In vivo micro-CT scanning of a rabbit distal femur: repeatability and reproducibility

Abstract

Before in vivo micro-CT scanning can be used to investigate femoral trabecular microarchitecture over time in rabbits, its repeatability and reproducibility must be demonstrated. To accomplish this, both distal femurs of two 6-month-old New Zealand white rabbits were scanned five times each in 1 day under different conditions (repeatability). Scanning was done at 28 microm isotropic voxel size to produce five image stacks of each femur. Three operators then followed a standard image processing protocol (reproducibility) to isolate two separate cubes from each anterior femoral condyle [total n = (8 cube sites)(5 scans)(3 operators) = 120]. Bone volume fraction (BV/TV) of the eight different cube sites (sample) ranged from 0.408 to 0.501 (mean: 0.453); trabecular thickness (Tb.Th) ranged from 158.1 to 185.5 microm (mean: 168.6 microm); and trabecular separation (Tb.Sp) ranged from 179.4 to 233.1 microm (mean: 204.7 microm). Using ANOVA and the variance component method, the total process variation was +/- 14.1% of the mean BV/TV of 0.453. The sample variation was +/- 13.9% (p < 0.001), the repeatability was +/- 2.1% (p < 0.001), and the reproducibility was +/- 0.1% (p > 0.05). Results were similar for Tb.Th and Tb.Sp. Though the contribution due to repeatability was statistically significant for each of the three indices, the natural sample differences were far greater than differences caused by repeated scanning under different conditions or by different operators processing the images. These findings suggest that in vivo micro-CT scanning of rabbit distal femurs was repeatable and reproducible and can be used with confidence to measure differences in trabecular bone microarchitecture at a single location in a longitudinal study design.

Figure 1

The custom designed open micro-CT system (ACTIS 150/225 Ffi-HR CT, BIR Inc., Lincolnshire, IL) was fitted with a specimen/small animal manipulator and a custom made rabbit holder. The rabbit holder allowed a single hind-limb to be extended into a 51 mm diameter tube for isolation and scanning of the distal femur. In an offset scanning mode, the three-dimensional isotropic voxel image size was 28 μm.

Figure 2

Representative slices from each of the five scanning conditions show that the rabbit was repositioned for each scan. The 86kVPlastic and 86kVTissue conditions contained extra material within the scan region to simulate soft tissue changes between scans. Image slices were 2048 × 2048 pixels with 28 μm size.

Figure 3

After image registration, the femoral cross sections were isolated and cropped to produce cubes from the anterior parts of the medial and lateral condyles. The “cubes” measured 125 × 125 × 125 or 125 × 125 × 100 voxels in size with 3.5 mm width, 3.5 mm length and either 3.5 or 2.8 mm height. Individual isotropic voxels were 28 μm. Cube from lateral side is shown.

Figure 4

The volume fractions (BV/TV) for each cube sample and for each scan condition are plotted. The error bars indicate the standard deviation for the three measurements by the different operators. The repeatability, or twice the standard deviation between different scan conditions, was 2.1% of the mean BV/TV, and contributed significantly to the total variance (p<0.001). The operator reproducibility, or twice the standard deviation between operators, was 0.1% of the overall mean, and did not contribute significantly to the total variance (p=0.06). For comparison, the sample variance component was 13.9% (p<0.001).

Figure 5

The mean trabecular thicknesses (Tb.Th) for each cube sample and for each scan condition are plotted. The error bars indicate the standard deviation for the three measurements by the different operators. The repeatability, or twice the standard deviation between different scan conditions, was 3.9% of the mean Tb.Th, and contributed significantly to the total variance (p<0.001). The operator reproducibility, or twice the standard deviation between operators, was 0.0% of the overall mean, and did not contribute significantly to the total variance (p=0.65). For comparison, the sample variance component was 11.6% (p<0.001).

Figure 6

The mean trabecular separations (Tb.Sp) for each cube sample and for each scan condition are plotted. The error bars indicate the standard deviation for the three measurements by the different operators. The repeatability, or twice the standard deviation between different scan conditions, was 3.2% of the mean Tb.Sp, and contributed significantly to the total variance (p<0.01). The operator reproducibility, or twice the standard deviation between operators, was 0.5% of the overall mean, and contributed significantly to the total variance (p<0.01). For comparison, the sample variance component was 19.3% (p<0.001).