Knowledge Base

Problem

Auto-classification in CatWalk XT fails to label any footprints, classifies them incorrectly, or the detection settings cause the system to start and stop acquisition automatically at unexpected times.

Details

Symptoms include: clicking 'Auto Classify' produces no labeled prints; prints are detected during acquisition but show no labels after classification; the system auto-triggers run start/stop erroneously during acquisition; some prints are detected but others are missed. Root causes are typically misconfigured detection parameters: gain too low (prints too faint to classify), noise threshold too low (too much background noise confuses classification), or noise threshold too high (real prints fall below detection). Distinct from the separate issue of footprint labels not *displaying* (see ticket #3494 below).

Solution

1. **Check Detection Settings:** In CatWalk XT, open the Detection Settings panel.
- If prints are too faint or not detected: increase the camera Gain, or lower the Green Intensity Threshold to make the system more sensitive.
- If there is too much background noise causing false triggers or misclassification: raise the Noise Threshold to filter out low-intensity noise, or reduce Gain slightly.
2. **Adjust Classification Noise Threshold:** In the Classification settings (Acquire > Classify Runs), adjust the noise threshold parameter to balance sensitivity vs. noise rejection for your specific animal model.
3. **Reset detection settings to defaults** if values have drifted: reset each parameter and re-optimize starting from recommended baseline values in the CatWalk XT Reference Manual.
4. **For auto-trigger issues (run starts/stops automatically):** Review the Run Detection settings (minimum/maximum duration, speed thresholds) and reset them to appropriate values for your corridor length and animal speed.
5. **For paralysis or low-pressure models** (e.g., spinal cord injury): lower the Green Intensity Threshold further, increase Gain, and control ambient light carefully to maximize signal from weak paw contacts.
6. After adjusting settings, perform a test run and verify classification output before proceeding with experimental data.