Evaluation of Bound and Pore Water in Cortical Bone Using Ultrashort Echo Time (UTE) Magnetic Resonance Imaging

Abstract

Bone water exists in different states with the majority bound to the organic matrix and to mineral, and a smaller fraction in ‘free’ form in the pores of cortical bone. In this study we aimed to develop and evaluate ultrashort echo time (UTE) magnetic resonance imaging (MRI) techniques for assessment of T₂*, T₁ and concentration of collagen-bound and pore water in cortical bone using a 3T clinical whole-body scanner. UTE MRI together with an isotope study using tritiated and distilled water (THO-H₂O) exchange as well as gravimetrical analysis were performed on ten sectioned bovine bone samples. In addition, 32 human cortical bone samples were prepared for comparison between pore water concentration measured with UTE MRI and cortical porosity derived from micro computed tomography (μCT). A short T₂* of 0.27 ± 0.03 ms and T₁ of 116±6 ms were observed for collagen-bound water in bovine bone. A longer T₂* of 1.84 ± 0.52 ms and T₁ of 527±28 ms were observed for pore water in bovine bone. UTE MRI measurements showed a pore water concentration of 4.7-5.3% by volume and collagen-bound water concentration of 15.7-17.9% in bovine bone. THO-H₂O exchange studies showed a pore water concentration of 5.9 ± 0.6% and collagen-bound water concentration of 18.1 ± 2.1% in bovine bone. Gravimetrical analysis showed a pore water concentration of 6.3 ± 0.8% and collagen-bound water concentration of 19.2 ± 3.6% in bovine bone. A mineral water concentration of 9.5 ± 0.6% was derived in bovine bone with the THO-H₂O exchange study. UTE measured pore water concentration is highly correlated (R² = 0.72, P < 0.0001) with μCT porosity in the human cortical bone study. Both bovine and human bone studies suggest that UTE sequences could reliably measure collagen-bound and pore water concentration in cortical bone using a clinical scanner.

Introduction

Cortical bone is a composite material with three major components: mineral (principally hydroxyapatite, ~40% of bone by volume), collagen (~35% by volume) and water (~25% by volume) ¹. Dual-energy X-ray absorptiometry (DXA) and computed tomography (CT) have been used for quantitative analysis of bone mineral density (BMD) ². However, the organic matrix and water, which together represent ~60% of bone by volume, make little contribution to the signal obtained with these techniques. BMD measurement alone predicts fractures with only a 30-50% success rate ^3,4. From ages 60 to 80 years fracture risk increases about 13-fold, however the decrease in BMD alone only explains a doubling in this fracture risk ⁵. A recent study of over 7800 patients reported that only 44% of all non-vertebral fractures occurred in women with a T-score below −2.5 (the World Health Organization definition of osteoporosis based on BMD), and this percentage dropped to 21% in men ⁶. The limitations of BMD measurement have sparked research into the use of other imaging modalities to assess bone quality ⁷, focusing on architectural deterioration as well as the contribution of organic matrix and water to the biomechanical properties of cortical bone.

Water in cortical bone is present at various locations and in different states ⁸. In normal bone a small fraction of water exists in ‘free’ form (pore water or PW) in Haversian canals (typical diameters ~50-200 μm) as well as in lacunae (~5 μm) and canaliculi (~0.1 μm) ^9,10. A larger portion of bone water exists in ‘bound’ form, either loosely bound to the organic matrix (collagen-bound water or CW) ¹¹ or tightly bound to mineral (mineral-bound water or MW) ^12,13. Recent nuclear magnetic resonance (NMR) spectroscopy studies suggest that the loosely bound water concentration reflects organic matrix density of bone ^14,15, while the “free” pore water concentration can potentially provide a surrogate measure of cortical porosity ¹⁶. This provides an opportunity to develop noninvasive techniques to measure bound and pore water concentration in cortical bone and potentially provides a more accurate assessment of bone quality.

As a complementary tool to DXA and CT, magnetic resonance imaging (MRI) detects signal from protons (mostly water) in bone rather than from mineral. However, water in cortical bone has a very short apparent transverse relaxation time (or T₂*) and it provides no detectable signal when examined with conventional clinical MRI sequences ¹⁷. Recently ultrashort echo time (UTE) sequences, with reduced nominal echo times (TEs) as short as 8 μs, which is about 100-1000 times shorter than the TEs of conventional clinical pulse sequences, have been used to detect signal from cortical bone ¹⁸. More recently, magnetization prepared UTE techniques including adiabatic inversion recovery (AIR) and double adiabatic full passage (DAFP) sequences have been developed for mapping of bound and pore water in cortical bone in vivo ¹⁹. This is important considering that bound and pore water make different contributions to the mechanical properties of cortical bone ²⁰.

In this study we aimed to assess the value of two-dimensional (2D) and 3D UTE techniques for determining bound and pore water in bovine and human cortical bone samples. An isotope study, namely tritiated and distilled water (THO-H₂O) exchange in sectioned bovine bone samples, together with gravimetrical analysis were used to validate UTE-based cortical bone water components quantification. A comparison study between UTE MRI measured pore water concentration and micro computed tomography (μCT) measured cortical porosity of cadaveric human cortical bone samples was also performed to further assess the accuracy of UTE MRI measurements of bone water components.

