Knowledge Base

Data Management

Network-Based Feature Extraction

The network was trained to extract features from video images of rodents of various colors and in various backgrounds. During tracking, the network analyzes a portion of the image that includes the detected subject, creates a map of probability of occurrence for both the nose- and tail-base points, and makes an estimate of the position based on the highest probability.

Contour-Based Body Point Detection

Any Species - Default

This detection method analyzes the contour of the area detected as subject at each sample to assign the nose-point and tail-base. Make sure in the detection settings that the tail is fully detected.

Rodents - Default

This detection method analyzes the varying shape of the contour of the area detected as subject and builds up a 'rodent model'. It is more robust because it does not require the nose and tail to be visible: it can 'predict' the position based on previous samples. Choose this method when you want to track a single rodent without occlusions.

Rodents - For Occlusions

This method can handle severe shape distortions, for example when the animal's body is occluded or when multiple animals touch. However, it requires a lot of computer performance. Choose this method for rodents in a social interaction context.

Other Species - For Occlusions

This method does not make any assumption about the shape of the subject. Choose this method to track insects, crustaceans and large animals. With this method you only track the center point.

Contour Adjustments

• Erosion – Decreases the animal's surface area with 1–10 pixels.
• Dilation – Increases the animal's surface area with 1–10 pixels.

Tracking Methods for Multiple Subjects Per Arena

Color Marker Tracking: The color marker itself is tracked and used for subject identification.

Contour-Based Body Point Detection offers two options:

• Center of gravity tracking only
• Center of gravity and nose point and tail-base tracking

Subject Identification Methods:

• Marker-Assisted Identification – Identities are distinguished based on a color marker.
• Unmarked Subjects – Identities may switch when paths cross.

Deep Learning Based Body Point Detection: This method requires the animals have uniform color but appear somewhat different in the camera image. When 2 identical mice or rats are used, it is recommended to use a tail marker, bleach the fur on the back of one animal, or shave part of the fur. Hooded rats (2-tone rodents) are not supported.

Detection Settings

EthoVision offers four different detection methods:

• Gray Scaling – Defines all connecting pixels with a gray value between two threshold values as a possible subject.
• Static Subtraction – Looks at differences between a reference image (without the animal) and the current video image (with the animal in the arena).
• Dynamic Subtraction – Like static subtraction, but updates the reference image on each sample.
• Differencing – Makes a statistical comparison between pixels in a reference image and the current image, using the variance in contrast to determine whether each pixel has changed enough to be considered part of the subject.

Trial List

The trial list consists of trials that are planned, carried out, or skipped, including system variables and user-defined independent variables. All are organized in a cross table and data can be copied to and from Excel sheets. You can schedule a list of trials based on pre-recorded video files which can be automatically acquired as a batch. The trial list also allows you to import trials from previously acquired experiments to analyze data collected in two or more identical experiments as one data set.

EthoVision offers the following acquisition methods:

1. Acquire Data Live – Live tracking requires no disk space for video storage, but creates no video backup either. Maximum trial duration (as tested) is 72 hours.
2. Acquire Data Live and Record Video – Live tracking while EthoVision records a backup video to an MPEG4 file. Maximum trial duration (as tested) is 72 hours depending on hardware.
3. Longer Trials – To acquire trials longer than mentioned above, you can split your multi-day testing into multiple trials.
4. Record Video and Acquire Data – EthoVision records a video to an MPEG4 file for data acquisition later. (Recommended if the computer is not fast enough for live acquisition and recording simultaneously.)
5. Acquire Data from Existing Video Files – Track data from video recorded with programs other than EthoVision.
6. Acquire Data from a Batch of Existing Video Files at Once – Track data from a series of pre-recorded video files.

In options 2–5, the video file is available for post-acquisition visualization. If you choose option 3, 4, or 5, it is possible to analyze samples at a rate faster than the actual sample rate. EthoVision supports MPEG2, MPEG4 and H.264 AVC video formats.

Series of Trials from Live Video Feed

You can use a live video feed to acquire a series of trials. You can predefine start and stop conditions and an inter-trial interval.

Data Acquisition

Acquire Additional Behavioral Data

You can score behaviors manually, adding scores live (while the test is running) or offline (by reviewing the video file). In addition, EthoVision is able to automatically recognize ten rat or mouse behaviors.

Acquire Data with DAQ System

External data acquired with a separate DAQ system can be synchronized with the tracking data. When you simultaneously acquire tracking and physiological data, EthoVision sends out a synchronization signal to the external DAQ system. After acquisition, you can import the physiological data and the two data streams are synchronized.

EthoVision offers import profiles for a number of DAQ systems, such as DSI's Ponemah. You can import:

• Signals sampled at a constant rate (with equidistant time stamps)
• Signals sampled at a non-constant rate (with non-equidistant time stamps)
• Physiological data stored in ASCII or EDF format
• Ultrasonic vocalization data recorded with UltraVox

Edit Data

In the Track Editor the stored video file can be used to evaluate the tracking result. This allows you to quickly and easily find wrong data points and correct them. EthoVision will find samples that match a specific criterion, for instance to find missing points or samples that are separated by a distance greater or smaller than a specific value. With the Track Editor, points can be deleted, moved to another position or interpolated. It can be used to swap subjects in case these were assigned incorrectly.

Source: EthoVision XT 19 Technical Specifications - Rodent, Noldus Information Technology