The SLAVV software (Supporting Functions) gives the user the ability to curate the vectorization output with live visual feedback. This Graphical Curator Interface (GCI) automatically opens at the appropriate times during the default execution of the Main Function. The SLAVV method consists of four workflow steps as described in the Methodology Manuscript Section: Automated Vascular Vectorization:
- Linear "Energy" Filtering
- Vertex Extraction
- Edge Extraction
- Network Organization
These steps must be executed sequentially, starting with the "energy" step. The input parameters for each specific workflow are described in the Documentation Section: Workflow-Specific Parameters. The SLAVV software automatically extracts vertices and edges from the Energy Image, and automatically organizes the edges and vertices into a graph representation of the network. The Graphical Curator Interface (GCI) opens after the Vertex and Edge Extraction steps by default as described in the Documentation Section: Logistical Parameters.
This tutorial demonstrates the use of the graphical curator interface on three large (~1 mm3), images of living adult mouse brain microvasculature. The three images (Image A, B, and C) are all from the same mouse, approximately the same field of view, and timed 2 weeks apart.
- Input Images and Parameters
- Energy Images
- Curator Interface Overview
- Projections
- Histograms
- Visualizations
| Image A are in the left column | Image B the center | and Image C the right |
|---|---|---|
-----------------A----------------- |
-----------------B----------------- |
-----------------C----------------- |
The three images (Image A, B, and C) are all from the same living mouse, approximately the same field of view, and timed 2 weeks apart.
To begin the vectorization process, input a raw TIF from a file location or a matrix from the MATLAB workspace into the Main Function, as described in the Documentation, Section: Optional Input. There is no pre-processing (interpolation, filtering, etc.) required. The software will prompt the user for all required inputs (e.g. size and shape of the voxels in microns, as well as processing parameters) if the inputs are not already input in NAME/VALUE pair format.
| A | B | C |
|---|---|---|
 |
 |
 |
Wrapper scripts (e.g. Example 1, Example 2, ...) are useful for recording and rerunning the input parameters that were used (image locations, resolutions, etc.), as well as re-running a subset (e.g. starting with 'vertices' or 'edges') of the workflows. The vectorization software allows the user to pick up in the middle or re-run parts of the vectorization with different parameters or curations. Wrapper scripts are not required, as all of the input parameters and image files and locations are automatically saved in the output batch_* folder.
| Image A Input File Location Selection | Starting Workflow Step Selection |
|---|---|
 |
 |
Default Energy Filtering processing parameters (purely Gaussian kernel) will work well for the images demonstrated here. If instead of the lumen, the vessel wall is illuminated, try starting with an 80/20 Gaussian/Ideal mixture, and a 50/50 annular/spherical ratio. After running the first step ('energy') of the SLAVV workflows, vessels in the requested size range should have dark centerlines in the "Energy" Image.
| A | B | C |
|---|---|---|
 |
 |
 |
The Graphical Curator Interface (GCI, shown below) has four windows (starting in top right and moving counter-clockwise):
| A | B | C |
|---|---|---|
 |
The Volume Map window shows the user:
- the current 3D field of view (FOV) by highlighting the 2D projections on the face of the box,
- The dimensions of the entire 3D volumetric image on the axes of the box in real spatial dimensions,
- The voxel coordinates and total number of voxels in all 3 dimensions of the current FOV in a table, and
- The volumes of influence of the previously applied threshold by painting red on the faces of the box.
The Volume Display window shows the user:
- A maximum intensity projection of the original image and overlayed vetors in the current 3D FOV,
- The relative depth and thickness of the 2D projection displayed using the two sliders ("Depth" and "Thickness"), and
- Which vector objects in the current field of view are status true (color:blue) or false (color:red).
The Volume Display window gives the user control over the current FOV:
- pan vertically and horizontally using the MATLAB panning tool for plots,
- zoom vertically and horizontally using the MATLAB magnifying glass for plots,
- navigate into/out-of the page using the depth scrollbar, and
- Increase/decrease the thickness of the FOV in the depth dimension.
The Volume Display window also gives the user control over the direction of projection for the displayed MIP, (i.e. permute X, Y, and Z dimensions; change X or Y to the depth dimension (instead of Z)).
The Volume Display window gives the user vector object classification ("Toggling") ability to:
- (for vertices and edges)
- point-and-click individual object to swap its true/false status (blue/red color) and
- (just for vertices)
- drag a box around a group of objects to change all of them to status:true/false (color:blue/red).
The Intensity Histogram window shows the user the distribution of pixel-intensities in the FOV, allowing the user to:
- set brightness and contrast for the underlying original image, and
- toggle the dispaly mode between "original" and "inverted" for the underlying original image.
The Energy Histogram window shows the user the distribution of vector-energies in the FOV, allowing the user to:
- set brightness and contrast of these objects accordingly,
- toggle the display mode between the "Graded" and "Binary" versions of the vector object brightnesses, and
- Set the Energy threshold value, assigning all vectors above that Energy value in the current FOV to false.
Note: the example curator screenshot shown in the "Anatomy of the GCI" image above is how the curator looks when the user first opens the edge objects for Image A. The vectors shown are the uncurated/unedited (no red in Volume Map or Display) automatic output of the new (unreleased) version of the Edge Extraction step of SLAVV.
After the vertices are automatically extracted (as local minima) from the Energy Image, the user can curate these vertex objects. Vertices have both point-location and radial-sizing components, and should be considered status:true (color:blue) when both the location and size match a vessel in the underlying original image. The user can curate (classify as true/false status) the vertices using local thresholding as well as point-and-click toggling.
