Three-dimensional adiabatic inversion recovery prepared ultrashort echo time cones (3D IR-UTE-Cones) imaging of cortical bone in the hip

Purpose: We present three-dimensional adiabatic inversion recovery prepared ultrashort echo time Cones (3D IR-UTE-Cones) imaging of cortical bone in the hip of healthy volunteers using a clinical 3T scanner. Methods: A 3DIR-UTE-Cones sequence, basedon a short pulse excitation followed bya 3D Conestrajectory, with anominalTEof32 μ s,wasemployedforhighcontrastmorphologicalimagingofcorticalboneinthehipofheathy volunteers. Signals from soft tissues suchas muscle and marrow fat weresuppressed via adiabaticinversion and signal nulling. T 2 ⁎ value of thecorticalbone wasalso calculatedbasedon 3DIR-UTE-Conesacquisitionswith a series of TEs ranging from 0.032 to 0.8 ms. A total of four healthy volunteers were recruitedfor this study. Average T 2 ⁎ values and the standard deviation for four regions of interests (ROIs) at the greater trochanter, the femoral neck, the femoral head and the lesser trochanter were calculated. Results: The 3D IR-UTE-Cones sequence provided ef ﬁ cient suppression of soft tissues with excellent image con- trast for cortical bone visualization in all volunteer hips. Exponential single component decay was observed for all ROIs, with averaged T 2 ⁎ values ranging from 0.33 to 0.45 ms, largely consistent with previously reported T 2 ⁎ values of cortical bone in the tibial midshaft. Conclusions: The3DIR-UTE-ConessequenceallowsinvivovolumetricimagingandquantitativeT 2 ⁎ measurement of cortical bone in the hip using a clinical 3T scanner.


Introduction
The femoroacetabular joint is a key skeletal element that provides a great deal of mobility and stability. Large loads are exerted on the hip during daily activities, such as walking or running, where forces can exceed five times body weight [1]. Unfortunately fractures can occur, and hip fractures are among the most frequent and devastating of all fractures, particularly in women and the elderly [2,3]. The mortality rate within the first year after a hip fracture may reach up to 30%, which is considerably higher than the mortality rate of breast cancer [4][5][6][7][8][9][10].
Hip fractures occur where the applied load is higher than the bone strength [11]. Cortical bone is an important contributor to overall bone strength [12]. Cortical bone strength and toughness, or simply the fracture resistance, can be decreased dramatically by microstructural changes such as reductions in thickness and density, or increases in porosity [13,14]. In vivo evaluation of cortical bone microstructure has been of great interest in both the orthopedic and radiologic communities [12,15,16]. Many research teams have attempted to quantify bone microstructure and assess the predictive value of various measures in determining fracture risk [17,18].
Magnetic resonance (MR) imaging is a non-invasive imaging modality that can be used to evaluate cortical bone. However, the current, clinically available MR sequences are not able to image cortical bone because the signal from the cortical bone decays rapidly. Therefore, the cortical bone in the clinical images usually appears black (pure signal void). Ultrashort echo time (UTE) techniques can acquire the rapidly decaying MR signal from cortical bone. In recent years, a variety of UTE techniques have been developed to image cortical bone [19][20][21][22][23][24]. Almost all the UTE techniques have focused on cortical bone in the tibial midshaft, with a few examples showing success in imaging the femur and the forearm [25]. Direct MR imaging of cortical bone in the hip with inversion recovery UTE (IR-UTE) technique has not yet been performed.
MR imaging of the cortical bone in the hip has traditionally been fraught with challenges. The main technical difficulties in this area are: 1) the thin cortex, 2) inefficient coils, 3) requirement for robust suppression of surrounding soft tissues to generate high contrast, and 4) long scan time. Also, the thin cortex requires high resolution, which in turn leads to limited signal-to-noise ratio (SNR). The SNR requirement can be moderated by utilizing efficient coils contoured for the Magnetic Resonance Imaging 44 (2017) 60-64 structure of interest, however, there are no commercially available coils for dedicated hip imaging. In addition, to create high contrast images, cortical bone signal must be preserved while undesired components are suppressed. This includes fat in the marrow and extramedullary locations as well as surrounding long T 2 structures such as muscle.
In this study, we report the use of three-dimensional adiabatic inversion recovery prepared UTE with Cones sampling (3D IR-UTE-Cones) to directly image and quantify cortical bone in the hip in vivo at 3 T.

