What's new about tracking in Volocity 3

You may have noticed that tracking in Volocity 3 has had some additional features added to improve your tracking experience. Let’s looks at the whole process from Classification right through to tracking with an example dataset.

Mission 1: Classification

Before you track you must first classify.

Volocity must be able to identify all the objects that you wish to track in all of the volumes of the time sequence, so it is important that you set up the Classifier correctly before you begin to track.

Remember - the tracking that you get is only as good as the Classifier that you make!

Above is a view from Volocity. The volume that you see is one in a sequence of 40 timepoints.

The Classifier will allow us to identify objects on the basis of Intensity, Size and Shape.

There are further tools to;

Reduce noise in selection

Fill holes within objects

Ignore objects that touch the edge of the sample

Separate touching objects.

First, let’s make a new Classifier - click on the New Classifier button on the Classifier palette to make one.

This will automatically open the "Edit Classifier" dialog.

For purposes of this demonstration we will use the "Intensity Mode". There are two modes namely HSI ( Hue , saturation and intensity) and RGB (Red, Green and Blue intensity) which you will probably already be familiar with.

The HSI mode has three histograms that allow the selection of Hue and Saturation as well as Intensity.

The RGB mode also has three histograms that allow the selection of Red, Green and Blue Intensities.

With the release of Volocity 3, two new modes were introduced, namely Percentage and Standard Deviation.

In Percentage mode, intensity selection is based on a percentage, 100% being the highest intensity value in the particular timepoint.

In Standard Deviation mode the selection is based on standard deviation(s) about the mean intensity.

Both of the above modes are useful in situations where it is difficult to apply an absolute intensity value to fit with all timepoints in a sequence, for instance if the intensity of your sample changes over the duration of the experiment.

Let’s have a closer look at the histogram.

See those red bars? The Classifier will only include those intensities that fall between the bars. In Volocity you can move the bars to adjust the range, this feature is pretty similar to Density Slicing in Openlab, if you are familiar with that. All pixel intensities below the Lower intensity will be excluded from classification, all pixel intensities above the Upper intensity will be excluded also.

But the Classifier has a whole bunch of extra stuff to make object identification easier.

Firstly there is a size threshold that will allow you to exclude large or small objects from the classification.

Shape recognition will allow you to select objects based on their shape. This tool will match objects to a series of models and exclude those that do not fall within the required parameters.

There is a whole host of shapes to pick from and if none of these seem to fit the bill you can create a custom shape or generate a shape from the current region of interest in the Image view. Remember you can still use the Image view at the same time as the Classification dialog.

If you have selected Custom shape, to edit it use the "Edit" button, see below. This will bring up the "Custom Shape" dialog from which you can modify your shape.

The "Fill Holes" option will fill holes in an object (i.e holes that are surrounded on all sides, within the object).

"Reduce Noise" will apply a 3x3x3 median filter to the selection.

"Ignore Objects Touching the Edge of the Sample" will exclude objects that touch the edge of the sample in either X,Y or Z.

"Separate touching objects" will look at the voxel distribution of one object and determine if it is constituted of two or more objects on the basis of the other selection criteria.

At any time when you are using the Classifier, clicking on the "Apply" button will apply the Classifier so that you can take a look at what it has found in the Image view.

If selected, the "Auto Apply" button will do this automatically each time you change one of the Classifier parameters.

When you are happy with the Classifier be sure to give it a name and save it.

Remember you don’t have to use all of the above criteria, think of them as tools to get the best "fit" to the objects that you want to track. Also it is a good idea to try your Classifier out with all of the volumes in the time series before you start, to ensure that it is finding the required objects in each successive volume.

Mission 2: Measuring

Go to the "Tools" menu and select "Measure Objects". The "Measure Objects" dialog will appear.

If you want to track you must apply a Classifier, ensure that you select "Track moving objects".

Bear in mind that some measurements are computationally more expensive than others; if you don’t need a particular measurement, disable it by clicking on the checkbox.

