This precision study was conducted in a medical imaging research centre setting and received academic institutional ethical approval from the University of British Columbia, Vancouver Canada. Community-dwelling adults were recruited from a large urban metropolitan setting. Participants received no financial remuneration for participation and provided informed consent to participate. With the exception of a physician diagnosis of inflammatory arthritis, participants were not screened for any other self-reported health (e.g. diabetes, osteoporosis) or lifestyle (e.g. smoking, alcohol consumption, physical inactivity) condition that may have affected their bone health. We specifically excluded individuals with a diagnosis of inflammatory arthritis as we were not be able to determine a priori if they may already have underlying macro-structural bone damage in the regions of bone we were examining. Participants were also excluded if they: 1) had any physical condition that would prevent them from sitting motionless with their arm in the scanner supported by a positioning device for up to 6 minutes, 2) had metal or surgical implants in the hand or forearm of interest, 3) were pregnant or possibly pregnant, 4) had sustained a fracture in their dominant arm hand or forearm in the previous 12 months, and 5) were unable to read or understand the consent form.
Prior to scanning we assessed height (cm) using a wall mounted stadiometer (SECA corp. Chino, CA) and weight (kg) using a medical grade digital floor scale (Tanita Corporation of America, Inc. Arlington Heights, Ill) using standard techniques. We derived body mass index (BMI) as wt/ht (kg/m). Following these anthropometric measures, the hand and forearm were positioned in a custom-made positioning device made of rigid thermoplastic splinting material. The forearm was aligned parallel to the long axis of the splint and the metacarpal phalangeal joints positioned in 0 degrees of flexion. The splint-supported hand and forearm were then positioned within a holder that was modified from manufacturer specifications to suit the hand (Scanco Medical AG, Switzerland). The hand and forearm were then stabilized with additional strapping (Figure 1A). Participants were positioned to face the imaging system. Pillows were placed behind participants' hips and in front of them so that the participant could lean forward and rest on the pillows with their opposite arm, upper body and head comfortably supported. The holder, with the arm correctly positioned within it, was then placed inside the HR pQCT unit for scan acquisition (Figure 1B).
Custom image acquisition positioning. A) Shows the standardized positioning of the hand and forearm (left or right) in a custom-made insert (top) with additional stabilization and placement in a modified manufacturer ex-vivo holder (bottom). B) Shows the modified positioning for imaging with an individual seated on a chair facing scanner with their head, upper body and opposite arm resting on pillows with the hand to be scanned in the holder and positioned inside the scanner for scanning.
A single trained operator (author LF) performed all scans using standard in vivo imaging parameters (82 μm nominal isotropic resolution, 60 kVp effective energy, 900 μA current, and 100 ms integration time). The training involved a rigorous and standardized training protocol developed by the facility for the safe operation of the scanner. Manufacturer specifications for the scanner define that for every 110 slices acquired the measurement time is 2.8 minutes with an effective dose of 3 μSv at distal extremity sites. This estimate of effective dose is based on a weighted computed tomography dose index (CTDIw) of 6.1 mGy and a local dose of 3.2 mGy using standard HR-pQCT in vivo image acquisition parameters. A trained operator also performed daily density calibrations and weekly geometry calibrations of the HR-pQCT imaging system using the manufacturer's calibration phantom.
Three scans of the dominant arm were completed in series during a single scanning session. The ROIs included the metacarpal head (MH), metacarpal mid-shaft (MS) and ultra-ultra-distal (UUD) radius sites. To assess short-term precision with repositioning, we acquired two additional series of three scans with repositioning between each series. The additional two series were completed during a single scanning session, three to seven days after the initial scans.
Prior to each scan, we performed a 150 mm length scout view of the hand and distal forearm which is the maximum available length for a scout view. The reference line for the radius scan was located at the medial edge of the distal radius; the scan region was 1 mm proximal to this reference line and extended 9.02 mm (110 slices) proximally. For the metacarpal head scan, the reference line was the tip of the most distal second or third metacarpal head; the scan started 2 mm distal to this reference line and extended 18.04 mm (220 slices) proximally. For the metacarpal shaft scan, the reference line was half (50%) the total length of the metacarpal shaft assessed on the scout view. The metacarpal shaft scan region of interest extended from 4.5 mm distal to the reference line to 9.02 mm (110 slices) proximal to the reference line (Figure 2 A, B, C).
Scan locations and cortical segmentation. Top Row (A,B,C) shows the reference line, scan location and Region of Interest (ROI) analyses overlaid on a 150 mm scout view for the Ultra-Ultra-Distal Radius (A), Metacarpal Head (B) and Metacarpal Shaft (C) scans. Bottom Row (D,E,F) shows examples of semi-automated cortical compartment segmentation in one HR-pQCT slice for the UUD radius (D), Metacarpal Head (E) and Metacarpal shaft (F) ROIs.
The operator visually assessed all images for motion artifact at the completion of the three-scan series. If motion artifact was apparent in only one image the operator repeated the scan. If there was motion artifact in two or more of the scans across the series, the operator repeated the scan at one site only. Our image order of priority was the distal radius followed by the metacarpal head.
