Author Topic: Naked mole rat training  (Read 18558 times)

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Naked mole rat training
« on: June 03, 2012, 01:42:40 PM »
Why am I here?

I'm a neuroscience grad student working with naked mole rats. We currently use RFID readers to keep track of individual animals as they run through tunnels and we're working on adding optical tracking to correlate social behaviors with location and identify animals in the colony hierarchy. Great stuff, but a pain in the neck to implement. More, it's expensive and an even greater pain to reconfigure a system once it's in place.
I've been writing Arduino sketches for a bit and we were intending to move to Arduino (or Lego Mindstorms) to operate feed hoppers based on RFID tags. Lots of other stuff - we're in the first week of a five year grant.

Where does Makey Makey fit in?

As a reconfigurable and low-cost sensor system that we can deploy both as proof-of-concept for certain experiments, and possibly to entirely substitute for the proposed finicky optical tracking. Naked mole rats are nearly blind and rely on their somatosensory system to navigate and for recognition. They investigate by touch and the rooms are unlit except when we dull scientists need to see.

Core project proposal:
A metal plate on the enclosure floor is wired to ground. Target objects are wired as keys - I won't use cursor keys, probably WASD to give props to my gaming roots. Any simple keylogger will give me the time and order that the "keys" were touched by the animals. Or just plain Notepad will record the order if timing is not an issue. We don't need temporal resolution finer than tenths of a second anyhow. Probably not even that. And the optical systems are not good at spotting fine manipulation. Heh. Right now they're not very good at spotting anything FWIW. We have some poor undergrads watching the videos for behaviors. I just showed one person how to watch videos at 2X speed in VLC, saving him hours...

For now, we want to just test it - see if we can get good input from an animal manipulating targets while standing on their grounded mat. After that, we intend to begin simple cooperative behavioral testing - two animals partitioned need to cooperate to release a treat for each. Finally, we're proposing operant conditioning, a chime played when a target is manipulated. This has not been done yet with naked mole rats and they're a fascinating animal to work with.

None of this yet interfaces with the RFID system but I've got five years to crack that and my lab director is willing to give Makey Makey and the Arduino family a good try.

What are the benefits to humanity?

You guys will get output and code from the very first test which we'll have in before the end of June, seeing if our little guys can properly input using your system. You can think of it as a pilot application to inspire other Makers in devising feeding systems for Fifi. And frankly, video of a bustling naked mole rat colony grabs eyeballs better than more Lolcats.

More, our undergrads work very hard at a boring job. If we can automate part of our behavioral work and also save grant money, you'll raise our research assistant's grades and allow us to divert some money to getting a coffee maker in the lab. And I'm definitely putting you guys in the acknowledgement of any related papers.

Lastly, the naked mole rat is an incredible animal. It lives in pitch-black and toxic conditions and navigates by touch. A better understanding of their adaptations is just good and responsible research.

Below is a link to some video of one of our colonies. We have roughly one hundred animals in three colonies. Nearly all of them are bionic rats with implanted RFID chips.
Untitled on Vimeo



SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #1 on: June 04, 2012, 05:54:59 PM »
 :-\ Video preview doesn't show up but it's there if you click through the link. FWIW I have experience coding as does my boss. And I've used Arduino before, so turnaround on this project will be very quick once we start.

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #2 on: June 07, 2012, 08:20:46 PM »
 :DWow, thanks for picking our project! This is going to be awesome!

