33–36), and adiabatic T_1ρ spin locking (37–39) have been combined with Cones sampling for quantitative UTE imaging, further demonstrating the efficacy and adaptability of the UTE-Cones sequence in morphological and quantitative imaging of short T₂ tissues such as the osteochondral junction (OCJ), menisci, ligaments, tendons, cortical bone, and hemosiderin.

Recently, a UTE-based DESS (UTE-DESS) sequence was proposed by Chaudhari et al. (40). In this approach, two balanced trapezoidal readout gradients were placed immediately before and after a short rectangular RF pulse to acquire S₊ and S₋ signals separated by an unbalanced, interleaved spoiling gradient. Their study showed the feasibility of using UTE-DESS to image both short (e.g., tendon) and long T₂ tissues (e.g., articular cartilage) in the human knee joint. A potential challenge posed by UTE-DESS imaging is the presence of a strong fat signal that needs to be suppressed as is the case in conventional DESS imaging. Unfortunately, the conventional water excitation technique based on composite RF pulses with a long duration (~10 ms) (41,42) cannot be applied in UTE-DESS sequences as they target TEs shorter than 0.1 ms. As a result, Chaudhari et al. achieved fat suppression with a flexible echo two-point Dixon technique instead (40), though it should be noted that their approach required a longer scan time to acquire the set of images with a different time delay.

Jang et al. recently proposed the single-point Dixon (spDixon) method in UTE-based MSK (UTE-MSK) imaging as a method for effectively suppressing lipid signal and was able to achieve high image contrast for short T₂ tissues such as tendons, ligaments, and menisci using this approach (43). The spDixon method was initially proposed by Ma (44) to target breath-hold body imaging, and used a single gradient recalled echo (GRE) image to separate fat and water by decomposing a complex MR signal. Jang et al. incorporated this idea in UTE-MSK imaging and utilized a dual echo UTE-Cones sequence to acquire both a UTE image and a GRE image (43). The GRE image was processed with spDixon to both estimate and suppress fat signal in the UTE image and demonstrated feasibility in morphological UTE-MSK imaging of human knee and ankle joints.

In this study, we proposed a novel UTE-Cones-based DESS (UTE-Cones-DESS) sequence for highly efficient morphological UTE-MSK imaging and explored the feasibility of applying spDixon-based fat suppression to this approach. In contrast to the previous spDixon method for UTE-MSK imaging which used two echoes (43), we demonstrated a new approach to robust fat-water separation which requires only a single image, removing the need for additional data acquisitions. The feasibility and efficacy of our approach were demonstrated in UTE-Cones-DESS imaging of human knee joints and lower legs.

METHODS

Pulse sequence

Figure 1 shows the pulse sequence diagram of the UTE-Cones-DESS sequence. Steady state of the transverse magnetization is achieved after a series of repeated rectangular RF pulses. A minimum, nominal TE of 84 μs (i.e., the interval between the end of RF pulse and the beginning of the readout) is achieved, which is slightly longer than that of our previous UTE-Cones sequence (i.e., 32 μs) due to the additional delays required for UTE-Cones-DESS imaging imposed by the RF and gradient systems of the targeted clinical MR system (MR750, GE Healthcare, Milwaukee, WI, USA). In UTE-Cones-DESS, two images are acquired: from S₊ with a center-out trajectory and S₋ with a fly-back trajectory. Figure 1B shows the 3D Cones trajectory which encodes the 3D k-space with rotating spiral trajectories on different conical surfaces.

Figure 1.

3D-UTE-Cones-DESS. (A) Pulse sequence diagram and (B) k-space trajectory. A spoiler gradient is applied after the first Cones read-out to spoil the remaining transverse magnetization from S₊. Since the center-out and fly-back spiral gradients are balanced, echoes (S₋) can be achieved in the subsequent TRs. The spoiler gradient is unchanged through the scan without rotation. A minimum TE of 84 μs is achieved which is slightly longer than the typical TE of 32 μs in the conventional UTE-Cones imaging due to an additional delay in the gradient and RF systems required for UTE-Cones-DESS imaging.

Echo subtraction

Echo subtraction (ES) was performed to achieve high contrast for tissues with short T₂s. Weighted echo subtraction (WES) further improved short T₂ contrast by suppressing both fat and long T₂ tissues, and was calculated as follows:

$$WES=S_{+}-\alpha S_{-},$$ (1)

where ⍺ is a weighting factor to control contrast in the ES image. In this study, ⍺ was empirically tuned to achieve the desired contrast.

Single-point Dixon based fat suppression

The spDixon method decomposes complex MR signal into individual fat and water signals with a known chemical shift-induced phase difference, represented as θ, as illustrated in Figure 2A. In our previous study of spDixon as applied in UTE-MSK imaging (43), an additional non-UTE echo acquisition (i.e., GRE) was required for fat suppression. This approach showed efficacy for fat suppression of a UTE image but was based on the indirect estimation of fat signal in a GRE image. Moreover, the acquisition of a B₀ field map was also necessary for this method, and both requirements increased the fat suppression method’s acquisition time. In this study, we propose a new approach to independently decompose the S₊ or S₋ signal to directly detect fat and water signal without the need for additional data acquisition.

Figure 2.

The single-point Dixon (spDixon) approach for fat suppression in 3D UTE-Cones-DESS imaging. (A) complex signal decomposition in spDixon to estimate fat and water signal, and (B) block diagram showing workflow of spDixon processing.

A prerequisite of spDixon is the removal of any phase errors which have been added to the complex MR signal including initial phase offset, ϕ₀, and phase error due to B₀ field inhomogeneity, ϕ_B0. In general, ϕ_B0 is determined by the field inhomogeneity, ΔB₀, and a time delay during FID, τ, such that

$${\varphi }_{B0}=\gamma \Delta B_0\tau ,$$ (2)

where γ is a gyromagnetic ratio of proton. ϕ₀ is more complicated to model as it involves many factors including electric conductivities of individual tissues, characteristics of the RF receiver coil, phase modulation during readout, and any imperfections in non-Cartesian image reconstruction. In UTE-DESS imaging, it is expected that ϕ_B0 will be much smaller than ϕ₀ because of the short τ used to acquire S₊ and S₋ signals, meaning that ϕ₀ will cause more phase errors in both the complex MR signal and the subsequent spDixon process, therefore necessitating its removal.

We propose to use the intrinsic signal properties of S₊ and S₋ in DESS to measure ϕ₀ without any additional acquisitions. The phase of S₊ in DESS can be written as

$${\varphi }_{S_{+}}={\varphi }_0+{\varphi }_{B0}+{\varphi }_c,$$ (3)

where ϕ_c is the phase of the complex signal which results from a combination of fat and water signals as illustrated in Figure 2A (between 0 and θ). S₋ is acquired at a TE of (2 × TR – τ) in the signal passage toward refocusing in the opposite direction on the transverse plane (1,45,46). Therefore, the phase of S₋ can be written as

$${\varphi }_{S_{-}}={\varphi }_0-{\varphi }_{B0}-{\varphi }_c+\pi .$$ (4)

Then, ϕ₀ can be solved for by using both equations such that

$${\varphi }_0=\left({\varphi }_{S+}+{\varphi }_{S-}-\pi \right)/2.$$ (5)

Figure 2B shows a block diagram for the proposed workflow to perform spDixon in UTE-Cones-DESS imaging. The summation of the phases of S₊ and S₋ cancels the phases due to chemical shift and B₀ inhomogeneity, and yields ϕ₀. Correction of field inhomogeneity induced phase error (ϕ_B0) is optional for rapid morphological UTE-Cones-DESS imaging.

Experimental setup

The proposed UTE-Cones-DESS sequence was implemented on a 3T clinical MR system (GE MR750). To evaluate the feasibility of UTE-Cones-DESS, six healthy volunteers (two females and four males, mean age of 35.3 ± 8.8) and five patients with osteoarthritis (OA) (five males, mean age of 51.0 ± 8.1) were recruited and underwent imaging of one knee. Additionally, two healthy volunteers (a 32-year-old male and a 37-year-old male) were recruited and underwent imaging of the lower leg. All human subjects research was performed in compliance with guidelines of the Human Research Protections Program at University of California, San Diego and written informed consent forms were collected from all subjects.

