文章大修,审稿人关于DPARSF预处理的问题,望请各位同道指教

Submitted by liuchen on

最近自己的一篇文章被PLOS ONE大修,方法学比较简单:一组病人与正常人之间ALFF自发脑活动的比较,数据的处理全部都是使用DPARSF进行的标准处理流程。

但是审稿人提了很多数据处理技术上的问题,鉴于自己的医学背景,望请各位高人指教,谢谢。

第一个问题:Q.2 The investigators excluded data that exceeded a specific motion threshold. However, it would be important to know if there were remaining differences in motion between groups? Was motion used as a covariate in the analyses performed?

在预处理阶段我已经把头动作为变量校正了,还需要在后续的两组比较把头动作为协变量吗?

如果两组头动参数在进行两组比较之后并没有显著性差异,还需要作为协变量吗?

 

第二个问题:It’s understood that the investigators used a standard value of 8mm smoothing of the structural data, however, it is not clear why such a large smoothing kernel is necessary (larger than the functional smoothing level)? Please provide justification for this.

数据分析我是把静息态功能数据配准到结构数据所创建的DARTEL模板上,审稿人的问题是静息态数据我使用的是6mm平滑核,但是结构数据却使用了更大的8mm平滑核,请问这样做会影响实验结果的准确性吗?

 

第三个问题:The global signal regressor could have major unintended influences on the results. Analyses should be performed with and without the global signal regressor to see how this processing step affects the results.

审稿人要求我做全脑平均信号去除与不去除之间的结果差异,怎样去和审稿人说做全脑信号去除的好处,而不用去做无全脑信号去除。

 

第四个问题:I am a bit uncertain whether it is sound to normalize the regional ALFF measures by the grand mean of each subject. Did the grand-mean values differ by group? This could indicate a generalized slowing. Perhaps it would be better to treat the grand mean value as a covariate in the analyses and report whether the covariate was significant interacting with group assignment。
 

ALFF到底需不需要去做全脑平均之后的mALFF,需要把平均的值作为协变量吗?有没有可靠的文献支持需要做平均?

望请各位专家,老师,同学给出意见,如果指出支持文献更好,谢谢!

最后是我的DPARSF设置参数图

Hi,

Here is my opnion about your questions. I did not answer your question about mALFF since I don't know much about it. But generally speaking, the reviewer's suggestions are reasonable, you should just do what the reviewer asked you to do.

在预处理阶段我已经把头动作为变量校正了,还需要在后续的两组比较把头动作为协变量吗?

如果两组头动参数在进行两组比较之后并没有显著性差异,还需要作为协变量吗?

--Yes and yes.

数据分析我是把静息态功能数据配准到结构数据所创建的DARTEL模板上,审稿人的问题是静息态数据我使用的是6mm平滑核,但是结构数据却使用了更大的8mm平滑核,请问这样做会影响实验结果的准确性吗?

-- Here, you need to write in your paper why you choose 8mm as your kernel of smoothing.

审稿人要求我做全脑平均信号去除与不去除之间的结果差异,怎样去和审稿人说做全脑信号去除的好处,而不用去做无全脑信号去除。

-- Please don't argue with reviewer with words, do what the reviewer asked you to do and argue with results/data.

And, I am looking at your DPARSF configuration, it seems that you did not regress out nuisance parameters? And I remembered that Chao-Gan mentioned in a post that you should use the advanced edition of DPARSF.

Good luck!

 

Yang

 

YAN Chao-Gan

Mon, 10/28/2013 - 19:02

Hi Chen,

1. You only performed realignment with DPARSF Basic edition. I will even suggest you to perform Friston 24 head motion regression (in Nuisance regression step) in preprocessing with DPARSF Advanced edition.

ALFF is pretty sensitive to head motion, thus I agree that you need to check if there is any group difference between the two groups, i.e., perform two-sample t-test on mean FD Jenkinson (an output under RealignParameters of DPARSFA). If so, including mean FD as a covariate is not a bad idea -- you results will be a little conservative, but at least the false positives can be controlled.

Please see: Yan, C.G., Cheung, B., Kelly, C., Colcombe, S., Craddock, R.C., Di Martino, A., Li, Q., Zuo, X.N., Castellanos, F.X., Milham, M.P., 2013. A comprehensive assessment of regional variation in the impact of head micromovements on functional connectomics. Neuroimage 76, 183-201.

2. I have an impression that structural studies (VBM) usually use a bigger smooth kernal, but to be honest, I am not sure the exact reason. I think you can find some papers who did this and cite them.

3. For the settings as you pasted, you didn't do global signal regression.

To my knowledge, functional connectivity is sensitive to global signal regression, but ALFF is not sensitive to that procedure.

4. Mean division is one method for standardizing you measures -- to control lots of unknowns, see 

Yan, C.G., Craddock, R.C., Zuo, X.N., Zang, Y.F., Milham, M.P., 2013. Standardizing the intrinsic brain: towards robust measurement of inter-individual variation in 1000 functional connectomes. Neuroimage 80, 246-262.

The method that the reviewer suggested (mean regression) is another standardization method, which is not bad. My personal favorite is mean reregression + SD division. There will be a standardization toolbox in future release of our software.

Best,

Chao-Gan

注1:该主题已移到论坛“我想用中文提问 (I want to post in Chinese)”。

注2:鉴于大家共同提高英语的目的,以及该邮件列表有很不少英语受众,我采用英语作为回复。如果你觉得确实有必要用中文回复,我也可以配合。

Thank doctor YAN's detail suggestion.

I have understood your explanation. However, I don’t know how to calculate mean value of alff mentioned in the fourth question. Could I get the specific mean value by rest software? Please help me go into details.

Hi,

You can use REST->Utilities->Extract ROI Signals to extract the brain mean values of all ALFF maps. Define the brain mask as an ROI.

You can organize the wc1, wc2 files and apply a smooth kernel to them, either use SPM's smooth, or use DPARSF (but pretend these files are functional files).

I will consider to add a separate control for VBM smooth kernel in next version.

Best,

Chao-Gan

Another tip, could you add the option which we can change the vbm smooth kernal other than the default 8mm smooth kernal in the next version DPARSF.

Attachment Size
QQ截图20131027120835.jpg 107.38 KB