Thomas

  • MaKey MaKer
  • Posts: 3
    • View Profile
Re: Naked mole rat training
« Reply #3 on: June 12, 2012, 01:39:30 AM »
Congrats on getting the grant!
I love the sound of your cooperation experiment- we are thinking along similar lines with mice.
I run a Animal Behaviour Facility in Sydney and we have some cutting edge and expensive equipment. This is great but ultimately it is very inflexible and sometimes not all that well designed (sometimes I wonder if behaviourists were even involved!). Ideally, I want to be cobbling my own automated equipment together and it seems like both of our projects will be in the same vein. Once I get started I will put some more info up. I would love to know of your progress and perhaps we can give each other tips as we go along.
All the best
Tom

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #4 on: June 12, 2012, 06:57:40 PM »
Thanks for the encouragement Thomas. After fiddling with a malfunctioning RFID reader for the umpteenth time, and getting frustrated that the optical couldn't tell an animal's nose from its rump, I just wished for a simpler way to set up at least the smaller tests. I'm champing at the bit and watching the mail for the first MaKey to arrive. I'm already filing IACUC paperwork for it tomorrow in anticipation.


SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #5 on: June 12, 2012, 07:31:36 PM »
And regarding designed by a behaviorist: The hall from our lab to the animal care facility is long and winding with very poor signage. Walking with someone once, he remarked that it's almost like I'm taking him through a maze. I stopped and shouted "Ok, guys! Forget it, he figured it out!" while the new guy looked around for cameras.

lightnin

  • Administrator
  • MaKey MaKer
  • Posts: 84
    • View Profile
Re: Naked mole rat training
« Reply #6 on: June 13, 2012, 09:09:26 PM »
And regarding designed by a behaviorist: The hall from our lab to the animal care facility is long and winding with very poor signage. Walking with someone once, he remarked that it's almost like I'm taking him through a maze. I stopped and shouted "Ok, guys! Forget it, he figured it out!" while the new guy looked around for cameras.

:) Awesome. I love stories like that.

Thomas

  • MaKey MaKer
  • Posts: 3
    • View Profile
Re: Naked mole rat training
« Reply #7 on: June 15, 2012, 02:14:26 AM »
Excellent! Great stuff. Keep us posted  :D

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #8 on: June 15, 2012, 11:32:01 AM »
Okay, so far I've sent in an addendum to our testing protocols to include the line "RFID, EEG, and optical instrumentation will be augmented by reconfigurable tactile sensors fabricated on the Arduino microcontroller platform."
It's deliberately vague since I don't quite know the final positioning of the sensors and such - floor? wall? An enrichment object? And it references the Arduino because review boards are more comfortable about an established technology. I'm not really worried about safety for the little guys but I do want to check just to be 100% sure. Also I really do hope to tie the Makey functionality in with some Arduino hacks to interface with existing systems.
But first of course I need to see to test the core functions. Do naked mole rats register on it? How do they respond to the new gadget in their colony? Those teeth are not just for show - they can chew through nearly anything and enjoy doing so. They investigate the world through their incisors as a sensory organ. So I'll need to find something they don't chew that can work with the Makey. And if they don't register, or register poorly I'll need to adjust the code behind this. Idk really, I need to see it first.
I'm still waiting for the first kit to arrive, I'm meeting a Netduino friend Saturday night and really want some test runs. There will be beer and cigars hopefully, so that should stimulate some creativity.
A prayer to the Postmaster General that his Mercury-heeled messengers bear my gift today!

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #9 on: June 21, 2012, 09:20:44 AM »
That didn't happen. There have been delays, but we have a ship date! The kits went out on Tuesday so I hope to see mine today or tomorrow.
Paperwork for using it with animals was filed yesterday. The thing is absurdly safe so I don't expect any snags in the approval process but still paperwork is paperwork and it does need to shuffle along at its own pace. I can test it outside the habitats until then.

Pity I didn't have it for the three vices shmooze on Saturday night. Met with friends and talked cigars, beer, and hardware hacking. Our host showed off his sprinkler system, run by a Netduino and controlled from his Android phone. Very nice setup. It can even check the weather report and decide whether to water the lawn.

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #10 on: June 26, 2012, 01:11:09 PM »
Just picked up my MaKey from the Post Office. Haven't hooked it up yet, will update soon.

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #11 on: July 02, 2012, 01:10:22 PM »
Warning: Casting area of effect Wall of Text now. tl;dr version: Played with it, cool and easy to use out of the box. At this moment I'm designing a means of testing it with the animals.