The knee imaging was performed using a transmit/receive 8-channel knee coil with the following parameters: 1) UTE-Cones-DESS: sagittal plane, a hard pulse with duration of 100 μs, flip angle (FA) = 10°, field of view (FOV) = 140×140×120 mm³, matrix = 256×256×60, TR = 4.9 ms, TE = 84 μs (S₊) or 9.7 ms (S₋), readout bandwidth (rBW) = 250 kHz, and scan time = 4 min 38 sec; 2) Dual echo Cones for field map acquisition: Parameters were the same as those used for UTE-Cones-DESS except: TR = 10 ms, TE = 2.2 and 4.4 ms, matrix = 128×128×60, and scan time = 1 min 42 sec; 3) Clinical T₁-weighted fast spin echo (T₁-FSE): sagittal plane, FA = 142 °, FOV = 140×140 mm², matrix = 352×320, slice thickness = 3 mm, number of slices = 28, TR = 527 ms, TE = 8 ms, rBW = 83.3 kHz, echo train length = 18, and scan time = 2 min 3 sec; 4) Clinical T₂-weighted fast spin echo (T₂-FSE): sagittal plane, standard chemical shift-based fat saturation, FA = 142°, FOV = 140×140 mm², matrix = 352×256, slice thickness = 3 mm, number of slices = 28, TR = 4894 ms, TE =70 ms, rBW = 83.3 kHz, echo train length = 18, and scan time = 2 min 3 sec.

To calibrate the chemical shift-induced phase, θ, used in signal decomposition for spDixon, a fat-water phantom comprised of saline and pure vegetable oil was made in a 20 mL syringe. The phantom was scanned twice using UTE-Cones-DESS (for estimation of ϕ₀) and dual echo Cones (for estimation of ϕ_B0) with parameters matching those used in the knee imaging. This process is required to compensate for discrepancy between the nominal TE and the effective TE arising from, for example, the duration of the RF pulse and time spent encoding the central region of k-space, thus achieving more accurate fat-water separation with spDixon (44).

UTE-Cones-DESS imaging of the lower leg was performed using the same knee coil with the following parameters: axial plane, a hard pulse with duration of 100 μs, FA = 10°, FOV = 130×130×120 mm³, matrix = 256×256×40, TR = 6 ms, TE = 84 μs (S₊) or 11.9 ms (S₋), rBW = 250 kHz, and scan time = 3 min 13 sec.

Data processing

Image reconstruction and data processing were performed using Matlab R2017b (Mathworks, Inc, Natick, MA, USA). Image reconstruction was performed using offline Matlab reconstruction codes developed in-house and based on Non-Uniform FFT (NuFFT) (47). The reconstructed images with data acquired by individual receive channels were combined to form a complex image (48). For spDixon and field map estimation, phase maps from S₊ and S₋ images were unwrapped using a 3D unwrapping algorithm provided by the FMRIB Software Library (v5.0) (49). spDixon was processed using the method described above. A field map was estimated based on the phase difference between two unwrapped phase images with TEs of 2.2 ms and 4.4 ms. The initial phase offset (ϕ₀) map was spatially filtered using a 3D Gaussian kernel with a standard deviation of 3, 3, and 0.75 pixels in the x, y, and z axes, respectively.

RESULTS

In vivo UTE-Cones-DESS imaging

The UTE-Cones-DESS technique showed high contrast for short T₂ tissues in WES images for all subjects who underwent knee and/or lower leg imaging.

Figure 3 shows results from UTE-Cones-DESS imaging of the knee joint of a representative healthy volunteer (29-year-old female). The FID-like S₊ image shows as a typical UTE image with T₁-weighting. The echo-like S₋ image shows more T₂ weighting due to the longer TE of 9.9 ms. The ES image delineates short T₂ tissues including tendons (red arrows), the osteochondral junction (yellow arrows), menisci (blue arrows), and ligaments (green arrow). The WES shows images with more suppressed fat signal and improved short T₂ contrast.

Figure 3.

In vivo UTE-Cones-DESS imaging of the knee joint of a 29-year-old healthy female volunteer. The weighted echo subtraction improved the image contrast for short T₂ tissues as indicated by arrows (red: tendon; yellow: osteochondral junction, blue: meniscus; green: ligament).

Figure 4 shows the results of lower leg imaging with a representative healthy volunteer (32-year-old male). The WES image provides high contrast for short T₂ tissues, such as the tibia and fibula cortical bone (yellow arrows), and the aponeurosis (red arrows).

Figure 4.

In vivo UTE-Cones-DESS imaging of the lower leg of a 32-year-old healthy male volunteer: (A) S₊ image, (B) S₋ image, and (C) weighted echo subtracted image. In the weighted echo subtracted image, short T₂ tissues such as cortical bone (yellow arrows) and aponeurosis (red arrows) are clearly depicted.

spDixon-based fat suppression in UTE-Cones-DESS

The physical phase difference between water and fat, θ, was empirically measured in an experiment using a fat-water phantom, where θ was measured as 0.44 radian at the targeted τ prescribed in this study (84 μs). Supporting Information Figure S1 shows the measurement of θ.

