Emory University

Georgia Tech Research Corporation

Children's Healthcare of Atlanta, Inc.

Universita′ Degli Studi Di Udine

Sanguina, Inc.

Gari Clifford

Amit Shah

Erik Reinertsen

Shamim Nemati

Wilbur A Lam

Cecelia Manzanares

Erika Amy Tyburski

LaVonda Brown

Robert Guy Mannino

Giulia Di Poian

Roberto Rinaldo

Ayse Cakmak

Supreeth Prajwal Shashikumar

Qiao Li

Rafi Haque

Allan Levey

The systems and methods can accurately and efficiently determine abnormal cardiac activity from motion data and/or cardiac data using techniques that can be used for long-term monitoring of a patient. In some embodiments, the method for using machine learning to determine abnormal cardiac activity may include receiving one or more periods of time of cardiac data and motion data for a subject. The method may include applying a trained deep learning architecture to each tensor of the one or more periods of time to classify each window and/or each period into one or more classes using at least the one or more signal quality indices for the cardiac data and the motion data and cardiovascular features. The deep learning architecture may include a convolutional neural network, a bidirectional recurrent neural network, and an attention network. The one or more classes may include abnormal cardiac activity and normal cardiac activity.

Methods and systems for determining abnormal cardiac activity

Techniques are provided for determining a qualitative, quantitative and/or categorical assessment of one or more users and/or images with respect to one or more populations. The eye movement data of the user may be obtained with respect to each image of the one or more images displayed for a period of time. One or more memory performance measures and/or one or more salience performance measures may be determined using the eye movement data with respect to the one or more regions of the one or more images for one or more of predetermined time ranges of the period of time. The quantitative, qualitative and/or categorical assessment of the user and/or images presented may be determined with respect to one or more populations, using the one or more memory performance measures and/or one or more salience performance measures.

Systems and methods for automated passive assessment of visuospatial memory and/or salience

A smartphone-based hemoglobin (Hgb) assessment application quantitatively analyzes pallor in patient-sourced photos using image analysis algorithms to enable a noninvasive, accurate quantitative smartphone app for detecting anemia. A user takes a photo of his/her fingernail beds using the app and receives an accurate displayed Hgb level. Since fingernails do not contain melanocytes, the primary source of color of these anatomical features is blood Hgb. At the same time, quality control software minimizes the impact of common fingernail irregularities (e.g. leukonychia and camera flash reflection) on Hgb level measurement. Metadata recorded upon capturing the image is leveraged for determining a users' Hgb level thereby eliminating the need for external equipment. A personalized calibration of image data with measured Hgb levels improves the accuracy of the application.

Systems and methods for quantitative diagnosis of anemia

Methods and systems for monitoring of sensor data for processing by machine-learning models to generate event predictions to estimate a risk a medical event are provided. An electronic device or wearable smart device may monitor the output of various sensors to collect data related to a person's activity level, location changes, and communications and may use this information as input to a personalized trained machine-learning model to predict a likelihood of an event.

Passive data collection and use of machine-learning models for event prediction

Techniques are provided for generating and processing compressed sensor data. Sensor signals can be collected using one or more sensors. The sensor signals can be compressed using a compression data structure. In some instances, the compressed signal corresponds to a sampling rate at or below the Nyquist sampling rate. The compressed signal can be compared to one or more templates. Events within the compressed signal can be detected and characterized based on the comparison.

Methods and systems for automatically detecting events based on ECG signals determined from compressed sensed measurements

Systems, methods, and computer-readable media for classifying a PTSD status. In an embodiment, an example method for using a classifier can comprise receiving information from electrocardiography performed on an individual; determining features from the information; comparing the features to the a logistic regression classifier trained using features determined from median quiescent segments of RR interval information from individuals with and without PTSD, wherein the median quiescent segments are non-overlapping time periods of lowest median HR for each individual, and the features include one or more of the following: deceleration capacity (DC), low frequency (LF) power, very low frequency (VLF) power, and standard deviation of all normal RR intervals (SDNN); and determining a posttraumatic stress disorder (PTSD) status of the individual based on the comparison of the features to the classifier, wherein the PTSD status is a severity of PTSD based on a probability of PTSD.

Using heartrate information to classify PTSD

Growing community of inventors

Atlanta, GA, United States of America

Gari Clifford