What is Live Mouse Tracker ?
LMT is designed to track mice on the long term. It has been created with on idea on mind: to provide behaviourist a reliable system to annotate automatically what mice are doing in their environment, and what are their social interactions. It is fully described in the publication available at bioRxiv. Now published in Nature biomedical engineering.
What behaviours can it extract automatically ?
Here is the list of basic behaviours the system can extract (without any manual intervention). Note that you can add your own (see data analysis section):
And then those labels can be watch directly in the integrated player of Live mouse tracker (full validation video available in validation section):
Is there somebody using it ?
LMT is mostly used in Europe, Japan, South Korea and U.S.
32 34 teams do have at least one LMT working. 46 52 systems are built (that we know about). ( Sep 2020) (Oct 2020)
The last version of Live Mouse Tracker can be downloaded here.
Shopping list and assembly instructions
Live mouse tracker is a DIY device ! Download the full PDF in IKEA style with shopping list and assembly instructions v3 here.
How does it work ?
This short video summarize how the system works:
For each experiment, LMT produces a set of data:
- A database file, in the SQLite format, which is handy as SQLite database are standalone files, which avoid using server. You can just copy/backup/restore/share your databases as easy as manipulating conventional files.
- A set of video, in the MP4 format, which store the view in LMT. The purpose of those video is to get back to the film if you find specific behaviour while digging in the data. To find back a specific event, we hard coded in the video the current frame and current date/time. MP4 files are split in 10 minutes segments so that you can play MP4 even during the experiment.
- A background view folder, which store the depth image of the cage, without the animals. This is great to study how mice are moving objects in the cage, how they organize their nest or how they move/eat their food.
To perform the analysis, we provide a framework in Python detailed in this presentation. We also provide code on GitHub. Examples of data analysis here that contains trajectory drawing, extraction of position, extracting timeline of events. You can also rebuild all the events available using this script. An example of interactive event exploration is also available here.
You can enrich/change the analysis with your own scripts.
To validate the system (for both the tracking quality and labelling), we annotated manually the following sequence, and we then compared the results to the automatic labelling.
Can I use Live Mouse Tracker to study ultrasonic vocalizations ?
Yes, please watch this video. (If you have no time, just watch the parts at 6:32 to analyse USVs and at 20:00 to watch the link between LMT and USVs). LMT associated with AviSoft Recorder allow you to record synchronized USVs with behavioral events. We also provide a website to test our USV detection method online, no installation needed: https://usv.pasteur.cloud
Please contact Elodie or me to get instructions to record and analyze USVs, and -if needed- to synchronize them with LMT.
The purpose of this website is to share the data acquired with LMT. Once registered, you can post link to your database, so that others can access and analyse them.
Frequently asked question
How much does it cost ?
The whole system cost less than 2000€. This includes the computer, the RFID antenna/readers, the Kinect v2, the Kinect adapter, cage and tubes. This is a DIY system, assembly instructions are available here.
Can I track elephants, monkeys, birds, ants, spiders, flies, mosquito, people, sheeps, cows, mantas, crocodiles or frogs with Live Mouse Tracker ?
I am sorry no. I would very much encourage you to find a dedicated solution for your species, there are trackers for all of them, with pros and cons.
If I want to track only 1 animal, do I need RFID ?
In that case you don’t need RFID. Just a Kinect.
On which OS does it work ?
Can I use another computer than the one you recommend ?
We recommend a Ryzen 7 2700X, 16GB RAM. All our systems are working with this kind of computer. It’s up to you to use something else. If we plan to add additional computation to provide better results, we assume that you are using this kind of computer.
Do I need a good GPU / graphic card ?
LMT does not use GPU, you can use any card.
How do I create the temperature/humidity/sound/light sensor ?
Download the kit here.
How do I create an event from another program or device using an UDP network protocol ?
If you have an external device that is creating event, and you want to synchronize them with LMT, you can use several solution. The most simple is the UDP protocol: you just send start and end of event on the fly, and they will be recorded as event in LMT.
How can I tune the antenna ?
Launch the plugin AntennaTuner in the Live Mouse Tracker tab.
Then in the output of Icy you will see the current tuning refreshed each second.
- It automatically detects all the COM port as you plug/unplug them.