Pulse sequence
The 3D UTE-Cones sequence is shown in Fig. 1A. The 3D UTE-Cones sequence employs a hard RF pulse for non-selective excitation, and center-out 3D spiral trajectories for k-space sampling [26], as shown in Fig.  1B. It also employs short readout times, which in combination with the centric trajectory, allows for signal acquisition from the rapidly decaying tissue components. The 3D UTE-Cones sequence is more time-efficient than radial trajectories in covering 3D k-space [26]. Also, the 3D UTE-Cones sequence resolves the sensitivity of the 2D UTE sequences to eddy currents by employing a hard RF pulse instead of a half RF pulse for signal excitation. Moreover, the 3D UTE-Cones sequence allows for anisotropic fields of view (FOVs) and spatial resolution (higher in-plane resolution with thicker slices), resulting in vastly reduced scan times. By using the 3D UTE-Cones sequence, 3D volumetric UTE imaging can be obtained in an SNR efficient way.
When followed by an adiabatic inversion recovery preparation pulse, the 3D UTE-Cones sequence can be utilized to acquire signal from cortical bone with high contrast [27,28]. This is done by inverting the longitudinal magnetization of the long T 2 signal components (i.e., muscle and bone marrow fat) while saturating the signal from cortical bone [27][28][29][30][31]. The Cones acquisition starts after an inversion time (TI) delay, which is used to null the long T 2 ⁎ components while permitting detection of recovered cortical bone signal (Fig. 1C).

T 2 ⁎ measurement with 3D IR-UTE-Cones
The steady-state 3D IR-UTE-Cones signal can be calculated as follows: Eq. (1) describes that T 2 ⁎ of cortical bone can be measured by mono exponential fitting of the acquired IR-UTE images at different TEs. It should be noted that TR/TI combination should be able to null the signal from the long T 2 components in the bone marrow and muscle.

MRI protocol
The 3D IR-UTE-Cones sequence was implemented on a 3 T scanner (Signa HDx, GE Healthcare, Milwaukee, WI) [27]. The sequence has a minimal TE of 32 μs and allows anisotropic field of view and spatial resolution for fast volumetric imaging. An adiabatic inversion pulse (duration = 8.64 ms) was used for robust inversion and suppression of the longitudinal magnetizations of long T2 water and fat. Four healthy volunteers (28,31,34, and 43 years old, male) were scanned by using a torso phased-array coil. The following scan parameters were used: TR = 116.7 ms, TI = 50 ms, four TEs (0.032, 0.2, 0.4, and 0.8 ms), BW = 250 kHz, FOV = 340 × 340 mm 2 , slice thickness = 3 mm, matrix = 128 × 128, flip angle = 18°, acquired voxel size = 2.6 × 2.6 × 3 mm 3 , and scan time = 4.5 min for each dataset. T 2 ⁎ was quantified with a single-component decay fitting of the multi-echo 3D IR-UTE-Cones images.

Data analysis
The code for the analysis was written in MATLAB (The MathWorks, Massachusetts) and was executed on the DICOM images obtained by the aforementioned protocols in the experimental setup section. The program allowed for the delineation of regions of interest (ROIs) on Fig. 1. The 3D UTE-Cones sequence employs a short rectangular pulse for signal excitation, followed by a 3D Cones trajectory (B) to allow time-efficient sampling with a minimal TE of 32 μs. The 3D UTE-Cones sequence combined with an adiabatic inversion recovery preparation pulse (3D IR-UTE-Cones) can invert and null the signal from long T 2 ⁎ components, including fat and muscle, allowing the cortical bone to be selectively imaged (C). In (C), the dashed curve line and the red solid line represent the long T 2 ⁎ components and the short T 2 ⁎ components, respectively. The dotted line illustrates that after time TI the signal from the long T 2 ⁎ components is nulled (reaching zero) while a significant amount of signal from the short T 2 ⁎ components remains. the UTE images. Four ROIs were drawn in the proximal femur for T 2 ⁎ analysis. As shown in Fig. 2, ROIs included the greater trochanter (ROI1), the lateral aspect of the femoral neck (ROI2), the femoral head (ROI3), and the lesser trochanter (ROI4). The average intensity of voxels within the ROIs was used for subsequent curve fitting. In addition, T 2 ⁎ of the cortical bone were calculated using Eq. (1).

