On the use of single-channels for sensing multisource activity in biomedical signals

This paper presents a framework which allows for the extraction of information from just single-channel measurements of biomedical signals. Such a method is vital as an intelligent preprocessing stage to the transmission of such information, for example, in the remote sensing of biomedical signals. The framework consists of two fundamental aspects; the first is a dynamical embedding step which provides a representation of multidimensional data from just a single channel recording; this is then followed by a variant of standard independent component analysis known as constrained independent component analysis. This latter method allows for the extraction of one of many sources underlying the measurement space, through the provision of a basic reference signal. The reference signal, or constraint, can be changed to extract different sources from the measured data - if they exist. Although the method can be applied to a whole range of biomedical signals, it is demonstrated here in the context of the extraction of seizure information from a single channel of pre-recorded (multichannel) epileptiform EEG. It can be seen that seizure information can be extracted from a single channel EEG recording and the sensitivity of the location of the measurement channel used (in relation to the focus of the epileptiform activity) is reduced considerably.