Materials and Methods

Sample Preparation

Fourteen bovine cortical bone samples of approximate dimensions 10×10×10 mm³ were prepared from mature bovine femoral midshafts obtained from a local slaughterhouse. These samples were initially sectioned using a low-speed diamond saw (Isomet 1000, Buehler) with constant saline irrigation, and stored in phosphate buffered saline (PBS) solution for 24 hours prior to study. Four of the samples were used to determine the time required for THO and H₂O exchange to reach equilibrium. Ten samples were used for bound and pore water measurement with UTE and THO-H₂O isotope exchange studies (details below).

Cadaveric human cortical bone samples (n = 32) from 12 donors (7 females, 5 males, 30-92 years old) were obtained from tissue banks, as approved by Institutional Review Board. Using the precision circular diamond-edge saw and saline irrigation, larger bone blocks were sectioned into rectangular slabs of cortical bone with dimensions of ~1×1×10 mm³. Individual samples were wrapped in saline-wet gauze and frozen at −70 °C in an ultralow freezer (Bio-Freezer; Forma Scientific, Marietta, OH, USA). The samples were allowed to thaw for 12 hours at 4 °C prior to UTE MR imaging and micro computed tomography (μCT).

Pulse Sequences

3D UTE sequences (Figure 1) were implemented on a 3T Signa TwinSpeed scanner (GE Healthcare Technologies, Milwaukee, WI). The basic 3D UTE sequence employed a short radio frequency (RF) rectangular pulse (duration = 26-52 μs) for signal excitation ²¹. The z-gradient could be turned off to allow non-selective 2D UTE imaging for fast imaging of cortical bone. Both collagen-bound and pore water are detectable with the basic UTE sequences (Figure 1B). Furthermore, adiabatic inversion recovery prepared UTE (IR-UTE) sequences were developed for selective imaging of collagen bound water (Figure 1C). In the IR-UTE sequence, a Silver-Hoult adiabatic inversion pulse (duration = 8.64 ms, bandwidth = 1.5 kHz) was used to invert the longitudinal magnetization of pore water ^18,22. The longitudinal magnetization of collagen-bound water which has a very short T₂* was not inverted but largely saturated by the adiabatic IR pulse ²². After an inversion time (TI) during which the inverted pore water magnetization approached the null point, the UTE acquisition was initiated to selectively detect signal from collagen-bound water (Figure 1D). 2D UTE and IR-UTE sequences were used for fast T1 and T2* quantification, while 3D UTE and IR-UTE sequences were used for bound and pore water concentration quantification (details below).

Figure 1.

The 3D UTE (A) and IR-UTE (C) sequences, as well as the contrast mechanisms for imaging of total water (B) and collagen-bound water (D). The UTE sequence employs a short rectangular pulse for signal excitation followed by 3D radial ramp sampling with a minimal nominal TE of 8 μs, which is short enough to detect signal from both collagen-bound water (CW) with very short T₂* (solid lines in B and D) and pore water (PW) with slightly longer T₂* (dashed lines in B and D). The IR-UTE sequence employs an adiabatic inversion pulse to invert and null the pore water magnetization. The collagen-bound water magnetization is not inverted and is detected by a subsequent UTE data acquisition (D). The z gradient can be turned off for fast non-selective 2D UTE and IR-UTE imaging.

Quantitative MR Imaging

Collagen-Bound and Pore Water T₂*s

Each sample was placed in perfluorooctyl bromide (PFOB) solution. This helped maintain the hydration of cortical bone and minimize susceptibility effects at tissue-air interfaces. A birdcage coil (diameter = 2.5 cm, length = 12.2 cm) was used for signal excitation and reception. 2D UTE and IR-UTE sequences were performed for fast T₂* measurement with the following parameters: field of view (FOV) = 4 cm, sampling bandwidth (BW) = 62.5 kHz, flip angle = 10°, pulse duration = 26 μs, TR = 100 ms for UTE and 300 ms for IR-UTE, TI = 110 ms (for IR-UTE only), reconstruction matrix size = 128×128, in-plane pixel size = 0.31×0.31 mm², 211 projections, number of excitation (NEX) = 1, scan time = 21 sec for UTE and 1 min for IR-UTE. UTE and IR-UTE images were acquired at a series of TE delays (TE = 8 μs, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 1, 1.2, 1.4, 1.6, 2, 2.5, 3, 3.5, 4, 5, 6, 7 ms) to separate collagen-bound and pore water using the following bi-component analysis model ²³:

$$SI\left(t\right)=S_{z0,\mathrm{cw}}\times e^{-t∕T_{2,cw}^{\ast }}+S_{z0,\mathrm{pw}}\times e^{-t∕T_{2,pw}^{\ast }}+\text{noise}$$ [1]

where S_Z0,cw and S_Z0,pw are the magnetization of the collagen-bound and pore water components, and T_2,cw* and T_2,pw* are their T₂* relaxation times.

Collagen-Bound and Pore Water T₁s

Collagen-bound and pore water may have different T₁ values. Both contribute to the UTE signal when a minimal TE of 8 μs is used. When a longer TE (e.g., 2.4 ms) is used, the UTE signal mainly comes from pore water since the signal from collagen-bound water with a T₂* ~0.3 ms very largely decays to zero. In this study we proposed to use a saturation recovery dual echo UTE imaging technique to measure the T_1s of collagen-bound and pore water components.

