The authors would like to thank Moumita Chatterjee and Alice Liu for technical support, and the Bloomington Drosophila Stock Centre (Indiana, USA) for making available the Drosophila strains used in this work. This work was supported by the Institute of Molecular and Cell Biology, Singapore; the Bioinformatics Institute, Singapore; the Agency for Science Technology and Research Joint Council Organization (grant number 15302FG149 to SA and LC); the Clinical Research Flagship Programme (Parkinson&#8217;s Disease) administered by the Singapore Ministry of Health&#8217;s National Medical Research Council (grant number NMRC/TCR/013-NNI/2014 to SA), the University of Alberta (startup grant to LC), and the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant (grant number RGPIN-2019-04575 to LC).

1. System setup
<ol>
	<li>Ensure that the recording station has a high-speed camera and a stage over it to hold the arena chamber (Figure 1). Adjust the camera to record at a minimum of 250 frames per second (fps), with an accordingly fast shutter speed (in this case, recording is performed at 1000 fps with a shutter speed of 1 ms). 
	NOTE: Check that the video is suitable for tracking by ensuring that there is minimal or no motion blur of moving legs in all frames. If the moving leg is so blurred that a human annotator cannot track it, then the camera recording speed and/or shutter speed must be increased.</li>
	<li>Place the infrared LED lights at the top of the stage with a diffuser (translucent sheet) between the camera and sample (Figure 1A,B).</li>
	<li>Make the recording chamber by cutting a 1.6 mm thick acrylic sheet. In this experiment, use a field-of-view of 11 mm x 11 mm. Place the chamber between two glass slides (Figure 1C).</li>
</ol>
2. Preparation of flies for recording
<ol>
	<li>Transfer the flies to be recorded into a new food vial 24 h prior to recording. 
	NOTE: Do not use CO2 (usually used to anesthetize the flies upon first collection) on the flies less than 24 h before recording.</li>
	<li>About 40 min before recording, transfer the flies into empty vials and keep on ice for 5-7&#160; min.&#160;</li>
	<li>In the meantime, wipe the arena and clear glass slides with water and a wipe. 
	NOTE: Do not use ethanol to clean the chambers and the slides.</li>
	<li>Prepare the recording chamber. Fix one of the microscopic glass slides below the chamber with tape.</li>
	<li>When the flies have been anesthetized on ice, transfer one fly in each chamber using a clean brush. 
	NOTE: Both male and female flies can be used in this setup, and, as far as possible, flies of both sexes should be analyzed to avoid sex-specific bias.</li>
	<li>Secure the chamber with another microscopic glass slide with tape (Figure 1C).</li>
	<li>Keep the chambered flies at room temperature for 15 - 20 min for acclimatization.</li>
</ol>
3. Generation of videos for FLLIT analysis
NOTE: This step is specific to the video camera used. In this case, a commercially available video camera is used (see Table of Materials).
<ol>
	<li>Turn on the power source. Wait for the green LED for power and the orange LED for the ethernet interface connection to stabilize. Switch on the power for the infrared LED. Ensure that the voltage remains at 12.5 V.</li>
	<li>Open the Viewer application on the connected computer system.</li>
	<li>Change the recording frame rate to 1000 fps. Set the shutter speed to 1/1000 s (1 ms).</li>
	<li>Place the chamber with the fly on the recording arena and select the LIVE button. Ensure that the camera is focused on the leg tips when the fly is walking upright on the floor of the chamber; the leg tips should be in sharp focus.</li>
	<li>Click Record (Figure 2).</li>
	<li>Record the fly walking, ensuring that:&#160; 
	The fly walked in a relatively straight trajectory without touching the edge of the arena. 
	The fly walked at least three strides per leg. 
	The fly does not pause during the walk. 
	The distance walked is equivalent to at least one body length.&#160; 
	 
