Hi experts,
I was wondering what kind of approaches come to your mind for segmenting task-fMRI data into trials. For example, If I would want to train a classifier to predict behavior on a trial-by-trial basis in a task-related design.
In literature I often see quite “basic” approaches of segmenting the data into individual trials and then shifting the indices for these segments by 4/6 seconds to account for the hemodynamic lag.
I guess that this is the most straightforward approach from a machine learning point of view, however from a neuroscientific point of view I wonder if there are less crude methods of doing this (e.g. by using more advanced statistical models), as I feel like this is not without risk.
I am not an expert in this field, so I would love to hear some thoughts and opinions from more experienced people (if you have any) 
Cheers!
Max