The AI Neuroscientist: An Interactive Agent for Neuroimaging Analysis

Summary

The paper presents the AI Neuroscientist, a language agent designed to make small-scale neuroimaging data exploration more accessible to researchers without specialized coding and statistical expertise. It combines a large language model with a neuroimaging toolset that supports quality control, modeling, and visualization. Researchers can ask questions about data quality and specify analysis parameters in natural language, creating a transparent, interactive alternative to conventional scripted pipelines. The authors demonstrate the system with functional near-infrared spectroscopy (fNIRS) data. They evaluate it on a custom fNIRS benchmarking suite and compare it with general-purpose LLM agents equipped with code sandboxes. The paper describes the system as an interactive analysis interface rather than a replacement for all conventional workflows, and focuses on small-scale data exploration. Future work will extend the architecture to other modalities, including functional magnetic resonance imaging (fMRI), and broaden the benchmark to cover additional fNIRS tasks.