In the saturation recovery UTE sequence a 90⁰ rectangular pulse is followed by a crusher gradient to dephase signals from both long and short T₂ species. Dual echo 2D UTE acquisitions with progressively increasing saturation recovery times (TSRs) are used to detect the recovery of the longitudinal magnetization of cortical bone. For simplification, we assume an effective T₁ (T₁^eff) of collagen-bound and pore water is measured using 2D saturation recovery UTE acquisitions with a minimal TE of 8 μs, and pore water T₁ (T₁^pw) is measured using saturation recovery UTE acquisitions with a longer TE of 2.4 ms. T₁^eff and T₁^pw were calculated using the following two equations:

$$S(TSR,TE=8\mu s)=S_0\times [1-(1-k)\times e^{-TSR∕T_1^{eff}}]+C$$ [2]

$$S(TSR,TE=2.4ms)=S_0\times [1-(1-k)\times e^{-TSR∕T_1^{pw}}]+C$$ [3]

where k accounts for the residual fraction of the longitudinal magnetization of cortical bone after a nominal 90° pulse. Similar imaging parameters were used except for a long TR of 3500 ms (which is long enough to ensure almost complete recovery of both bound and pore water longitudinal magnetizations after each 90° pulse), TE = 8 μs and 2.4 ms, a series of TSRs (7, 25, 50, 100, 200, 400, 600, 800, 1000, 1200, 1600, 2000, 2400, 2800, 3400 ms) with a total scan time of 110 min.

The collagen-bound water signal obtained with IR-UTE imaging can be described by the following equation ²⁴:

$$S_{IR-UTE}(\mathrm{TR},\mathrm{TI},\mathrm{TE}=8\mu \mathrm{s})\propto \frac{[1-(1-Q)\times e^{-TI∕T_1^{CW}}-Q\times e^{-TR∕T_1^{CW}}]}{1-Q\times \cos \left(\theta \right)\times e^{-TR∕T_1^{CW}}}\times \sin \left(\theta \right)\times e^{-TE∕T_{2,\mathrm{cw}}^{\ast }}$$ [4]

where $T_1^{CW}$ is the T₁ of collagen-bound water, and Q is the inversion efficiency of the adiabatic IR pulse. For collagen-bound water with a T₂* of ~0.3 ms, Q approximates 0 (i.e., < 0.04) according to Bloch equation simulation ¹⁸. As a result, the IR-UTE signal can be simplified as follows ²⁴:

$$S_{IR-UTE}\propto 1-e^{-TI∕T_1^{CW}}$$ [5]

here $T_1^{CW}$ can be measured by fitting the IR-UTE signal acquired with a series of TR and TI combinations, under the condition that each TR/TI combination satisfies the inversion and nulling condition necessary to suppress the signal from pore water. We used imaging parameters similar to the 2D IR-UTE T₂* analysis above, but with a series of TR/TI combinations (e.g., TR = 50, 100, 200, 300, 400, 500 ms, TI was chosen for each sample based on the criteria to null pore water with a measured T₁^pw). The total scan time was about 6 min.

Collagen-Bound and Pore Water Concentration Measurement

Total water concentration (WC_Total) was measured by comparing the 3D UTE signal intensity of cortical bone with that from a doped water phantom ²⁵. Accurate estimation of bone water requires consideration of relaxation times and coil sensitivity. In our experiments the external reference was a mixture of distilled water (20%) and D₂O (80%) doped with 26 millimolar MnCl₂, resulting in a short T₂* of ~300 μs and a T₁ of ~ 5 ms (measured with standard UTE acquisitions wit variable TEs or TSRs, as shown in Ref 18). With the use of a short excitation pulse (14 μs), ultrashort TE (8 μs), relatively long TR (50 ms) and a small flip angle (5°), T₁ and T₂* relaxation effects can be ignored, and $W{C}_{Total}^{UTE}$ can be simplified as follows:

$$W{C}_{Total}^{UTE}\approx \frac{I_{bone}^{UTE}}{I_{H_{2}O-D_{2}O}^{UTE}}\times \eta \times 20\%$$ [6]

where $I_{bone}^{UTE}$ and $I_{H_{2}O-D_{2}O}^{UTE}$ are the 3D UTE signal intensities of cortical bone and H₂O-D₂O water phantom, respectively, and η is a coil sensitivity correction. η was corrected by dividing the 3D UTE signal from cortical bone or doped water phantom by the 3D UTE signal obtained from a separate scan of a bottle of water, which was large enough to cover the region occupied by both the cortical bone sample and doped water phantom. Other imaging parameters were similar to those used for 2D UTE T₂* measurements except for 3D UTE acquisitions with 20,000 projections and a voxel size of 0.31×0.31×0.31 mm³. The total scan time was 16.7 min.

Collagen-bound water concentration (WC_Collagen) was measured by comparing the 3D IR-UTE signal intensity of cortical bone with that from the water phantom. Based on Eq.4, $W{C}_{Collagen}^{IR-UTE}$ can be measured as follows:

$$W{C}_{Collagen}^{IR-UTE}\approx \frac{1-e^{-TI∕T_1^{H_{2}O-D_{2}O}}}{1-e^{-TI∕T_1^{CW}}}\times \frac{I_{bone}^{IR-UTE}}{I_{H_{2}O-D_{2}O}^{IR-UTE}}\times \eta \times 20\%\approx \frac{I_{bone}^{IR-UTE}}{I_{H_{2}O-D_{2}O}^{IR-UTE}\times (1-e^{-TI∕T_1^{CW}})}\times \eta \times 20\%$$ [7]

where $I_{bone}^{IR-UTE}$ and $I_{H_{2}O-D_{2}O}^{IR-UTE}$ are the IR-UTE signal intensities of cortical bone and H₂O-D₂O water phantom, respectively. After each IR pulse the longitudinal magnetization of H₂O-D₂O is almost fully recovered since its T₁ is much shorter than TI. Imaging parameters were similar to those used for 2D IR-UTE T₂* measurements, except for 3D IR-UTE with 20,000 projections and a voxel size of 0.31×0.31×0.31 mm³. The total scan time was 100 min.

