Fully Automated Leg Tracking in Freely Moving Insects using Feature Learning Leg Segmentation and Tracking (FLLIT)

FLLIT is an algorithm that calculates the body and leg compositions of a tracked animal. And this is useful for characterizing the circuitry that underlies both normal and disease walking behaviors. The unique thing about FLLIT is that it does not require any user annotated data sets for training and is hence fully automated.

But also automatically analyzes the track data producing a series of gait measurements, gait plots for visualization and a track video. Demonstrating the procedure will be Animesh Banerjee, a postdoctor colleague from my laboratory. Before beginning an experiment, confirm that the recording station has a high speed camera with a stage over it to hold the arena chamber.

Place infrared LED lights at the top of the stage with a diffuser between the camera and the sample. About 40 minutes before beginning the recording, transfer the flies into an empty vile on ice for five to seven minutes. While the flies are being sedated, use water and a wipe to clear the arena and glass slides of the chamber and tape one of the glass slides below the chamber.

Using a clean brush, transfer one anesthetized fly into the chamber and secure the chamber with a second slide and tape. Then allow the chambered fly to acclimate at room temperature for 15 to 20 minutes. To generate videos for FLLIT analysis, first, turn on the power source and open the viewer application on the connected computer system.

Change the recording frame rate to 1, 000 frames per second and set the shutter speed to one millisecond. Place the chamber with the fly on the recording arena and click the LIVE button. Confirmed that the camera is focused on the leg tips when the fly is walking upright on the floor of the chamber and click Record.

While recording, ensure that the fly walks in a relatively straight trajectory without touching the edge of the arena, walks at least three strides per leg, does not pause during the walk, and that the distance walked is the equivalent of at least one and a half body lengths. At the end of the analysis click Record Done to stop the recording and save the files in mraw or tiff format into the appropriate folders. To install FLLIT, first download the program onto any operating system.

Extract the contents of the zip file and download sample data sets. To install FLLIT in Ubuntu, navigate to the FLLIT/Compiled directory and right-click to select Open in Terminal. Download a sample data set and create a folder data under FLLIT-master/Compiled into which the data set folders will be placed.

Install the MATLAB runtime libraries to HOME/MCR. When the MATLAB runtime library installation has completed, issue the command to ensure that the executable rights are recorded to FLLIT. Then enter the command to execute FLLIT.

For segmentation of the video, convert the file into individual TIFF files and save the converted files in the FLLIT data folder. When all of the files have been transferred, run FLLIT. In the popup window, select 0 for carrying out the leg segmentation only or select 1 to include leg tracking with leg segmentation.

Select the folder containing the frame-by-frame TIFF images of the video to be tracked and the file of interest and click Add. Then click Done to initiate the segmentation and tracking of the selected video. To check the accuracy of the tracking and to carry out error corrections, click Select Data Folder, then select the folder to be tracked.

Click View Tracking and check the labeling for all of the legs in the first frame. If a leg is incorrectly labeled and a correction is required, click Pause viewing and adjust prediction. In the Leg to Adjust panel, select the leg that requires correction and double click on the correct position of this leg in the image window.

Then click Save and Exit. To process the segmented and tracked data, click Data Process. First, calculate the actual field of view of the captured video so that the gate parameters can be measured.

In the popup window enter the number of frames per second at which the video is recorded and enter the calculated field of view in millimeters. To access the tracking results, open the results and tracking folders. To generate a video of the tracked fly, select Make Video.

Following leg segmentation, tracking and data processing, FLLIT automatically generates raw data for the positions of the body, and each leg claw, 20 gait parameters, and accompanying plots, and tracked videos. SCA3 is typified by an ataxic gait with body veering, erratic foot placement, and short lurching steps. To characterize the gait of mutant SCA3 flies and to investigate whether they display a similar gait to that of human patients, the relevant gait parameters generated by FLLIT can be analyzed.

SCA3-Q84 flies exhibit more turns, erratic foot placement as exhibited by a low footprint regularity, and increased leg domain overlap, enlarged leg domains in length and area, and decreased stride length. FLLIT can also generate a video showing the tracked fly and legs in the arena-centered and body-centered views, body trajectory, heading direction, and vertical and lateral displacements of each leg. Compared to SCA3-Q27 flies, SCA3-Q84 flies exhibit irregular intersecting leg domains of different sizes, indicative of a lurching, ataxic gait.

The raw data for the body and leg claw positions can further be analyzed for specific movements of interest. For example, the use of FLLIT allowed our group to determine that the rigid phenotypes displayed by Drosophila models of Parkinson's disease are caused by dopaminergic neuron dysfunction.