	NOTE: Having the background subtracted cleanly is critical for accurate segmentation. The automated background subtraction algorithm employed by FLLIT requires that the imaged fly moves at least one body length in distance.</li>
	<li>Click Rec Done to stop recording (Figure 2).</li>
	<li>Crop the video to ensure that the recording encompasses only a straight walk of the fly (as described in Step 3.6).</li>
	<li>Click Save (Figure 2). Save the files in &#8216;.mraw&#8217; or &#8216;.tiff&#8217; format in the respective folders. 
	NOTE: &#8216;.mraw&#8217; format gives greater flexibility to change the file name (if required) and for storage of videos compared to &#8216;.tiff&#8217; file format.</li>
</ol>
4. Installation of FLLIT program
NOTE: Up-to-date instructions can be found at: https://github.com/BII-wushuang/FLLIT/blob/master/Compiled/Readme.pdf
<ol>
	<li>Download FLLIT on any operating system
	<ol>
		<li>Download FLLIT from the following Github link: https://github.com/BII-wushuang/FLLIT/archive/master.zip. Extract the contents of the zip file.</li>
		<li>Download sample datasets from the following Google Drive link: https://bit.ly/2EibvNY. Create a folder Data under FLLIT-master/Compiled and place dataset folders in this Data directory.</li>
	</ol>
	</li>
	<li>Install FLLIT In Ubuntu
	<ol>
		<li>Navigate to the FLLIT/Compiled directory.</li>
		<li>Right-click and select Open in Terminal.</li>
		<li>Issue the following command to download and install the MATLAB runtime libraries to $HOME/MCR: 
		bash MCR_2016a.sh</li>
		<li>After installation of the MATLAB runtime libraries has been completed, issue the following command to ensure that executable rights are accorded to FLLIT: 
		chmod +x FLLIT</li>
		<li>Open a terminal in the FLLIT/Compiled directory and issue the following command to execute FLLIT: 
		bash run_FLLIT.sh $HOME/MCR/v901</li>
	</ol>
	</li>
	<li>Install FLLIT in Windows
	<ol>
		<li>For 7 and 10th Home Edition, install Docker Toolbox at: 
		(https://github.com/docker/toolbox/releases/download/v19.03.1/DockerToolbox-19.03.1.exe).</li>
		<li>For Windows 10 Pro or Enterprise Edition, install Docker Desktop for Windows at: (https://download.docker.com/win/stable/Docker\%20Desktop\%20Installer.exe).</li>
		<li>To enable execution of GUI applications in a Docker container on Windows, first install VcXSrV (https://sourceforge.net/projects/vcxsrv). On starting VcXsrv, configure the settings as in Figure S1. 
		NOTE: Make sure Docker and VcXsrv are running before starting FLLIT.</li>
		<li>Double click FLLIT.bat to run FLLIT. 
		NOTE: When executing for the first time, it will take some time to pull the Docker image from Docker Hub.</li>
	</ol>
	</li>
	<li>Install FLLIT in MacOS
	<ol>
		<li>Download Docker Desktop for MacOS at https://download.docker.com/mac/stable/Docker.dmg</li>
		<li>Install socat by opening a terminal and issuing the following command: 
		brew install socat</li>
		<li>Start socat with: 
		socat TCP-LISTEN:6000,reuseaddr,fork UNIX-CLIENT:\&#34;$DISPLAY\&#34; &#38; disown</li>
		<li>Install XQuartz (https://www.xquartz.org) to enable execution of GUI applications in a Docker container on MacOS. Start XQuartz and change the preferences by checking the Allow connections from network clients in the Security tab as shown in Figure S2. 
		NOTE: Make sure that Docker, socat and XQuartz are all running before starting FLLIT.&#160;</li>
		<li>Open a terminal in the FLLIT/Compiled directory and execute FLLIT with the following command: 
		bash FLLIT_Mac.sh 
		NOTE: When executing for the first time, it will take some time to pull the Docker image from Docker Hub.</li>
	</ol>
	</li>
</ol>
5. Running FLLIT for automated leg tracking
<ol>
	<li>Segmentation
	<ol>
		<li>Convert the video into individual TIFF files and copy into the FLLIT data folder.</li>
		<li>Run FLLIT (In Ubuntu, right click to open FLLIT in Terminal).</li>
		<li>Select the folder containing the frame-by-frame TIFF images of the video to be tracked and click the Add button.</li>
		<li>In the pop-up window choose 0 for carrying out leg segmentation only, or 1 to include leg tracking with leg segmentation.</li>
		<li>Click Done to initiate segmentation and tracking of the selected video.&#160;</li>
	</ol>
	</li>
	<li>Tracking
	<ol>
		<li>To check the accuracy of tracking and carry out error corrections (if any), click on Select Data Folder. Select the folder to be tracked and click Open.</li>
		<li>Click on View Tracking. 
		NOTE: Check that Viewer Mode remains selected while viewing the tracked leg positions. Otherwise, any previous corrections made will be over-ridden.</li>
		<li>Check labeling for all the legs in the first frame 
		NOTE: Since the leg labels are placed according to the position on the screen, if the fly is walking upright, the fly&#8217;s RIGHT side is labelled as L1 (fore-leg), L2 (mid-leg), L3 (hind-leg) and&#160;the fly&#8217;s LEFT side is labelled as R1 (fore-leg), R2 (mid-leg), R3 (hind-leg), respectively (Figure 3). If the fly is walking&#160;upside, down the leg labels will be correctly annotated.</li>
		<li>If a leg is wrongly labeled and a correction is required, click Pause viewing, followed by Adjust Prediction (Figure 3).</li>
		<li>From the right panel headed Leg to Adjust, select the leg that requires correction.&#160;</li>
		<li>Double click on the correct position for this leg in the image window, click on Save, and then Exit. To go to the previous frame or subsequent frame, click Pause viewing followed by the forward and backward &#60;I and I&#62; buttons, respectively (Figure 3).</li>
		<li>To correct any incorrectly tracked videos, open the Data folder of the video to be retracked and select Manually Initiate Tracking.</li>
		<li>Click the Tracking button, which will then change its label to Initial.</li>
		<li>Click Adjust Prediction and correct the leg labels by double-clicking each leg tip and then assigning it with the correct leg label. Click Save and Exit.</li>
		<li>Click on Resume to initiate tracking.</li>
	</ol>
	</li>
	<li>Data processing and video generation
	<ol>
		<li>Click on Data Process. In the popup window, type in the number of&#160;frames per second (fps) at which the videos were recorded (e.g., 1,000 fps).&#160;</li>
		<li>Use the following equation to calculate the actual field of view of the captured video so that gait parameters can be measured in millimeters: 
		 
		<img alt="Equation 1" src="/files/ftp_upload/61012/61012eq1.jpg" /> 
		 
		NOTE: For example, if the actual chamber size is 7 mm, the width of the image frame is 137 mm, the width of the chamber in the image frame on the computer screen is 81 mm, and the width of the field of view was 11.83 mm (Figure S3).</li>
		<li>To see the tracking results, go to the Tracking folder under the Results folder.</li>
		<li>To generate a video of the tracked fly, select Make video. The video will be saved in the same Results folder as that of the original video data analyzed. 
		NOTE: The Start&#160;(first) and End (last) frames of the video can be selected.</li>
	</ol>
	</li>
	<li>Normalization to body length of each fly. 
	NOTE: As each fly may be slightly different in size, certain gait parameters should be normalized to the body length of each fly to facilitate comparisons (e.g., stride length may be longer in larger flies, and shorter in smaller flies).
	<ol>
		<li>Open three still images from the video of each fly (usually first, middle and last frames) using image software.</li>
		<li>Magnify each image frame to 800% and label the anterior-most pixel of the head and posterior-most pixel of the abdomen at the midline using a bright color (e.g., yellow).</li>
		<li>Open the labeled images in ImageJ.</li>
		<li>Use Set Scale to input the scale accordingly: Set Distance in pixels: 512; Known distance: Actual Field of view (mm) as measured in step 5.3.2.; Unit of length: mm.</li>
		<li>Draw a straight line between the labelled head and abdomen tip pixels to obtain the body length.</li>
		<li>Open Set Scale again to obtain the value in a known distance, which is the body length in mm.</li>
		<li>Take an average of the length determined in each of the three images to obtain the average body size in mm.</li>
	</ol>
	</li>
</ol>

