The rapid growth of the functional neuroimaging literature presents significant challenges for synthesizing findings across studies. Although automated meta-analysis platforms like Neurosynth.org facilitate large-scale literature exploration, their capacity for addressing nuanced research questions was limited. As a result, manual meta-analyses remain prevalent despite being highly time-intensive, often suffering from limited reproducibility, and frequently resulting in the loss of valuable expert-curated data due to idiosyncratic workflows. To address these limitations, we introduce Neurosynth Compose, a web-based platform designed to streamline the creation of transparent, reproducible, and high-quality neuroimaging meta-analyses. Neurosynth Compose features a user-friendly interface for study curation and annotation, adhering to PRISMA guidelines, and is integrated with NeuroStore, a centralized database containing over 30,000 studies with pre-extracted activation coordinates. Meta-analytic models are specified using the Neuroimaging Meta-Analysis Data Standard (NiMADS) and executed via the comprehensive NiMARE Python library, which supports diverse coordinate- and image-based algorithms. Analyses can be executed locally or in the cloud using portable execution bundles, and results are uploaded back to the online platform, facilitating interactive review and rapid sharing to colleagues and the community. By combining automated data collection with expert-guided study selection and powerful analysis tools in an open and flexible system, Neurosynth Compose streamlines the process of creating high-quality neuroimaging meta-analysis. At the same time, this open and collaborative framework encourages users to share their valuable annotations and meta-analyses, fostering a valuable crowdsourced knowledge-base, and enabling users to update existing meta-analyses, paving the way for "living" syntheses that can be updated as new research emerges.
Kent et al. (Mon,) studied this question.