The sample sizes are very low (around 20-30), but the brain areas they’re studying are about 200,000, meaning there’s about 200,000 variables
Furthermore, you need a lot of subjects if the relationship between the brain regions and constructs is diffused and weak