Experimental Assessment of Mouse Sociability Using an Automated Image Processing Approach

The overall goal of these behavioral experiments paired with automatic analysis is to quantify the sociability of mice in a time-efficient manner. This method can help with the rapid screening of drugs to increase sociability in validated mouse models of autism spectrum disorders. The main advantage of introducing automated analysis is that it accelerates the process and makes screening more efficient.

The need for an automated method to assess mouse social behavior arose because the real-time visual screening of a single dose of a candidate compound takes an excess of 50 hours. Later in the video my graduate student Frency Varghese will demonstrate the automated analysis. Begin with setting up the three-compartment black-walled cage.

For the first session place an empty inverted wire cup in each end compartment of the cage. Now prepare the video camera to view the cage from above. All three compartments should be seen with a resolution of at least 720 by 480 pixels.

Then illuminate the cage with indirect incandescent lighting so that the cage interior, when measured from the floor, has no more than 3.5 Lux. Also provide infrared light to illuminate the cage at a level which the camera can detect. Take a single-frame image of the empty cage for reference.

For the first session place the test mouse in the cage. Then to queue the automated analysis of the video wave a hand over the cage just once. Let the mouse explore the test cage for 10 minutes.

Then return it to its home cage. For the second session put the stimulus mouse under one of the two inverted cups. Select which cup in a counterbalance fashion.

In the software indicate which cup the stimulus mouse was placed under. For naming conventions the cup containing the enclosed mouse is the social cup whereas the other cup is the nonsocial cup. The compartments are named in a similar fashion.

Queue the beginning of the second session with a hand wave like the first session. At the end of the second session remove the inverted cups to release the second mouse. Use another hand wave to indicate the beginning of the third session and let the two mice interact in the test cage for 10 minutes.

After 10 minutes return both mice to their home cages. This section explains how to process the movies using the user interface. However a rapid batch mode analysis without user input is also possible.

Begin by loading the movie using the Load button. Then select the cage configuration. The analysis program locates the cage boundaries, compartments and cups.

Select one of the four preset configurations or manually adjust the coordinates. Now for each frame of the movie execute the following processing steps. First convert the frames to grayscale.

Second subtract the grayscale reference frame made during the preparations thus zeroing out the pixel values everywhere except at the mouse position. Third binarize the frames by setting all the pixels above a threshold brightness to white and making all the others black. Adjust this setting to match the lighting conditions.

Now remove the tail from the images by using the Erode/Dilate function. First erode the binarized picture three times. Then dilate the frame six times.

Then erode the frames three more times. This may require adjustment if a different pixel resolution is used. The hand waves are easily identified as huge leaps in the volume of white pixels.

This allows for automatic detection of session starts and session order. Within each session detect the largest component of white pixels using the L/Comp command. This eliminates the pixels that do not belong to the mouse.

Then use the Find Ends function to determine the two most distant pixels within the largest component of white pixels. These are the nose and tail. The software determines the center of gravity or COG for the white pixel object and calculates its direction of motion from frame to frame.

For sessions with two mice two objects must be tracked. Select and orient the two largest white pixel objects using the same command set. The software automatically checks the COGs from frame to frame.

If either object changes in size by more than 20%or a COG moves farther than possible for a mouse then the binarization threshold is automatically increased until these criteria are met. The software automatically checks the centers of gravity from frame to frame. If either object changes in size by more than 20%or a center of gravity moves further than possible for a mouse then the binarization threshold is automatically adjusted until these criteria are met.

Interactions are automatically detected when the objects are within two centimeters. These segments are played back for user input. For each segment label the test mouse's perspective as either Social avoidance, Social approach, or Neither.

To verify the automatic analysis load the movie using the Load button. Then click on Start Session 1 to display the first frame of that session. Use the Play/Stop command to control playback.

The processed movies show each mouse with a different color circle at their heads and tails. Also the sociability measures are displayed on screen and updated during playback. Correct any errors encountered in the automated analysis.

The most common error is a head-tail flip which can be fixed with a tool on the interface. This occurs in about 20%of videos. In the rare instance of a more complicated error they can usually be solved by manually indicating the location of the mouse's nose across a certain series of video frames.

Once verified export the sociability data in a spreadsheet format using the Export function. To export data from multiple movies as a single spreadsheet the movies must be bundled in one folder and the Compile Data function does the operation. In session one a four-week-old BALB/c test mouse was tracked moving between compartments 59 times over 10 minutes.

The time spent in each compartment is considered a baseline for compartment preference without social stimulus. In session two a social and nonsocial cup were presented with a hidden four-week-old ICR mouse. There was a clear bias toward spending time in the social compartment.

Specifically the mouse spent more time with the social cup than the nonsocial cup. In the third session the cups were lifted and the mice were allowed to interact. In total 35 social approaches were made over 10 minutes.

After its development this technique facilitates more rapid screening of promising medications to improve the core symptom domain of impaired sociability in validated mouse models of autism spectrum disorders. The automated processing can be adapted to different cage designs and species. Any behavior of interest can be evaluated as long as an algorithm to reliably detect it can be developed.