Sensitivity analysis of the balloon model parameters in functional near-infrared spectroscopy simulation

Background: Accurate modeling of the hemodynamic response is critical for fNIRS data interpretation. While the Balloon model is a cornerstone for this, the quantitative impact of its key parameters on the fNIRS signal, particularly in the presence of realistic artifacts, remains under-characterized. New method: We developed an end-to-end fNIRS simulation pipeline. It incorporates a neural activity model, the Balloon model for hemodynamics, convolution for signal generation, and realistic motion, cardiac, and respiratory artifacts. We performed a sensitivity analysis by systematically varying Grubb's exponent (α) and transit time (τ). Results: Both α and τ significantly influence the simulated fNIRS response. α shows a non-linear relationship with peak amplitude, while τ has a more linear effect on signal timing. Regression models quantifying these effects demonstrated a strong statistical fit (p < 0.05, R² > 0.9 for α). Comparison with existing methods: Unlike prior fMRI-focused studies, this is the first quantitative sensitivity analysis specifically for fNIRS signals that incorporates a realistic noise model. Our framework characterizes the forward model's behavior, providing parameter-specific insights not previously available for fNIRS simulations. Conclusions: The fNIRS hemodynamic response is highly sensitive to the Balloon model's α and τ parameters. These findings highlight the importance of accounting for physiological variability in fNIRS analysis and provide a robust framework for generating synthetic data to test signal processing algorithms.