ASTEROID, A Validated Annotated Dataset of Analytical Decisions in Graph Theory Based fMRI
ASTEROID, A Validated Annotated Dataset of Analytical Decisions in Graph Theory Based fMRI
Imam, S. H.
AbstractGraph theory-based functional magnetic resonance imaging (fMRI) studies use complex analytical pipelines, but methodological decisions are often reported in unstructured text, limiting systematic comparison and the design of multiverse analyses. Here, we describe ASTEROID, a structured annotation dataset derived from 156 graph theory-based human fMRI study publications. Its schema contains 125 labels, comprising 61 methodological labels and 64 labels for study characteristics and experimental design. An LLM-assisted workflow generated initial annotations, which underwent automated quality-control checks and expert human review. The dataset captures preprocessing, functional connectivity, and graph-analysis decisions. Within these categories, steps and their sequence, software, algorithms, parameters, and supporting text spans are included where available. This accompanying paper provides prompt specifications, processing scripts, and documentation, and describes an interactive viewer that supports exploration of the annotation schema and the study records. ASTEROID enables comparison of reported analytical choices, the informed design of multiverse-based robustness analysis and supports the development and evaluation of automated methodological information-extraction tools and educational resources for fMRI data analysis.