Full version:
Okay, an update. I'm in the lab right now, just demoed the MaKey for my PI. He's excited and we're trying to figure out logistics and deployment.

Backing up a bit: I took this thing out for a spin Saturday night. It's amazing how transparent it can be in its core function. I hooked it up so my porch railing was a down arrow and plugged the other end into a borrowed laptop. Officially I was supposed to be working, but... Saturday night?
Anyhow, half an hour later finds me browsing thru Erfworld comics when I discover I've been tapping the railing to scroll down instead of using the real keyboard. That's what I mean by transparent - once it was hooked up, I interacted with the railing and forgot about the hardware.

To use this in a solid scientific manner I need to tackle a few topics, devise a proof-of-concept, then test its limits and build an experiment that leverages the MaKey.
The topics are:
  • Logistics - how do I put this system together with targets into place in an animal enclosure? Lots of wires to run and naked mole rats are sausages with teeth - very big teeth. They like to chew stuff as much as I like to read comics. Where do put the ground? Will their regular enrichment toys work? More likely I'll need to fab some plates and embed them in toys. If that happens, the plates might become targets for teeth. Other stuff, having to do with where to place hardware in the room and minimize drilling and construction on their habitats.
  • Feasibility/Usability - Does it work? Can it detect touch from an NMR? How easily can I modify scripts or adapt hardware? What was that other thing I forgot to list?
  • Limits - Temporal resolution - does it get the keys in correct order every time? Is there a race condition where one input may be lost or flipped in front of an earlier input? Key rollover - I know it can handle four at once, but how does it show up? Time stamping - I intend to use a keylogger with time stamps but is there a better solution? One of those "catch your cheating spouse" programs seems like a kludge. Anyone have a better idea for how to capture and timestamp inputs? I'd prefer not to go to Linux if possible.
We already blue-skied some behavioral experiments that a MaKey can help with but I first need to resolve the issues in that list.
Last question for those of you out there: Help a decaffeinated fellow out? Was there any topic I missed? Or do you have a solution to propose? Love to hear from you.

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #12 on: July 02, 2012, 05:21:54 PM »
Wow, okay. Just hit an astonishing snag and I'm hoping it's because I'm tired.
Apparently naked mole rats have an incredibly high electrical resistance. Seriously. Beside making for the best context-free quote of the day ("Hey! Did you know that naked mole rats have an incredibly high electrical resistance?") this also poses a stumper. See, humans have high resistance on our hairy skin but not glabrous skin, so grabbing a multimeter probe in each hand gave me a high but not unmanageable resistance. IOW, MaKey works for peeps.

But I probed a naked mole rat all over and couldn't find conduction. Didn't try the mouth, but just about everywhere else that's reasonable. I had a little guy stand on aluminum foil and tapped him with MaKey clips in different places. Got an occasional response from the top of a paw. Intermittent, not good enough. Nowhere else worked at all! Tried with two different multimeters, too. Nothing.

Here are my options: Raise the sensitivity, either in hardware or software. Problems: I don't know the hardware well enough to do it without guidance. And I don't have the software (original sketch) at all. Potential problem: Boosting the sensitivity might generate noise. Stray air currents, humidity - all might be enough to set it off. I don't want spurious results, obviously. So what now?

Another option: Flip the outputs so it prints a character when the circuit is interrupted instead of closed. Lots of problems with this, but I want to head home so enough for now.

Last option: Test the mouth and hope that the MaKey can sense when an animal stands on a grouding mat and bites into a sweet potato clipped to the MaKey. Lots of problems with this too, not least of which is that hopefulness is not the same as reality. Also, how much food can I stand there and clip to a MaKey? Especially since these guys don't eat when I'm looming over them. And anyhoo, using only dentition (aka via food) severely limits my experiment options.