With total water and collagen-bound water known, pore water concentration (WC_Pore) can be calculated as their difference:

$$W{C}_{Pore}=W{C}_{Total}^{UTE}-W{C}_{Collagen}^{IR-UTE}$$ [8]

μCT Imaging

The human cortical bone samples were imaged using a μCT scanner (1076, Skyscan, Kontich, Belgium) with the following parameters: 0.5 mm Aluminum filter, 72 kV, 140 μA, 720 views collected at 0.5° increments corresponding to one full rotation of each bone specimen, FOV 25 mm, isotropic 9 μm voxels. The total scan time was ~ 3 hours.

THO-H₂O exchange study

THO is radioactive, with a half life of 12.3 years. The relatively long half life makes it ideal for isotope study. Four bovine cortical samples were used to determine the time required for THO-H₂O exchange to reach equilibrium. Each bone sample was blotted dry and put in a glass bottle filled with 2 ml of THO for times of 1, 2, 4, 7 and 10 days and maintained at room temperature. After each exchange stage, 0.1 ml of the mixed solution (a mixture of THO and H₂O) was mixed by vortex with 5 ml of scintillation fluid (Ecoscint, National Diagnostics, Atlanta, GA) in scintillation vials and placed into a liquid scintillation counter (LS6000SC, Beckman Coulter Inc, Brea, CA) for beta radiation measurement in counts per minute (CPM). CPM is defined by the total number of photons counted divided by the count time. CPM was plotted against isotope exchange time to determine the time required to reach equilibrium.

Each glass vial containing an individual bone sample was filled with 2 ml of THO. The glass vial was tightly sealed for THO-H₂O exchange. After equilibrium (~7 days) 0.1 ml of the mixed solution was prepared for beta radiation detection. Meanwhile a calibration curve was generated by measuring the radiation of solutions with eight different ratios (1:0, 1:0.05, 1:0.1, 1:0.15, 1:0.2, 1:0.3, 1:0.5 and 1:1) of THO and H₂O. CPM was plotted against the THO dilution factor and then used as the calibration curve. Based on the radiation calibration curve a dilution factor for the THO-H₂O exchange was calculated, and used for accurate measurement of the volume of H₂O in each bovine cortical bone. Bone water concentration was calculated by dividing H₂O volume by bone volume from μCT (more details are given in the Gravimetrical Analysis section below).

Gravimetrical Analysis

Previous literatures suggest that air-drying at room temperature mainly results in loss of pore water, while subsequent oven-drying at 60-100 °C mainly results in loss of water bound to the organic matrix ^26-30. In this study we employed air-drying (in an oven with controlled humidity at room temperature) of blot-dried bovine cortical bone for 48 hours to remove pore water, followed by oven-drying at 100 °C for 72 hours to remove collagen-bound water, after which water remaining in the bone is assumed to be tightly bound to mineral. The air-drying and oven-drying times were empirically determined based on weight change. About two days were needed for air-drying and three days were needed for oven-drying to reach equilibrium mass (weight change of less than 0.1% in the last five hours of drying).

Mankin et al. used an approach combining THO-H₂O exchange and drying course to measure bound and free water in articular cartilage ³¹. We adopted a similar approach in this study. First, all bovine cortical bone samples (n=10) were blotted dry and subjected to UTE and IR-UTE imaging to quantify T₁^pw, T₁^cw, T_2,pw*, T_2,cw*, WC_Collagen and WC_Pore, and μCT imaging to measure bone volume ($V_{bone}^{bolt-dry}$), as well as use of a digital precision balance (Mettler Toledo AL104) to measure bone weight ($W_{bone}^{bolt-dry}$). Then half of the bone samples (n=5) were subjected to air-drying to remove pore water, followed by measurement of bone weight ($W_{bone}^{air-dry}$). The volume of H₂O from each blot-dried ($V_{water}^{blot-dry}$) and air-dried ($V_{water}^{air-dry}$) bovine cortical bone sample was measured via THO-H₂O exchange. Each bone sample was then immersed in saline (50 ml), changed daily until the CPM was reduced to background level (<40), indicating that all THO had been replaced by H₂O. Oven-drying for 72 hours at 100 °C was used to remove collagen-bound and pore water. Each bovine bone sample was cooled down in a sealed glass tube to minimize reabsorbtion of water. The oven-dry weight was measured as $W_{bone}^{oven-dry}$. Finally each bone sample was placed into THO-H₂O for a second THO-H₂O exchange to measure the volume of mineral-bound water ($V_{water}^{oven-dry}$) which had not been removed by the oven-drying process. Total bone water concentration from the THO-H₂O exchange study ($W{C}_{Total}^{THO-H_{2}O}$) was calculated as follows:

$$W{C}_{Total}^{THO-H_{2}O}\approx \frac{V_{water}^{blot-dry}}{V_{bone}^{blot-dry}}\times 100\%$$ [9]

The concentration of water bound to collagen ($W{C}_{Collagen}^{THO-H_{2}O}$) was calculated as follows:

$$W{C}_{Collagen}^{THO-H_{2}O}\approx \frac{V_{water}^{air-dry}-V_{water}^{oven-dry}}{V_{bone}^{blot-dry}}\times 100\%$$ [10]