Figure 5 shows the results of spDixon in UTE-Cones-DESS imaging of a healthy volunteer (32-year-old male). Figures 5A and 5B show raw phase maps of S₊ and S₋. A low-frequency bias gradient is observed in the two raw phase maps, mainly due to the initial phase error, ϕ₀. Figure 5C shows the initial phase offset, ϕ₀, estimated using the proposed method based on the intrinsic signal property of DESS. Figures 5D and 5E show the ϕ₀-corrected phase images for S₊ and S₋. Figures 5F–J show the results with S₊ signal. Without correction of ϕ₀, spDixon yields largely misestimated water and fat signals (Figures 5G and 5H), whereas with correction of ϕ₀, spDixon provides more reasonable fat and water images (Figures 5I and 5J). Figures 5K–O show the results of spDixon with S₋ signal, demonstrating the efficacy of ϕ₀ correction in spDixon being comparable to such correction with S₊.

Figure 5.

spDixon with a healthy volunteer (37-year-old male). Phase of S₊ and S₋ images (A, B), the estimated phase offset (C), and the phase of S₊ and S₋ images after correction (D, E) are shown in the top row. Magnitude of S₊ image (F) and the resultant water and fat images without ϕ₀ correction (G, H) and with ϕ₀ correction (I, J) are shown in the middle row. Magnitude of S₋ image (K) and the resultant water and fat images without ϕ₀ correction (L, M) and with ϕ₀ correction (N, O) are shown in the bottom row.

Correction of phase error due to B₀ field inhomogeneity, ϕ_B0, was optional in this study. Supporting Information Figure S2 shows a ϕ_B0 map estimated with an additional field map acquisition. The estimated ϕ_B0 for the prescribed τ was negligibly small (Supporting Information Figure S2C) compared to the phase of targeted complex MR signal, ϕ_c (Figures 5D and 5E). Supporting Information Figure S3 shows the comparison between spDixon without (Supporting Information Figure S3A and S3D) and with (Supporting Information Figure S3B and S3E) correction of B₀ field inhomogeneity: the difference was very small and did not impair morphological information (Supporting Information Figure S3C and S3F). Supporting Information Figure S4 shows another example of UTE-Cones-DESS with spDixon in the lower leg of a healthy volunteer (27-year-old male).

UTE-Cones-DESS knee imaging with spDixon

For all subjects who underwent UTE-Cones-DESS knee imaging, the proposed spDixon method successfully separated water and fat.

Figure 6 shows the results of UTE-Cones-DESS imaging with spDixon on a representative healthy volunteer (32-year-old male), in which fat-suppressed water images were obtained for both S₊ and S₋ by utilizing the proposed spDixon approach. Both WES images with and without spDixon achieve high contrast specific to the short T₂ tissues for the OCJ (yellow arrows), tendons (red arrows), menisci (blue arrows) and ligaments (green arrows).

Figure 6.

UTE-Cones-DESS imaging with spDixon of the knee joint of a healthy volunteer (37-year-old male). Both weighted echo subtractions with and without spDixon-based fat suppression achieved high contrast specific to short T₂ tissues including the OCJ (yellow arrows), tendons (red arrows), menisci (blue arrows), and ligaments (green arrows).

Figure 7 shows UTE-Cones-DESS imaging with spDixon of a representative patient with OA (52-year-old male). Long T₂ water signal from fluid is well-detected in the water image of S₋ (Figure 7F), which corresponds well with the T₂-FSE image (Figure 7C) as indicated by red arrows. Sclerosis of subchondral bone is also well-depicted as bright contrast in the S₊ water image (Figure 7E) and the WES water image (Figure 7G), which correspond with dark contrast in the T₁-FSE image (Figure 7D) as indicated by yellow arrows. Note that the lesion is not well-depicted in the WES image without spDixon (Figure 7H) since the unsuppressed fat signal coming from the bone marrow obscures detection of sclerosis on the image.

Figure 7.

UTE-Cones-DESS imaging with spDixon in a representative OA patient (52-year-old male) with osteoarthritis. (A) Magnitude image of S₊, (B) magnitude image of S₋, (C) T₂-FSE image, (D) T₁-FSE image, (E) water image from S₊, (F) water image from S₋, (G) weighted echo subtraction of water images from spDixon, and (H) weighted echo subtraction of S₊ and S₋ without spDixon. Sclerosis of subchondral bone is well depicted in the S₊ water image (E) and the weighted echo subtracted water image with spDixon (G) which corresponds well with the T₁-FSE image (D) as indicated by yellow arrows. Long T₂ water is also detected in the S₋ water image from spDixon (F) which corresponds well with the clinical T₂-FSE image (C).

