Dear brainweb manager,
Hello, I am a graduate student at a Vanderbilt university, Electrical Engineering.
I am researching about comparison between two segmentation algorithms which are SPM from MATLAB and FSL. For this, I’d like to use your dataset, so I have tried to use them.
I set the gold-standard groundtruth with the Grey Matter from 20 anatomical models of 20 normal brains.
And I set the data for implementing each algorithm with ‘Normal Brain Database’. And I have tried to get the segmentation results from this whole-brain image to compare with the GM image from previous model.
The problem is that the sizes of the groundtruth and whole-brain dataset is different.
The groundtruth is 362x434x362, and the latter one is 181 x 217 x 181.
I found that the voxel size is double, but don’t know well how can I change the size of it.
I had tried to imresize for double it, but when I did it, the image was totally messed up.
So I am sending this email for getting any advice from you.
Thank you!
Sincerely,
Hyeonsoo Moon
M.S in Electrical Engineering
Featheringill Hall
Vanderbilt University, 400 24th Ave S, Nashville, TN 37212
Phone: (615) 648 – 4508
Hyeonsoo.moon@vanderbilt.edu
Hi,
It seems like you require same voxel sizes for both volumes, rather than needing to scale the image.
If so, and using MINC, you may re-sample the volume to your reference image with mincresample (http://bic-mni.github.io/man-pages/man/mincresample):
mincresample wholebrain_volume.mnc wholebrain_volume_resampled.mnc -like ground_truth.mnc
You could convert the voxel to int type:
mincresample wholebrain_volume.mnc wholebrain_volume_resampled.mnc -like ground_truth.mnc -byte
Hope that helps,
Kelvin
Thank you for your support.
I had tried the function imresize() first as you said, but it didn’t work well so I tried some bunch of ways.
first one was making my own function, and it works not bad, but when I compare with the ground truth, it has a lot of difference. another attempt was ‘flirt’ command. it worked well, so I implemented this one.
I will also try the function you said me at this time.
Thank you so much.
Sincerely,
Hyeonsoo Moon
M.S in Electrical Engineering
Featheringill Hall
Vanderbilt University, 400 24th Ave S, Nashville, TN 37212
Phone: (615) 648 – 4508
Hyeonsoo.moon@vanderbilt.edu