Activation likelihood estimation (ALE) is a very good method for task fMRI meta-analysis studies (including both within-group between-condition activation difference and between-group activation difference). Although the design of task fMRI studies varies a lot from study to study, most task fMRI studies use very similar GLM method and focus on the local activity of, rather than the relationship between, brain regions. There have been many meta-analysis task fMRI studies.
The design of resting-state fMRI is very similar across studies, therefore it is should be better for meta-analysis, especially on brain disorders. There have been many many resting-state fMRI papers on brain disorders. Unfortunately, few meta-analysis papers have been published.
One big problem is that the analytic methods are very different for most resting-state functional connectivity studies. This is especially a problem for seed-region-based functional connectivity studies. Virtually, I can write hundreds of papers on a single resting-state fMRI dataset by using seed-region-based functional connectivity analysis because I have many many seed regions, multiplying the various nuisance regressors. And actually, I have had a few resting-state functional connectity papers on a single dataset of ADHD. Each sounds like an interesting story. The situation seems better for complex network analyses, e.g., ICA and graph. While we know the complex intergration of brain function is critical, more attention should be paid on simpler local activity, e.g., amplitude of low frequency fluctuation (almost the same as root mean square and standard deviation, equal to the square root of power) and regional homogeneity (local synchronization, similar to short range functional connectivity density). Although papers on local activity seem to have less novelty than complex network, it is more helpful to meta-analysis, and therefore to provide strong evidences for clinical medicine.