<ol>
	<li>Wu, S., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Fully+automated+leg+tracking+of+Drosophila+neurodegeneration+models+reveals+distinct+conserved+movement+signatures.">Fully automated leg tracking of Drosophila neurodegeneration models reveals distinct conserved movement signatures.</a> PLoS Biology. 17 (6), 3000346 (2019).</li><li>Mathis, A., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=DeepLabCut:+markerless+pose+estimation+of+user-defined+body+parts+with+deep+learning.">DeepLabCut: markerless pose estimation of user-defined body parts with deep learning.</a> Nature Neuroscience. 19, 1281-1289 (2018).</li><li>Pereira, T. D., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Fast+animal+pose+estimation+using+deep+neural+networks.">Fast animal pose estimation using deep neural networks.</a> Nature Methods. 16 (1), 117-125 (2019).</li><li>Machado, A. S., Darmohray, D. M., Fayad, J., Marques, H. G., Carey, M. R. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=A+quantitative+framework+for+whole-body+coordination+reveals+specific+deficits+in+freely+walking+ataxic+mice.">A quantitative framework for whole-body coordination reveals specific deficits in freely walking ataxic mice.</a> eLife. 4, (2015).</li><li>Lu, B., Vogel, H. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Drosophila+models+of+neurodegenerative+diseases.">Drosophila models of neurodegenerative diseases.</a> Annual Reviews in Pathology. 4, 315-342 (2009).</li><li>McGurk, L., Berson, A., Bonini, N. M. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Drosophila+as+an+In+Vivo+Model+for+Human+Neurodegenerative+Disease.">Drosophila as an In Vivo Model for Human Neurodegenerative Disease.</a> Genetics. 201 (2), 377-402 (2015).</li><li>Dionne, H., Hibbard, K., Cavallaro, A., Kao, J. -. C., Rubin, G. M. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Genetic+reagents+for+making+split-GAL4+lines+in+Drosophila.">Genetic reagents for making split-GAL4 lines in Drosophila.</a> bioRxiv. , (2017).</li><li>Cande, J., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Optogenetic+dissection+of+descending+behavioral+control+in+Drosophila.">Optogenetic dissection of descending behavioral control in Drosophila.</a> eLife. 7, (2018).</li><li>Nern, A., Pfeiffer, B. D., Rubin, G. M. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Optimized+tools+for+multicolor+stochastic+labeling+reveal+diverse+stereotyped+cell+arrangements+in+the+fly+visual+system.">Optimized tools for multicolor stochastic labeling reveal diverse stereotyped cell arrangements in the fly visual system.</a> Proceedings of the National Academy of Sciences of the United States of America. 112 (22), 2967-2976 (2015).</li><li>Xie, T., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=A+Genetic+Toolkit+for+Dissecting+Dopamine+Circuit+Function+in+Drosophila.">A Genetic Toolkit for Dissecting Dopamine Circuit Function in Drosophila.</a> Cell Reports. 23 (2), 652-665 (2018).</li><li>Warrick, J. M., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Ataxin-3+suppresses+polyglutamine+neurodegeneration+in+Drosophila+by+a+ubiquitin-associated+mechanism.">Ataxin-3 suppresses polyglutamine neurodegeneration in Drosophila by a ubiquitin-associated mechanism.</a> Molecular Cell. 18 (1), 37-48 (2005).</li><li>Ebersbach, G., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Comparative+analysis+of+gait+in+Parkinson's+disease,+cerebellar+ataxia+and+subcortical+arteriosclerotic+encephalopathy.">Comparative analysis of gait in Parkinson's disease, cerebellar ataxia and subcortical arteriosclerotic encephalopathy.</a> Brain. 122 (7), 1349-1355 (1999).</li><li>Stolze, H., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Typical+features+of+cerebellar+ataxic+gait.">Typical features of cerebellar ataxic gait.</a> Journal of Neurology, Neurosurgery and Psychiatry. 73 (3), 310-312 (2002).</li><li>Mendes, C. S., Bartos, I., Akay, T., Marka, S., Mann, R. S. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Quantification+of+gait+parameters+in+freely+walking+wild+type+and+sensory+deprived+Drosophila+melanogaster.">Quantification of gait parameters in freely walking wild type and sensory deprived Drosophila melanogaster.</a> eLife. 2, 00231 (2013).</li><li>DeAngelis, B. D., Zavatone-Veth, J. A., Clark, D. A. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=The+manifold+structure+of+limb+coordination+in+walking+Drosophila.">The manifold structure of limb coordination in walking Drosophila.</a> eLife. 8, (2019).</li><li>Gunel, S., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=DeepFly3D,+a+deep+learning-based+approach+for+3D+limb+and+appendage+tracking+in+tethered,+adult+Drosophila.">DeepFly3D, a deep learning-based approach for 3D limb and appendage tracking in tethered, adult Drosophila.</a> eLife. 8, (2019).</li><li>Isakov, A., et al. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=Recovery+of+locomotion+after+injury+in+Drosophila+melanogaster+depends+on+proprioception.">Recovery of locomotion after injury in Drosophila melanogaster depends on proprioception.</a> The Journal of Experimental Biology. 219, 1760-1771 (2016).</li><li>Aw, S. S., Lim, I. K. H., Tang, M. X. M., Cohen, S. M. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=A+Glio-Protective+Role+of+mir-263a+by+Tuning+Sensitivity+to+Glutamate.">A Glio-Protective Role of mir-263a by Tuning Sensitivity to Glutamate.</a> Cell Reports. 19 (9), 1783-1793 (2017).</li><li>Eberl, D. F., Duyk, G. M., Perrimon, N. <a target="_blank" href="http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed&cmd=Search&doptcmdl=Citation&defaultField=Title+Word&term=A+genetic+screen+for+mutations+that+disrupt+an+auditory+response+in+Drosophila+melanogaster.">A genetic screen for mutations that disrupt an auditory response in Drosophila melanogaster.</a> Proceedings of the National Academy of Sciences of the United States of America. 94 (26), 14837-14842 (1997).</li></ol>

The authors have nothing to disclose.

In this manuscript, we describe in detail the steps involved in using FLLIT, an automated machine-learning program1, to analyze gait in freely walking Drosophila. After tracking and data analysis, FLLIT automatically generates raw data for the positional information of the body and leg claws, producing twenty body and gait features as well as a video of the tracked fly to enable gait visualization.
There are now a number of methods for leg movement tracking of Drosophila and other animals1,2,3,4,14,15,16, giving researchers a wide range of options depending on the goals of the experiment. Some of these are foot printing-based approaches, which are highly accurate but which report only claw contact points with the detection surface4,14. On the other hand, recent deep learning approaches2,3,16 are highly versatile, allowing analysis of behaviors that require tracking of leg joints and other body parts in any animal, with the caveat that the algorithms need to first be trained with user annotated datasets. A third type of approach uses morphology or image-contrast-based methods1,15,17 to find the outline of each leg to identify claw positions. In general, these methods deal poorly with behaviors where the legs cross over (e.g. during grooming). FLLIT combines the second and third approaches, using morphological parameters to train a boosting algorithm for leg segmentation. This allows FLLIT to bypass the laborious task of user annotation to generate the training dataset, while enhancing accuracy using machine learning. Future improvements to FLLIT will have to deal with instances where legs cross over, to allow for analysis of more complex behaviors.
FLLIT is robust to slight changes in illumination, recording resolution and frame speed1. However, frame speed of recorded videos should not fall below 250 fps, and FLLIT runs optimally for videos recorded at 1000 fps. If there is motion blur in the images, such that it is challenging for a human annotator to identify the leg position, FLLIT will not be able to accurately identify leg tips in those frames. In the light of this, it is essential that the camera be focused sharply on the leg tips. To prevent segmentation artifacts, the arena should be cleaned thoroughly, and should not be moved during the recording. For accurate background subtraction and clean segmentation, the fly should move at least one body length during the recording, without pausing. After automatic segmentation and tracking the labeling of the all legs must be checked. If the fly gait is not tracked or tracked wrongly, the file should be retracked manually using the Manually Initiate Tracking option (step 5.2.7 &#8211; 5.2.10).
Neurodegenerative diseases and movement disorders are increasingly prevalent in our aging societies. Fly models of neurodegeneration have been studied for more than 2 decades, during which advances have been made regarding the molecular and cellular aspects of disease pathophysiology. However, specific behavioral consequences of disease have been technically difficult to assess. For example, while reports of trembling movements in the fly have been made18,19, these had not been quantitatively studied until recently1. The climbing assay has been a useful and quantitative, yet relatively coarse measure6. This technical deficit has similarly hampered high-resolution movement analysis in other animal models. The advent of new tools for behavioral analysis, hence, has promise to rejuvenate the field of movement disorders to enable researchers to study how molecular and cellular mechanisms of neuromuscular diseases lead to specific behavioral outcomes in animal models. In this paper and in our previous work1, we showed using FLLIT that fly models of SCA3 exhibit a hyperkinetic ataxic gait, while PD fly models exhibit a hypokinetic rigid gait, recapitulating movement hallmarks of the respective human diseases1. Gait analysis also enabled us to identify distinct neuronal populations underlying specific movement dysfunctions. Going forward, detailed movement analysis, combined with the powerful imaging and functional tools available in the fly, will allow us to gain novel insight into mechanisms of locomotor dysfunction, illuminating our understanding of neurodegenerative diseases with respect to circuit mechanisms.
FLLIT should be widely applicable to study gait in other small arthropods, as it was previously demonstrated to be highly accurate for tracking spider leg movements1. While we focus here on the use of detailed movement phenotyping for quantifying pathogenic gait and its underlying circuitry, advances in movement tracking have already revolutionized, and will have continuing impact on, the understanding of normal walking coordination and gait and its underlying circuits, especially in myriad different branches of the evolutionary tree.&#160;