Figure 8 shows UTE-Cones-DESS imaging with spDixon of another representative patient with OA (56-year-old male). Meniscal tear is detected in the S₋ water image (as bright contrast in Figure 8F), WES water image with spDixon (as dark contrast in Figure 8G), and WES image without spDixon (as dark contrast in Figure 8H), which corresponds well with the T₂-FSE and T₁-FSE images (Figures 8C and 8D).

Figure 8.

UTE-Cones-DESS imaging with spDixon in a representative OA patient (56-year-old male) with osteoarthritis. (A) Magnitude image of S₊, (B) magnitude image of S₋, (C) T₂-FSE image, (D) T₁-FSE image, (E) water image from S₊, (F) water image from S₋, (G) weighted echo subtraction of water images from spDixon, and (H) weighted echo subtraction of S₊ and S₋ without spDixon. Meniscal tear (yellow arrows) is clearly depicted as bright contrast in the S₋ water image (F) and as dark contrast in the weighted echo subtracted water images (G, H), which corresponds well with clinical FSE images (C, D).

DISCUSSION

In this study, we implemented UTE-Cones-DESS on a clinical 3T MR scanner and showed its feasibility for rapid morphological imaging of short T₂ tissues. Furthermore, we proposed a novel, spDixon-based approach for fat suppression in UTE-DESS. The proposed method was evaluated in human knee joints with or without OA symptoms, as well as lower legs. In the knee imaging, UTE-Cones-DESS with or without spDixon provided high contrast imaging specific to short T₂ tissues such as the OCJ, tendons, menisci, and ligaments. In lower leg imaging, short T₂ tissues, such as tibia and fibula, and the aponeurosis were well-depicted.

The UTE-Cones-DESS sequence allows more time-efficient data acquisition than the conventional UTE-DESS sequences based on a purely radial trajectory. According to our previous experiments, the Cones trajectory more than halves the length of scan time of a UTE-DESS sequence using radial trajectory (23,30). Scan time can be further accelerated by slightly stretching the spiral arms (23,30). However, the optimal readout strategy (i.e., radial or Cones with and without stretching) depends on the tissues being targeted. For example, when imaging tissues with high lipid contents or extremely short T₂* decay, radial-mode UTE-DESS may be beneficial for avoiding chemical shift artifacts or short T₂* blurring artifacts. Note that the radial trajectory can also be easily implemented by straightening the readout trajectory of UTE-Cones-DESS.

We demonstrated that fat suppression overall improved lesion detection of long T₂ water and subchondral bone sclerosis (Figure 7). Because our method utilizes intrinsic information derived from the UTE-DESS signal, whereas the two-point Dixon-based approach requires additional images to be acquired at a different τ for each S₊ and S_-, our proposed fat suppression strategy has the advantage of shorter scan time over the two-point Dixon-based approach. Combined with the efficient Cones trajectory, the UTE-Cones-DESS spDixon technique can be expected to reduce total scan time to a quarter of the previously proposed UTE-DESS sequence using a radial trajectory and two-point Dixon-based fat suppression (40). Furthermore, spDixon directly decomposes the fat and water signals at the targeted τ, which gives a more accurate snapshot of the signal at τ, while the fat and water signals estimated by two-point Dixon may not reflect the actual signal at τ if there is significant T₂* decay between the two τs used in the Dixon algorithm, a principle also demonstrated in our previous work (43).

To realize the efficiency of spDixon in UTE-Cones-DESS, we proposed a new approach to estimate ϕ₀ using a combination of S₊ and S₋ phases. As described above, ϕ₀ is affected by several different factors and it is therefore difficult to accurately estimate either analytically or empirically. The proposed approach may provide a fast and simple way to estimate ϕ₀ based on DESS imaging (and is not only limited to UTE-DESS imaging). However, there may be a significant error in the estimated ϕ₀ if the phase information has a low signal-to-noise ratio, which may happen in S- signal for short T₂ tissues (e.g., tendon and cortical bone). To mitigate this error, Gaussian filtering can be applied to the estimated ϕ₀ map. Iterative non-linear optimization to estimate ϕ₀ combined with advanced image processing approaches (e.g., region growing method) (44,50) may allow more accurate estimation of ϕ₀, a method which will be investigated in future work. Notably, this proposed ϕ₀ estimation method could be used for other applications beyond fat-water separation, such as tissue electric property mapping (EPM) (51,52), so long as any phase errors other than the B₁-related phase could be removed.

