Water proton density in human cortical bone obtained from ultrashort echo time (UTE) MRI predicts bone microstructural properties

Abstract

Purpose:

To investigate the correlations between cortical bone microstructural properties and total water proton density (TWPD) obtained from three-dimensional ultrashort echo time Cones (3D-UTE-Cones) magnetic resonance imaging techniques.

Materials and Methods:

135 cortical bone samples were harvested from human tibial and femoral midshafts of 37 donors (61±24 years old). Samples were scanned using 3D-UTE-Cones sequences on a clinical 3T MRI and on a high-resolution micro-computed tomography (μCT) scanner. TWPD was measured using 3D-UTE-Cones MR images. Average bone porosity, pore size, and bone mineral density (BMD) were measured from μCT images at 9 μm voxel size. Pearson’s correlation coefficients between TWPD and μCT-based measures were calculated.

Results:

TWPD showed significant moderate correlation with both average bone porosity (R=0.66, p<0.01) and pore size (R=0.58, p<0.01). TWPD also showed significant strong correction with BMD (R=0.71, p<0.01).

Conclusions:

The presented 3D-UTE-Cones imaging technique allows assessment of TWPD in human cortical bone. This quick UTE-MRI-based technique was capable of predicting bone microstructure differences with significant correlations. Such correlations highlight the potential of UTE-MRI-based measurement of bone water proton density to assess bone microstructure.

1. Introduction

Standard evaluation of cortical bone in clinics has been focused on x-ray-based techniques such as dual-energy X-ray absorptiometry (DEXA) and quantitative computed tomography (QCT) [1,2] to measure bone mineral density (BMD).

Magnetic resonance imaging (MRI)-based cortical bone evaluation has recently been increasing since MRI avoids the potential harm associated with x-ray-based techniques [1,3–5]. MRI-based bone evaluation may also provide excellent assessment of the surrounding soft tissue such as tendons [6] and muscles, a benefit which is not available in x-ray-based techniques. Notably, conventional MRI is not able to detect considerable signal of cortical bone because the apparent transverse relaxation time (T2*) of bone is very short. Among the recently developed MRI techniques, ultrashort echo time (UTE) MRI is capable of imaging cortical bone with high signal and can also offer quantitative measurements [3–5,7–9]. Specifically, the bone signal in UTE-MRI techniques can be acquired in less than a tenth of millisecond after radiofrequency (RF) excitation.

Total water proton density (TWPD) in cortical bone can be estimated by comparing the UTE-MRI signals of bone and of an external reference with known water content [8,10–15]. A mixture of distilled water and heavy water (e.g., 20% H₂O and 80% D₂O) has been used routinely as the external reference [8,11,12,14,16]. It should be noted that any material with known apparent proton density and a range of MRI properties similar to bone, such as rubber, may be used as the external phantom. Bound and pore water proton densities (BWPD and PWPD) in bone can be estimated using UTE-MRI techniques combined with different preparation techniques. Inversion recovery UTE-MRI (IR-UTE-MRI) signal in cortical bone compared with an external reference has been used to estimate BWPD [8,14–17]. PWPD in cortical bone can be calculated indirectly by subtracting BWPD from TWPD [11,16,18]. Alternatively, PWPD can be estimated using a double adiabatic full passage pulse (DAFP) preparation to saturate the bound water signal followed by a UTE acquisition to selectively image pore water protons [8,14,19,20].

Gradual increase in the porosity of human bone due to aging or osteoporosis (OP) is hypothesized to return an increased PWPD proton density in bone. If we assume that independent changes in the bone’s organic matrix during normal bone aging and osteoporosis development are limited, alterations in PWPD should be seen in TWPD. Therefore, investigating only TWPD in cortical bone, which merely requires a quick UTE-MRI scan, may be enough for initial bone evaluation. An earlier feasibility study on eight tibial cortical bone samples demonstrated significant correlations between cortical bone microstructural parameters and TWPD [16]. Previous study was performed on a clinical MRI scanner using a standard eight channels knee coil and demonstrated that TWPD technique can be potentially translated to clinical investigation. However, such correlations have not been investigated in a large number of samples from a considerable number of donors, which is necessary before embarking on clinical translational studies.