Also when tracking please be mindful of the option “Measure only from current region”. If you use this option it will only measure from the current selection you have onscreen, so only enable this if you want measurements from a specific region in the image.

When you click on the Measure button you will advance to your next mission which is tracking.

Mission 3: Tracking

If you have reached this point you should see the dialog below.

First of all you have to decide which tracking model to choose from your options are Shortest Path or Trajectory Variation.

The Shortest Path model matches measured objects between successive timepoints and tracks them based on a maximum distance moved between each timepoint.

The Trajectory Variation model uses the Shortest Path model to establish a trajectory based on the measured objects from timepoints 1 and 2. It also takes into consideration a maximum distance moved between each timepoint. This model is sometimes useful where tracked objects cross paths.

Just to let you know, in this example I am going to use the “Shortest Path” model.

There are the familiar options to Ignore “Static Objects” and “Ignore New Objects” but in Volocity 3 there is also a new option to “Automatically Join Broken Tracks”

This new option is sometimes useful where an object is lost by the classifier for one or two timepoints then reappears. This process is only effective for small gaps. For this reason, an object must be present in at least three consecutive timepoints, be absent for a maximum of two consecutive timepoints, then be present for at least three consecutive timepoints. If these criteria apply, Volocity will attempt to join the tracks when this feature is enabled.

Next is the all important distance between objects where you can choose to “Estimate automatically” or “Use this distance” where you can enter a specific distance.

If you pick the option to “Estimate automatically” Volocity will estimate a maximum distance based on the first two timepoints. It will obtain an average distance moved between matched objects, add one standard deviation to this and use this figure as the maximum distance for the tracking model.

Alternatively if you know the maximum distance that you expect your objects to move you can choose “Use this distance” and adjust the slider or just type the value in. To give yourself more of an idea, you could even measure this in Volocity by measuring the movement of a object between two timepoints using the line measurement tool. I think this is quite handy so I’ll cover this at the end of the note if that’s OK.

If you are using the “Shortest Path” model you are now ready to track so click on the Track button.

But just before I go on, you have probably noticed the “Trajectory Variation” slider at the bottom of the “Track Objects” dialog.

This is only used if you choose the “Trajectory Variation” model. Adjusting the slider alters the sensitivity of the tracking algorithm. Using a low value would restrict the tracking to only consider smooth trajectories, whereas if a higher value was used the tracking would consider more erratic trajectories.

OK, measuring and tracking has now been performed, it is time to take a closer look at the tracks.

Switch to the Measurements view by clicking on the Measurements tab.

Select the appropriate measurement session if you have more than one then view the tracking measurements by clicking on the button as shown below.

Select the tracks to be viewed by clicking on them, if you want to select more than one remember to hold down the “shift” key as you click.

Now switch back to the image view to take a look.

This looks a bit cluttered if you just want to see the track.

If you don’t like the look of the way things are being displayed switch back to the Measurements view and adjust the feedback options, this can be accessed from the Measurements menu.

Now back to the Image view again, that’s better.

If you can use the HR Renderer in Volocity Visualization you can see your tracks in this view too.

Bonus Mission : How to measure using the Line tool in Volocity

You might remember earlier that I mentioned that you can get an idea of the distance an object moves by using the line measurement tool in Volocity. Well this is how it’s done.

Firstly I switched to the Image view and selected the Line tool from the measurements tool bar.

I then located an object I was interested in, zoomed in and clicked on what I deemed to be its centre.

I advanced one timepoint using the Navigation palette.

Holding down the shift key I then clicked on the object’s new centre. This will produce a line as shown in the image below.

Now I measured the line by going to the “Tools” menu … “Measure Objects”

The measurement dialog should be set up as below, you want to measure the current timepoint and the current selection to get the Line Length.

After this has been done you can switch to the Measurements view to check the length of the line and hence get an idea of how far the object is moving between the two timepoints.

I’m glad I got all that off my chest, see you next month for a spot of charting.