C-arm CT for histomorphometric evaluation of lumbar spine trabecular microarchitecture: a study on anorexia nervosa patients

Abstract

Bone histomorphometry measurements require high spatial resolution that may not be feasible using multidetector CT (MDCT). This study evaluated the trabecular microarchitecture of lumbar spine using MDCT and C-arm CT in a series of young adult patients with anorexia nervosa (AN). 11 young females with AN underwent MDCT (anisotropic resolution with a slice thickness of ∼626 μm) and C-arm CT (isotropic resolution of ∼200 µm). Standard histomorphometric parameters the of L1 vertebral body, namely the apparent trabecular bone volume fraction (BV/TV), trabecular thickness (TbTh), trabecular number (TbN) and trabecular separation (TbSp), were analysed using MicroView software (GE Healthcare, Piscataway, NJ). Bone mineral density (BMD) was measured using dual-energy X-ray absorptiometry. Trabecular parameters derived from MDCT and C-arm CT were compared, and their association with BMD parameters was evaluated. Histomorphometric parameters derived from C-arm CT, namely TbTh, TbN and TbSp, were significantly different from the corresponding MDCT parameters. There were no significant correlations between C-arm CT-derived parameters and the corresponding MDCT-derived parameters. C-arm CT-derived parameters were significantly (p<0.001) correlated with anteroposterior L1 spine BMD and Z-scores: TbTh (r=0.723, r=0.744, respectively), TbN (r=−0.720, r=−0.712, respectively) and TbSp (r=0.656, r=0.648, respectively). BV/TV, derived from C-arm CT, was significantly associated with body mass index (r=0.636) and ideal body weight (r=0.730) (p<0.05). These associations were not present in MDCT-derived parameters. This study suggests that the spatial resolution offered by C-arm CT more accurately captures the histomorphometric parameters of trabecular morphology than MDCT in patients with AN.

Anorexia nervosa (AN) is associated with substantial bone loss [1,2] and low bone mineral density (BMD) [3]. Additionally, both the spatial arrangement of trabeculae and the mechanical properties of bone are significantly impaired in AN [4,5]. Even though the parameters of trabecular microarchitecture do not include information about mechanical properties or tissue-level strain, they have been shown to distinguish healthy from diseased bone better than BMD alone [6,7]. In fact, Bredella et al [4] recently demonstrated that parameters of trabecular microstructure are frequently impaired in mild-to-moderate AN, despite normal dual-energy X-ray absorptiometry (DXA) measures of BMD.

Although previous studies have reported trabecular microarchitecture of the peripheral skeleton in adult and adolescent patients with AN [5,8], no study to date has fully evaluated the trabecular structure of the axial skeleton such as spine. Bone histomorphometry requires high spatial resolution imaging that may be limited using standard multidetector CT (MDCT). C-arm CT (DynaCT; Siemens Medical Solutions, Erlangen, Germany) is an imaging modality, where a wide-angle cone-beam X-ray tube and a high resolution flat-panel detector are integrated with a C-arm gantry [9]. By providing an open platform for imaging, this system offers a greater flexibility in orienting the detector around the patient than a closed CT gantry, therefore allowing for imaging of the entire body [10,11]. Furthermore, C-arm CT can be utilised to estimate the parameters of trabecular microarchitecture of the axial skeleton, most importantly of the spine and hip, which are the common locations of osteoporotic fractures [12,13]. The spatial resolution of C-arm CT is excellent and is able to resolve the thickness of trabeculae with an isotropic spatial resolution of 200 μm [14,15]. In comparison, the conventional MDCT system (GE Discovery™ CT750 HD; GE Healthcare, Waukesha, MI) has an in-plane resolution of 230 μm, with a slice thickness of 625 μm (vendor specifications). Because the size range of the bone trabeculae (50–200 μm) is finer than the resolution offered by MDCT, this conventionally used imaging technique tends to overestimate the trabecular thickness (TbTh) and spacing and underestimate the trabecular number (TbN) [11].

In this study, we aim to evaluate the trabecular microarchitecture of the lumbar spine using C-arm CT in a series of young adult patients with AN. We compared the C-arm CT-derived histomorphometric parameters and the corresponding MDCT-derived parameters with DXA measures of BMD to determine which set of CT parameters has stronger associations with DXA-BMD.

