FIELD: data processing; medicine.
SUBSTANCE: invention relates to image processing, and more specifically to support of medical decisions based on analysis of medical images. Method of detecting cardiovascular diseases by ECG comprises steps of: obtaining a data file for ECG analysis, performed in 12 leads, containing an array of signal values for each lead; and analysing the obtained file using a pretrained model of classification of cardiovascular diseases by ECG, wherein the model is based on a neural network, which is assembled on a ResNET architecture with a small number of layers, the activation function used is the Gelu function, regularizer used is a combination of dropout regularizer with linear combination of l1 and l2, wherein the model is trained using contrastive training and a triple loss function, the reference metric used is an F1-measure.
EFFECT: reduced computational complexity, as well as high accuracy and speed of training.
5 cl, 2 dwg
Title | Year | Author | Number |
---|---|---|---|
METHOD OF GENERATING DATA FILE FOR ECG ANALYSIS | 2023 |
|
RU2822304C1 |
METHOD OF PROCESSING AND ANALYSING ELECTROCARDIOGRAM (ECG) DATA | 2023 |
|
RU2823433C1 |
METHOD FOR PREDICTING THE STATE OF CARDIO-VASCULAR SYSTEM | 2004 |
|
RU2257838C1 |
HARDWARE AND SOFTWARE COMPLEX FOR ELECTROCARDIOGRAPHIC MEASUREMENTS | 2020 |
|
RU2759404C1 |
METHOD AND SYSTEM FOR NONINVASIVE SCREENING PHYSIOLOGICAL PARAMETERS AND PATHOLOGY | 2016 |
|
RU2657384C2 |
METHOD AND SYSTEM FOR AUTOMATIC ECG ANALYSIS | 2020 |
|
RU2767157C2 |
METHOD OF FORMING TWO-DIMENSIONAL IMAGE OF BIOSIGNAL AND ANALYSIS THEREOF | 2013 |
|
RU2538938C2 |
METHOD AND SYSTEM FOR SUPPORTING MEDICAL DECISION MAKING USING MATHEMATICAL MODELS OF PRESENTING PATIENTS | 2017 |
|
RU2703679C2 |
METHOD OF STUDYING HEART FUNCTION USING WAVELET TRANSFORM OF ELECTROCARDIOGRAMS AND RAMAN SPECTROSCOPY | 2023 |
|
RU2820398C1 |
METHOD FOR COMPLEX ANALYSIS OF CARDIOVASCULAR BIOMECHANICS | 2018 |
|
RU2727748C2 |
Authors
Dates
2025-01-14—Published
2024-05-30—Filed