Glad to sharing our new paper addressing timing issues between video features and BOLD responses during movie watching. And we showed that deconvolution can be done using continuous feature time series as inputs to derive hemodynamic response functions (HRFs).
New paper in Imaging Neuroscience by Xin Di, George B. Hanna and Bharat B. Biswal: Beyond the canonical HRF: Flexible temporal modeling reveals unconstrained BOLD profiles during naturalistic viewing doi.org/10.1162/IMAG...