Q-VAT: Quantitative Vascular Analysis Tool

Abstract

As our imaging capability increase, so does our need for appropriate image quantification tools. Quantitative Vascular Analysis Tool (Q-VAT) is an open-source software, written for Fiji (ImageJ), that perform automated analysis and quantification on large two-dimensional images of whole tissue sections. Importantly, it allows separation of the vessel measurement based on diameter, allowing the macro- and microvasculature to be quantified separately. To enable analysis of entire tissue sections on regular laboratory computers, the vascular network of large samples is analyzed in a tile-wise manner, significantly reducing labor and bypassing several limitations related to manual quantification. Double or triple-stained slides can be analyzed, with a quantification of the percentage of vessels where the staining's overlap. To demonstrate the versatility, we applied Q-VAT to obtain morphological read-outs of the vasculature network in microscopy images of whole-mount immuno-stained sections of various mouse tissues.

Keywords: density; immunohisthochemistry; morphometric analysis; quantification; vasculature.

Figure 1

User interface of the Q-VAT tool. Through this interface the user selects the input directory containing the data to be analyzed, the spatial calibration (µm/pixel) as well as several input parameters: the vascular compartment separation threshold (µm), the close label radius (µm) and the prune ends threshold (µm). The user can choose whether or not to save the output figures. If multiple channels are to be analyzed, this can also be indicated.

Figure 2

Morphological read-outs for 4 animals. Mean vascular density (average per tile) (A), mean vessel diameter (B), mean branch length (C) and tortuosity index (D) of the entire vasculature for each acquisition tile from the entire brain sections of each of the four mice (M1, M2, M3, M4).

Figure 3

Validation of Q-VAT by comparison with existing methods for vascular feature quantification in microscopy images. (A) Representative image segmented (bottom) and quantified (top) using manual analysis, Q-VAT, AngioTool and REAVER. (B) Tissue masking of a representative images for all three methods. (C) Evaluation of the segmentation performance of the tissue area. (D) Evaluation of the segmentation performance of each quantification tool using Dice similarity (left), sensitivity (middle), and specificity (right). (E) Evaluation of absolute error comparing morphological read-outs (cluster density (#/mm²), vascular density (%), branch density (#/mm²) and endpoint density (#/mm²) to those obtained from manual segmentation.

Figure 4

Region-wise analysis. (A) Manually delineated Regions Of Interest (ROIs) of the cortex (red), corpus callosum (blue) and the hippocampus (yellow) (n = 4). (B) Mean vessel diameter and (C) Average vascular density within the different ROIs for all vessels or vessels above/below a threshold (10 µm). All data was analyzed with a one-way ANOVA and Tukey's HSD post hoc test. *p < 0.05, **p < 0.01, ***p < 0.0001.

Figure 5

Overview of the automated quantification of the vascular network in immuno-stained microscopy images of different types of tissue using Q-VAT. (A) The original stitched high resolution images of the entire tissue section (first column, scalebars 1000 µm) are pre-processed and used to create vascular masks (second column) and and tissue masks (third column). These masks are divided into the original acquisition tiles (last column; (top) half of an original acquisition tile; (bottom) segmented vascular mask of half of the acquisition tile) and are used to perform automated quantification of the vasculature. (B-C) Average morphological read-outs for the mean vessel diameter (B) and vascular density (C) (n = 4 animals per organ).

Figure 6

Example of a tile-wise morphological output map. (A) Original stitched high resolution images (scalerbar 1000 µm), (B) Vascular mask, (C) Tile-wise output map of vascular density expressed in percent of tissue of the entire rat brain section. Area outside the brain tissue was set to pure black.