In the last few decades, neurodegenerative diseases and movement disorders have grown more prevalent in our aging populations. Although our understanding of many neurodegenerative diseases has advanced at the molecular and cellular level, fundamental features of the affected neuronal circuitry underlying disease remain poorly understood. Recently developed behavioral tracking tools1,2,3,4 now allow us to study movement abnormalities in animal disease models in order to identify molecular, cellular and circuit dysregulation underlying disease.
Molecular pathways involved in many neurodegenerative diseases are conserved in the fruit fly&#160;Drosophila&#160;melanogaster, and Drosophila disease models have helped to elucidate fundamental mechanisms underlying neurodegeneration5,6. We recently showed that fly models of Parkinson&#8217;s Disease (PD) and Spinocerebellar ataxia 3 (SCA3) exhibit distinct, conserved gait signatures that resemble those of the respective human diseases1, demonstrating that the fly model can be used to understand circuit mechanisms underlying movement dysfunction in specific movement disorders. The rich and continually growing arsenal of tools in the fly model for targeted manipulation and visualization of neurons at the single gene and single cell level7,8,9,10 makes the fly an ideal model one to probe the relationship between disease pathways, neuronal circuitry and behavioral phenotypic manifestation in vivo. To enable precise, automated insect gait analysis, we recently developed a machine learning method, Feature&#160;Learning-based&#160;LImb segmentation and Tracking (FLLIT)1.
FLLIT consists of a fully automated multi-stage algorithm that first segments the leg pixels, which are subsequently used to locate and track the corresponding leg claws. FLLIT employs a boosting algorithm for segmentation, in contrast to deep learning algorithms used in recent work2,3. There are some similarities with convolutional neural networks in that for both frameworks, feature extraction is done automatically through learning convolutional kernels. The first step in FLLIT involves using morphological operations (edge and skeletonization) to automatically generate positive (pixels on the legs) and negative (background or pixels on the fly body) training samples with high confidence. Hence, FLLIT is fully automated and does not require user annotated training samples. Using the above training samples, a classifier is then trained in the framework of a boosting algorithm. An ensemble of weak classifiers is iteratively learnt, with each consisting a set of convolutional kernels for feature extraction and a decision tree. The final learnt classifier is then used for leg segmentation and is able to better discern difficult regions/hard samples better than morphological operations, producing an overall much more accurate segmentation for tracking1. From the segmented legs, we locate the tips and track them using the Hungarian algorithm: by matching tips across frames such that the sum of the distance moved by each tip is minimized. FLLIT can handle occlusion cases by remembering the last seen location (in fly centered coordinates) so that a leg tip is recovered once it is no longer under occlusion.
We previously showed that FLLIT can automatically and accurately track leg movements and analyze gait in an unmarked, freely moving fly or spider from high-speed video1; FLLIT should hence be broadly applicable for arthropod leg tracking. By extracting machine learning training sets using morphological parameters, FLLIT automatically trains itself to segment and track insect legs without the need for laborious manual annotation, which is required for most deep learning methods. FLLIT is hence fully automated. After leg segmentation and tracking, FLLIT automatically produces raw tracked body and leg positions in every video frame, 20 gait parameters, 5 plots and a tracked video for gait analysis and visualization of gait movements. This protocol provides a step-by-step guide to using FLLIT.

<table><tbody><tr><td>Acrylic Sheets</td><td>Dama</td><td>1.6 mm thickness, Clear sheets</td><td>Singapore, Singapore</td></tr><tr><td>Clear Glass slides</td><td>Biomedia</td><td>BMH 880101</td><td>Singapore, Singapore</td></tr><tr><td>High speed camera</td><td>Photron</td><td>Fastcam MC2.1</td><td>Tokyo, Japan. A shutter speed of 1 msec or faster is ideal to reduce motion blur of captured images</td></tr><tr><td>Infra Red LED</td><td>Any - Generic from hardware store</td><td>940nm Infrared Light Board</td><td>Singapore, Singapore</td></tr><tr><td>Kimwipe</td><td>Kimberly Clark</td><td>34155-01LS</td><td>Irving, Texas, USA</td></tr></tbody></table>

fully automated leg tracking in freely moving insects using feature learning leg segmentation and tracking (fllit)

The Drosophila model has been invaluable for the study of neurological function and for understanding the molecular and cellular mechanisms that underlie neurodegeneration. While fly techniques for the manipulation and study of neuronal subsets have grown increasingly sophisticated, the richness of the resultant behavioral phenotypes has not been captured at a similar detail. To be able to study subtle fly leg movements for comparison amongst mutants requires the ability to automatically measure and quantify high-speed and rapid leg movements. Hence, we developed a machine-learning algorithm for automated leg claw tracking in freely walking flies, Feature Learning-based Limb segmentation and Tracking (FLLIT). Unlike most deep learning methods, FLLIT is fully automated and generates its own training sets without a need for user annotation, using morphological parameters built into the learning algorithm. This article describes an in depth protocol for carrying out gait analysis using FLLIT. It details the procedures for camera setup, arena construction, video recording, leg segmentation and leg claw tracking. It also gives an overview of the data produced by FLLIT, which includes raw tracked body and leg positions in every video frame, 20 gait parameters, 5 plots and a tracked video. To demonstrate the use of FLLIT, we quantify relevant diseased gait parameters in a fly model of Spinocerebellar ataxia 3.

