Deriving cardiac waveforms from fMRI data using slice selective averaging and a deep learning filter.

This work is a new technique to find Cardiac waveform from the fMRI data. For that purpose, a three stage data analysis is performed. In the first two stages, a candidate signal is derived by averaging over the voxels in every slice and combining them with proper time delays and resampling to 25Hz. As the third stage a deep learning architecture is used to improve the signal quality. The reconstructed signal is a good estimate of the plethysmogram data which is collected simultaneously with fMRI data.