The Xenon project seems to have reached a point where I can start getting excited again. I had been mucking around with the images for weeks, trying to register them so the strong lines caused by mismatching between the bone and skin structure in the images would not affect the FFT— we only want to observe changes in the FFT of the brain tissue. Today, I convinced myself that is impossible— the registration of the raw data set seems to be as good as it’s going to get, because moving an image by even as little as one pixel in any direction causes the number and width of the extraneous lines to increase.

Papadakis suggested that I use an image processing program to remove the bones and skin, and smooth out the remaining edges of the brain tissue, to get rid of the extraneous high frequency content. I started to do that reluctantly, because to do so, I would have to convert the DICOM 0-65536 data to 0-255, and lose information in the process— but then I realized something obvious: I can use GIMP to get a selection mask for the brain tissue in one frame, save that mask as a grayscale image, load it back into MATLAB, and use the mask on all the raw DICOM data in Matlab.

By the time I realized that, my time in lab was up, but I’m definitely going to try that on Monday.

Possibly relevant posts:

• Gimp and Scheme in One Defun (7/19/2005)
• Volume Visualization (12/27/2005)
• Project: Automated Curve-Fitting for Font Design (5/26/2003)