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, 5 plots and a tracked video (Table 1).
Here, we demonstrate these analyses using a fly model of Spinocerebellar ataxia 3 (SCA3). The pan-neuronal driver Elav-GAL4 was used to drive either the full-length wildtype human SCA3 with 27 glutamines in the polyQ tract (UAS-SCA3-flQ27), or a full-length mutant human SCA3 with 84 glutamines in the polyQ tract (UAS-SCA3-flQ84)11. SCA3 is typified by an ataxic gait with body veering, erratic foot placement and short, lurching steps12,13 (Table 2). To characterize gait of mutant SCA3 flies and investigate whether they display a similar gait to that of human patients, we analyzed relevant gait parameters generated by FLLIT, namely: Number of body turns, footprint regularity, leg domain overlap and sizes, and leg stride lengths (Table 2).
We found that SCA3-Q84 flies exhibited more turns (Figure 4A,A&#8217;), erratic foot placement as exhibited by low footprint regularity (enlarged standard deviations of the AEP14) (Figure 4B), increased leg domain overlap (Figure 4C-D), enlarged leg domains in length and area (Figure 4E,F), and decreased stride length (Figure 4G).
FLLIT also generates a video showing the tracked fly and legs in the arena-centered and body-centered views, body trajectory and heading direction, and vertical and lateral displacements of each leg (Figure 5). The tracked videos allow side-by-side comparison of leg movements in different flies. Representative videos of Elav-GAL4&#62;SCA3-flQ27 (Video 1) and Elav-GAL4&#62;SCA3-flQ84 (Video 2) flies demonstrate that compared to Elav-GAL4&#62;UAS-SCA3-flQ27 flies (Figure 5A), Elav-GAL4&#62;UAS-SCA3-flQ84 flies (Figure 5B) exhibit irregular, intersecting leg domains of different sizes, indicative of a lurching, ataxic gait.
<img alt="Figure 1" class="xfigimg" src="/files/ftp_upload/61012/61012fig1.jpg" /> 
Figure 1. Setup of the recording station and arena. Recordings from the (A) front and (B) side views. (C) An example of an arena used for making fly recordings for FLLIT tracking. <a href="https://www.jove.com/files/ftp_upload/61012/61012fig1large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Figure 2" class="xfigimg" src="/files/ftp_upload/61012/61012fig2.jpg" /> 
Figure 2: View of the active window during fly gait recording using a dual head camera, which allows simultaneous recording of two flies. <a href="https://www.jove.com/files/ftp_upload/61012/61012fig2large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Figure 3" class="xfigimg" src="/files/ftp_upload/61012/61012fig3.jpg" /> 
Figure 3: Active FLLIT window showing the button panel and labelled legs after segmentation and tracking. <a href="https://www.jove.com/files/ftp_upload/61012/61012fig3large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Figure 4" class="xfigimg" src="/files/ftp_upload/61012/61012fig4.jpg" /> 
Figure 4: Representative FLLIT-generated data for relevant gait parameters of flies expressing wildtype (SCA3-flQ27) vs. mutant (SCA3-flQ84) SCA3. (A) Number of turns in the body trajectory. (B) Mid-leg footprint regularity normalized to body length. (C-C&#8217;) Traversed leg domains of each leg. (D) Domain overlap between legs. (E) Mid-leg domain length normalized to body length. (F) Mid-leg domain area normalized to body length2. (G) Mid-leg stride length normalized to body length. <a href="https://www.jove.com/files/ftp_upload/61012/61012fig4large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Figure 5" class="xfigimg" src="/files/ftp_upload/61012/61012fig5.jpg" /> 
Figure 5: Snapshot of representative FLLIT-generated videos. (A) Elav-GAL4&#62;UAS-SCA3-flQ27 and (B) Elav-GAL4&#62;UAS-SCA3-flQ84 flies. <a href="https://www.jove.com/files/ftp_upload/61012/61012fig5large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Video 1" src="/files/ftp_upload/61012/61012video1.jpg" /> 
Video 1: Representative video of a fly expressing pan-neuronal wild-type human full-length SCA3 (Elav-GAL4&#62;UAS-SCA3-flQ27). <a href="https://www.jove.com/files/ftp_upload/61012/Video_1_Elav-GAL4_SCA3-flQ27_22d.mp4" target="_blank">Please click here to view this video. (Right-click to download.)</a>
<img alt="Video 2" src="/files/ftp_upload/61012/61012video2.jpg" /> 
Video 2: Representative video of a fly expressing pan-neuronal mutant human full-length SCA3 (Elav-GAL4&#62;UAS-SCA3-flQ84). <a href="https://www.jove.com/files/ftp_upload/61012/Video_2_Elav-GAL4_SCA3-flQ84_22d.mp4" target="_blank">Please click here to view this video. (Right-click to download.)</a>
<img alt="Supplemental Figure 1" src="/files/ftp_upload/61012/61012supfig1.jpg" /> 
Supplemental Figure 1: Configurations for VcXSrv. <a href="https://www.jove.com/files/ftp_upload/61012/61012supfig1large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Supplemental Figure 2" src="/files/ftp_upload/61012/61012supfig2.jpg" /> 
Supplemental Figure 2: Configuration for Xquartz. <a href="https://www.jove.com/files/ftp_upload/61012/61012supfig2large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<img alt="Supplemental Figure 3" src="/files/ftp_upload/61012/61012supfig3.jpg" /> 
Supplemental Figure 3: Image labelled with the dimensions needed for calculating Field of view. <a href="https://www.jove.com/files/ftp_upload/61012/61012supfig3large.jpg" target="_blank">Please click here to view a larger version of this figure.</a>
<table border="1" fo:keep-together.within-page="1">
	<tbody>
		<tr>
			<td>Category</td>
			<td>Parameters</td>
			<td>Description</td>
			<td>File/Plot (if applicable)</td>
		</tr>
		<tr>
			<td rowspan="4">Raw data</td>
			<td>Body position</td>
			<td>Positional coordinates of the body centroid in each frame</td>
			<td>First two columns of CoM.csv</td>
		</tr>
		<tr>
			<td>Body trajectory</td>
			<td>Angle of rotation of the body axis in degrees (relative to the y-axis)</td>
			<td>Third column of CoM.csv</td>
		</tr>
		<tr>
			<td>Arena-centred leg claw 
			positions</td>
			<td>Positional coordinates of each leg claw in each frame based on arena coordinates</td>
			<td>trajectory.csv</td>
		</tr>
		<tr>
			<td>Body-centred leg claw 
			positions</td>
			<td>Positional coordinates of each leg claw in each frame 
			based on arena coordinates</td>
			<td>norm_trajectory.csv</td>
		</tr>
		<tr>
			<td rowspan="3">Body motion</td>
			<td>Body length (mm)</td>
			<td>Length of the sample animal estimated in each frame (anterior-most position on head to posterior-most 
			position on the wings)</td>
			<td>bodylength.csv</td>
		</tr>
		<tr>
			<td>Instantaneous body velocity 
			(mm/s)</td>
			<td>Instantaneous velocity of the body (centroid) in the sample animal</td>
			<td>BodyVelocity.csv; BodyVelocity.pdf</td>
		</tr>
		<tr>
			<td>Turning points of the body 
			trajectory</td>
			<td>To locate the turning points, the trajectory is reduced to a piecewise-linear curve using the Dougl as&#226;ASPeucker 
			algorithm, following which a turning event is identified as involving an angle &#62; 50 deg between two neighbouring 
			linear segments constituting the simplified trajectory</td>
			<td>BodyTrajectory.pdf</td>
		</tr>
		<tr>
			<td rowspan="9">Individual stride parameters</td>
			<td>Stride duration (ms)</td>
			<td>The duration of a stride event</td>
			<td>StrideParameters.csv</td>
		</tr>
		<tr>
			<td>Stride period (ms)</td>
			<td>The duration from one stride event to the next</td>
			<td></td>
		</tr>
		<tr>
			<td>Stride displacement (mm)</td>
			<td>Displacement of the leg claw during a stride event</td>
			<td></td>
		</tr>
		<tr>
			<td>Stride path covered (mm)</td>
			<td>Total path covered by the leg claw during a stride event</td>
			<td></td>
		</tr>
		<tr>
			<td>Anterior extreme position 
			(mm)</td>
			<td>Landing position (relative to the body) of a leg claw at the end of a stride event</td>
			<td></td>
		</tr>
		<tr>
			<td>Posterior extreme position (mm)</td>
			<td>Take-off position (relative to the body) of a leg claw at the start of a stride event</td>
			<td></td>
		</tr>
		<tr>
			<td>Stride amplitude (mm)</td>
			<td>Displacement along the direction of motion for a stride event</td>
			<td></td>
		</tr>
		<tr>
			<td>Stance linearity (mm)</td>
			<td>Defined as the deviation of the stride path from a curve smoothed over (at 20ms intervals) the corresponding anterior 
			and posterior extreme positions of the stride</td>
			<td></td>
		</tr>
		<tr>
			<td>Stride stretch (mm)</td>
			<td>Distance of the leg claw position from the body centre in the middle of a stride event</td>
			<td></td>
		</tr>
		<tr>
			<td rowspan="10">Leg motion</td>
			<td>Leg speed (mm/s)</td>
			<td>The instantaneous speed of each leg</td>
			<td>LegSpeed.csv; Gait.pdf</td>
		</tr>
		<tr>
			<td>Gait index</td>
			<td>This measures the type of gait coordination exhibited by the (six-legged) sample animal during its motion. A gait 
			index of 1 corresponds to a tripod gait, _1 corresponds to a tetrapod gait while 0 constitutes an non-canonical gait. 
			In our implementation, the gait index is obtained by a moving average over a 120 ms window</td>
			<td>GaitIndex.csv; GaitIndex.pdf</td>
		</tr>
		<tr>
			<td>Movement percentage</td>
			<td>Percentage of the time that a leg is in motion</td>
			<td>LegParameters.csv</td>
		</tr>
		<tr>
			<td>Mean stride period (ms)</td>
			<td>Average duration from one stride event to the next</td>
			<td>LegParameters.csv</td>
		</tr>
		<tr>
			<td>Footprint regularity (mm)</td>
			<td>Measured as the standard deviations of the posterior and</td>
			<td>LegParameters.csv</td>
		</tr>
		<tr>
			<td></td>
			<td>anterior extreme positions of a leg</td>
			<td></td>
		</tr>
		<tr>
			<td>Leg trajectory domain area 
			(mm2)</td>
			<td>The area of the minimal convex hull that contains the entire leg trajectory in the body-centred frame of reference</td>
			<td>LegParameters.csv; LegDomain.pdf</td>
		</tr>
		<tr>
			<td>Length and width of the 
			leg trajectory domain (mm)</td>
			<td>Obtained via the maximum projected distance of the claw positions onto the major (domain length) and minor 
			(domain width) principal axes</td>
			<td>LegParameters.csv</td>
		</tr>
		<tr>
			<td>Leg domain intersection/overlap 
			(mm2)</td>
			<td>The intersection/overlap between each possible</td>
			<td>LegDomainOverlap.csv</td>
		</tr>
		<tr>
			<td>Stance width (mm)</td>
			<td>Average of the distance between the AEP and PEP of the left and middle legs</td>
			<td>StanceWidth.csv</td>
		</tr>
	</tbody>
</table>
Table 1: Gait parameters generated by FLLIT.
<table border="1" fo:keep-together.within-page="1">
	<tbody>
		<tr>
			<td></td>
			<td colspan="4">Gait feature</td>
		</tr>
		<tr>
			<td>Gait features of Spinocerebellar ataxia 3 (SCA3)</td>
			<td>Veering</td>
			<td>Erratic foot placement and leg crossing over</td>
			<td>Lurching steps</td>
			<td>Short strides</td>
		</tr>
		<tr>
			<td>Measurement Parameter</td>
			<td>Number of body turn events</td>
			<td>Footprint regularity</td>
			<td>Size of leg domains, degree of domain overlap</td>
			<td>Stride length</td>
		</tr>
		<tr>
			<td>FLLIT File</td>
			<td>BodyTrajectory.pdf</td>
			<td>LegParameters.csv</td>
			<td>LegDomainOverlap.csv</td>
			<td>StrideParameters.csv</td>
		</tr>
	</tbody>
</table>
Table 2: Table showing hallmark SCA3 gait features in human patients with their corresponding FLLIT parameters and output files.