METHODS AND MATERIALS

Study design

This study followed the Health Insurance Portability and Accountability Act guidelines and was approved by the institutional review board of Massachusetts General Hospital, Harvard Medical School, Boston, MA. Written informed consent was obtained from all subjects. Between February 2009 and November 2010, 11 young females [mean age 22.7±2.3 years; mean body mass index (BMI) 17.0±1.5 kg m⁻²] were studied (Table 1). All subjects fulfilled Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) criteria for AN [16]. Exclusion criteria included pregnancy, inability to undergo a 10- to 15-min CT examination of the lumbar spine and history of lumbar trauma with bone injury and/or orthopaedic hardware instrumentation. The subjects underwent MDCT system (GE Discovery CT750 HD) and C-arm CT (DynaCT) scans of the L1 vertebral body to determine histomorphometric parameters. In addition, patients were assessed using DXA (QDR 4500; Hologic®, Waltham, MA) and had their BMD (g cm⁻²) and Z-scores measured at their lumbar spine, hip and distal radius.

Table 1.

Clinical characteristic of the study participants (n=11)

Variables	Mean ± standard deviation	Range
Age (years)	22.69±2.31	19–26
Menarche (years)	14.38±1.37	12–16
Onset of AN (years)	18.27±3.09	13–23
Duration of AN (months)	52.78±49.46	2–144
BMI (kg m−2)	17.01±1.53	14.63–19.04
%IBW	77.42±7.23	65.92–88.91
AP L1 BMD (g cm−2)	0.76±0.09	0.58–0.91
AP L1 Z-score	−1.28±0.73	−2.60–0.0
AP tot. L-spine BMD (g cm−2)	0.83±0.09	0.65–0.99
AP tot. L-spine Z-score	−1.81±0.90	−3.60 to −0.30
Lat tot. L-spine BMD (g cm−2)	0.58±0.06	0.48–0.69
Lat tot. L-spine Z-score	−2.79±0.79	−3.90 to −1.38

AN, anorexia nervosa; AP, anteroposterior; BMD, bone mineral density; BMI, body mass index; %IBW, percent of ideal body weight; Lat, lateral; tot., total.

C-arm CT image acquisition

C-arm CT was performed to obtain 200 μm³ resolution CT images of the first lumbar vertebral body, which were used for assessing the trabecular microarchitecture of the vertebral spine (Figure 1). Subjects were imaged in the supine position with the spine in the centre of the gantry. The L1 vertebral body was localised via an anteroposterior and a lateral scout image. After localisation, three-dimensional (3D) imaging of approximately 4 cm centred on the mid-L1 vertebral body was obtained using a low-dose protocol with the following parameters: 80 kV, automatic tube current modulation ranging from 70 to 125 mA, 20 s rotation time leading to an approximate radiation dose of 0.7 mSv that is comparable with the background source radiation that one normally receives over a 2.5-month period [17]. In order to increase the signal-to-noise ratio, the scanner was operated in a 2×2 binning mode, providing a 200 μm isotropic spatial resolution [14]. Acquired images were then transferred to a dedicated post-processing workstation (Leonardo; Siemens Medical Solutions), where a volume data set was reconstructed with a voxel size of 200 μm and an optimised kernel for visualisation of the bony structure (sharp bone kernel).

Figure 1.

Multiplanar reconstruction of L1 vertebral body obtained with C-arm CT (a–c) and multidetector CT (d–f). The trabecular details, which can be used for histomorphometry, are better depicted in C-arm CT, especially in sagittal and coronal reconstruction. The panels present axial (a, d), sagittal (b, e) and coronal (c, f) orientation.

Each subject was scanned with a calibration phantom to standardise the greyscale values and maintain consistency. The calibration phantom was placed as close as possible to the L1 vertebral body. It consists of three test tubes embedded within a foam cushion. The first test tube contained normal saline to provide calibration for 0 Hounsfield Units (HU), whereas the other two test tubes contained calibrated solutions of Ca₂(HPO₄)₃ that measured 130 and 200 HU on a conventional MDCT. The in-house calibration phantom has been used for previous studies [4,5].

Multidetector CT image acquisition

For MDCT imaging, patients were also asked to wear a lead vest and a lead skirt to minimise radiation to other parts of their body. A 4 cm extent of the lumbar anatomy centred at L1 was imaged using low-dose spiral mode scanning at 100 kV, 220 mA and 0.5 s rotation time. Images were reconstructed with 0.625-mm section thickness, using a sharp kernel, a field of view (FOV) of 100 mm and a 512×512 matrix. The estimated radiation dose was approximately 0.9 mSv (CT dose index 15.74). Each subject was scanned with the calibration phantom to standardise greyscale values and maintain consistency.

