SiGn: An Open-Source Signal-Generating fMRI Phantom for Dynamic Quality Assurance

Abstract

Functional magnetic resonance imaging (fMRI) quality assurance has traditionally relied on static, geometrically regular phantoms that cannot generate the dynamic signal changes fMRI analysis pipelines are designed to detect. Here we present the Signal Generating (SiGn) anthropomorphic brain phantom, a 3D-printed cortical model derived from an individual participant's structural MRI, filled with tissue contrast-mimicking agar gels and coupled to a hemin-based infusion system that produces controlled, time-varying $T_2^*$-weighted signal changes. We validated the phantom across two scanning sessions on a 3\,T Siemens MAGNETOM Vida scanner, demonstrating that hemin infusion produced spatially localised activation detectable by standard general linear model analyses. Because the phantom's geometry is derived from real participant anatomy, its functional data can be coregistered and spatially normalised to standard brain templates through the same pipeline applied to human data, enabling end-to-end assessment of how each preprocessing step affects a known ground-truth signal. To support adoption and reproducibility, we openly release the full resource, including 3D-printable STL model files, tissue-mimicking gel recipes, the BIDS-formatted dataset, preprocessing and analysis scripts, and a containerised reproducibility workflow; the corresponding archival container image is also deposited on Zenodo. This framework is intended to lower the barrier for other groups to fabricate, scan, and analyse an equivalent device on their own hardware, adapt it to specific research questions, and iteratively improve the design, thereby supporting more rigorous and transparent fMRI quality assurance practices across the neuroimaging community.