Update: Intrinsic Imaging
I recently posted everything you need to know about building an intrinsic imaging rig. Check out that post here. In the post, I included some of the very first results from the rig. Since then, I've improved the images a bit. This actually very little to do with the rig itself - it’s great! These changes were essentially tweaks to the experimental process and image analysis.
During imaging across many trials, it’s incredibly important for the reflectance to not change over time. You’re attempting to pick up a change of about 0.1% in the optical signal! If you use saline over the skull or cranial window, evaporation over time can swamp out your signal. I’ve gotten best results by filling the reservoir over the skull with mineral oil, and also making the entire fluidic surface planar by capping it with a coverslip.
As far as the analysis, here is algorithm I use for trials that consist of 4 seconds of baseline and 4 seconds of whisker simulation at a frame rate of 31Hz:
Convert 16-bit images to double and average all baseline and stimulation frames
Subtract the average baseline frame from the average stimulation frame
In this subtracted image, subtract the median to account for baseline fluctuations due to isoflurane
Gaussian filter the frame
Rescale the matrix (a double) to the bit depth of the image. Ex: an 8-bit image can have pixel values in the range 0-255. For me, I rescale my matrices to a 16-bit image which have pixel values in the range 0-65535.
This is NOT the same thing as converting a double to a uint16. If you do a direct conversion, any matrix values less than zero (due to previous image processing) will be given a pixel value of zero. Essentially, by doing a direct conversion, you lose all benefit of having a high bit depth image.
Do all of the above image processing for each trial, then average across trials to get the final optical signal
Here are the images I’m getting today with only 10-15 trials. I can begin to see a semblance of the barrels with as few as 5 trials. These are the C2 and D2 barrels in mouse S1.
That’s all y’all, happy imaging.
-Daniel