- Make the measurement when you are not touching the coil and put the coil far from metallic masses or wires.
- To increase the value up to 134.2kHz, remove some coil. If you removed more than 10cm of coil, you can cut it and re-solder (note that this will increase a bit the tuning).
- If you see 0.0 / 134.2kHz it means the wire is broken, or you have a problem with the soldering.
- To solder the coil, turn it around your wire, and burn (as you solder) the thin translucid isolation plastic that is around the coil.
- You can also test the detection of your glass RFID probe with the Antenna Tuning Tester plugin. It will cycle through antenna for 5 seconds each, and perform 10 reading per second for each cycle. It will display the RFID number detected in the output.
How much the tuning of the antenna will affect its reading range ?
Robert Warren, from Helmholtz Zentrum München answered this question with this graph (you will use 4.2mm of range per kHz difference with 134.2kHz) (This graph was with previous RFID tags from BioLogID)
How do I insert the RFID probe ?
The RFID should be gently put in the animal using, using gas anesthesia, and local subcataneous analgesia. The RFID should be inserted in the neck and gently pushed on the side of the animal.
Where and which RFID probe should I order ?
We asked RFID suppliers to provide samples for test. The current best supplier is Biomark: www.biomark.com . According to our tests, their RFIDs are 1.54x better than the one we recommended while publishing the paper !
The RFID reference is RFID APT12 PIT tag (FDX)
You may also ask for an injector
You can mention that you are an LMT user (we don’t have financial agreements !).
Can I get back the RFID chip to use them later on ?
You should definitely not. RFID chip are provided with unique number bought by the company which sells it to you. If you do re use number, you will compute databases with several times the same animal, and this will wreck the analysis. Avoid reprogramming the chip with your own number, those number will jam others if you share the database on this website.
How should I setup the 63cm height of the camera ?
You should not try to measure with a ruler, but directly from the application, with the program called “LiveMouseTrackerCalibration”. It provides a view of the field with red, orange and green dots. When all are green, you are calibrated. You maximize the frame and zoom-in/out with the mousewheel.
Can I use several setup in the same room ?
More and more labs are using several setup in the same room (typically 4 setups) to run parallel experiments. To do this, you should consider the following points:
- Kinects should not send direct light to other setup. If it happends, you will see that the image is slightly blinking. To avoid this, you should use kind of matte box (homemade is fine!) attached to the front of the kinect. Using isolating boxes per setup is also a solution.
- If several RFID are activated, they do jam the signal. On one system, LMT ensures that only one antenna is started at a time. If you start multiple setup in the same room, several antenna will be active at the same time. So you should keep a minimal distance of 1 meter between setup. If a setup is not used at all, disconnect the power of RFID reader (disconnect the hub power) since the default behavior of readers when started is to start reading. In that case they will be all active and will jam the signal up to several meters and other system will not read any tag (this one was tricky to find out !).
RFID probe are not detected at all, or very badly.
- Check if the USB are correctly alimented. If your USB Hub needs a separated power source, be sure to have it connected.
- Use A+/A- connector on the RFID reader board to connect antenna instead of AR/A-.
RFID probe were correctly detected 1 month ago, and not anymore.
- Check if the power of your USB hub is still working. Try another hub to check this.
The real time and the time recorded is not the same (and I see red flashes at the bottom of the recording window)
- You may have launched the program with icy.exe. You must start it with “live mouse tracker for java 8.bat”, which will launch Icy for you with a specific configuration of the memory dedicated to real time applications. You should see no more than 1 second of time drift per 3 hours or recording.
Tracks are cut for one frame unexpectedly or images hangs for a few seconds to up to 20 seconds.
This is Kinect related issue. Several solutions are listed here:
- Get to Control Panel -> Hardware and Sound -> Sound -> Recording -> Kinect Microphone and enable it.
- Open the Microsoft Store Camera app and under Setting -> Change privacy settings -> Microphone -> Allow apps to access your microphone -> enable the microphone.
- Some RealTech sound driver seems to create problems (they can consume a lot of memory (more than 1GB) and CPU (1 Full proc). Uninstall those drivers.
- The Kinect may be connected on a USB3 that is not fast enough. Connect the Kinect to a USB3 that is on the back of the computer (avoid front panel).
doc by Fabrice de Chaumont