The segmentation nets typically output one-hot encoded values. So if you have 30 labels, the output layer has 30 outputs, which can be loosely interpreted as p-values for each label (they will add to 1). To turn it into a segmentation you take the argmax over the 30-label dimension and toss the underlying probabilities
From: freesurfer-boun...@nmr.mgh.harvard.edu <freesurfer-boun...@nmr.mgh.harvard.edu> On Behalf Of Douglas N. Greve Sent: Friday, October 30, 2020 11:39 AM To: 李梦君 <meng...@csu.edu.cn>; Freesurfer support list <freesurfer@nmr.mgh.harvard.edu> Subject: Re: [Freesurfer] Statistic Difference Region on the Fsaverage Turned into No Statistic Difference after Mapping onto the Subjects sorry, I don't understand your question. can you elaborate? On 10/28/2020 11:57 PM, 李梦君 wrote: External Email - Use Caution Thanks for your help! I finished the -qcache for every subject before group analysis with Qdec/mri_label2label. I want to do the correlation analysis between the significant cortex area and the behavior score. Should I compare the group analysis without the -qcache before Qdec/mri_label2label? -----原始邮件----- 发件人:"Douglas N. Greve" <dgr...@mgh.harvard.edu><mailto:dgr...@mgh.harvard.edu> 发送时间:2020-10-28 05:08:03 (星期三) 收件人: freesurfer@nmr.mgh.harvard.edu<mailto:freesurfer@nmr.mgh.harvard.edu> 抄送: 主题: Re: [Freesurfer] Statistic Difference Region on the Fsaverage Turned into No Statistic Difference after Mapping onto the Subjects When you run your original analysis, you probably smoothed the map before doing the group analysis. This smoothing would not be in your ROI analysis, and the difference could be pretty big if the label is small. Also, the area map will have a jacobian correction that will not be in the raw data. On 10/19/2020 11:29 PM, 李梦君 wrote: External Email - Use Caution Dear FreeSurfer Developers: I compared the thickness, area, volume in the mild traumatic brain injury patients with health control by FreeSurFer, and found a statistic difference region in the cortex area, I defined the statistic difference region as a label on the fsaverage and mapped this label to subjects by Qdec/mri_label2label. The label defined on the fsaverage was mapped onto all subjects successfully, and the label in subjects matched well with the inflated cortex of subjects in Freeview Then the area value of the label in every single subjects was extracted, and I found that there is no statistic difference in the area value between the mild traumatic brain injury patients and health control. I have checked the segmentation of every subjects and it was a good job done by FreeSurFer. FreeSurfer version: freesurfer-Darwin-OSX-stable-pub-v6.0.0-2beb96c Platform: macOS High Sierra. Best wishes Mengjun Li Second Xiangya Hospital Central South University _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer