Dear Jana,
design looks fine. The contrast should be set to the 2nd column.
If you have a scanner change between a subject's scans, you can never correct for that, ever. You can only harmonize the effects if you have multiple sites/scanners, but the site/scanner is the same for a subject's intra-individual data. If you change the scanner between time points, you have no idea where the differences come from. Are these real long-term effects or just scanner effects? The latter cannot be harmonized. Sorry to stress this so much, but it is simply not possible. Pooling different scanners (if the scanner is the same for all time points of a subject) is easy and recommended, but not between scans.
Best,
Christian
On Tue, 29 Nov 2022 19:41:56 +0000, Schill, Jana <[log in to unmask]> wrote:
>Dear Christian,
>
>Thank you very much for your explanations! I have now been able to successfully set up the model.
>Just to verify: The first column (time) is fully white and does not have different entries for the two time points (see attached). Is this correct?
>
>Concerning your question on scanner change: These are pooled data from different institutes and scanners. We intend to address this problem by harmonizing the preprocessed images using neuroCombat, which eliminates site effects while retaining biological variation within the data. Please let me know if you think this is insufficient for this kind of analysis.
>
>Thank you!
>Jana
>
>
>-----------------------------------------
>
>Dear Jana,
>
>>Dear all,
>>
>>I am trying to run a retrospective longitudinal VBM analysis in cat12. I am looking at approximately 50 subjects with two MRI scans - however, the time interval between scans varies widely from 1 to 15 years.
>15 years are quite impressive. Was there any change in scanner or head coil? Just curious about this, because it's quite hard to keep everything stable over such a long time and scanner change would prevent that you can use that data for a longitudinal analysis.
>
>>
>>I set up the statistical model according to the guidelines in the cat12 manual, using age as a covariate (as it changes between scans and captures the differences in time intervals). However, I run into problems when estimating the model: The estimated design looks mostly black and I cannot define contrasts on it using the contrast manager (error message: 'invalid contrast'). Please see attached pictures.
>
>Designs with varying time between the scans are not that straightforward to model: Please use that design as starting point and skip the factor group and it's definition for the conditions and also the interaction:
>https://secure-web.cisco.com/1xq5-_THeZGyg4gl-0dNeDLlMJZzFTqwDUXXw6qRLnlzc8Zvs0rSXertQGDsTlQW6ay4jNzev3B9PK1iQVHkGdbUcPAB-nuEifpVhyhqywfqkcj7cfe3CTVO3_Zp5dv6rxQPCeWiDVsJAuq9XDRYVoGm9hHqslArP7Mlm1E_gnfCNWAB_9MtypGC7-wKg8vLi_3Y3ABTK81dH3RNCZ-Oo7TgYktmqrPWyJhpbtbkO_Nly-5enngwB46qGSm3KplvmDHOu1jFBEi0UhutTJmBuECIeT_NgyLxFbI9JDAgH7Zdp1kepojpXafuRZklBtuXoHjdDyFd0nU_Ej-ryJprM1g/https%3A%2F%2Fneuro-jena.github.io%2Fcat12-help%2F%23long_two_interact
>
>Design → "Longitudinal data (Flexible factorial)"
>
>Factors → "New: Factor; New: Factor; New: Factor"
>Name → [specify text (e.g. "subject")]
>Independence → Yes
>Variance → Equal
>Factor
>Name → [specify text (e.g. "time")]
>Independence → No
>Variance → Equal
>
>Subject
>Scans → [select files (the smoothed GM data of the 1st Subject)]
>Conditions → " 1 1 " [two time points]
>...
>Scans → [select files (the smoothed GM data of the 1st Subject)]
>Conditions → " 1 1 " [two time points]
>
>Main effects & Interaction → "New: Main effect"
>Main effect
>Factor number → 2
>Main effect
>Factor number → 1
>
>Covariates
>define here age for each time point and subject...
>
>The final design should look similar to the last example here, except that there is just one group and the first column defines the time points and the 2nd age as covariate with two varying values for each subject:
>https://secure-web.cisco.com/18sulLDySbNJeQgUiz1sGHM1tDrne9Heesmda6GDJZ2pdKSONAe3bGCnZnDWckqMsmV2Vuv4fQEMUcN4u0q1oJDVvFGSd13MHqXEwwM77VHpakqpHOYZBbJK0cFXlqU3LhWNvPIl4A0b3vh37SaWoQBpIEayyzlGO-zICGPdWmjn5sqrfqksj3e7Yb9Vcw29FPKK276YtXql0Hm92ZpZ6RI3leO_in2vvUy0sYdb5eUsCkSEOKbIaUEvou0ZZzyJ3wiPoYdZrypdXovzBF6RK5P4LlysXKQWjdrmyjxpZpXE9sWfuGBNeL-Fq4E-oEIwFrfgRbrY5eaU607F777W5cw/https%3A%2F%2Fneuro-jena.github.io%2Fcat12-help%2F%23long_contrast
>
>The contrast should be then set:
>0 -1 (i.e. atrophy over time)
>
>Best,
>
>Christian
>
>>
>>My questions are:
>>1. Can I use age as a covariate in this setup and will it achieve what I want to achieve by using it (accounting for inter-scan time intervals).
>>2. If not, what would be the correct way to set up the statistical model?
>>3. What are possible reasons for the mostly black design matrix?
>>
>>Thank you,
>>Jana
>
>
>
>
>The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Mass General Brigham Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline <https://www.massgeneralbrigham.org/complianceline> .
>Please note that this e-mail is not secure (encrypted). If you do not wish to continue communication over unencrypted e-mail, please notify the sender of this message immediately. Continuing to send or respond to e-mail after receiving this message means you understand and accept this risk and wish to continue to communicate over unencrypted e-mail.
>
|