Measurement of trabecular microarchitecture

Trabecular structure parameters were determined using MicroView software (GE Healthcare, Piscataway, NJ). For each L1 vertebral body, a 3D oval region of interest (ROI) was defined within the vertebral body to cover a maximum area of trabecular bone without including any cortical bone within the ROI. ROIs were customised to the cross-sectional area of each scan. To improve reproducibility, all ROIs were placed by one observer (CMP, with 10 years of experience in radiology). The interreader reproducibility of C-arm CT [coefficient of variation (CV) 2.66–4.55%] is reported to be comparable with that of MDCT (CV 2.41–4.08%) [11]. Trabecular bone was segmented from bone marrow using the automatic threshold level function of the software [18–20]. This function provides automated segmentation by generating the attenuation histogram of the volume of interest and fitting the data by using the Otsu algorithm [21]. By using standard methods from histomorphometry [22], the following measures of trabecular structure were calculated: apparent trabecular bone volume fraction (BV/TV) as a percentage, apparent TbTh in millimetres, apparent TbN in 1 per millimetre and apparent trabecular separation (TbSp) in millimetres. The parameters of trabecular microarchitecture are labelled as being “apparent” owing to the fact that the spatial resolution we used for our analysis is lower than that required for standard bone histomorphometry [22,23].

Statistical analysis

Statistical analysis was performed with R software, v. 2.13 (R Foundation for Statistical Computing, Vienna, Austria). The required sample size to detect a significant association at α=0.05 and with a power of 80% was estimated to be 11. Continuous variables are expressed as mean ±standard deviation. Kolmogorov–Smirnov test was used to test the normality of the C-arm CT, MDCT and DXA parameters. All variables were normally distributed. Different trabecular parameters were compared between C-arm CT and MDCT using the independent sample t-test. The strength of association of C-arm CT and MDCT variables with DXA variables was measured using Pearson’s correlation coefficients. p<0.05 was considered to be statistically significant.

RESULTS

Characteristics of the study participants are presented in Table 1. All C-arm CT and MDCT scans were of sufficient quality for trabecular structure analysis (no relevant reconstruction artefacts were noted). As presented in Table 2, apart from BV/TV, other trabecular parameters (including TbTh, TbN and TbSp) measured using C-arm CT differed significantly from similar parameters measured by MDCT.

Table 2.

Differences between trabecular parameters derived from C-arm CT vs MDCT

Trabecular parameters	C-arm CT	MDCT	p
BV/TV (%)	0.49±0.02 (0.46–0.51)	0.49±0.02 (0.43–0.51)	0.234
TbTh (mm)	0.35±0.01 (0.32–0.36)	0.46±0.05 (0.40–0.54)	<0.001
TbN (1 per mm)	1.34±0.27 (0.53–1.50)	1.06±0.14 (0.83–1.26)	0.004
TbSp (mm)	0.41±0.16 (0.34–0.95)	0.50±0.08 (0.40–0.68)	0.049

BV/TV, apparent trabecular bone volume fraction; MDCT, multidetector CT; TbN, trabecular number; TbSp, trabecular spacing; TbTh, trabecular thickness.

Variables are presented as mean ± standard deviation and range is given in parentheses.

There were no significant correlations between C-arm CT and the corresponding MDCT-derived parameters. The correlation among parameters derived from C-arm CT and MDCT is presented in Table 3. C-arm CT-derived parameters were significantly correlated with anteroposterior L1 spine BMD and Z-scores (p<0.05): TbTh (r=0.723, r=0.744, respectively), TbN (r=−0.720, r=−0.712, respectively) and TbSp (r=0.656, r=0.648, respectively). These associations were not present for MDCT-derived parameters.

Table 3.

Correlation coefficients between C-arm CT, MDCT and BMD parameters

Trabecular parameters	For C-arm CT parameters		For MDCT parameters
	AP L1 BMD	AP L1 Z-score	AP L1 BMD	AP L1 Z-score
TbSp	0.656a	0.648a	0.461	0.400
TbTh	0.723a	0.744a	0.595	0.537
TbN	−0.720a	−0.712a	−0.535	−0.471
BV/TV	−0.004	0.062	−0.156	−0.112

AP, anteroposterior; BMD, bone mineral density; BV/TV, apparent trabecular bone volume fraction; MDCT, multidetector CT; TbN, trabecular number; TbSp, trabecular spacing; TbTh, trabecular thickness.

^ap<0.05.

