The Statistics of EEG Unipolar References: Derivations and Properties

• DOI: 10.1007/s10548-019-00706-y
• Publication Date: 2019-04-10T19:14:41Z
• Authors: Shiang Hu; Dezhong Yao; Maria L. Bringas-...; Yun Qin; Pedro A. Valdes-Sosa
• Keywords: Rest (music); Reference model

In this brief communication, which complements the EEG reference review (Yao et al. in Brain Topogr, 2019), we provide mathematical derivations that show: (1) any admits general form of a linear transformation ideal multichannel potentials with to infinity; (2) average (AR), electrode standardization technique (REST), and its regularized version (rREST) are solving inverse problems can be derived from both maximum likelihood estimate (MLE) Bayesian theory; however, REST is based on more informative prior/constraint volume conduction than AR; (3) show for first time also unipolar (UR), allowing us define family URs unified notations; (4) some notable properties 'no memory', 'rank deficient by 1', 'orthogonal projector centering'; (5) point out here, time, rREST provides optimal interpolating function used when channel missing or 'bad' channels rejected. The imply that: (a) two transform each other referencing multiple times will not accumulate artifacts; (b) whatever data was previously transformed with, minimum norm solution problem AR without modeling conduction, respectively; (c) MLE theory theoretical optimality REST. advantages limitations discussed guide readers their proper use.