 |
 |
Hi Alexa, this is quite simple to do in matlab, though there is a bit of fiddling around involved. 1. Read the image in a matrix (use spm_read_vols) 2. and create a binary mask using a threshold that you think is suitable. 3. You can identify the different clusters by using the function "bwlabeln". This function basically labels connected components in binary images (your mask) and you can also specify the connectivity, e.g. 8, 26 etc. Look at the help information in matlab. If I am not mistaken SPM uses 18 connectivity (but I may be remembering wrong. Please correct me). The output of the functions is a matrix (same dimensions as the input) with voxels labeled according to the cluster they belong to. 4. You can then use "find" to get the voxel coordinates for each cluster, e.g. find(mask == 1) for cluster 1. This way you can also count the number of voxels per cluster (the length of the index) and you can also use the output from "find" to extract the statistical values from your original statistical image. To get an index that is for 3-D images, you can use "ind2sub". To get the peak value just use "max" on the extracted values and use the index from "max" to get the voxel coordinates. Does that make any sense? Sorry for not sending a script, but I don't have one for exactly what you are asking. Let me know if you have trouble writing it though. I'll be happy to help out. I am still in Edinburgh. I hope you are well, Marina Alexa Morcom wrote: > Hello list people > > Has anyone out there written code to extract image values - ideally > cluster sizes and peak values/ locations - from a 'raw' statistical > image? I have some images processed outside SPM that I'd like to explore > in this way; they're in .nii format. > > Thanks! > > Alexa Morcom > > > ------- > > Dr. Alexa Morcom > Department of Psychology/ Centre for Cognitive and Neural Systems > University of Edinburgh > 1 George Square > Edinburgh EH8 9JZ > > Tel: +44131 9511907 (ext. 511907) > Fax: +44131 650 4572 > > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336 > > > >