Feasibility of a novel 3-dimensional mathematical algorithmic computation of feline bladder volumes using point-of-care longitudinal and transverse cysto-colic ultrasonographic images

Abstract

Conventional geometric formulas for estimating bladder volume assume that bladders have a perfectly uniform spheroid geometry. Bladders are often irregularly shaped, however, especially when under-filled or distorted by a full colon, which results in inaccurate ultrasonographic linear measurements and volume estimation. This pilot study investigates the feasibility, inter-observer reliability (reproducibility), robustness, and agreement of a novel 3-dimensional bladder volume computation method using bladder circumference tracing compared to a published feline linear bladder dimension formula. Paired sets of longitudinal and transverse B-mode bladder ultrasound images (n = 228) were acquired by 2 observers with different point-of-care ultrasonography skills using 10 healthy purpose-bred cats positioned in dorsal recumbency at various time points. Using strict criteria for Lin's concordance correlation coefficient, inter-observer agreements (n = 223) were found to be substantial (0.95 to 0.99) with statistically significant but clinically non-significant median differences (biases) of 0.96 mL [interquartile range (IQR): 0.16 to 2.46, P < 0.001] and 0.23 mL (IQR: 0.88 to 1.97, P = 0.006) when bladder circumference tracings were made on similar sets of ultrasound images respectively. Inter-observer agreements improved from substantial (0.95 to 0.99) to almost perfect (> 0.99) strength-of-agreement as the quality of ultrasound images improved. The bladder circumference tracing method showed moderate (0.90 to 0.95) strength-of-agreement with the recently published feline linear bladder dimension formula, with significant additive median differences (biases) of -6.76 mL (IQR: -9.06 to -3.88, P < 0.001) and -6.44 mL (IQR: -11.41 to -3.81, P < 0.001) recorded by each observer (n = 111, n = 83), respectively. Data obtained from orthogonal ultrasonographic bladder circumference tracings justify further investigation into use of this method for estimating bladder volume in cats.

Figure A1

Bladder circumference tracings. a — Longitudinal tracing, b — Transverse tracing.

Figure A2

Aligned bladder circumference tracings.

Figure A3

Assumed shape of bladder z-cross sections. Longitudinal tracing and transverse tracing intersections with the z-cross section are blue and red stars, respectively. The assumed shape of the bladder away from the axes is the green region. In the following, the notation (m, n) indicates the number of positive longitudinal, m, and transverse, n, intersections with the z-cross section. (a) (1,0) The region is a half rose petal. (b) (2,0) The region is a half ellipse, not extending above the 45-degree line. (c) (3,0) The region is a half rose petal less half an ellipse whose height is no higher than about half the rose petal height at x_m. (d) (1,1) This is the most common case. The region is a quarter ellipse. (e) (2,1) The region is a quarter ellipse less a half rose petal. (f) (3,1) The region is a large quarter ellipse less a smaller half ellipse, which remains below the 45-degree line. (g) (2,2) The region is a large quarter ellipse less a small quarter ellipse. (h) (3,2) The region is a large quarter ellipse less a small half ellipse centered on the x-axis and less a half rose petal on the y-axis. (i) (3,3) The region is a large quarter ellipse less 2 small half ellipses on the x- and y-axes.

Figure 1

A — Longitudinal B-mode image showing a bladder with poorly defined bladder margin with loss of fluid-tissue interface near the trigone region and extension of bladder apex beyond the ultrasound field of view. B — Same image demonstrating temporary removal of individual points (blue dots) at the edge of distinguishable bladder margin to allow formation of an automated boundary curve during bladder circumference tracing (green outline). C — Replacement of individual points (blue dots) on the automated boundary curve to provide sufficient coordinates for alignment and generation of a 3-dimensional model.

Figure 2 A–B

Example of a set of poor-quality ultrasound images. A — Longitudinal B-mode image with bladder extending beyond the ultrasound field of view. B — Corresponding transverse image with poorly defined bladder margins. C–D — Example of a set of intermediate quality ultrasound images. C — Longitudinal B-mode image showing bladder with clearly distinguishable bladder margins and contained entirely within the ultrasound field of view. D — Corresponding transverse image with poorly defined bladder margins. E–F — Example of a set of excellent quality ultrasound images. Longitudinal and transverse B-mode images displaying bladder with clearly distinguishable border and contained entirely within the ultrasound field of view.

Figure 3

A — Longitudinal B-mode image showing a large distended bladder extending beyond the ultrasound field of view with acquisition of maximal bladder longitudinal length and height (red lines) following boundary demarcation by bladder circumference tracing (green curve). B — Corresponding transverse image showing the same bladder extending beyond the ultrasound field of view with acquisition of maximal bladder transverse width (red line) following boundary demarcation by bladder circumference tracing (green curve).

Figure 4

Representative scatter plot graphs and Bland-Altman plots. A — Scatter plot graph of estimated bladder volumes using the bladder circumference tracing method between observers. Line of best fit drawn through individual observation points almost overlaps on the 45-degree line through origin, which represents perfect agreement. B — Bland-Altman plot demonstrating the pattern of agreement and absolute difference (bias) in estimated bladder volumes using the bladder circumference tracing method between observers. C — Scatter plot graph of estimated bladder volumes between the novel bladder circumference tracing method and recently published feline linear bladder dimension formula. Line of best fit drawn through individual observation points departs from the 45-degree line through origin as the bladder volume increases. D — Bland-Altman plot demonstrating the pattern of agreement and absolute difference (bias) between the novel bladder circumference tracing method and recently published feline linear bladder dimension formula. LOA — limits of agreement; CI — confidence interval.

Figure 5

Feline bladders of various shapes. A — Ellipsoid-shaped bladder. B — Spheroid-shaped bladder. C — Cuboidal-shaped bladder. D — Triangularshaped bladder. E — Pear-shaped bladder. F — Undefined bladder shape distorted by adjacent full colon.