There are major discrepancies in methods and reporting practices across papers.
Seed-based studies may be the most common, but there are many papers using ICA, graphs.
These techniques can appear in combination in a single article.
Understanding what seed was used, and matching them across studies, is a major pain for a human, and will be very difficult to achieve automatically.
We went through that process with Dr Badhwar for this paper, https://www.ncbi.nlm.nih.gov/pubmed/28560308
We ended up grouping seeds based on an atlas of large parcels.
There is no consensus about how to best approach this problem.
For this reason there are few systematic meta analysis of rs-connectivity.
Maybe neurosynth have found a quick-dirty-but-still-useful workaround, but I doubt it.