Parrot: a subject-specific whole-brain EEG-fMRI simulator

EEG provides unparalleled temporal resolution for whole-brain imaging, yet the scarcity of ground-truth data limits the development of robust machine learning models for source reconstruction. To address this limitation, we developed a high-fidelity simulation engine designed as a data-generation infrastructure for next-generation AI-based inverse models. While established platforms such as The Virtual Brain (TVB) [1] model large-scale brain dynamics, their EEG simulations primarily focus on neocortical sources. Building on TVB, our framework expands this scope by automating the integration of cortical, deep, and cerebellar structures into a unified simulation pipeline. Using modern neuroimaging workflows, we combine high-density cortical parcellations with detailed subcortical atlases and incorporate high-resolution structural references—including BigBrain 100 µm histology and a submillimeter cerebellar atlas—to support anatomically and electrically precise modeling. Our system automatically generates biophysically grounded whole-brain models exceeding 1000 interconnected nodes. Structural connectivity is estimated using MRtrix3 [2] probabilistic tractography applied to the Human Connectome Project young adult dataset [3], and dynamically coupled with TVB neural mass models. The pipeline further constructs high-resolution tetrahedral meshes for DUNEuro-based forward modeling [4], enabling accurate projection of neural activity to scalp electrodes while simulating realistic spatiotemporally correlated noise. This scalable framework enables the generation of large synthetic EEG datasets with known ground truth. We use these datasets to train machine-learning-based inverse models aimed at recovering deep and cerebellar sources that are traditionally difficult to resolve with scalp EEG. The source code and a containerized implementation will be released publicly. [1] Jirsa et al., 2013 [2] Smith et al., 2025 [3] WU-Minn HCP Consortium, 2013 [4] Vorwerk et al., 2021