Watch this Scientific Journal Video about Fully Automated Leg Tracking in Freely Moving Insects using Feature Learning Leg Segmentation and Tracking FLLIT at JoVE.com

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

We describe detailed protocols for using FLLIT, a fully automated machine learning method for leg claw movement tracking in freely moving Drosophila melanogaster and other insects. These protocols can be used to quantitatively measure subtle walking gait movements in wild type flies, mutant flies and fly models of neurodegeneration.

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

The Drosophila model has been invaluable for the study of neurological function and for understanding the molecular and cellular mechanisms that underlie ...

fully-automated-leg-tracking-freely-moving-insects-using-feature

Research

JoVE Journal

Behavior

6.7K Views.  Agency for Science, Technology and Research. FLLIT is een algoritme dat de lichaams- en beensamenstellingen van een bijgehouden dier berekent. En dit is handig voor het karakteriseren van de circuits die ten grondslag liggen aan zowel normaal als ziekte loopgedrag. Het unieke aan FLLIT is dat het geen geannoteerde gegevenssets van de gebruiker nodig heeft voor training en dus volledig geautomatiseerd is.Maar analyseert ook automatisch de track data produceren van een reeks van gang metingen, gang percelen voor visualisatie en een...

Video: Fully Automated Leg Tracking in Freely Moving Insects using Feature Learning Leg Segmentation and Tracking FLLIT

Quantifying Learning in Young Infants: Tracking Leg Actions During a Discovery-learning Task

Task-specific actions emerge from spontaneous movement during infancy. It has been proposed that task-specific actions emerge through a discovery-learning process. Here a method is described in which 3-4 month old infants learn a task by discovery and their leg movements are captured to quantify the learning process. This discovery-learning task uses an infant activated mobile that rotates and plays music based on specified leg action of infants. Supine infants activate the mobile by moving their feet vertically across a virtual threshold. This paradigm is unique in that as infants independently discover that their leg actions activate the mobile, the infants&#8217; leg movements are tracked using a motion capture system allowing for the quantification of the learning process. Specifically, learning is quantified in terms of the duration of mobile activation, the position variance of the end effectors (feet) that activate the mobile, changes in hip-knee coordination patterns, and changes in hip and knee muscle torque. This information describes infant exploration and exploitation at the interplay of person and environmental constraints that support task-specific action. Subsequent research using this method can investigate how specific impairments of different populations of infants at risk for movement disorders influence the discovery-learning process for task-specific action.

