Biosignal quality detection: An essential feature for unsupervised telehealth applications

We propose a system architecture for unsupervised telehealth applications in which routine, remote monitoring of patient clinical measurements are performed. It is argued that biosignal quality detection is a fundamental process that must be adapted from existing supervised recording environments and added to telehealth architectures in order for such systems to provide usable longitudinal records of vital sign parameters. Biosignal detection approaches in unsupervised environments must examine both overall waveform quality, which could be associated with excessive artifact contamination of the recorded clinical measurement data, and also signal features that would indicate poor measurement technique. As illustrative examples, data are taken from previous studies, including 4751 single lead-I electrocardiogram (ECG) recordings and similar raw data waveforms of pulse oximetry, and auscultatory and oscillometric blood pressure. These approaches can then be used to attach a confidence weighting to the parameters that are reported in the longitudinal electronic health record, in order to help in the rejection of outliers and false trends.