The goal of this study was to investigate the correlations between cortical bone microstructural properties and TWPD obtained from a single UTE-MRI sequence performed ex vivo on a large number of human bone samples. This study highlights the potential applications of a quick UTE-MRI scan to assess human cortical bone.

2. Materials and methods

2.1. Cortical bone strips preparation

135 cortical bone samples were harvested from the femoral and tibial midshafts of 37 donors (61±24 years old, 52 samples from 19 females, 83 samples from 18 males). Fresh frozen cadaver legs were obtained from a non-profit whole-body donation company (United Tissue Network, Phoenix, AZ, USA). The cortical bone midshafts were cut into 40 mm segments using a commercial band saw. Two approximately rectangular bone strips were harvested from each midshaft segment using a low-speed diamond saw (Isomet 1000, Buehler, IL, USA). The final size of the samples was approximately 4×2×40mm³. Samples were soaked in phosphate-buffered saline (PBS) solution overnight at room temperature before the MRI scans. Then, 15–20 samples were placed into 30-mL syringes (8 groups) filled with perfluoropolyether (Fomblin, Ausimont, Thorofare, NJ) to minimize susceptibility artifacts and to avoid sample dehydration. One rubber phantom with a known apparent proton density of 33 mol/L H¹ was placed in each syringe and scanned together with the bone strips. The rubber phantom proton density was measured previously by comparing the UTE-MRI signals with the signal of a water phantom (20% H₂O and 80% D₂O, 22 mol/L H).

2.2. UTE-MRI protocol

The UTE-MRI scans were performed on a 3T clinical scanner (GE Healthcare, Waukesha, WI, USA) using a transmit/receive birdcage coil (25.4 mm inner diameter). Two 3D-UTE-Cones MRI sequences were performed. First, to measure TWPD, a single UTE sequence was performed with the following acquisition parameters: rectangular RF excitation pulse with a duration of 26 μs, repetition time (TR)=100 ms, echo time (TE)=0.032 ms, flip angle (FA)=10˚, field of view (FOV)= 40mm, matrix size=160×160, in-plane pixel size=0.25mm, slice thickness=2 mm, receiver bandwidth=±62.5 kHz. Second, a 30-mL syringe filled with pure water was imaged using the UTE-MRI protocol to generate the coil sensitivity map (η) over the selected FOV. The total scan time was approximately 7 minutes for UTE-MRI and coil sensitivity. The details of the 3D-UTE-Cones sequence have been given in the literature [21–23].

2.3. Calculating TWPD in bone strips

Water proton density calculation in bone samples was performed by MRI signal comparison between bone samples and the rubber phantom with a known apparent proton density (33 mol/L H¹, T2≈1.3 ms, T1equals to280 ms) [3,8,14–17,20]. UTE MRI signal can be described using the Ernst equation, as presented in Eq.1 [11].

$$S{I}_{\mathit{\text{Bone}}}\left(\text{TE}\right)\propto \frac{1-e^{-TR/T_1}}{1-\text{cos}\theta \times e^{-TR/T_1}}\times e^{-TE/T_2^{\text{*}}}\times \mathit{\text{TWPD}},$$ Eq.1

where TR, θ, and T₁ are repetition time, FA, and bone total water longitudinal relaxation time, respectively. A proton density-weighted UTE acquisition was used in this study to simplify the TWPD calculation and minimize potential errors (e.g., a relatively long TR of 100 ms, a short TE of 32 μs, and a low FA of 10° employing a short rectangular excitation pulse of 26μs). Since T2*_bone and T2*_REF are much higher than the ultrashort TE and the rectangular excitation pulse duration, the T2* and T1 effects in Eq.1 can be neglected; thus, the TWPD can be estimated by comparing the UTE signals of bone and external reference using Eq.2.

$$\mathit{\text{TWPD}}\approx \frac{S{I}_{\mathit{\text{Bone}}}^{UTE}}{S{I}_{Rub}^{UTE}}\times \eta \times {\rho }_{Rub},$$ Eq.2

where η and ρ_Rub are coil sensitivity and proton density in the rubber phantom, respectively. It should be noted that average signal intensity in each bone specimen was used for TWPD calculation.