To select a threshold, use the middle text-entry box in the Energy Histogram window labeled "Threshold." Try to choose a threshold that ensures high sensitivity (i.e. leave most vertices blue, even if they are false positives), but removes the many vertices in the extravascular regions of the image (owing to image noise).
| A | B | C |
|---|---|---|
| Before Thresholding | ||
 |
||
| After Thresholding | ||
 |
Extend the thickness of the current field of view using the "Thickness" slider to span the entire depth of the image stack in order to apply the threshold to the entire image (globally).
| A | B | C |
|---|---|---|
| Full-Depth FOV | Partial-Depth FOV | |
 |
 |
Use the display options on the Intensity and Energy Histogram windows to better inspect the effect of the threshold.
| A | B | C |
|---|---|---|
 |
| A | B | C |
|---|---|---|
 |
After making a low specificity global threshold, navigate to the brighter regions of the original image to apply more specific local thresholds.
| A | B | C |
|---|---|---|
 |
 |
Extending the depth in this FOV allows the user to apply this local threshold across all image slices.
| A | B | C |
|---|---|---|
| Before Thresholding | After Thresholding | |
 |
 |
The Sweep button on the Volume Display removes the false objects from the display and histogram making it easier to see other objects.
| A | B | C |
|---|---|---|
| Before Sweeping | ||
|
||
| After Sweeping | ||
 |
Some vertices cannot be easily removed by thresholding and need to be selected individually by point-and-click or dragging a box over them.
| A | B | C |
|---|---|---|
| Before Toggling | ||
 |
||
| After Toggling | ||
 |
These are the final curated vertex sets that were passed to the Edge Extraction step.
| A | B | C |
|---|---|---|
| Full FOV | Full FOV | |
 |
 |
| A | B | C |
|---|---|---|
| Partial FOV | ||
 |
After the edges are automatically extracted (by tracing the Energy Image with the vertices as termini), the user can curate these edge objects. Edges consist of ordered lists of locations and sizes, which connect one vertex to another along the path of the edge. Edges should be considered status:true (color:blue) when both the locations and sizes match a vessel segment in the underlying original image. The user can curate (classify as true/false status) the edges using local thresholding as well as point-and-click toggling.
Use the "Depth" slider to navigate deeper into the volume. Select in the margin of the slider to move the current FOV to the next adjacent, non-overlapping, FOV.
Fly-through of the (uncurated/unedited) output of the new (unreleased) version of the automated Edge Extraction step of SLAVV.
Note: red-penned circles are pointing out missing parts of the vectorization that need to be manually added.
| A | B | C |
|---|---|---|
 |
||
 |
 |
|
 |
 |
|
 |
 |
|
 |
||
 |
||
 |
||
 |
Use the "Threshold" text-entry box in the Energy Histogram window to set a threshold. Try to choose a threshold that labels edge objects outside of vessels as status:false (color:red).
| A | B | C |
|---|---|---|
| Before Thresholding | ||
 |
||
| After Thresholding and Sweeping | ||
 |
| A | B | C |
|---|---|---|
 |
Use the "Z-Depth" or "Z-Thickness" button in the Volume Display window to change the direction of projection of the displayed MIP. Use an X or Y projection to get a better perspective on the penetrating vessels that are aligned with the Z-axis.
Note: red-penned circles are pointing out missing parts of the vectorization that need to be manually added.
| A | B | C |
|---|---|---|
| Thin Y-Projected FOV | ||
 |
||
| Thick Y-Projected FOV | ||
 |
||
| Thin X-Projected FOV | ||
 |
||
| Thick X-Projected FOV (annotated) | ||
 |
Use the "Add" Button in the Volume Display window to add edges to existing vertices (shown in gold during edge curation). Select two vertices, one after the other, to connect them with a straight edge, which interpolates the vertex sizes across its length. Note: after selecting the add button, any vertices in the FOV that do not yet have connected edges (and are normally hidden to the user) will also be shown to the user in the Display Volume to serve as possible termini for the added edge.
| A | B | C |
|---|---|---|
 |
||
 |
||
 |
||
 |
Use the "Toggle" button in the Volume Display window to point-and-click edges to toggle their true/false status.
| A | B | C |
|---|---|---|
 |
||
 |
This is the final curated edge set that was passed to the Network Organization step and represented in the histograms and visualizations.
| A | B | C |
|---|---|---|
 |
Maximum intensity projection outputs from the middle (1/3 in all 3 dimensions) of the vectors overlaying the image.
Note: from left to right, the color coding is: strands (random color for each), depth, and direction.
| A | B | C |
|---|---|---|
 |
| A | B | C |
|---|---|---|
 |
| A | B | C |
|---|---|---|
 |
Lateral area- or strand-weighted histograms of various statistics of interest from the vectorized vascular network in the input image can be used to analyze the anatomy in question.
Note: Strands are vessel segments that connect bifurcations/endpoints to bifurcations/endpoints. The left histograms are strand-weighted, the right lateral area-.
| A | B | C |
|---|---|---|
 |
| A | B | C |
|---|---|---|
 |
| A | B | C |
|---|---|---|
 |
Use the network_*.vmv file output in the vectors directory of the batch_* folder to export the vascular vectorization to the VessMorphoVis plugin to Blender to visualize the network in 3D with lighting and color-coding.
Note: colorful images are coded by vessel radius.
| A | B | C |
|---|---|---|
 |
 |
|
 |
||
 |
 |
|
 |
 |
|
 |
||
 |
| A | B | C |
|---|---|---|
 |
||
 |
||
 |
||
 |
||
 |
||
 |
| A | B | C |
|---|---|---|
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |
||
 |