Cardio-kinetic cross-spectral density for assessment of sleep physiology

Inventors

Olivier, Laurence Richard

Assignees

Lifeq Global Ltd

Abstract

The invention comprises methods and systems capable of processing simultaneously measured cardiac and motion signals, selecting from said signals the optimal cardiac and motion signal combination for cross spectral density calculations and, hereafter, itemizing the shared information in terms of the individual constituent frequencies and the effect exerted upon said frequencies by common underlying physiological states. The information is passed through a classifier, where said information is used to quantify the underlying physiological states, exemplified by but not limited to, sleep stage, quality and/or disorders that affect sleep stability. Inferences made by the methods described, for example, can be employed to aid home based sleep analysis and automated screening of patients with sleep disorders.

Core Innovation

The invention relates to a computer-implemented method to assess sleep physiological states of a subject using simultaneous measured signals that contain degrees of cardiac activity and motion data. The method selects from a plurality of inputs at least two measured signals, performs a cross-spectral density calculation on the at least two measured signals, and determines the degree of shared activity over a range of frequencies in the at least two measured signals. The sleep physiological states include classified underlying physiological states of the subject, including different sleep stages and sleep-related conditions described in the document.

The method calculates a quality metric based on the quality of each of the at least two measured signals, where the quality metric is based on each signal itself. Using shared activity features derived from the cross-spectral density, the method classifies different underlying physiological states of the subject. The quality metric allows for conditional prediction of the different underlying physiological states to limit predictions to periods when signals of sufficient quality are measured or collected.

The invention further includes displaying or transmitting the classified underlying physiological states of the subject, implemented as a non-invasive system. In this system, a data acquisition device selects from a plurality of inputs at least two measured signals containing degrees of cardiac activity and motion data, and a computing device performs the cross-spectral density calculation, calculates the per-signal quality metric, and uses the shared activity features to classify underlying physiological states with quality-conditioned conditional prediction. Dependent aspects describe using spectral coherence with cross-spectral density and using the approach for specific sleep disorder assessment including narcolepsy by identifying an early or direct transition from wakefulness to REM sleep.

Claims Coverage

The document includes two independent claims, a method claim and a system claim. Across these claims, the core inventive coverage centers on cross-spectral density-based shared activity across frequencies between cardiac and motion signals, a per-signal quality metric used for conditional prediction, and classification with display or transmission of the classified sleep physiological states.

Overall, the independent claims are directed to selecting at least two measured signals representing cardiac activity and motion data, computing cross-spectral density to quantify shared activity across frequencies, computing a quality metric based on each signal itself to enable conditional prediction that limits classification to sufficient-quality periods, and displaying or transmitting the classified underlying physiological states.

Stated Advantages

Documented Applications