BV/TV, derived from C-arm-CT, was significantly (p<0.05) associated with the BMI (r=0.636) and percent ideal body weight (%IBW, r=0.730). This association was not significant for the MDCT-derived parameters. There was no significant association between the BMI (or %IBW) and L1 vertebral BMD and Z-scores.

DISCUSSION

Previous research has shown a high prevalence of osteoporosis [24,25] with decreased BMD Z-scores [26] at the lumbar spine of adolescent and adult AN patients. Therefore, visualisation of trabecular structure of the spine in these patients is of particular clinical interest. Conventional MDCT does not have adequate resolution for this purpose and could only provide information about trabecular architecture, after complex analysis, while specialised scanners such as micro-CT and Xtreme-CT (SANCO Medical AG, Brüttisellen, Sweden) do not have the FOV to image the axial skeleton [27]. Our study suggests that C-arm CT can be used for assessing the trabecular microstructure of the axial skeleton in patients with AN.

In our study, trabecular parameters (TbTh, TbN and TbSp) derived from MDCT were significantly different from the corresponding C-arm CT parameters. This is an interesting point, and one may justify it as over- or underestimation of the trabecular measures by MDCT owing to its lower spatial resolution. Considering the higher resolution of C-arm CT, it is expected that C-arm CT could offer a better depiction of the trabecular structure, interconnectivity and the edge sharpness of the cortical rim than MDCT. Another finding that points to the higher accuracy of C-arm CT is that its derived parameters had more significant associations with DXA estimates of BMD in comparison to the corresponding MDCT parameters.

Recent studies have utilised flat-panel Volume CT (fpVCT) to assess the trabecular microarchitecture and to calculate the mechanical properties (failure load and stiffness) by using finite-element analysis of the distal radius in adolescent girls, as well as adult females with AN [5,8,28]. C-arm CT provides similar isotropic spatial resolution to the fpVCT resolution of 200 μm, and thus both modalities are able to adequately differentiate the thickness of the trabeculae. The main advantage of C-arm-CT over fpVCT is its open platform, which allows imaging of the axial skeleton [11].

In patients with AN, the axial skeleton is a key area for monitoring disease. These patients have an increased risk of low BMD and an increased fracture risk, including spinal compression fractures, compared with age-matched healthy individuals [2,29]. Therefore, the primary site of interest in this study was the lumbar spine. The potential advantage of analysing the parameters of trabecular microstructure of the axial skeleton is that it is the most frequent location of osteoporosis in adolescent and adult patients with AN [1,2,26], as well as a frequent location of osteoporotic fractures [2,29]. Therefore, trabecular microarchitecture analysis by C-arm CT may prove to be a valuable new technique for evaluating fracture risk of the axial skeleton. Further research is needed to determine the feasibility of C-arm CT for the stratification of fracture risk.

For interpretation of our results, some points should be considered. First, in this study, owing to ethical limitations of scanning healthy individuals with ionising radiation, we did not recruit a healthy control group. Therefore, differences in trabecular microstructure between patients with AN and healthy young adult females could not be assessed. Nevertheless, this point may not be a major limitation, because the main aim of this study is to compare MDCT and C-arm CT estimates of trabecular structure in patients with AN. Second, it should be noted that the use of C-arm CT as a technique for assessment of trabecular microarchitecture may be limited to specific subgroups of patients owing to FOV limitations of this modality (the size and weight of the patient may influence this). However, in all patients with AN included in our cohort, we were able to successfully quantify trabecular structure parameters despite this theoretical limitation. Future studies could be done in older patients with primary osteoporosis, and the feasibility of this technique assessed in these patients who may not be as slim as patients with AN and the scatter correction might be more challenging. The study was designed and conducted based on a pre-determined sample size of 11 to have a power of 80% and α=0.05. Nevertheless, longitudinal and larger studies in AN patients with osteoporotic fractures could provide a more comprehensive assessment of the role of C-arm CT in patients with AN. We believe that our results and conclusions are a statement about the spatial resolution of MDCT and C-arm CT. Recent advances in MDCT spatial resolution (ultra high resolution [27]) may improve in vivo assessment of trabecular bone structure in patients with AN. We encourage future studies to compare newer generations of MDCT and C-arm CT for evaluation of bone microstructure in patients with AN.

In conclusion, this study suggests that the spatial resolution offered by C-arm CT (approximately 200 μm, isotropic) more accurately captures the histomorphometric parameters of trabecular morphology than conventional MDCT in patients with AN (anisotropic spatial resolution of approximately 230×230×625 μm³).

FUNDING

This work was supported by the following grants from the National Institute of Health—DK052625 and 2 UL1 RR025758.