Images were then independently analysed by 1 of 2 trained and experienced operators, one of whom was the same person as the image acquisition operator in this study (first author LF), the other a study research assistance. Before conducting any image analysis in this study, each operator was required to obtain an intra-rater reliability coefficient (Pearson R) of ≥ 0.90 for measures of UUD trabecular bone fraction from at least 10 images assessed twice by the same operator within 7 to 10 days.
Prior to analysis, each image was graded visually for motion artifact using the 5-point manufacture grading system. We included images graded 3 or less by both operators for final data analysis; any disagreement was resolved by consensus. Image analyses were conducted based on operator availability; operators did not use image registration to evaluate repeated scans. Operators were blinded to previous image analyses data; we allowed at least 10 days between image analyses of a repeated scan in any individual by the same operator. Both operators assessed the same numbers of scan images.
Using the manufacturer evaluation software (V 6.0), the operator analyzed five sub-regions of interest [1 - UUD radius (110 slices); 2 - MH2 & MH3 (110 slices); 2 - MS2 and MS3 (110 slices)] (Figure 2, A,B,C). They performed semi-automated contouring of the periosteal bone surface and segmented bone from surrounding soft tissue using standard manufacturer evaluation script protocols. The operator extracted cortical and trabecular regions using the semi-automated segmentation method, but applied a modified boundary condition for analysis of the metacarpal head.
Following initial segmentation, the operator made minor adjustments to endosteal and periosteal contours as needed. This step included a visual inspection of the computer generated lines for delineation of the cortical region segmentation in all slices, making minor manual corrections to any deviations from accurate periosteal or endosteal surface delineation (Figure 2, D,E,F). Manual correction at this step was rarely indicated; usually only required for the correction of the endosteal edge delineation in a limited number of slices in any image. The most common reason for the need for any manual correction was in instances when there were very larger intra-cortical pores or large bi-cortical breaks created by vascular channels. These manual adjustment procedures have been described in further detail by Burghardt et al.,.
The operator then ran a series of evaluation scripts using the manufacturer evaluation software for assessment of the full, cortical and trabecular bone regions using direct transformation image analyses scripts adapted from standard microCT evaluation scripts recently developed for cortical bone and described in more detail by Nishiyama KK et al., and Liu XS et al.,. These adopted direct transformation evaluation scripts for HR-pQCT are now included in current upgrades of manufacturer evaluation software.
For the periarticular UUD Radius, MH2 and MH3 regions we examined apparent volumetric bone mineral density (vBMD) for the full (vBMDfull - mgHA/cm), cortical (vBMDCort - mgHA/cm) and trabecular (vBMDTrab - mgHA/cm) bone regions. We also examined selected microstructural morphometric bone parameters, including:
•Cortical bone: thickness (CtTh - mm) and porosity (CtPo - %).
•Trabecular bone: volume fraction (BV/TVtrab - %), number (TbN – 1/mm), thickness (TbTh - mm) and separation (TbSp - mm).
At the extra-articular MS2 and MS3 mid-shaft sites we examined full and cortical bone apparent volumetric BMD (vBMDfull & vBMDcort - mgHA/cm), as well as, cortical bone material bone mineral density (vTMDcort - mgHA/cm). In addition we examined the following selected micro- and macro-structural morphometric parameters:
•Full bone: volume (BVfull - mm), volume fraction (BV/TV full - %), section modulus – major direction (SMfull - mm), polar moment of inertia (pMOIfull - mm), and marrow space diameter (MSdia - mm).
•Cortical bone: thickness (CtTh - mm), porosity (CtPo - %), volume (BVcort - mm), volume fraction (BV/TVcort - %), section modulus – major direction (SMcort - mm), polar moment of inertia (pMOIcort - mm).
Direct transformation evaluation methods applied to images acquired using HR-pQCT, in vivo tend to overestimate some trabecular bone outcomes (TbTh, TbSp and BV/TVtrab). Therefore, the standard manufacturer HR-pQCT evaluation script applies a correction factor to these parameters to adjust for known differences. We also applied this correction factor to variables acquired at the UUD Radius, MH2 and MH3 sites so as to directly compare our data with values acquired using standard image evaluation methods at other bone regions. Trabecular bone volume fraction (BV/TVtrab_s) was derived using a standard approach [trabecular bone apparent volumetric bone mineral density (vBMDtrab) divided by 1200 mg/cm)]. Trabecular thickness (TbThs) and trabecular separation (TbSps) were derived using a standard approach; BV/TVs and 1 – BV/TVs divided by TbN, respectively. Standard evaluation of HR-pQCT images uses direct transformation methods to determine trabecular number (TbN) and full bone and trabecular bone apparent volumetric bone mineral density (vBMDfull and vBMDtrab). Therefore we did not apply conversion factors to these variables.
We assessed short-term precision of repeated measures with repositioning using intraclass correlational coefficient (ICC), mean coefficient of variation (CV%), root mean square coefficient of variation (RMSCV%) and least significant change (LSC%95). Participant tolerance to the imaging protocol and rates of excessive image motion artifact were assessed by percentage of scan reacquisition due to discomfort or motion during imaging and percentage of final images graded as higher than 3 on a 5 point scale respectively.