Full-head fNIRS data generator calibrated via simulation-based inference

Due to the growing use of data-driven approaches such as deep learning, there is an increasing demand for large, labeled fNIRS datasets. To address this, we have developed a tool for generating synthetic fNIRS data. To preserve spatiotemporal information across channels, we utilize Mesh-Based Monte Carlo simulations. To increase fidelity of the synthetic data, amortized simulation-based inference is used to further calibrate the data generator parameters to the experimental data being modeled. With this, large datasets may be generated which can be tailored to a wide range of experimental paradigms, making this a broadly usable tool across fNIRS studies.