CT measurement of prostate volume using OsiriX® viewer is reliable, repeatable, and not dependent on observer, CT protocol, or contrast enhancement in dogs

Abstract

Computed tomography (CT) is an established method for evaluating dogs with suspected prostatic disease; however, publications assessing the effects of varying factors on prostate volume measurements are lacking. The objectives of this two-part, observer agreement, methods comparison study were to assess observer agreement and the effects of varying CT technical parameters for volume measurements of canine prostate glands on CT images using OsiriX^® DICOM viewer software. In the first retrospective study, two observers measured prostate volumes of 13 client-owned dogs thrice on noncontrast and contrast CT images. In the second prospective study, two observers measured the prostate volume of 10 cadavers using five different CT protocols and eight cadavers using three slice thicknesses. Observer agreement analyses were performed, and prostatic CT volume measurements were compared with water displacement volume measurements. Intra- and interobserver variability and the effect of contrast enhancement were found to be minimal when a one-way analysis of variance model and intraclass correlation coefficients were used. No significant differences emerged between different protocols and slice thicknesses using a linear mixed effects model. When the prostate CT volume was compared using a Bland-Altman plot with the reference volume acquired by the water displacement method, agreement without consistent bias between the methods was shown, and over 90% of measurements were located within the 95% limits of agreement. The findings supported using OsiriX^® software for CT prostatic volume measurements in dogs.

Keywords: canine; interobserver; intraobserver.

FIGURE 1

Flow chart diagram illustrating how measurements for this study were implemented. Boxes with bold margins indicate the image series that were used for measurements

FIGURE 2

Transverse noncontrast (A) and contrast (B) computed tomography images of the pelvic area of the same dog in dorsal recumbency demonstrating prostate margins traced on images with a freehand image tracing software tool. Computed tomographic images were reconstructed with a soft tissue algorithm. A slice thickness of 0.625 mm, window level of 40 HU, and window width of 400 HU were used

FIGURE 3

Scatter plot of linear regression analysis to demonstrate the agreement of the two observers. Measurements performed from all image series acquired using five different CT protocols were included. The regression line has a slope of 1.004 (0.993 to 1.014) and an intercept of 0.115 (‐0.192 to 0.421). On the x‐axis, prostate volumes (cm³) measured by observer #1 and on the y‐axis, prostate volumes (cm³) measured by observer #2 are shown

FIGURE 4

Bland–Altman diagram to demonstrate the difference between CT volume and water displacement method volume against the average of these two measurements. Markers o and + are used for the measurement of observers #1 and #2, respectively, and each of five CT protocols are represented with identical markers. A solid horizontal line with a y‐axis intercept of 0 represents perfect agreement, and the average bias is located at the same level. The 95% LoAs are also displayed as solid horizontal lines. The y‐axis shows the difference between the two paired measurements, and the x‐axis represents the average of these measurements

FIGURE 5

3D ROIs from computed tomography images of a cadaver prostate demonstrating measured volumes. ROIs were drawn on the image series reconstructed with a soft tissue algorithm. On the left (A and B), 2.5 mm thick slices and on the right (C and D), 0.625 mm thick slices were used. On the top (A and C) transverse view, dorsal recumbency, caudal aspect. On the bottom (B and D), oblique view from the same 3D ROI as on the top. ROI; region of interest, defined with a freehand image tracing software tool