Techniques for detecting atrial fibrillation

Inventors

Kenttä, Tuomas Viljami • Kent, Steve • Zhang, Xi

Assignees

Oura Health Oy

Abstract

Methods, systems, and devices for atrial fibrillation (AFib) detection are described. A method may include acquiring physiological data collected from a user via a wearable ring device, the physiological data including temperature data and motion data, and identifying that a measurement trigger condition for AFib has been satisfied based on the temperature data satisfying a temperature threshold and the motion data satisfying a motion threshold. The method may include sampling photoplethysmogram (PPG) data for the user via the wearable ring device based on identifying satisfaction of the measurement trigger condition, and classifying one or more AFib states based on the sampled PPG data. The method may further include causing a GUI of a user device to display an indication of the one or more AFib states.

Core Innovation

A wearable ring device includes a ring-shaped housing having an inner curved surface and an outer curved surface configured to contact tissue of a finger. One or more temperature sensors and one or more photoplethysmogram (PPG) sensors are arranged on the inner curved surface, and one or more processors disposed at least partially within the ring-shaped housing generate physiological data based at least in part on light received by the PPG sensors. A communication module transmits the physiological data.

The device identifies that a measurement trigger condition for atrial fibrillation is satisfied based on one or more values of temperature data being above a temperature threshold and one or more values of motion data being below a motion threshold. Based on satisfying the measurement trigger condition, the processors supply a first power level to the PPG sensors so that the PPG sensors emit light using the supplied first power level to sample PPG data for one or more time periods, and the sampled PPG data is provided as PPG pulses to one or more first machine learning classifiers.

The processors classify one or more atrial fibrillation states based at least in part on one or more characteristics of a set of PPG pulses, where the atrial fibrillation states are indicative of a relative risk of atrial fibrillation occurrences for the user. A user device receives an indication of one or more behaviors of the user associated with triggering of the atrial fibrillation occurrences, and one or more second machine learning identifiers predict one or more atrial fibrillation occurrences trained according to the indication of the one or more behaviors. A graphical user interface causes display of an indication of the one or more atrial fibrillation states, an indication of the one or more atrial fibrillation occurrences, or both.

Claims Coverage

The independent claims include three categories: a method, a system, and a non-transitory computer-readable medium. Across these independent claims, four main inventive features are used: threshold-based measurement triggering using temperature and motion, power-controlled PPG sampling, machine-learning classification of atrial fibrillation states from PPG pulse characteristics, and behavior-based prediction with graphical user interface output.

The independent claim set covers a wearable ring acquisition and sensing architecture using temperature and PPG with motion-based triggering, followed by power-controlled PPG sampling and machine-learning classification of atrial fibrillation states from PPG pulse characteristics. It further covers behavior-input-based prediction of atrial fibrillation occurrences and graphical user interface display of the atrial fibrillation state and/or occurrence indication.

Stated Advantages

Documented Applications