US-12616422-B2 - Prediabetes detection system and method based on combination of electrocardiogram and electroencephalogram information

US12616422B2US 12616422 B2US12616422 B2US 12616422B2US-12616422-B2

Abstract

A prediabetes detection system and method based on combination of electrocardiogram and electroencephalogram information are provided. The system includes: a signal obtaining module, configured to obtain an electrocardiogram signal and an electroencephalogram signal of a user in a noninvasive manner; a feature extraction module, configured to: perform dimension reduction processing on a combined feature set composed of an electrocardiogram feature and an electroencephalogram feature to obtain a plurality of dimension-reduced combined feature sets, and select an electrocardiogram feature and an electroencephalogram feature meeting a preset criteria of correlation by analyzing a correlation between the plurality of dimension-reduced combined feature sets and a blood glucose concentration value to constitute an optimized combined feature set; and a multimodal fusion module, configured to input the optimized combined feature set into a plurality of trained neural network models, to obtain a detection result by fusing results of the plurality of neural networks.

Inventors

Zedong NIE
Jingzhen Li
Yuhang Liu

Assignees

SHENZHEN INSTITUTES OF ADVANCED TECHNOLOGY CHINESE ACADEMY OF SCIENCES

Dates

Publication Date: 20260505
Application Date: 20201113
Priority Date: 20200529

Claims (5)

1 . A non-invasive prediabetes detection method based on combination of electrocardiogram and electroencephalogram information, comprising steps of: S 1 , during an oral glucose tolerance test, synchronously obtaining an electrocardiogram (ECG) signal and an electroencephalogram (EEG) signal of a user in a noninvasive manner by utilizing a wearable device; S 2 , performing dimension reduction processing on a combined feature set composed of ECG features and EEG features in various ways to obtain a plurality of dimension-reduced combined feature sets, and select the ECG features and the EEG features meeting a preset criteria of correlation by analyzing a correlation between the plurality of dimension-reduced combined feature sets and a blood glucose concentration value to constitute an optimized combined feature set; and S 3 , respectively inputting the optimized combined feature set into a plurality of trained neural network models, obtaining from each trained neural network model a classification output that classifies the user as prediabetic or non-prediabetic, and determining that the user is prediabetic when a majority of the classification outputs classify the user as prediabetic, and Outputting, via voice or text, a determination indicating whether the user is prediabetic or non-prediabetic; wherein S 1 comprise substeps of: placing six ECG electrodes V 1 to V 6 configured to monitor the ECG signal onto the user's chest, wherein the ECG electrode V 1 is placed in a fourth intercostal space at a right border of a sternum, the ECG electrode V 2 is placed in a fourth intercostal space at a left border of the sternum, and the ECG electrode V 3 is placed in a midpoint of a connecting line between the ECG electrode V 2 and the ECG electrode V 4 , the ECG electrode V 4 is placed at an intersection between a left mid-clavicular line and a fifth intercostal space, the ECG electrode V 5 is parallel to an anterior axillary line, and the ECG electrode V 6 is parallel to a midaxillary line; wearing an EEG electrode cap on the user's head, six EEG electrodes configured to monitor the EEG signal being provided in the EEG electrode cap, wherein the six EEG electrodes are respectively corresponding to a frontal lobe, an occipital lobe and a parietal lobe of a left hemisphere of a brain, and a frontal lobe, an occipital lobe and a parietal lobe of a right hemisphere of the brain; and carrying out the oral glucose tolerance test, and starting an ECG collection device and an EEG collection device to synchronously obtain the ECG signal and the EEG signal of the user: wherein S 2 comprising substeps of: extracting, from the ECG signal, feature information of a plurality of different segments, and respectively extract, from the EEG signal, EEG feature information of different frequency bands corresponding to different positions of a brain, to constitute the combined feature set; performing dimension reduction processing on the combined feature set based on a principal component analysis to obtain a first combined feature set; performing dimension reduction processing on the combined feature set based on an independent component analysis to obtain a second combined feature set; performing dimension reduction processing on the combined feature set based on a lasso regression analysis to obtain a third combined feature set; and analyzing the correlation between the blood glucose concentration and the first combined feature set, the second combined feature set, and the third combined feature set respectively, and then screen out the ECG features and the EEG features in the first combined feature set, the second combined feature set, and the third combined feature set that meet the preset criteria of correlation, to constitute the optimized combined feature set.
2 . The non-invasive prediabetes detection method based on combination of electrocardiogram and electroencephalogram information according to claim 1 , wherein the correlation is analyzed based on a Pearson correlation analysis, and the preset criteria of correlation is set as correlation k>0.2 and P≤0.05, P representing a probability of performing hypothesis testing on a correlation coefficient.
3 . The non-invasive prediabetes detection method based on combination of electrocardiogram and electroencephalogram information according to claim 1 , wherein the performing dimension reduction processing on the combined feature set based on the principal component analysis comprises: calculating a covariance matrix of a feature point of each feature in the combined feature set; and calculating eigenvectors of the covariance matrix and eigenvalues corresponding to the eigenvectors: sorting the eigenvectors according to magnitudes of the eigenvalues to form a matrix u=[u 1 , u 2 , u 3 , . . . , u n ], the corresponding eigenvalues being λ 1 , λ 2 , λ 3 , . . . , λ n in descending order, and intercepting, from the matrix u, a certain proportion of top-ranked eigenvalues as new feature points of each feature to achieve data dimension reduction.
4 . The non-invasive prediabetes detection method based on combination of electrocardiogram and electroencephalogram information according to claim 1 , wherein the plurality of types of trained neural network models comprise at least two types of a convolutional neural network, a long-short term memory network, and a recurrent neural network.
5 . The non-invasive prediabetes detection method based on combination of electrocardiogram and electroencephalogram information according to claim 1 , wherein the determination that the user is prediabetic, further classifies the user as suitable for intervention measures to reduce the risk of progression to diabetes.