The concentration of water bound to mineral ($W{C}_{Mineral}^{THO-H_{2}O}$) was calculated as follows:

$$W{C}_{Mineral}^{THO-H_{2}O}\approx \frac{V_{water}^{oven-dry}}{V_{bone}^{blot-dry}}\times 100\%$$ [11]

Pore water concentration ($W{C}_{Pore}^{THO-H_{2}O}$) was calculated as follows:

$$W{C}_{Pore}^{THO-H_{2}O}=W{C}_{Total}^{THO-H_{2}O}-W{C}_{Collagen}^{THO-H_{2}O}-W{C}_{Mineral}^{THO-H_{2}O}$$ [12]

Both pore water ($W{C}_{Pore}^{Gravimetry}$) and collagen-bound water ($W{C}_{Collagen}^{Gravimetry}$) can also be estimated from the gravimetric measurements using the following equations ²⁹:

$$W{C}_{Pore}^{Gravimetry}\approx \frac{W_{water}^{blot-dry}-W_{water}^{air-dry}}{V_{bone}^{blot-dry}}\times 100\%$$ [13]

$$W{C}_{Collagen}^{Gravimetry}\approx \frac{W_{water}^{air-dry}-W_{water}^{oven-dry}}{V_{bone}^{blot-dry}}\times 100\%$$ [14]

Mineral-bound water is inaccessible with gravimetric analysis due to the difficulty in separating it from collagen gravimetrically.

Data Analysis

A semi-automated MATLAB (The Mathworks Inc. Natick, MA, USA) program was developed for single- and bi-component T₂* and T₁ analysis. Regions of interest (ROIs) were drawn in the central part of cortical bone for data analysis. Goodness of fit statistics including the R-squared value and standard error or fitting confidence level were computed. Fit curves along with their 95% confidence intervals (CI) and residual signal curves were created. Total, collagen-bound, mineral-bound and pore water concentrations were calculated using the equations shown in [6-14].

For each human cortical bone sample, 12 μCT images every 2.5 mm along the sample long axis were selected and imported into Matlab. A custom program was used to determine the global histogram for all the images and this was used to determine a local minimum value used for thresholding. Binarized images were despeckled to remove noise, and regions of interest corresponding to the outer boundary of the sample as well as to bone were automatically generated. Cortical porosity was determined as one minus the ratio of the area of the bone to that of the outer boundary of the sample.

Results

UTE and IR-UTE sequences were used to access different water components in cortical bone using a clinical MR scanner. Figure 2 shows selected UTE and IR-UTE images of a bovine cortical bone sample with different TEs. Systematic residual signal is seen with single-component fitting of UTE signal decay, suggesting the existence of another water component. Excellent curve fitting is achieved with a bi-component model which shows two distinct water components: one with a short T₂* of 0.26 ms accounting for 72.4% of the total UTE signal, and the other with a longer T₂* of 1.56 ms accounting for 27.6% of the signal decay. The shorter and longer T₂* values are consistent with those of collagen-bound and pore water respectively based on recent NMR spectroscopic studies by Ni et al., who reported a short T2* of 0.21 ms and a longer T2* of 2.70 ms ¹⁶, suggesting that UTE bi-component analysis can access bound and pore water using a whole-body clinical MR scanner. The IR-UTE images show a single component T₂* decay of 0.31 ms which is close to the short T₂* value of 0.26 ms obtained from the bi-component fitting of UTE T₂* signal decay, suggesting that water bound to the organic matrix is selectively detected in IR-UTE imaging. The pore water component with a longer T₂* is largely suppressed by the IR preparation pulse through adiabatic inversion and signal nulling.

Figure 2.

Selected 2D UTE imaging of a bovine cortical bone sample with TEs of 8 μs (A), 0.2 ms (B), 0.4 ms (C), 0.6 ms (D), 0.8 ms (E), 1.6 ms (F), 2 ms (G), 3 ms (H), 4 ms (I), 5 ms (J), 6 ms (K) and 7 ms (L) as well as single component (M) and bi-component (N) fitting of the UTE images. Single component fitting of the corresponding IR-UTE images with a TR of 300 ms and TI of 90 ms was also shown (O). An ROI was drawn in (A). Single component fitting of the UTE images shows significant residual signal (> 10%) (M), which is reduced to less than 0.5% by bi-component fitting (N), which shows a shorter T₂* of 0.25 ms and a longer T₂* of 2.20 ms with respective fractions of 81.2% and 18.8% by volume. Excellent single component fitting of the IR-UTE images is achieved with residual signal less than 0.3% (O), and a T₂* of 0.30 ms which is close to that of the short T₂* component in UTE images, suggesting that bound water was detected by the IR-UTE sequence with pore water largely suppressed by the long adiabatic inversion pulse.

Table 1 shows the mean and standard deviation of T₂* values of 10 bovine bone samples. An effective T₂* of 0.41 ± 0.05 ms was demonstrated from single-component fitting of UTE signal decays with contribution from both collagen-bound and pore water components. Bi-component fitting shows a mean pore water T₂*^pw of 0.28 ± 0.03 ms and collagen-bound water T₂*^cw of 1.84 ± 0.52 ms, with a fraction of 25.4 ± 4.2% and 74.6 ± 4.2%, respectively. The IR-UTE signal shows an excellent single-component decay behavior, with a mean T₂*^cw of 0.32 ± 0.02 ms, which is comparable to that derived from bi-component fitting of UTE signal decay.

Table 1.

Measurement of pore water T₂* (T₂*^pw), collagen-bound water T₂* (T₂*^cw) and total water effective T₂* (T₂*^eff) using single-component and bi-component fitting of UTE images, and single-component fitting of IR-UTE images, respectively.