A method is described in which 3-4 month old infants learn a task by discovery and their leg movements are captured to quantify the learning process.

Task-specific actions emerge from spontaneous movement during infancy. It has been proposed that task-specific actions emerge through a discovery-learning ...

Studying the Neural Basis of Adaptive Locomotor Behavior in Insects

Studying the neural basis of walking behavior, one often faces the problem that it is hard to separate the neuronally produced stepping output from those leg movements that result from passive forces and interactions with other legs through the common contact with the substrate. If we want to understand, which part of a given movement is produced by nervous system motor output, kinematic analysis of stepping movements, therefore, needs to be complemented with electrophysiological recordings of motor activity. The recording of neuronal or muscular activity in a behaving animal is often limited by the electrophysiological equipment which can constrain the animal in its ability to move with as many degrees of freedom as possible. This can either be avoided by using implantable electrodes and then having the animal move on a long tether (i.e. Clarac et al., 1987; Duch &amp; Pfl&uuml;ger, 1995; B&ouml;hm et al., 1997; Gruhn &amp; Rathmayer, 2002) or by transmitting the data using telemetric devices (Kutsch et al, 1993; Fischer et al., 1996; Tsuchida et al. 2004; Hama et al., 2007; Wang et al., 2008). Both of these elegant methods, which are successfully used in larger arthropods, often prove difficult to apply in smaller walking insects which either easily get entangled in the long tether or are hindered by the weight of the telemetric device and its batteries. In addition, in all these cases, it is still impossible to distinguish between the purely neuronal basis of locomotion and the effects exerted by mechanical coupling between the walking legs through the substrate. One solution for this problem is to conduct the experiments in a tethered animal that is free to walk in place and that is locally suspended, for example over a slippery surface, which effectively removes most ground contact mechanics. This has been used to study escape responses (Camhi and Nolen, 1981; Camhi and Levy, 1988), turning (Tryba and Ritzman, 2000a,b; Gruhn et al., 2009a), backward walking (Graham and Epstein, 1985) or changes in velocity (Gruhn et al., 2009b) and it allows the experimenter easily to combine intra- and extracellular physiology with kinematic analyses (Gruhn et al., 2006).
We use a slippery surface setup to investigate the timing of leg muscles in the behaving stick insect with respect to touch-down and lift-off under different behavioral paradigms such as straight forward and curved walking in intact and reduced preparations.

We describe a method to record motor activity, timed to the electrically recorded tarsal contact signal in a tethered insect, walking on a slippery surface. This is used to study the neural basis of adaptive behavior under reduced influence of mechanical interaction between legs through the substrate.

Studying the neural basis of walking behavior, one often faces the problem that it is hard to separate the neuronally produced stepping output from those ...

Neuroscience

High-Resolution Video Tracking of Locomotion in Adult Drosophila Melanogaster

Flies provide an important model for studying complex behavior due to the plethora of genetic tools available to researchers in this field. Studying locomotor behavior in Drosophila melanogaster relies on the ability to be able to quantify changes in motion during or in response to a given task. For this reason, a high-resolution video tracking system, such as the one we describe in this paper, is a valuable tool for measuring locomotion in real-time. Our protocol involves the use of an initial air pulse to break the flies momentum, followed by a thirty second filming period in a square chamber. A tracking program is then used to calculate the instantaneous speed of each fly within the chamber in 10 msec increments. Analysis software then compiles this data, and outputs a variety of parameters such as average speed, max speed, time spent in motion, acceleration, etc. This protocol will discuss proper feeding and management of flies for behavioral tasks, handling flies without anesthetization or immobilization, setting up a controlled environment, and running the assay from start to finish.

The study of complex locomotor behavior in Drosophila melanogaster is dependent upon the ability to quantify changes in a given fly's movement. This article demonstrates how to do this using a high-resolution tracking system.

Flies provide an important model for studying complex behavior due to the plethora of genetic tools available to researchers in this field.  Studying ...

Biology

C. elegans Tracking and Behavioral Measurement

We have developed instrumentation, image processing, and data analysis techniques to quantify the locomotory behavior of C. elegans as it crawls on the surface of an agar plate. For the study of the genetic, biochemical, and neuronal basis of behavior, C. elegans is an ideal organism because it is genetically tractable, amenable to microscopy, and shows a number of complex behaviors, including taxis, learning, and social interaction1,2. Behavioral analysis based on tracking the movements of worms as they crawl on agar plates have been particularly useful in the study of sensory behavior3, locomotion4, and general mutational phenotyping5. Our system works by moving the camera and illumination system as the worms crawls on a stationary agar plate, which ensures no mechanical stimulus is transmitted to the worm. Our tracking system is easy to use and includes a semi-automatic calibration feature. A challenge of all video tracking systems is that it generates an enormous amount of data that is intrinsically high dimensional. Our image processing and data analysis programs deal with this challenge by reducing the worms shape into a set of independent components, which comprehensively reconstruct the worms behavior as a function of only 3-4 dimensions6,7. As an example of the process we show that the worm enters and exits its reversal state in a phase specific manner.

C. elegans Tracking and Behavioral Measurement

We have developed a video-rate tracking microscope system that can record and quantify C. elegans behavior at high resolution and high speeds. We have also developed computational methods to reduce the dimensionality of the worm images to a fundamental set of measurements that completely describe the shape of the worm.

We have developed instrumentation, image processing, and data analysis techniques to quantify the locomotory behavior of C. elegans as it crawls on the ...

Extracellular Wire Tetrode Recording in Brain of Freely Walking Insects

Increasing interest in the role of brain activity in insect motor control requires that we be able to monitor neural activity while insects perform natural behavior. We previously developed a technique for implanting tetrode wires into the central complex of cockroach brains that allowed us to record activity from multiple neurons simultaneously while a tethered cockroach turned or altered walking speed. While a major advance, tethered preparations provide access to limited behaviors and often lack feedback processes that occur in freely moving animals. We now present a modified version of that technique that allows us to record from the central complex of freely moving cockroaches as they walk in an arena and deal with barriers by turning, climbing or tunneling. Coupled with high speed video and cluster cutting, we can now relate brain activity to various parameters of the movement of freely behaving insects.

We previously developed a technique for implanting tetrode wires into the central complex of cockroach brains that allows us to monitor activity in individual units of tethered cockroaches. Here we present a modified version of that technique that allows us to also record brain activity in freely moving insects.

Increasing interest in the role of brain activity in insect motor control requires that we be able to monitor neural activity while insects perform ...