Results
Fig . 2 shows a representative image of the hip of a 34-year-old healthy volunteer imaged with the 3D IR-UTE-Cones sequence. Cortical bone in the greater trochanter, the femoral neck, and the femoral head, as well as lesser trochanter is depicted with excellent image contrast. Muscle and marrow fat about the hip, which typically have far higher signal than that of cortical bone, were efficiently suppressed by the adiabatic inversion pulse. SNR for cortical bone in the femoral head is relatively low due to the thin structure and limited coil sensitivity from the clinical torso phased array coil.
Cortical bone in the hip at four different TEs of 0.032, 0.2, 0.4 and 0.8 ms is shown in Fig. 3. As can be seen, excellent contrast can be achieved by using the 3D IR-UTE-Cones sequence with rapid decay of signal. Fig. 4 shows representative signal decay curves for the greater trochanter, the femoral neck, the femoral head, and the lesser trochanter in a 34-year-old male volunteer. Excellent single-component exponential decay was observed for all the ROIs, with a short T 2 ⁎ of 0.41 ± 0.11 ms for the greater trochanter, 0.33 ± 0.08 ms for the femoral neck, 0.34 ± 0.05 ms for the femoral head, and 0.33 ± 0.04 ms for the lesser trochanter. Table 1 summarizes the mean and standard deviation for T 2 ⁎ values for the greater trochanter, the femoral neck, the femoral head, and the lesser trochanter, respectively, between the four volunteers. The average T 2 ⁎ values ranged from 0.33 to 0.45 ms, and were largely consistent with previously reported T 2 ⁎ values of bound water in the tibial midshaft.

Discussion
In this study, a direct MR-based imaging technique, based on IR-UTE technique, for cortical bone in the hip was reported for the first time in vivo. Although direct imaging of cortical bone has previously been presented at different locations such as the midshaft of the tibia [28,[32][33][34], direct imaging of the hip is considered of higher clinical significance since fractures at this location are more devastating [3]. In addition, imaging at this location is considered more technically challenging.
Preliminary results from this study have shown that the cortical bone in the hip can be imaged using the 3D IR-UTE-Cones sequence (Fig. 3). The adiabatic inversion pulse provides robust suppression of  long T 2 signal. Despite long T 2 muscle and fat containing distinct T 1 values, both can be adequately suppressed with the single IR pulse, resulting in a high contrast, selective imaging technique for cortical bone. T 2 ⁎ can then be estimated via single-component fitting of the 3D IR-UTE-Cones acquisitions with a series of TEs (Fig. 4).
Using the 3D IR-UTE-Cones technique, cortical bone may be visualized with high contrast and T 2 ⁎ may be quantified, providing a single method to measure thickness and estimate porosity, respectively. Cortical bone thickness and porosity are both considered key elements for fracture risk [12,[15][16][17][18]. Specifically, cortical bone thinning and increasing porosity are important features for fracture initiation and propagation [15,16]. Different image processing techniques [35,36] can be used for accurate thickness calculation and to pinpoint thin and weak cortical bone locations.
The UTE and IR-UTE sequences can potentially evaluate cortical porosity (pore water content) and organic matrix (bound water content) in the hip [27,33,[37][38][39][40][41]. Previous studies employing the UTE sequence have shown that both bound water (short T 2 ⁎ component) and pore water (long T 2 ⁎ component) in cortical bone can be quantified using a bi-component T 2 ⁎ analysis technique [18,[25][26][27][28]. Moreover, T 2 ⁎ decay time for bound water in cortical bone can be measured with the IR-UTE technique by suppressing the signal from the pore water and then using a single component analysis of the remaining signal [27,29,38]. A single component decay pattern of the cortical bone in the hip suggests that the 3D IR-UTE-Cones signal is likely from bound water.
One limitation of the 3D IR-UTE-Cones technique is that it requires an adiabatic inversion pulse, which can increase the specific absorption ratio (SAR). This problem can be moderated by using lower flip angles at a cost of sub-optimal contrast, or longer TR at a cost of longer scan time. Another limitation is that UTE with an inversion pulse requires a wide spectral profile to cover both water and fat peaks for robust inversion. Also, the 3D IR-UTE-Cones sequence requires more time to cover kspace compared to 2D UTE imaging techniques. However, the 3D IR-UTE-Cones sequence is less prone to partial volume effects.
Further optimization of the imaging protocol, including in-plane resolution, slice thickness and scan time will be performed in future studies. In addition to further optimization techniques, 3D IR-UTE-Cones sequence should be employed in multiple cohorts. These future studies will help in defining the limitations and thresholds for direct MR imaging of cortical bone.

Conclusion
Our pilot study has demonstrated the feasibility of selective MR imaging of cortical bone in the hip using the 3D IR-UTE-Cones technique. Long-T 2 species (e.g., fat in bone marrow and muscle) were robustly suppressed, providing high image contrast for the cortical bone in clinically compatible imaging times.