help
help > RE: CONNToolbox: Changing nGauss for custom TPM.nii/TPM.mat
Nov 26, 2021 09:11 PM | Chidi Ugonna - University of Arizona
RE: CONNToolbox: Changing nGauss for custom TPM.nii/TPM.mat
Hi Naomi,
So I am not fully conversant with the intricacies of Gaussian Mixture modelling but based on a quick perusal of this paper from Ashburner and Friston [url=https://www.fil.ion.ucl.ac.uk/~karl/Unified%20segmentation.pdf]https://www.fil.ion.ucl.ac.uk/~karl/Unified%20segmentation.pdf[/url] it looks like the intensity distribution of each tissue class is modeled as a mixture of gaussians with varying means and variances. The choice of the number of gaussians to use for each class can vary. It looks like the defaults for SPM are [2 2 2 4] which i think refers to [graymatter whitematter csf everything_else]
Since you have 7 tissue classes then your batch.Setup.preprocessing.tpm_ngaus will have 7 values to match the volumes in your TPM.nii file. Something like batch.Setup.preprocessing.tpm_ngaus=[2 2 2 N1 N2 N3 N4]. I presume that you will want to keep 2 for gm, wm and csf but dont quote me on that -its possible that these numbers may need to be tweaked for your data set. I am not sure what your N1, N2, N3 and N4 should be.
So I am not fully conversant with the intricacies of Gaussian Mixture modelling but based on a quick perusal of this paper from Ashburner and Friston [url=https://www.fil.ion.ucl.ac.uk/~karl/Unified%20segmentation.pdf]https://www.fil.ion.ucl.ac.uk/~karl/Unified%20segmentation.pdf[/url] it looks like the intensity distribution of each tissue class is modeled as a mixture of gaussians with varying means and variances. The choice of the number of gaussians to use for each class can vary. It looks like the defaults for SPM are [2 2 2 4] which i think refers to [graymatter whitematter csf everything_else]
Since you have 7 tissue classes then your batch.Setup.preprocessing.tpm_ngaus will have 7 values to match the volumes in your TPM.nii file. Something like batch.Setup.preprocessing.tpm_ngaus=[2 2 2 N1 N2 N3 N4]. I presume that you will want to keep 2 for gm, wm and csf but dont quote me on that -its possible that these numbers may need to be tweaked for your data set. I am not sure what your N1, N2, N3 and N4 should be.
Threaded View
Title | Author | Date |
---|---|---|
naomi g | Nov 24, 2021 | |
Alfonso Nieto-Castanon | Nov 26, 2021 | |
naomi g | Nov 30, 2021 | |
Chidi Ugonna | Nov 26, 2021 | |
naomi g | Nov 30, 2021 | |