Assessing Polycystic Kidney Disease in Rodents: Comparison of Robotic 3D Ultrasound and Magnetic Resonance Imaging

Abstract

Polycystic kidney disease (PKD) is an inherited disorder characterized by renal cyst formation and enlargement of the kidney. PKD severity can be staged noninvasively by measuring total kidney volume (TKV), a promising biomarker that has recently received regulatory qualification. In preclinical mouse models, where the disease is studied and potential therapeutics are evaluated, the most popular noninvasive method of measuring TKV is magnetic resonance imaging (MRI). Although MRI provides excellent 3D resolution and contrast, these systems are expensive to operate, have long acquisition times, and, consequently, are not heavily used in preclinical PKD research. In this study, a new imaging instrument, based on robotic ultrasound (US), was evaluated as a complementary approach for assessing PKD in rodent models. The objective was to determine the extent to which TKV measurements on the robotic US scanner correlated with both in vivo and ex vivo reference standards (MRI and Vernier calipers, respectively). A cross-sectional study design was implemented that included both PKD-affected mice and healthy wild types, spanning sex and age for a wide range of kidney volumes. It was found that US-derived TKV measurements and kidney lengths were strongly associated with both in vivo MRI and ex vivo Vernier caliper measurements (R ²=0.94 and 0.90, respectively). In addition to measuring TKV, renal vascular density was assessed using acoustic angiography (AA), a novel contrast-enhanced US methodology. AA image intensity, indicative of volumetric vascularity, was seen to have a strong negative correlation with TKV (R ²=0.82), suggesting impaired renal vascular function in mice with larger kidneys. These studies demonstrate that robotic US can provide a rapid and accurate approach for noninvasively evaluating PKD in rodent models.

Figure 1.

An overview of the acquisition workflow for the robotic ultrasound system. (A) Multiple parallel sweeps are acquired, usually three, resulting in thousands of individual two-dimensional images from across the animal’s body. (B and C) These images are then stitched together to produce a single wide-field three-dimensional (3D) image volume, which can be viewed in different orientations. The 3D data can be seen in (B1) frontal, (B2) transverse, and (B3) sagittal planes, respectively. (C) Each kidney can then be segmented in 3D to assess total kidney volume. Scale bar, 1 cm.

Figure 2.

A side by side comparison of an US scan versus an MRI scan of a kidney. Slices at the same location in both imaging modalities show cysts (yellow arrows) in the same regions. The inferior vena cava is also visible (red arrow).

Figure 3.

Regression and Bland–Altman plots comparing robotic ultrasound–based in vivo measurements of kidney volume with in vivo MRI. CV, coefficient of variation; LOA, limits of agreement; r, Pearson correlation coefficient; r², Pearson coefficient squared; R², coefficient of determination; rho, Spearman correlation coefficient.

Figure 4.

Vernier calipers versus US calipers. (A) An image showing how Vernier calipers were used to measure kidney length (dashed line) and width (solid line). (B) US calipers drawn in software over a coronal in vivo US image. (C) US calipers drawn over a US image of ex vivo kidneys. Lines in (A, B, and C) correspond to the same axes. (D–F) Bland–Altman plots comparing (D) in vivo US calipers, (E) ex vivo US calipers, and (F) ex vivo Vernier calipers. CV, coefficient of variation; LOA, limits of agreement; r, Pearson correlation coefficient; r², Pearson coefficient squared; R², coefficient of determination; rho, Spearman correlation coefficient.

Figure 5.

Inter-reader variability for total kidney volume measurements for cohort 1. Readers are indicated by the marker, with mean and SDs plotted. ID, identifier.

Figure 6.

Correlation between kidney vascularity and kidney size. (A) Vessel morphology as visualized via maximum intensity projections of acoustic angiography (AA) volumes. A single two-dimensional (2D) slice of the volume with an exemplary cyst (white arrow) is shown. After thresholding at the optimal threshold of 46 counts, the voxels within the cyst, and elsewhere, have been set to zero and are not included in the percent positivity metric for vascularity. (B) Mean histogram of all AA volumes of cohort 1. SDs of the histogram bins are shown with error bars. (C) The percent of bright voxels in a kidney during microbubble perfusion (positivity %) negatively correlates with kidney volume. In this plot, the kidneys were thresholded at 46 counts. (D) The coefficient of determination of the linear regression in (C) depends on the thresholding level. Thresholding at 46 counts gives the highest R² value. This threshold excludes cysts but includes most vasculature (seen on the right in [A]).