US-20260123873-A1 - SYSTEM FOR PREDICTING HEALTH CONDITION BASED ON DEEP LEARNING BY USING PLURALITY OF ELECTROCARDIOGRAMS

Abstract

The present invention discloses a system for predicting a health condition based on deep learning by using a plurality of electrocardiograms according to an embodiment of the present invention includes: an electrocardiogram measurement unit ( 110 ) configured to acquire N pieces of electrocardiogram data by measuring electrocardiograms of the same examinee at different times; a prediction unit ( 120 ) configured to generate disease information predicting the presence/absence, progress and degree of a disease by integrating semantic feature values or output values from the N pieces of electrocardiogram data provided from the electrocardiogram measurement unit ( 110 ) through a diagnostic algorithm previously constructed by being trained on datasets of N pieces of electrocardiogram data at different times and diseases corresponding to the electrocardiogram data; and an input unit ( 130 ) configured to convert the N pieces of electrocardiogram data input from the electrocardiogram measurement unit ( 110 ) and input them to the prediction unit ( 120 ).

Inventors

• Joon Myoung KWON

Assignees

• MEDICAL AI CO., LTD.

Dates

Publication Date

20260507

Application Date

20220818

Claims (6)

1. 1 . A system for predicting a health condition based on deep learning by using a plurality of electrocardiograms, the system comprising: an electrocardiogram measurement unit configured to acquire N pieces of electrocardiogram data by measuring electrocardiograms of a same examinee at different times having a time difference therebetween; a prediction unit configured to generate disease information predicting a presence/absence, progress and degree of a disease by integrating semantic feature values or output values from the N pieces of electrocardiogram data provided from the electrocardiogram measurement unit through a diagnostic algorithm previously constructed by being trained on datasets of N pieces of electrocardiogram data at different times and diseases corresponding to the electrocardiogram data; and an input unit configured to convert the N pieces of electrocardiogram data input from the electrocardiogram measurement unit and input them to the prediction unit.
2. 2 . The system of claim 1 , wherein the input unit integrates the N pieces of electrocardiogram data, input from the electrocardiogram measurement unit, into a single piece of electrocardiogram data.
3. 3 . The system of claim 2 , wherein the input unit integrates the N pieces of electrocardiogram data on a per electrocardiogram data unit basis.
4. 4 . The system of claim 2 , wherein the input unit classifies the N pieces of electrocardiogram data for respective leads, and integrates the electrocardiogram data, classified for the respective leads, for the respective leads.
5. 5 . The system of claim 2 , wherein the input unit classifies the N pieces of electrocardiogram data for respective leads, divides them into bit units, and pairs and integrates bit units for each same lead.
6. 6 . The system of claim 1 , wherein the prediction unit generates the disease information by extracting semantic feature values from the N pieces of electrocardiogram data provided from the input unit, analyzing the semantic feature values in an integrated manner, and comparing and analyzing the N pieces of electrocardiogram data.

Description

TECHNICAL FIELD The present invention relates to a system for predicting a health condition based on deep learning by using a plurality of electrocardiograms that may improve the accuracy of measurement, diagnosis, examination, and prediction of a health condition by comparing and analyzing a plurality of electrocardiograms of the same examinee, taken at different times, based on deep learning and may analyze and predict the presence/absence, progress, and degree of disease in a time-series manner. BACKGROUND ART As is well known, an electrocardiogram is a graphical record of electrical potentials related to heartbeat on the surface of a human body. Electrocardiograms include not only standard 12-lead electrocardiograms but also exercise stress electrocardiograms and Holter electrocardiograms. Such electrocardiograms are used for the examination and diagnosis of circulatory diseases, and have the advantages of being simple, relatively inexpensive, non-invasive, and easily recorded repeatedly. Meanwhile, as for standard 12-lead electrocardiograms used in hospitals, six electrodes are attached to the front of the chest, three electrodes are attached to the limbs, 12-lead electrocardiogram information is all collected and determined in an integrated manner, and then a disease is diagnosed. In this case, 12-lead electrocardiograms record electrical potentials of the heart in 12 electrical directions centered on the heart. Through this, the disease of the heart confined to one area can be diagnosed. However, changes in the electrocardiogram do not apply equally to all examinees. In other words, the shape of an electrocardiogram may appear like an examinee has a disease even when he or she does not have the disease at all. Although another examinee has a disease, the shape of his/her electrocardiogram may appear like he or she does not have the disease. Therefore, there is a demand for technology that can improve the accuracy of measurement, diagnosis, examination, and prediction of a health condition by comparing and analyzing a plurality of electrocardiograms of the same examinee taken at different times. DISCLOSURE Technical Problem The technical problem to be overcome by the spirit of the present invention is to provide a system for predicting a health condition based on deep learning by using a plurality of electrocardiograms that may improve the accuracy of measurement, diagnosis, examination, and prediction of a health condition by comparing and analyzing a plurality of electrocardiograms of the same examinee, taken at different times, based on deep learning. Technical Solution In order to overcome the above-described problem, an embodiment of the present invention provides a system for predicting a health condition based on deep learning by using a plurality of electrocardiograms, the system including: an electrocardiogram measurement unit configured to acquire N pieces of electrocardiogram data by measuring electrocardiograms of the same examinee at different times having a time difference therebetween; a prediction unit configured to generate disease information predicting the presence/absence, progress and degree of a disease by integrating semantic feature values or output values from the N pieces of electrocardiogram data provided from the electrocardiogram measurement unit through a diagnostic algorithm previously constructed by being trained on datasets of N pieces of electrocardiogram data at different times and diseases corresponding to the electrocardiogram data; and an input unit configured to convert the N pieces of electrocardiogram data input from the electrocardiogram measurement unit and input them to the prediction unit. In this case, the input unit may integrate the N pieces of electrocardiogram data, input from the electrocardiogram measurement unit, into a single piece of electrocardiogram data. Furthermore, the input unit may integrate the N pieces of electrocardiogram data on a per electrocardiogram data unit basis. Furthermore, the input unit may classify the N pieces of electrocardiogram data for respective leads, and may integrate the electrocardiogram data, classified for the respective leads, for the respective leads. Furthermore, the input unit may classify the N pieces of electrocardiogram data for respective leads, may divide them into bit units, and may pair and integrate bit units for each same lead. Moreover, the prediction unit may generate the disease information by extracting semantic feature values from the N pieces of electrocardiogram data provided from the input unit, analyzing the semantic feature values in an integrated manner, and comparing and analyzing the N pieces of electrocardiogram data. Advantageous Effects According to the present invention, there are effects in that the accuracy of measurement, diagnosis, examination, and prediction of a health condition may be improved by comparing and analyzing a plurality of electrocardiograms of the same examinee, ta