ICA-Based Blind Source Separation (BSS) Recovery of the Pattern Electroretinogram (PERG) from Single Channel Records with Poor SNR

A method of signal recovery of the pattern electroretinogram (PERG) from recordings contaminated with electrical artefacts resulting from eye movements and blinks is described. The single channel data are decomposed into statistically-independent components using the jadeR ICA algorithm preceded by a dynamical embedding matrix with dimensions dictated by the stimulus rate and sampling frequency. The PERG signal of interest is identified and scaled using a marker feature incorporated into the original record. The performance for artificial and real clinical data is presented.