Hemodynamic Signals Reshape Biological Inference in Widefield Calcium Imaging

Abstract

Widefield calcium imaging is widely used to study cortex-wide neural dynamics, yet fluorescence signals are strongly influenced by hemodynamic fluctuations arising from blood volume and oxygenation changes. Although hemodynamic correction is frequently applied, it remains unclear whether vascular contributions represent a modest preprocessing concern or systematically bias biological interpretations of cortical activity. Here, we used dual-wavelength imaging to determine how hemodynamic correction reshapes inference of cortical dynamics across mouse lines expressing GCaMP6s, GCaMP6f, and jGCaMP8m, across multiple analytical domains, and under healthy and glioblastoma conditions. We systematically compared uncorrected and corrected signals using analyses spanning functional parcellation, connectivity, spectral structure, brain-behavior coupling, and low-dimensional network-state dynamics. Hemodynamic correction consistently reduced global functional connectivity, increased network modularity, redistributed spectral power away from slow-frequency dominance, and reorganized multivariate representations of cortical state space. These findings demonstrate that vascular signals do not behave as unstructured measurement noise but instead introduce organized variance that propagates across analytical pipelines and influences inference of cortical dynamics. The consequences of this bias were particularly relevant in glioblastoma, where tumor-associated vascular remodeling amplifies the mismatch between fluorescence signals and underlying neuronal activity. In this disease setting, correction revealed hemispheric asymmetries, reduced network-state entropy, and constrained trajectories within cortical state space that were obscured in uncorrected recordings, demonstrating that vascular remodeling can fundamentally alter interpretation of tumor-associated brain dynamics. More broadly, vascular signals systematically biased estimates of functional organization, network architecture, brain-behavior relationships, and disease-associated phenotypes. Together, these findings establish hemodynamic correction as a critical determinant of biological interpretation in mesoscale calcium imaging rather than a simple preprocessing refinement.