Need help building a Near Infrared Spectroscopy Device

[stanford]

Hi,
I'm a computer scientist w/ very little hardware experience. I've been playing with the idea of using NIRS for a brain computer interface.
The basic principle is pretty simple: You have a laser diode and a detector that is 2-3 cm away from the diode. Unfortunately I have almost no experience with any of these electronic components so was wondering if I could get some guidance.

I've found the lasers here:
http://search.digikey.com/us/en/products/D8505I/38-1030-ND/287034

Now I need a detector which will basically measure how much light is bounced back out of the skull. I would like to sample this at somewhere around 100-500 Hz. It is important that there be as little noise and as much precision as possible. I was looking at using a "photodetector" (is that the right part?)
http://search.digikey.com/us/en/products/PNZ335/PNZ335-ND/274305

The photodetector would be wired to a fiberoptic cable that would be placed on the scalp a few cm away from the emitter. There would probably be 4-6 of these detector/emitter pairs.

A few key questions:
1) How do I couple the photodetector to the fiberoptic cable.
2) What sort of hardware would you recommend for connecting the photo detector to a computer and reading it in at ~100-500hz
3) Would it be possible to modulate the laser diode some how (i.e make it change power at a certain frequency).
4) Any general advice on how to go about building this or problems you think I might face.
5) Anyone interested in helping me build this/design this? I am a graduate student in Computer Science at Stanford and have a fair bit of machine learning code already written for analyzing the data (have already used the code to explore data from a NIRS machine we have on campus -- but that one only has a sample rate of 10hz) ... this project is mostly for fun and would love to collaborate with people interested in this stuff.

 

[Mike_In_Plano]

I'm not quite sure of the mechanism by which the brain would interact with the light. As far as physical density, etc, isn't it just a lump?

In any cases, I can give you some tidbits related to sending light out, getting it back and reporting how much made the trip.

Firstly, you need a good linear detector with a wide dynamic range, so photo transistors are out. What you need instead is a reverse biased photodiode. It will give you an output current proportional to the incident light.

You'll need a photodiode amplifier. There are a bunch of app notes on these at linear.com. These are also called transimpedance amplifiers. It is a simple circuit built around an op amp.

If you're not overly picky about the focus of the beam, there are LEDs that work as good as a laser and are cheaper. It's helpful to vary the drive current to your light source so that you can get a wider dynamic range.

Ambient light is a real issue. Filters can help some, and you can get photo diodes with a dark filter. However, these will still sense some ambient light - particularly if it's from incandescent or halogen lamps.

By quickly toggling the light source on and off, you can get a baseline reading and subtract some of the influence of ambient light. However, you really need to perform this trick quickly (on the order of kHz) to ditch the lesser effects of fluorescent lighting.

Narrowband color filters can help, and glass ones are not as difficult to cut as you might think. Use a glass scribing tool and scribe a line adjacent to an edge. Tap it from behind to create a break along the line, thus forming a rectangle. Scribe and tap little squares off of this rectangle. Using about 600 grit, wet paper, sand the edges off of each to form the approximations of little circles. (Be sure to wear gloves). Get some black PVC or Delrin collets made that will snuggly fit you LED / photo diode on one end and have a larger opening for the filter on the other. Drop your filter in, and retain it with a rubber o-ring.
You can do this with collimating lenses as well.

 

[stanford]

Thanks for the detailed response. I'm wondering if photodiodes will have the sensitivity necessary -- generally avalanche photodiodes or photo multiplier tubes are used although I've heard diodes have worked too. If I use a diode which would be best in terms of sensitivity to small amounts of light (I would shield the diode and the source light would be delivered via a fiberoptic cable).