Hello machine-learning experts,
I have two groups of subjects (100 controls and 100 patients) and cortical volume (CV) calculated from 75 brain regions from all subjects (i.e., 200x75 matrix). I was wondering if I should normalize CV of each region wrt to head size (ICV) e.g., get a new matrix 200x75 of normalized CV (i.e., raw CV divided by ICV) before converting to zscore, or I do not need to normalize raw CV wrt ICV and just converting raw CV to zscore is enough.
I found mixed literature with some papers normalizing raw CV wrt ICV before converting to zscore, and some just converting raw CV to zscore.
I would really appreciate any suggestions/tips.
Best,
Sahil
Hi Sahil, Your first option is mostly preferred in ml. Alternatively, do option and also covary Icv like you do other covariates eg. age sex etc.
Hi Ganesh Chand, thanks for your response. Could you please give a little more insight into how can I add ICV as a covariate while doing performing ML stuff? I import processed data (i.e., after correcting CV wrt ICV) to MATLAB and further use MATLAB GUI for the analysis, but not sure can I add covariates?
Covariates can be corrected outside ml workflow, then zscore and use SVM. I think, your first option would be good enough.
- Badges
- Science topic