Using the simplified Eq.2 instead of Eq.1 may result in some errors in calculation of TWPD. Based on our earlier measurements, T1 of TW bone is 350 ms and T1 of the rubber phantom is 280 ms. Therefore, the T1 correction term in Eq.1 equals 0.957 and 0.966 for TW bone and rubber phantom, respectively, when TR=100ms and FA=10˚. This may lead to an error of less than 1% in TWPD calculation. The main difference between the rubber phantom and bone will be in their T2* values. Based on our earlier measurements, T2* of TW bone is 0.4 ms and T2* of the rubber phantom is 1.3 ms, which results in the T2* correction terms of 0.975 and 0.923, respectively. This may lead to an error of around 5% in TWPD calculation. Consequently, the total combined potential error in TWPD estimation will average 6 % when neglecting the T1 and T2* terms in Eq.1. The expected errors would be slightly different if separate contributions of PW and BW pools were assumed in the calculations. Based on our earlier measurements, T1 of PW is 500 ms and T2* of PW is 3 ms, which results in the correction terms of 0.9357 and 0.989, respectively. Comparing these correction values with those of the rubber phantom leads to an error of 5% in PWPD calculation. On the other hand, T1 of BW is 150 ms and T2* of BW is 0.3 ms, which results in the correction terms of 0.985 and 0.898, respectively. Comparing these correction values with those of the rubber phantom leads to an error of 6% in BWPD calculation. Therefore, the maximum expected error for measured water proton density can reach up to 6%.

2.4. Bone microstructure with μCT

A Skyscan 1076 (Kontich, Belgium) μCT scanner at 9 μm isotropic voxel size was used to scan the bone strips. The μCT scan time was approximately 12 hours. Two cylindrical hydroxyapatite phantoms (0.25 and 0.75 g/cm³) were placed in the field of view enabling BMD measurements. A 0.05-mm aluminum filter in addition to a 0.038-mm copper filter were utilized in the μCT scans. Other scanning parameters were as follows: 100 kV, 100 mA, 0.3˚ rotation step, and 5 frame-averaging.

The μCT images were segmented using a gray intensity threshold. Gray intensity histograms were used to select the gray intensity threshold for the segmentation process. Intracortical bone porosity was estimated as the number of voxels in pores divided by the total number of voxels covering each bone sample. The pore size at each pore voxel was defined as the diameter of the largest covering sphere. Local BMD value at each voxel was assumed as a linear function of its gray intensity, which was derived from the average gray intensities of the scanned hydroxyapatite phantoms. Finally, the average BMD for each sample was calculated by averaging the local BMD values over all corresponding voxels.

2.5. Statistical correlations

Pearson’s correlation coefficients were calculated between TWPD and μCT-based measures including porosity, pore size, and BMD. Correlations with p-values below 0.05 were considered significant. All image processing and statistical analyses were performed using MATLAB (version 2018, The Mathworks Inc., Natick, MA, USA) codes developed in-house.

3. Results

Figure 1a shows the UTE-MRI image (TE=0.032 ms) of a set of bone samples in the 30-ml syringe scanned in axial plane, showing the average 4mm × 2mm cross sections of samples. The rubber phantom with the known proton density placed in the syringe for TWPD measurement is indicated with a yellow arrow in the UTE-MRI image. Figure 1b shows the μCT image of the same set of samples at 9-micron isometric voxel size.

