Vascular Tracking

From Spin Help
Jump to: navigation, search
Contain two algorithms for segmenting/tracking either arteries or veins. Designed to work with MRA or SWI data.

Algorithm Description

Dynamic Thresholding

A dynamic thresholding algorithm designed to track arteries in MRA images, it works in a region growing fashion by adding adjacent voxels that are within σ standard deviations of the mean of the marked region (marked as arteries). This algorithm performs an initial global threshold manually set by the user to form a seed region. Adjacent pixels at the edge of the seed region are added if they lie within σ standard deviations of the nearest 15-20 voxels already marked as arteries. The process is iterated until no more voxels meet this criterion.

Statistical Thresholding

A thresholding algorithm designed to track arteries in MRA images or veins/hemorrhage in SWI images. It works by marking voxels that lie σ standard deviations above or below (for MRA and SWI respectively) the mean of the background voxels. Generally performs better than the dynamic thresholding algorithm. First, a global threshold is performed to mark the large obvious structures. Second, a local mean and standard deviation is calculated for a region that is 40x40x3 voxels, no voxels previously marked will be included in this calculation. The calculated threshold is then applied to the center 5x5x3 voxels of this region. The region is moved and this process is repeated until the entire image has been covered. If multiple iterations are specified the entire image is run again. This can improve results as in each iteration you get a more accurate calculation for the mean and standard deviation of the background as less vessel voxels are included in the calculation.

Before running this on SWI data the black background must be removed. This can be done by running the complex thresholding (recommended) or by clicking the remove black background box to run an erosion dilation algorithm (not recommended).

Vascular Tracking Input

Select First File - Select the first file of the series that you want to process
Select Number of Files - Select the number of files in the series you would like to process. Disable if process entire series is selected
Process Entire Series - Automatically determines the number of files in the series and will process all of them (default).
Maximum Intensity (MRA) – Select this to process MRA datasets (looks for high intensity objects)
Minimum Intensity (SWI) – Select this to process SWI datasets (looks for low intensity objects)
Eliminate Black Background – This runs an erosion dilation algorithm to pull out just the brain and remove the skull and background. When checked this will run when “Create MIP” is selected. This processing takes a long time (many minutes) so be patient, there is no progress bar. This is not the recommended way to do this, use complex thresholding instead. This algorithm can give good results but it is untested and may have bugs.
Create MIP – Creates and then opens either a minimum or maximum intensity projection of the files selected above; for use in the initial global threshold, and visualization of results.
Open MIP – If a MIP has already been generated it can be opened here for use in the initial global threshold.
Simple Threshold Value (Seed Region) – This applies a global threshold to the entire dataset. The results are displayed on the MIP to help select the proper value, this will be applied in 3D to the original data. This should be selected to mark the large obvious structures while including minimal noise. The later steps will mark the lower contrast vessels.
Sigma Factor – Sets the number of standard deviations from the mean the threshold is set at. Increasing and decreasing this has the opposite effect for dynamic thresholding and statistical thresholding. For dynamic thresholding low values (0-1) are very conservative and high values (3-5) are very aggressive, for statistical thresholding low values (1-2) are very aggressive and high values (4-5) are very safe.
Background Value – This is the minimum of maximum value that is included in any of the calculations. This is set to exclude the background from affecting the mean and standard deviation calculations.
Connectivity Filter Size – Sets the minimum number of connected voxels to be included in the final results. This is set to exclude a large number of isolated points in the final results. Set to a conservative value as small objects can be removed afterwards in the object information dialog.
Number of Iterations – This sets how many times the statistical thresholding is run. Improves edge definition and low contrast objects near high contrast objects as the high contrast objects are removed from the calculations in the first iteration.
Dynamic Thresholding – Runs the dynamic thresholding algorithm
Statistical Thresholding – Runs the statistical thresholding algorithm
Keep/Remove Objects – Pull up the object information dialog after the thresholding has been performed

Vascular Tracking Output

The results will display as a colored overlay on the original slices and projected onto the MIP. Each connected region is referred to as a single “object” and have separate colors assigned to them. In this was a single color represents a unique connected structure, in practice the colors are reused as there are only a finite number of easily differentiable colors.

The objects can be saved and loaded using right click or the tracking menu. You can view object information and export it to excel using object information and delete individual objects with right click.