help > Selecting QC co-variate
Showing 1-4 of 4 posts
Mar 3, 2020 02:03 AM | Ramesh Babu MG
Selecting QC co-variate
Dear Alfonso,
I did preprocessing steps as mentioned in the conn manual for fMRI analysis. After preprocessing, there are 7 co-variate included automatically as shown below.
1. All subjects
2. QC Valid scan
3. QC Invalid scan
4. QC Max motion
5. QC Mean Motion
6. QC Mx GS Change
7. QC Mean GS Change
I have a question here. Why there are 2 co-variate? For motion parameters, there are mean and maximum values. Can I select any one of this? or both should be there as default co-variate? Two of my subjects showed 2.01 and 2.04 as max motion. Should I include those subject or remove?
What is the cut-off point for global signal change? One of my subject showed highest GS while comparing the rest of the subjects? Should I remove that subject or not?
Please advise me
Thanks
Ramesh
I did preprocessing steps as mentioned in the conn manual for fMRI analysis. After preprocessing, there are 7 co-variate included automatically as shown below.
1. All subjects
2. QC Valid scan
3. QC Invalid scan
4. QC Max motion
5. QC Mean Motion
6. QC Mx GS Change
7. QC Mean GS Change
I have a question here. Why there are 2 co-variate? For motion parameters, there are mean and maximum values. Can I select any one of this? or both should be there as default co-variate? Two of my subjects showed 2.01 and 2.04 as max motion. Should I include those subject or remove?
What is the cut-off point for global signal change? One of my subject showed highest GS while comparing the rest of the subjects? Should I remove that subject or not?
Please advise me
Thanks
Ramesh
Mar 3, 2020 08:03 PM | Alfonso Nieto-Castanon - Boston University
RE: Selecting QC co-variate
Dear Ramesh,
Yes, in my experience mean motion is often more representative of potential residual BOLD effects than max motion, but you may explore that in your own dataset, for example in the Quality Assurance plots section of the GUI looking at the "QC-FC association" plots (see https://web.conn-toolbox.org/fmri-method... for details).
Regarding cutoff for subject removal, there is no global recommended cutoff value (in part because many of these QC measures are rather sensitive to details of your experimental acquisition and subjects sample). Generally you may based this decision from the violin/histogram plots in the Quality Assurance section of the GUI ("distribution of subject-level QC measures" plots). Those plots will display the distribution of motion and GS change values in your sample together with reasonable outlier thresholds (3rd quartile + 1.5 * interquartile range) to determine which subjects may be potential outliers (and consider controlling their potential effect either by removing them or using statistical control techniques).
Best
Alfonso
Originally posted by Ramesh Babu:
Yes, in my experience mean motion is often more representative of potential residual BOLD effects than max motion, but you may explore that in your own dataset, for example in the Quality Assurance plots section of the GUI looking at the "QC-FC association" plots (see https://web.conn-toolbox.org/fmri-method... for details).
Regarding cutoff for subject removal, there is no global recommended cutoff value (in part because many of these QC measures are rather sensitive to details of your experimental acquisition and subjects sample). Generally you may based this decision from the violin/histogram plots in the Quality Assurance section of the GUI ("distribution of subject-level QC measures" plots). Those plots will display the distribution of motion and GS change values in your sample together with reasonable outlier thresholds (3rd quartile + 1.5 * interquartile range) to determine which subjects may be potential outliers (and consider controlling their potential effect either by removing them or using statistical control techniques).
Best
Alfonso
Originally posted by Ramesh Babu:
Dear Alfonso,
I did preprocessing steps as mentioned in the conn manual for fMRI analysis. After preprocessing, there are 7 co-variate included automatically as shown below.
1. All subjectsj
2. QC Valid scan
3. QC Invalid scan
4. QC Max motion
5. QC Mean Motion
6. QC Mx GS Change
7. QC Mean GS Change
I have a question here. Why there are 2 co-variate? For motion parameters, there are mean and maximum values. Can I select any one of this? or both should be there as default co-variate? Two of my subjects showed 2.01 and 2.04 as max motion. Should I include those subject or remove?
What is the cut-off point for global signal change? One of my subject showed highest GS while comparing the rest of the subjects? Should I remove that subject or not?
Please advise me
Thanks
Ramesh
I did preprocessing steps as mentioned in the conn manual for fMRI analysis. After preprocessing, there are 7 co-variate included automatically as shown below.
1. All subjectsj
2. QC Valid scan
3. QC Invalid scan
4. QC Max motion
5. QC Mean Motion
6. QC Mx GS Change
7. QC Mean GS Change
I have a question here. Why there are 2 co-variate? For motion parameters, there are mean and maximum values. Can I select any one of this? or both should be there as default co-variate? Two of my subjects showed 2.01 and 2.04 as max motion. Should I include those subject or remove?
What is the cut-off point for global signal change? One of my subject showed highest GS while comparing the rest of the subjects? Should I remove that subject or not?
Please advise me
Thanks
Ramesh
Mar 5, 2020 07:03 AM | Ramesh Babu MG
RE: Selecting QC co-variate
Dear Alfonso,
Thanks for this wonderful toolbox. On considering the mean motion for outliers detection, one subject showed mean motion more than outlier threshold (3rd quartile + 1.5 * interquartile range) another subjects showed mean GS change slightly more than the outlier threshold. Many subjects have valid scan less than the threshold level (1st quartile + 1.5 * interquartile range). If mean motion considered as potential outliers, should I remove that subject? When looking other plots with all subjects it looks fine. It will be very helpful for me if you go through the attached QC measure file of ll subjects and give your suggestions.
Thanks
Ramesh
Thanks for this wonderful toolbox. On considering the mean motion for outliers detection, one subject showed mean motion more than outlier threshold (3rd quartile + 1.5 * interquartile range) another subjects showed mean GS change slightly more than the outlier threshold. Many subjects have valid scan less than the threshold level (1st quartile + 1.5 * interquartile range). If mean motion considered as potential outliers, should I remove that subject? When looking other plots with all subjects it looks fine. It will be very helpful for me if you go through the attached QC measure file of ll subjects and give your suggestions.
Thanks
Ramesh
Sep 29, 2022 09:09 AM | Renzo Torrecuso - Max Planck Institute for Human Cognitive and Brain Sciences
RE: Selecting QC co-variate
Dear Alfonso,
I am using CONN on a multi center cohort study and therefore searching a way to quantify homogeneity of data using CONN's QC features.
Could you please clarify the following?
1. How does CONN compute QC Max motion, QC Mean Motion, QC Mx GS Change, QC Mean GS Change?
2. Given that each cohort arrives to different QC Quality Assurance thresholds, could you also please clarify how does CONN compute and suggest these outliers thresholds?
3. Is there any way to access Epi SNR with CONN?
Thank you very much for your attention,
All best
Renzo
I am using CONN on a multi center cohort study and therefore searching a way to quantify homogeneity of data using CONN's QC features.
Could you please clarify the following?
1. How does CONN compute QC Max motion, QC Mean Motion, QC Mx GS Change, QC Mean GS Change?
2. Given that each cohort arrives to different QC Quality Assurance thresholds, could you also please clarify how does CONN compute and suggest these outliers thresholds?
3. Is there any way to access Epi SNR with CONN?
Thank you very much for your attention,
All best
Renzo