Figure 1:

UTE-MRI image and μCT image of a set of twenty cortical bone strips harvested from different donors possessing different levels of porosity. (a) UTE-MRI (TE=0.032ms) image of a set of twenty cortical bone strips harvested from different donors with 4×2 mm² cross-sections (250 μm in plane pixel size) soaked in fomblin, which has no signal in MRI. UTE scans of the samples were performed in the presence of a rubber phantom with known proton density (33 mol/L). (b) μCT image of the same set of bone samples at 9 μm voxel size.

Figures 2a, 2b, and 2c demonstrate the scatter plots and linear regressions of TWPD on μCT-based average bone porosity, pore size, and average BMD, respectively. TWPD demonstrated significant moderate correlation with both average bone porosity (R=0.66, p<0.01) and pore size (R=0.58, p<0.01). TWPD demonstrated significant strong with BMD (R=0.71, p<0.01).

Figure 2:

Scatter plots and linear regression analyses with significant correlations (p<0.01) of total water proton density (TWPD) on μCT-based microstructural properties. TWPD versus (a) porosity, (b) average pore size, and (c) bone mineral density (BMD).

It should be noted that a set of inversion recovery (IR) UTE sequences were performed to measure BWPD and consequently PWPD of the studied bone specimens using the described equations in an earlier study [16]. Since the focus of this article was employing a relatively fast MRI scan for TWPD estimation to be used for bone microstructural assessment, the correlations of PWPD and BWPD versus microstructural parameters are presented in Supplementary Data, Figure 1. PWPD correlations with microstructural parameters were slightly higher than the TWPD correlations with bone microstructure. BWPD did not show significant correlations with microstructural parameters.

4. Discussion

We have measured TWPD in cortical bone and investigated its correlations with cortical bone microstructure in a large number of bone samples (i.e., 135) from a considerable number of donors (i.e., 37). Earlier feasibility studies on TWPD and bone microstructure correlations were performed on a limited number of tibial bone samples [16]. UTE-MRI-based techniques for bone assessment can be considered non-invasive and readily translatable to in vivo studies. Moreover, the bone MRI-based assessment can provide invaluable evaluation of the surrounding soft tissues such as tendons and muscles, which can improve diagnostic procedures.

For accurate estimation of bone water protons we must consider: i) the difference between relaxation times of cortical bone and the rubber phantom, ii) the spatial variation of coil sensitivity in scanned FOV, and iii) the duration of RF pulse and its inhomogeneity (or actual flip angles) [11]. Due to short T₁ in cortical bone and due to the use of a relatively low FA and relatively high TR in the PD-weighted 3D-UTE-Cones sequence, the T1 effect on the TWPD calculation could be neglected in this study. The actual FA is a function of RF transmit field (B1). Since FA used in this study has been relatively low (FA=10˚), a regular B1 variation within the studied tissue (e.g., ≈5%) may not lead to considerable errors in TWPD calculation. Since the T2*s of the rubber phantom and bone were significantly higher than ultrashort TE of 32μs, the T2* term in the proton density measurement could also be neglected in this study (Eq. 1).

To validate the correlations between the TWPD values and the bone microstructural properties, μCT was employed for bone porosity and BMD measurements. TWPD showed significant moderate correlation with bone porosity and pore size (R=0.66 and 0.57, respectively, p<0.01, Fig.2). TWPD also showed significant strong correction with BMD (R=0.71, p<0.01, Fig.2). Presented correlations were in the range of previously reported correlations for eight tibial bone cross-sections imaged in clinically used knee coils [16]. This study complements the earlier feasibility study and validates earlier obtained correlations. As presented in the supplementary figure, PWPD correlations with microstructural parameters were higher compared with TWPD correlations. However, BWPD did not show significant correlations with microstructural parameters in this study for three potential reasons. First, the range of BWPD was relatively restricted, potentially limiting the results of the regression analysis. Second, because of the cadaveric bone specimens used in this study, BW restoration may not have completely occurred in some specimens even after overnight soaking in PBS. Third, the rubber phantom used in this study was not pure short T2 material and contained some impurity which make it improper for BWPD measurement.

In addition to TWPD estimation [8,10–16], other UTE-MRI techniques have also demonstrated significant correlations with intracortical bone porosity. Bicomponent T2* fitting technique [7,24] can estimate short-T2* (i.e., BW pool) and long-T2* (PW pool) proton fractions and their average T2* values in cortical bone [7,24,25]. Estimated BW and PW fractions can predict bone microstructure indirectly [7,24,25]. Tricomponent T2* fitting technique [9,25] improves bicomponent analysis by estimating the fat proton fraction in addition to the water proton fractions. Tricomponent fitting avoids the common BW overestimation by bicomponent analysis in endosteal side of cortex [9,25]. Bi- and tricomponent T2* fitting techniques do not estimate absolute water proton contents. Such techniques need a series of MRI images with differing TEs which can extend the scanning process over 30 mins. Quantitative susceptibility mapping (QSM) [26] is another method to indirectly estimate bone mineral content as an index for bone microstructure. The larger the QSM, the higher the mineral content. QSM requires a set of MRI images with various TRs and TEs that can lead to scan times over 30 mins. Dual-echo UTE imaging [27] is faster technique (can be performed in 7 minutes) that calculates the signal ratio between two MRI images, one with TE<0.05ms and one with TE≈2ms. Although, this technique does not estimate the absolute water proton contents, it can estimate porosity indirectly. Direct pore water imaging after nulling bound water [8,14,19,20] is another relatively fast MRI scan that can estimate the PWPD. This technique requires an excellent nulling of bound water signal which can be challenging. Magnetization transfer modeling technique [7,16,28,29] has been also reported for bone microstructural assessment. This technique can estimate bone’s macromolecular proton fraction (MMF) relative to water fraction using a series of MT pulse saturation powers and frequency offsets. MMF is an index related to bone’s collagen matrix which can be used to detect bone stress injuries [30]. MT modeling technique may require approximately 15 mins scan time for the setup used in this study.

This study was performed ex vivo on bone samples cut from pure cortical bone layers of femoral and tibial midshafts where there was negligible bone marrow and no surrounding muscle. The samples were scanned in fomblin using a small coil that resulted in high SNR. In the light of the abovementioned facts, the performance of the presented technique will likely drop when translated to in vivo studies due to the presence of soft tissues, lower spatial resolutions, lower SNR, temperature differences [31], and subject motion during scan. Thus, further in vivo studies are still necessary to compare cortical bone TWPD and its measurable microstructural parameters. Human cortical bone often has considerable amount of fat due to the bone marrow infiltration into intracortical pores. The fat presence in cortical bone results in an increased UTE-MRI signal (Figure 2a) which eventually lead to TWPD overestimation. Such TWPD overestimations can be reduced by employing lower FAs (e.g., FA≈2˚). Employing tricomponent T2* fitting model has been proposed in previous studies for accurate measurement of water proton fraction by considering the fat-water chemical shift contribution in UTE-MRI signal [9,25]. Nevertheless, tricomponent T2* modeling will require a relatively long scan time which is not aligned with the main objective of this study; proposing a relatively fast UTE-MRI technique for bone microstructural assessment.

5. Conclusion

A quick 3D-Cones-UTE MRI technique was presented to measure total water proton density in a large number of cortical bone samples. Water proton density showed significant moderate to strong correlations with bone microstructural properties, as measured with high resolution μCT. Such significant correlations demonstrated the potential of UTE-MRI-based measurement of bone water proton density to assess bone microstructure.

Abbreviations:

MR

magnetic resonance

MRI

magnetic resonance imaging

3D

three-dimensional

3D-UTE

three-dimensional ultrashort echo time imaging

RF

radio frequency

FOV

field of view

ROI

region of interest

TE

echo time

TR

repetition time

CT

computed tomography

μCT

micro-computed tomography

FA

flip angle

TWPD

total water proton density

PWPD

pore water proton density

BWPD

bound water proton density

BMD

bone mineral density

PBS

phosphate-buffered saline

DEXA

dual-energy X-ray absorptiometry

QCT

quantitative computed tomography

HR-pQCT

peripheral quantitative computed tomography