Knowledge Base

Classify Runs

6.1 Classify the Paw Prints

To open the classification screen, do one of the following:

• In the Experiment Tasks window, click Classify Runs.
• Choose Acquire > Classify Runs.
• In the Experiment Explorer:
  • Right-click a trial (containing runs) and select Classify Runs.
  • Right-click a run and select Classify.
  • Double-click a run.

Before you start classifying paws, first check the quality of the detected paw prints by playing back the runs. If necessary, you can adjust the Green Intensity Threshold to increase the quality of the paw prints or to remove noise.

If you use the Automatic Footprint Classification module, there are two main classification scenarios:

• You only want to label paw prints – Use automatic classification to let CatWalk label the paw prints. Next, use manual classification to label paw prints that CatWalk was not able to classify automatically.
• You want to label both paw and non-paw prints – First, manually label the non-paw prints. Next, use automatic classification to let CatWalk label the paw prints.

Adjust the Green Intensity Threshold

The Green Intensity Threshold is set in the Detection Settings before acquisition. It is the minimum pixel intensity detected by CatWalk; all pixels below this minimum are not processed. In CatWalk XT 11, you can increase or decrease the Green Intensity Threshold of a run before classification, after acquisition.

You may need to adjust the Green Intensity Threshold when the value set in Detection Settings is either too low or too high. When it is too low, a lot of noise appears (for example, part of the animal's body). When it is too high, paws are not detected completely (for example, toes or parts of the sole are missing).

To adjust the Green Intensity Threshold:

1. In the Classification screen, under Video, click the Adjust the Green Intensity Threshold button on the right. The Adjust the Green Intensity Threshold window appears and shows the following:
   • Video window – Displays prints based on the Current Green Intensity Threshold, which initially matches the Original Green Intensity Threshold set in Detection Settings.
   • Original Green Intensity Threshold – The value set in Detection Settings.
   • Current Green Intensity Threshold – The value you can adjust here.
   • Show filtered out signals – When selected, prints filtered out by the Current Green Intensity Threshold value are displayed in red.
2. Move the slider to the desired value. Increasing the threshold removes prints; decreasing it adds prints. Clicking to the left or right of the slider on the slider bar moves it in steps of 5. When adjusting the threshold for a run recorded with a previous version of CatWalk XT, the lowest possible value is the value with which the run was originally acquired.
3. When the Current Green Intensity Threshold has the desired value, click OK. The run is processed with the new value, which may take a few seconds depending on run length and the threshold set.
4. You can now classify the run with the new Green Intensity Threshold value.

Please note that changing the Green Intensity Threshold affects Run Statistics. We recommend using the same Green Intensity Threshold for all runs within one experiment to ensure results can be compared between runs.

Automatic Classification

You need the Automatic Classification module to classify paw prints automatically. With the base version of CatWalk XT 10, labels can only be assigned manually.

The Automatic Footprint Classification (AFC) module uses the body contour of the animal to determine the position of the paw prints.

Performance of Automatic Classification

Automatic Footprint Classification cannot be guaranteed to work in all situations. Tests have shown that the AFC module gives good results when used with an animal walking with a normal gait across a clean glass plate. It is likely to function less well when the glass plate is dirty, resulting in background noise, or when the animal touches the glass plate with body parts other than the paws.

How to Improve Automatic Classification

The following ground rules apply for the AFC module:

• Minimize background noise. Dim ambient light as much as possible and ensure the glass plate is clean before acquisition. If noise persists in the video image, reconfigure the Detection Settings.
• Remove junk prints. If acquired runs contain a lot of noise, increase the Green Intensity Threshold before classification. If noise prints remain, manually label them as Junk and remove them.
• The animal should walk in a straight line and in one direction. If the animal turns around during a run, AFC will not work. Place corridor walls as close together as possible without restricting the animal, to encourage straight runs.
• The animal should not rear during a run. Install the diffusion plate on top of the corridor walls to help prevent this.
• The animal should only make contact with the glass plate with its paws. If other body parts contact the glass plate, manually label those prints before running the AFC.
• The corridor should be positioned in the middle of the glass plate. Center the corridor over the camera so that the animal, on average, walks along the middle line of the video image.
• Minimize the area of the video image covered by corridor walls, as walls may interfere with detection of the body contour.

Classification Settings

Choose Setup > Classification Settings. In the Classification Settings window, you can select the following options:

• Reset all assigned labels when applying the AFC – Do not select this option if you have already manually labeled part of a run and want to keep those labels.
• Only assign a label when the Max Intensity is higher than... – Noise such as dirt on the glass plate has a much lower Max Intensity than paw prints. Selecting this option with the default value of 100 causes most noise prints to be ignored.

Automatic Classification Procedure

1. Under Trial & Run Selection, select a trial and one of the runs in that trial, or click the Previous/Next button to navigate to the trial and run you want to classify. The trial list displays animal ID, time point, acquisition status, and classification status. The run list displays compliant/non-compliant status and classification status.
2. Click the Auto Classify button to automatically classify the complete run.

After automatic classification, do one of the following:

• If the AFC module classified the complete run, the Classification Finished Tasks window appears. Select one of the options in this window to continue.
• If some prints could not be classified, the video automatically jumps to the first frame with an unlabeled print. Continue manually labeling the unlabeled prints.
• Review the run and, if necessary, correct any errors.