Dear Colleagues,
My question is related to second level analysis of language tasks in tumor patients. There will be 40 patients total with 20 tumors in/neighboring the anterior language areas and 20 in/neighboring posterior language regions. How similar should tumor location be in each subgroup to run reliable second level analyses?
Tumor size and malignancy will obviously be confounding factors. Most of our patients have aggressive tumors.
Thank you so much.
Kind regards,
Monika
Hi Monika,
Unfortunately, I don't think that there is a clear "correct" answer to this question. First, the answer will depend on how you define "similarity" for tumors -- this could be measured in multiple ways (e.g. Dice similarity coefficient, spatial correlation, tumor volume, correlation of % damage to anatomical structures, etc.). Nonetheless, there is no objective similarity threshold for any of these metrics that would guarantee reliable second-level analyses, although one would expect greater reliability for regions less proximal to the affected areas. One potential solution would be to use a data fusion approach that considers within-group variability in the tumor location (e.g. parallel ICA, mCCA+joint ICA, etc. -- see Article A Review of Multivariate Methods for Multimodal Fusion of Br...
for a review). This would have the added benefit of providing information about how different tumor topologies relate to the language task activation.Best,
Joseph
Hi Joseph,
Yes, I was not expecting a straightforward answer but your reply is very helpful. Great thanks!
Best wishes,
Monika
- Badges
- Science topic
- Similar topics
- Biological Science
- Neuroscience