I am wondering if a voxel-wise linear mixed-effects model via 3dLME is appropriate for my dissertation dataset.

I want to examine how response to uncertain threat might vary as a function of trauma symptom severity (PCL-5). I am using an uncertain picture task that examines valence (neutral, negative) and predictability (predictable, unpredictable), for a total of 4 within-subject conditions (predictable-neutral, unpredictable-negative, etc.). I also want to control for age and sex. I think that I could use 3dLME with valence (2) and predictability (2) as within-subjects factors, age and sex as between-subjects factors, and PCL-5 scores as a within-subject quantitative variable.

Does this seem appropriate? I read through the Chen 2013 paper but I still can't quite figure out if this fits or not. The other option I considered was a hierarchical linear regression analysis.

Thanks,
Ken