Correction of ϕ_B0 can be beneficial but is not critical for robust fat suppression in morphological UTE-Cones-DESS spDixon imaging. That being said, if there is strong B₀ field inhomogeneity which cannot be compensated for through B₀ shimming, ϕ_B0 may become so large that correction is needed to avoid any significant errors in spDixon. Moreover, correction of ϕ_B0 may be essential for quantitative fat-water separation (e.g., to calculate fat fraction) and, in these cases, rapid acquisition of a low-resolution B₀ field map could be performed to correct ϕ_B0. Note that the proposed spDixon approach could also be applied to conventional DESS (i.e., non-UTE-DESS) imaging if ϕ_B0 is corrected.

In general, the downside of the spDixon over the two-point Dixon method is the compromised signal-to-noise ratio (SNR) associated with the former. Theoretically, amplification of noise in spDixon depends on the chemical shift-induced phase difference, θ, where the standard deviation of noise increases on the order of 1/sinθ. For example, if θ is π/2, SNR will not be affected. Therefore, θ close to π/2 is desired in terms of SNR. With the UTE-Cones-DESS imaging in this study, θ was measured as 0.44 radian, which was estimated to decrease SNR by 2.3x. To mitigate this issue, a phase map can be low pass-filtered to allow better noise performance (44), or a slightly delayed TE can be used to achieve larger θ. Two-point Dixon that utilizes a combination of two images tends to have a higher SNR.

This study only focused on morphological UTE-Cones-DESS imaging, though quantitative UTE-Cones-DESS imaging is also feasible for simultaneous estimation of T₁, T₂, and diffusivity if the acquisition is repeated with different spoiling gradient amplitudes, flip angles, or TRs. It has been shown that a simplified spin echo model or a more comprehensive SSFP model could be used to fit the parameters in DESS imaging (4,40,45,53–55). We will investigate our quantitative UTE-Cones-DESS imaging method based on these different signal models in future studies. It is expected that our investigations will continue to demonstrate that UTE-Cones-DESS imaging can provide rapid quantitative mapping of T₁, T₂, and diffusivity for short T₂ tissues including tendons, ligaments, and menisci.

CONCLUSIONS

In this study, we reported a UTE-Cones-DESS sequence for fast morphological MSK imaging on a 3T clinical MR system, which provided high contrast images specific to short T₂ tissues in human lower extremities. We also showed the feasibility of an spDixon-based fat suppression approach for UTE-Cones-DESS imaging. The proposed method provides a new imaging tool to assess short T₂ tissues such as the OCJ, tendons, ligaments, menisci, and cortical bone.

Supplementary Material

Supporting Informati

Supporting Information Figure S1. Fat-water phantom. (A) Estimated phase offset, ϕ₀, (B) estimated field map and (C) the resultant ϕ_B0 map, (D) magnitude image and (E) phase image at TE of 84 μs, and (F) the final phase image after correction of ϕ₀ and ϕ_B0. The phase θ of fat was measured as 0.44 radian in (F).

Supporting Information Figure S2. Phase error in UTE-Cones-DESS imaging caused by B₀ inhomogeneity. (A) A phase image corrected with an estimated initial phase offset, (B) an estimated B₀ field map with an external acquisition, and (C) the calculated phase error at S₊ caused by B₀ inhomogeneity based on the estimated field map in (B).

Supporting Information Figure S3. Impact of phase error caused by B₀ inhomogeneity in spDixon-based fat suppression in UTE-Cones-DESS imaging. spDixon-based fat and water images of S₊ without (A, D) and with (B, E) correction of phase error due to B₀ inhomogeneity, and the corresponding absolute difference map (C, F). Images are displayed in a pixel intensity normalized by maximum intensity in water or fat images with B₀ phase correction (B, E). In the region indicated by a yellow circle, the signal intensity in the difference map was 0.050 ± 0.035 in fat (C) and 0.055 ± 0.038 in water (F).

Supporting Information Figure S4. UTE-Cones-DESS imaging with Single Point Dixon in the lower leg (37-year-old male). An example of (A) phase in S₊ after correction of ϕ₀, estimated B₀ field map, and estimated ϕ_B0, and (B) the resulting fat and water images of S₊ and S₋.