So just a final closing line: HELP!!!! :'(

tigertrainer

  • MaKey MaKer
  • Posts: 11
    • View Profile
Re: Naked mole rat training
« Reply #13 on: July 03, 2012, 01:00:52 AM »
Hi, I'm the one trying to make a DAS system with the Makey Makey concept: http://www.makeymakey.com/forums/index.php?topic=433.0

In terms of software, I think you should give Max/MSP or PD (Pure Data) a try. Both work well with the Arduino platform and both are high level languages so programming is not too difficult (You can also program low-level if you wish!). A big advantage with both these languages is that they are interpretive so you can change stuff around without having to compile first, and both mix the interface with the code. Max/MSP is a bit expensive, but there are university prices and it could be worth it. You can download a 30 day free trial version. Pure Data is similar to Max/MSP and it is free. The big advantage with Max/MSP is the interface which is more productive and more beautiful.

I use Max/MSP and it's very simple. You just load the Arduino with a Standard Firmata so that it can communicate directly with Max/MSP. Now you don't need to open the Arduino software again, just open Max/MSP and you'll be able to grap your sensors in there. You can use the Maxuino patch that was made for this. You'll be able to use all kinds of graphic interfaces to show what is happening in with your sensors.

For timing, Max/MSP uses a scheduler that is set with a top speed of 1ms. That means you get a resolution of 1000 events per second. You can then filter these events using quantization (a sort of time grid). For example, if you only need the precision of 4 events per second, then you divide your grid in 250 ms. So, anything happening from 0-250 ms is logged as event 1, anything from 250-500ms as event 2, etc...

If the rats do not conduct, then you probably won't be able to use Makey Makey. Although, since you have a lot of time, I would suggest you try many setups with various resistances. The easiest way would be to build your own little circuit with an arduino and replace the resistors with a potentiometer. This way it will be easy to calibrate the resistance. Obviously, you don't want to give them a shock they will notice as this will completely destroy your experiment!!!! But if you try many resistances, you might be able to find the sweet spot.

Another idea would be to have mechanical triggers set for the various events you want to log. For example, if you want to log a rat chewing on an apple, you could have the apple setup in a way that when the rat chews on it the apple gets moved and hits a trigger to note the event. Let's say the apple is attached to a stick which has a sort of spring at the base. When the rat eats on the apple, the stick moves a bit, then a connection is made at the base of the stick. In this way, you don't need the rats to be carriers at all.

What are the events you are trying to monitor exactly? Can you make a list.... Or better yet, can you upload a drawing of your rat maze (or photo) and show us how the stations are designed and so on... That way, I could give you more ideas.

I think you'll definitely have to build your own hardware. And, really, this should be your goal since you have 5 years of time and you are at a university. Get someone from the engineering department to help!
« Last Edit: July 03, 2012, 01:02:40 AM by tigertrainer »

SJester

  • MaKey MaKer
  • Posts: 36
    • View Profile
Re: Naked mole rat training
« Reply #14 on: July 03, 2012, 10:52:05 PM »
Thanks. I'll look at Max/MSP and PD. I've already considered using other sensors instead of touch and treating the MaKey as a pure Arduino. But it's not a good solution for a number of reasons, not least of which is that I really do need a touch sensor that works on contact alone, not pressure or tripping a beam. I've worked with Arduinos and fabbed equipment before so I don't think I'll have a problem expanding the MaKey's capability, but I have to get that core function working first. I think I'm stuck until I get access to the loaded sketch and see how the MaKey actually works.

Still, assuming it's similar to the Drawdio (same developers) and similar projects, I'm guessing it works through measuring R/C decay. But that ground wire throws me off...

So a possible hardware target would be the pull-up resistors. Or the software low pass filter, if I have access to the code.

As an aside, I don't have five years. The grant is for five years and like many large science grants there's primary goals and then side projects. This is a side project with the hopeful outcome of augmenting our sensor suite for behavioral studies. In terms of time, I have about a month to really work at this. After that, I don't know.

Any thoughts on ? I'd love to hear from Eric and Jay. The MaKey is their baby, and I'm just a biologist.