T2* Measurements	T2*PW	T2*CW	PW Fraction	CW Fraction	T2*eff
UTE Single-Component Analysis	-	-	-	-	0.41 ± 0.05 ms
UTE Bi-Component Analysis	1.84 ± 0.52 ms	0.28 ± 0.03 ms	25.4 ± 4.2%	74.6 ± 4.2%	-
IR-UTE Bi-Component Analysis	-	0.32 ± 0.02 ms	-	-	-

Figure 3 shows T₁ measurements using saturation recovery dual echo UTE acquisitions as well as IR-UTE acquisitions with variable TR/TI combinations. Single-component fitting of the saturation recovery curve shows a T₁^eff of 268 ms, which is the effective T₁ of collagen-bound and pore water since both water components contribute to the UTE signal with a minimal TE of 8 μs, and a T₁^pw of 509 ms, which is the T₁ of pore water only since signal from collagen-bound water decays to near zero with a TE of 2.4 ms. Exponential fitting of the IR-UTE curve shows a T₁^cw of 122 ms, which is likely the T₁ of collagen-bound water since signal from pore water is selectively suppressed with each of IR-UTE acquisitions with variable TR/TI combinations.

Figure 3.

T₁ measurement using saturation recovery dual echo UTE acquisitions with a TE of 8 μs (A) and 2.4 ms (B), and IR-UTE acquisitions with variable TR/TI combinations (C). A T₁^eff of 268 ms and fitting error of ±22 ms was measured with saturation recovery UTE acquisitions with a TE of 8 μs (A). A T₁^pw of 509 ms and fitting error of ±22 ms were measured with saturation recovery UTE acquisitions with a TE of 2.4 ms (B). A T₁^CW of 122 ms and fitting error of ±12 ms were measured with IR-UTE acquisitions with variable TR/TI combinations (C).

Table 2 shows the mean and standard deviation of T₁ values of 10 bovine bone samples. A mean effective T₁^eff of 243 ± 37 ms was demonstrated from single-component fitting of saturation recovery UTE signal recovery with a TE of 8 μs, where both collagen-bound and pore water components contributed to the UTE signal recovery. A mean pore water T₁^pw of 527 ± 28 ms was demonstrated from single-component fitting of saturation recovery UTE signal recovery with a TE of 2.4 ms, where only pore water contributed to the UTE signal recovery. A mean collagen-bound water T₁^cw of 116 ± 6 ms was demonstrated from single-component fitting of IR-UTE signal recovery with different TR/TI combinations. T₁^cw is significantly shorter than T₁^pw.

Table 2.

Measurement of pore water T₁ (T₁^pw), collagen-bound water T₁ (T₁^cw) and total water effective T₁ (T₁^eff) using single-component fitting of saturation recovery dual echo UTE acquisitions with TEs of 8 μs and 2.4 ms, as well as IR-UTE acquisitions with variable TR/TI combinations and a TE of 8 μs.

T1 Measurements	T1PW	T1CW	T1Effective
Saturation Recovery UTE (TE1 = 8 μs)	-	-	243 ± 37 ms
Saturation Recovery UTE (TE2 = 2.4 ms)	527 ± 28 ms	-	-
Inversion Recovery UTE, variable TR/TI	-	116 ± 6 ms	-

Measurement of bone water concentration using THO-H₂O isotope exchange requires calibration. Figure 4 shows the calibration curves generated for water concentration measurement in the THO-H₂O experiment. An excellent linear relationship (R² = 0.9997) between CPM and THO fractions in THO-H₂O solutions was observed, suggesting that dilution factor, and thus water concentration from bovine cortical bone can be accurately measured by linearly interpolating CPM of the THO-H₂O mixed solution.

Figure 4.

Calibration curve used for THO dilution factor measurement was generated by measuring CPM as a function of THO fraction in THO/H₂O solutions. The excellent linear behavior suggests that THO dilution factors can be accurately measured via linear interpolation.

Total, bound and pore water concentration in bovine cortical bone measured with UTE MRI, THO-H₂O isotope exchange and gravimetrical analysis are shown in Table 1. UTE measured collagen-bound and pore water concentration approximates that determined by THO-H₂O isotope exchange and gravimetrical analysis, suggesting that UTE sequences can reliably measure collagen-bound and pore water in cortical bone. Furthermore, THO-H₂O study of oven-dried bone suggests the existence of a significant portion of bone water (~9.5% by volume), which is likely to be water tightly bound to mineral that has not been identified previously.

A high correlation (R² = 0.72; P < 0.0001) was observed between μCT porosity and pore water concentration in 32 cadaveric human cortical bone samples, as shown in Figure 5. Water residing in the microscopic pores of cortical bone is expected to behave more like ‘free’ water. The high correlation between μCT porosity and ‘free’ water concentration further confirmed the accuracy of UTE techniques in assessing bound and ‘free’ water in cortical bone. μCT porosity is consistently lower that ‘free’ water content assessed by UTE MRI, likely due to “free” water in smaller pores being detected by UTE MRI but not μCT imaging.

Figure 5.

Correlation between UTE measured pore water concentration and μCT porosity in cadaveric human cortical bone samples (n=32). A high correlation (R² = 0.72; P < 0.0001) was observed between pore water concentration and μCT porosity, suggesting that UTE sequences can reliably access water in cortical bone using a clinical MR scanner.

