General Electric Company

Ge Precision Healthcare LLC

Aalborg Universitet

Joel Quizhen Xue

Gordon Ian Rowlandson

Robert Edward Henderson

Long Yu

Chepur Sudakar Rao

Brian Joseph Young

Johannes Jan Struijk

Claus Graff

Mads Peter Andersen

Weijia Lu

Thomas Bork Hardahl

Egon Toft

Jie Shuai

Mark Gilbert Langer

Robert Farrell

Shuyan Gu

Lifei Hu

Methods and systems are provided for automatically diagnosing a patient based on a reduced lead electrocardiogram (ECG), using one or more deep neural networks. In one embodiment, a method for automatically diagnosing a patient using a reduced lead ECG comprises, acquiring reduced lead ECG data, wherein the reduced lead ECG data comprises less than twelve lead signals, determining a type of each of the less than twelve lead signals, selecting a deep neural network based on the type of each of the less than twelve lead signals, and mapping the less than twelve lead signals to a diagnosis using the deep neural network. In this way, reduced lead ECG data may be mapped to a diagnosis using an intelligently selected deep neural network, wherein the deep neural network was trained on reduced lead ECG data comprising a same set of ECG lead types as the acquired reduced lead ECG data.

Systems and methods for reduced lead electrocardiogram diagnosis using deep neural networks and rule-based systems

Methods and systems are provided for automatically diagnosing an electrocardiogram (ECG) using a hybrid system comprising a rule-based system and one or more deep neural networks. In one embodiment, by mapping ECG data to a plurality of features using a convolutional neural network, mapping the plurality of features to a preliminary diagnosis using a decision network, and determining a diagnosis based on the ECG data and the preliminary diagnosis using the rule-based system, a more accurate diagnosis may be determined. In another example, by incorporating both a rule-based system and one or more deep neural networks into the hybrid system, the hybrid system may be more easily adapted for use in various contexts/communities, as the one or more deep learning networks may be trained using context/community specific ECG data.

Systems and methods for electrocardiogram diagnosis using deep neural networks and rule-based systems

A system for identifying arrhythmias based on cardiac waveforms includes a storage system storing a trained deep neural network system, wherein the trained deep neural system includes a trained representation neural network and a trained classifier neural network. A processing system is communicatively connected to the storage system and configured to receive cardiac waveform data for a patient, identify a time segment in the cardiac waveform data, and transform the time segment into a spectrum image. The processing system is further configured to generate, with the representation neural network, a latent representation from the spectrum image, and then to generate, with the classifier neural network, an arrhythmia classifier from the latent representation.

System and method for identifying cardiac arrhythmias with deep neural networks

A system for processing ECG recordings from multiple patients includes a preprocessing database containing unprocessed ECG records from multiple patients, a reporting database containing processed ECG records from multiple patients, a processor, and a triage module executable on the processor to assess each of the unprocessed ECG recordings from the multiple patients. The triage module is executable to detect a presence or absence of one or more known abnormalities and determines at least one abnormality identifier based on the detected known abnormality. One or more abnormality groups are then identified based on the abnormality identifiers. A normal group of ECG recordings from multiple patients is then identified from those ECG recordings that are not in the abnormality group. The normal group of ECG recordings is then stored in the reporting database then associated a normal identifier.

System and method for processing ECG recordings from multiple patients

A system for processing ECG to detect atrial fibrillation includes three software modules. A beat module is executable on a processor to receive a time series of ECG data, identify heart beats, and determine a beat AFIB value based on a timing of each identified heart beat. The beat AFIB value represents a presence or absence of AFIB based on variability in the timing of each identified heart beat. A segment module is executable to receive the time series of ECG data, divide the time series of ECG data into two or more time segments, and determine a segment AFIB value for each time segment. The segment AFIB value indicates a presence or absence of AFIB in the time segment based on whether any of a set of rhythms are identified. The AFIB detection module is executable to determine an AFIB identification value for each time segment based on the beat AFIB value during that time segment and the segment AFIB value for that time segment.

System and method for detecting atrial fibrillation

A cluster database includes existing ECG datasets organized into clusters, wherein each existing ECG dataset includes an existing ECG waveform with at least one corresponding existing feature and existing interpretation. Each cluster is comprised of existing ECG datasets having a common existing feature. The cluster training module is executable by the processor to receive a new ECG waveform and a feature extracted from the new ECG waveform. The cluster training module then selects a cluster interpretation module based on the feature, wherein the cluster interpretation module is trained on one of the clusters from the cluster database. The cluster training module processes the new ECG waveform and/or the feature to provide a cluster interpretation output. The cluster interpretation output is then displayed on the user interface, and the cluster training module receives clinician input via the user interface accepting or rejecting the cluster interpretation output.

System and method of adaptive interpretation of ECG waveforms

A method of directing positioning of ECG electrodes on a patient includes receiving at a processor an image of the patient with one or more electrodes and determining with the processor an actual location of each of the electrodes on the patient based on the image. The method further includes determining with the processor whether the actual location of each of the electrodes is correct and providing information via a user interface regarding the actual location of the electrodes.

Method and system of directing positioning of ECG electrodes

A method of analyzing electrocardiograph (ECG) data includes receiving a first representative ECG of a patient and isolating a first principal component, a second principal component, and a third principal component of the first representative ECG. The principal components are isolated by selecting a portion of the first representative ECG relating to depolarization, calculating a covariance matrix based on the portion of the first representative ECG, conducting a principal component analysis of the covariance matrix, and selecting a first component of the principal component analysis as the first principal component, the second component of the principal component analysis as the second principal component, and the third component of the principal component analysis as the third principal component. A depolarization subspace is then formed based on the first principal component, second principal component, and the third principal component of the first representative ECG.

Method and apparatus for comparing and visualizing serial ECG data

A system and apparatus for obtaining physiological data from a patient. The system and apparatus comprising at least one electrode disposed to collect physiological data from the patient and an electrode connection device having a conductive array formed by a plurality of conductive regions and a plurality of nonconductive regions. The conductive regions are suitable to be connected to the at least one electrode. The electrode may be connected to a first conductive region of the electrode communication device and wherein upon movement by the patient or the conductive array, the electrode is connected to a second conductive region of the conductive array.

System and apparatus for collecting physiological signals from a plurality of electrodes

A system and a method for analysis of ECG curvature, which system involves a mathematical analysis comprising the following features where a number of different parameters are isolated and stored in a computer, where a first number of parameters are selected from at least one main group, where the selected parameters are combined in at least a first mathematical analysis. It is the object of the invention to improve mathematical analysis of ECG curvature particular for complex parts of the curvature such as notches or concavities. This can be achieved by selecting parameters from groups of symmetry, flatness, duration and/or complexity, using the parameters as input to an algorithm, where concavity intervals are evaluated on subsegments of the ECG segment and form the basis of up and downwards concavity quantification, where the system based on an algorithm detects and quantifies concavity on ECG curvatures.

Growing community of inventors

Germantown, WI, United States of America

Joel Quizhen Xue