Radio Frequency Identification and Motion-sensitive Video Efficiently Automate Recording of Unrewarded Choice Behavior by Bumblebees

We present two methods for observing bumblebee choice behavior in an enclosed testing space. The first method consists of Radio Frequency Identification (RFID) readers built into artificial flowers that display various visual cues, and RFID tags (i.e., passive transponders) glued to the thorax of bumblebee workers. The novelty in our implementation is that RFID readers are built directly into artificial flowers that are capable of displaying several distinct visual properties such as color, pattern type, spatial frequency (i.e., &#8220;busyness&#8221; of the pattern), and symmetry (spatial frequency and symmetry were not manipulated in this experiment). Additionally, these visual displays in conjunction with the automated systems are capable of recording unrewarded and untrained choice behavior. The second method consists of recording choice behavior at artificial flowers using motion-sensitive high-definition camcorders. Bumblebees have number tags glued to their thoraces for unique identification. The advantage in this implementation over RFID is that in addition to observing landing behavior, alternate measures of preference such as hovering and antennation may also be observed. Both automation methods increase experimental control, and internal validity by allowing larger scale studies that take into account individual differences. External validity is also improved because bees can freely enter and exit the testing environment without constraints such as the availability of a research assistant on-site. Compared to human observation in real time, the automated methods are more cost-effective and possibly less error-prone.

This video describes Radio-Frequency Identification (RFID) and motion-sensitive video recording methods to monitor choice behavior by bumblebees.

We present two methods for observing bumblebee choice behavior in an enclosed testing space. The first method consists of Radio Frequency Identification ...

Insect-machine Hybrid System: Remote Radio Control of a Freely Flying Beetle (Mercynorrhina torquata)

The rise of radio-enabled digital electronic devices has prompted the use of small wireless neuromuscular recorders and stimulators for studying in-flight insect behavior. This technology enables the development of an insect-machine hybrid system using a living insect platform described in this protocol. Moreover, this protocol presents the system configuration and free flight experimental procedures for evaluating the function of the flight muscles in an untethered insect. For demonstration, we targeted the third axillary sclerite (3Ax) muscle to control and achieve left or right turning of a flying beetle. A thin silver wire electrode was implanted on the 3Ax muscle on each side of the beetle. These were connected to the outputs of a wireless backpack (i.e., a neuromuscular electrical stimulator) mounted on the pronotum of the beetle. The muscle was stimulated in free flight by alternating the stimulation side (left or right) or varying the stimulation frequency. The beetle turned to the ipsilateral side when the muscle was stimulated and exhibited a graded response to an increasing frequency. The implantation process and volume calibration of the 3 dimensional motion capture camera system need to be carried out with care to avoid damaging the muscle and losing track of the marker, respectively. This method is highly beneficial to study insect flight, as it helps to reveal the functions of the flight muscle of interest in free flight.

Insect-machine Hybrid System: Remote Radio Control of a Freely Flying Beetle Mercynorrhina torquata

This protocol describes the process of constructing an insect-machine hybrid system and carrying out wireless electrical stimulation of the flight muscles required to control the turning motion of a flying insect.

The rise of radio-enabled digital electronic devices has prompted the use of small wireless neuromuscular recorders and stimulators for studying in-flight ...

Insect-controlled Robot: A Mobile Robot Platform to Evaluate the Odor-tracking Capability of an Insect

Robotic odor source localization has been a challenging area and one to which biological knowledge has been expected to contribute, as finding odor sources is an essential task for organism survival. Insects are well-studied organisms with regard to odor tracking, and their behavioral strategies have been applied to mobile robots for evaluation. This "bottom-up" approach is a fundamental way to develop biomimetic robots; however, the biological analyses and the modeling of behavioral mechanisms are still ongoing. Therefore, it is still unknown how such a biological system actually works as the controller of a robotic platform. To answer this question, we have developed an insect-controlled robot in which a male adult silkmoth (Bombyx mori) drives a robot car in response to odor stimuli; this can be regarded as a prototype of a future insect-mimetic robot. In the cockpit of the robot, a tethered silkmoth walked on an air-supported ball and an optical sensor measured the ball rotations. These rotations were translated into the movement of the two-wheeled robot. The advantage of this "hybrid" approach is that experimenters can manipulate any parameter of the robot, which enables the evaluation of the odor-tracking capability of insects and provides useful suggestions for robotic odor-tracking. Furthermore, these manipulations are non-invasive ways to alter the sensory-motor relationship of a pilot insect and will be a useful technique for understanding adaptive behaviors.

The capability to localize an odor source is necessary for insect survival and is expected to be applicable to artificial odor-tracking. The insect-controlled robot is driven by an actual silkmoth and enables us to evaluate the odor-tracking capability of insects through a robotic platform.

Robotic odor source localization has been a challenging area and one to which biological knowledge has been expected to contribute, as finding odor ...

SwarmSight: Real-time Tracking of Insect Antenna Movements and Proboscis Extension Reflex Using a Common Preparation and Conventional Hardware

Many scientifically and agriculturally important insects use antennae to detect the presence of volatile chemical compounds and extend their proboscis during feeding. The ability to rapidly obtain high-resolution measurements of natural antenna and proboscis movements and assess how they change in response to chemical, developmental, and genetic manipulations can aid the understanding of insect behavior. By extending our previous work on assessing aggregate insect swarm or animal group movements from natural and laboratory videos using the video analysis software SwarmSight, we developed a novel, free, and open-source software module, SwarmSight Appendage Tracking (SwarmSight.org) for frame-by-frame tracking of insect antenna and proboscis positions from conventional web camera videos using conventional computers. The software processes frames about 120 times faster than humans, performs at better than human accuracy, and, using 30 frames per second (fps) videos, can capture antennal dynamics up to 15 Hz. The software was used to track the antennal response of honey bees to two odors and found significant mean antennal retractions away from the odor source about 1 s after odor presentation. We observed antenna position density heat map cluster formation and cluster and mean angle dependence on odor concentration.

This protocol describes steps for using the novel software, SwarmSight, for frame-by-frame tracking of insect antenna and proboscis positions from conventional web camera videos using conventional computers. The free, open-source software processes frames about 120 times faster than humans and performs at better than human accuracy.

Many scientifically and agriculturally important insects use antennae to detect the presence of volatile chemical compounds and extend their proboscis ...

C. elegans Movement Tracking: A Method to Assess Locomotion in Worms

Source: Davies, A. G. et al. <a href="https://www.jove.com/video/52681/an-assay-for-measuring-effects-ethanol-on-locomotion-speed">An Assay for Measuring the Effects of Ethanol on the Locomotion Speed of&#160;Caenorhabditis elegans.</a>&#160;J. Vis. Exp.&#160;(2015).
This video describes the use of video and tracking software to record worm movement in an arena.&#160; The example protocol measures the effect of ethanol exposure on worm locomotion.

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

Samenvatting

Meer video's verkennen

Hoofdstukken in deze video