Discussion

UTE imaging has the potential to assess different water components in cortical bone in vivo. Our data indicates that UTE sequences detect both collagen-bound and pore water in cortical bone when using a clinical whole-body 3T scanner. The excellent bi-component fitting suggests that no more than two components are needed to explain the UTE signal decay behavior. The majority of the bone water (70-80%) exists in the form of collagen-bound water in cortical bone. The IR-UTE signal decay shows excellent single-component decay behavior with an untrashort T₂* of around 0.3 ms, suggesting that only collagen-bound water is detected by the IR-UTE sequence with pore water effectively suppressed by the adiabatic IR pulse. UTE measured pore water is about 10~20% lower than that measured with THO-H₂O isotope exchange, and 15~25% lower than that measured with gravimetric analysis, partly because of the loss of loosely bound water during air-drying at room temperature ^26-29. Furthermore, air-drying may not fully remove all pore water, further complicating the comparisons ²⁶. The high correlation (R² = 0.72) between UTE measured pore water concentration and μCT cortical porosity further suggests that UTE and IR-UTE sequences can reliably access total and collagen-bound water components in cortical bone.

Previous NMR spectroscopy studies have confirmed the existence of multiple water components in cortical bone. Proton NMR spin grouping and exchange in dentin, a bone-like hard tissue, shows that 30% of the water is strongly bound to mineral with a T₂* of ~ 12 μs, 52% of the water is loosely bound to the organic matrix with a T₂* of ~200 μs, and 18% of the water is trapped in the dentinal tubules with a T₂* of ~1000 μs ¹². Considering that mineral-bound water has a too short T₂* to be detected by clinical MR scanners, loosely bound water accounts for ~75% of the total UTE detectable signal. Inversion relaxation analysis of FID signal of bone samples also showed three distinct signal components with T₂*s of ~11μs, ~210 μs, and ~2.7 ms, respectively ¹⁶. Multiple component analysis of FID and CPMG signals of cortical bone samples identified five signal sources, including collagen-bound water, pore water, mineral-bound water, collagen methylene, and lipid methylene ¹¹. Mineral-bound water has an extremely short T₂* of ~11.8 μs, accounting for ~8% of total water in bone. Collagen-bound water accounts for ~73% of total water, with the other ~18% for pore water. Furthermore, our measured mean T₁^pw of 527 ± 28 ms for bovine cortical bone samples at 3T are comparable with the recent published results by Horch, et al ²², who reported a mean T₁^pw of 551 ± 120 ms for human cortical bone samples at 4.7 T. Our measured mean T₁^cw of 116 ± 6 ms at 3 T is significantly shorter than the mean T₁^cw of 357 ± 10 ms at 4.7 T reported by Horch et al ²². This might be partly due to the field dependence of T₁. More recently Seifert et al reported a mean T₁^pw of 880 ± 281 ms and a mean T₁^cw of 145 ± 25 ms at 3 T²⁸. T₁^cw value from our study is very close to that reported by Seifert et al ²⁸. Meanwhile, the mean T₁^pw values of 880 ± 281 ms at 3T and 1790 ± 470 ms at 7T from the Seifert study are significantly higher than the mean T₁^pw values of 527 ± 28 ms at 3T from our study and 551 ± 120 ms at 4.7T from the Horch study ^{22, 28}. The difference might be due to multiple factors, including different field strengths (T₁ is field strength dependent) and type of specimen (T₁^pw in human cortical bone with bigger pores is expected to be longer than in bovine cortical bone with smaller pores).

Bulk water in the pores has a long T₂ (> 100 ms) but a short T₂* (< 10 ms), and can be detected with conventional clinical fast spin echo (FSE) sequences with TEs of 10 ms or longer ³². FSE-determined porosity is highly correlated (R² = 0.83) with μCT porosity ³³. However, it should be noted that not all “free” water in pores has T2* long enough to be detectable by conventional FSE sequences. There is a broad distribution of T2 and T2* values in pore water ^{11, 30}. Water loosely bound to the organic matrix has very short T₂ and T₂*, and remains invisible to conventional clinical MR sequences. Wu et al developed water- and fat-suppressed proton projection MRI (WASPI) for bone imaging in vitro ¹⁵. Gravimetric and amino analyses showed that the WASPI signal is highly correlated (R² = 0.98) with collagen content ¹⁴. Water tightly bound to mineral has an extremely short T₂*. Bone mineral crystals contain hydroxyl (OH⁻) ions which can be observed with proton magic angle spinning (MAS) NMR spectroscopy ³⁴. More recently, using the 2D Lee-Goldberg cross-polarization under magic-angle spinning (2D LG-CPMSA) pulse sequence, Wilson et al observed a structured water layer at the surface of the mineral in cortical bone ³⁵. In a later study, they identified three types of structurally-bound water ³⁶, which serves to stabilize these defect-containing crystals or mediating mineral-organic matrix interactions. Tightly-bound water was also observed by Ivanova et al. using proton MAS NMR spectroscopy ³⁷. However, water bound to mineral is largely neglected in commonly used drying techniques ^26-29. The difference between oven-dried weight and ash weight is typically considered to be the weight of the organic phase ²⁹. This simplification is expected to overestimate the organic matrix content, and underestimate the total water content in cortical bone. As our study indicated, approximately 28% of bone water (i.e., mineral-bound water) would be misclassified as organic matrix content according to the conventional gravimetric techniques.

