Improved decoding of neural activity from fMRI signals using non-separable spatiotemporal deconvolutions

► Most fMRI analyses assume space-time separable hemodynamic response functions (HRF). ► We predict intracranial neural signals using non-separable deconvolutions of fMRI signals. ► Neural signal prediction is consistently better when accounting for non-separable HRFs. ► We identify the spatiotemporal features that give rise to this improvement.