Systems and methods for linear-time clustering for bounded, repeatable, rare events in physiological signals

Abstract

Described herein are methods and systems for clustering events in temporal signals, such as physiological signals. In some embodiments, a method for identifying events of interest from noisy physiological signal data is provided, the method comprising: receiving physiological signal data; generating filtered signal using a filter defined by a spectral band; generating signal peak data by identifying, for each of a plurality of time windows, a signal peak in the filtered data; grouping waveforms corresponding to the time windows based on the signal peak data to generate clustered event data; determining parameters for a noise beta distribution; identifying a cluster of waveforms that does not fall within the noise beta distribution; generating a graphical representation based on the waveforms in the cluster; and displaying the graphical representation.