The strength of bone is determined by its composition and structure. Recent work by Zebaze et al. suggested that accurate assessment of bone structure, especially cortical porosity, could improve identification of individuals at high risk of fracture and assist treatment ³⁸. Age related adverse changes in the collagen network may lead to the decreased toughness of bone ³⁹. A recent work done by Gallant et al. ⁴⁰ found that collagen-bound water, rather than pore water, was highly correlated with canine bone biomechanics following raloxifene treatment, while cortical porosity, BMD and bone mineral concentration (BMC) show no significant correlation. Those results suggest the importance of simultaneous assessment of bound and pore water in cortical bone. More recently, two other UTE-derived indices, the porosity index ⁴¹ and suppression ratio ⁴², were introduced. However, these indices only indirectly measure bone quality without direct information on cortical porosity and collagen content. The 3D UTE and IR-UTE techniques introduced in this study have the potential to assess organic matrix via bound water measurement and cortical porosity via pore water measurement, thus allowing direct comparison of cortical porosity and collagen matrix across different groups of people (e.g., aging and disease related changes in cortical bone). UTE together with magnetization transfer (MT) imaging may provide further information on mineral-bound water, and potentially bone mineral content ⁴³. More importantly, these techniques allow accurate volumetric mapping of different water components in cortical bone in vivo in a time-efficient way using clinical whole-body scanners. This may significantly advance the study of bone diseases including OP, osteomalacia, osteopenia, Paget disease, renal osteodystrophy and insufficiency fractures in the setting of biophosphonate therapy or raloxifene treatment ⁴⁰.

There are several limitations in this study. First, the 2D and 3D UTE sequences are time consuming, limiting their clinical applications. However, more advanced UTE sequences employing a 3D Cones trajectory have shown promise in measuring bound and pore water in cortical bone in vivo ⁴⁴. Second, the 2D and 3D UTE sequences are based on the use of T₂* than T₂ to separate bound water from pore water. Susceptibility and surface relaxation shortening of the pore water components, making it challenging to achieve robust separation of bound water and pore water especially at higher field strengths ²⁸. Third, reference techniques including THO-H₂O isotope exchange and gravimetrical method have limitations ^26-29. During air-drying bound water is partly removed. Oven-drying cannot completely remove all the bound water, especially water tightly bound to bone mineral. T₂ spectral analysis is a more accurate technique in separating bound and pore water ²⁸. However, it is technically difficult or impossible to acquire multi-echo spin echo data for T₂ analysis in vivo, due to the limitations in RF power and specific absorption ratio (SAR) associated with clinical MR scanners. Fourth, no comparison has been done between our 2D and 3D UTE techniques and several other short T₂ imaging techniques, such as zero echo time (ZTE) imaging ⁴⁵, pointwise encoding time reduction with radial acquisition (PETRA) ⁴⁶, double adiabatic full passage (DAFP) ¹⁹, and sweep imaging with Fourier transformation (SWIFT) ⁴⁷. Imaging techniques such as ZTE and SWIFT have potential advantages over 2D and 3D radial UTE sequences when imaging extremely short T2 species due to their shorter effective TEs. However, the contrast mechanisms associated with ZTE and SWIFT, and their applications in measuring bound and pore water T₁s, T₂*s and water concentrations remain to be established. Fifth, computed tomography (CT) is an established imaging technique for studying bone mineral density and content. Comparison between UTE MRI and CT techniques would be of considerable interest and will be performed in future studies.

In conclusion, we have demonstrated that T₁, T₂* and concentration of collagen-bound and pore water components can be measured with UTE and IR-UTE sequences. Further clinical studies will be necessary to evaluate the diagnostic power of this method.

Table 3.

Mean and standard deviation of pore water concentration (WC_Pore), collagen-bound water concentration (WC_Collagen) and mineral-bound water concentration (WC_Mineral) using UTE imaging techniques, THO-H₂O isotope exchange study and gravimetric analysis, respectively.

	Pore Water (WCPore)	Collagen-Bound Water (WCCollagen)	Mineral-Bound Water (WCMineral)	WCPore + WCCollagen	WCPore + WCCollagen + WCMineral
UTE Measured WCTotal	-	-	-	21.0 ± 2.8%	-
IR-UTE Measured WCCollagen	-	17.9 ± 1.6%	-	-	-
Pore Water (WCTotal – WCCollagen)	4.7 ± 0.5%	-	-	-	-
UTE + Bi-Component Analysis	5.3 ± 0.4%	15.7 ± 2.3%	-	-	-
THO-H2O Exchange	5.9 ± 0.6%	18.1 ± 2.1%	9.5 ± 0.6%	-	33.5 ± 0.6%
Gravimetric Analysis	6.3 ± 0.8%	19.2 ± 3.6%	-	25.5 ± 0.6%	-

Abbreviations used

2D

two-dimensional

BMC

bone mineral concentration

BMD

bone mineral density

CPM

counts per minute

CW

collagen-bound water

DXA

Dual-energy X-ray absorptiometry

FSE

fast spin echo

FID

free induction decay

FOV

field of view

IR

inversion recovery

LG-CPMSA

Lee-Goldberg cross-polarization under magic-angle spinning

MAS

magic angle spinning

MRI

magnetic resonance imaging

MT

magnetization transfer

NEX

number of excitations

NMR

nuclear magnetic resonance

PFOB

perfluorooctyl bromide

PBS

phosphate buffered saline

PW

pore water

RF

radio frequency

ROI

region of interest

SNR

signal to noise ratio

TE

echo time

TI

inversion time

TSR

saturation recovery time

μCT

micro computed tomography

UCSD

University of California, San Diego

UTE

ultrashort echo time

WASP

water- and fat-suppressed proton projection MRI