Description

CROSS REFERENCE TO THE RELATED APPLICATIONS This application is the national stage entry of International Application No. PCT/CN2020/128559, filed on Nov. 13, 2020, which is based upon and claims priority to Chinese Patent Application No. 202010475003.6, filed on May 29, 2020, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD The present disclosure relates to the field of medical & healthcare technologies, and more particularly, to a prediabetes detection system and method based on combination of electrocardiogram and electroencephalogram information. BACKGROUND There exists a period named as prediabetes in the process of gradually developing from the healthy population with normal blood glucose levels into the diabetic population. The prediabetes refers to a period of impaired blood glucose regulation functions, including impaired fasting blood glucose and impaired glucose tolerance, but having not yet reached diagnostic criteria for diabetes. According to statistics, about 25% of young people and about 20% of adolescents have the prediabetes. About 10% of prediabetic will progress to diabetes every year if no intervention is made. However, if corresponding measures are adopted to intervene in time in the prediabetes phase by, for example, taking drugs, controlling diets, and strengthening exercise, the risk of developing into the diabetes can be reduced by 30%-75%, and the probability of returning to the normal blood glucose levels can rise up to about 70%. Therefore, detection of the prediabetes has great significance. Existing methods for determining the prediabetes include: 1) Obtaining a blood glucose concentration by blood sampling on an empty stomach. It may be determined as the prediabetes if a fasting blood glucose value ranges between 5.6 mmol/L and 7.0 mmol/L. 2) Carrying out oral glucose tolerance test. Two hours after taking oral glucose, the blood glucose concentration is obtained by blood sampling, and it is determined as the prediabetes if the blood glucose value ranges between 7.8 mmol/L and 11.1 mmol/L. However, all the existing technical solutions need to collect venous blood or fingertip blood, which may cause greater pain and risk of infection to patients, and detection costs are relatively high. SUMMARY An objective of the present disclosure is to overcome the above defects of the existing technologies by providing a prediabetes detection method based on combination of electrocardiogram and electroencephalogram information. According to this method, detection of prediabetes is implemented by carrying out oral glucose tolerance tests, by synchronously obtaining electrocardiogram and electroencephalogram information using a wearable device, and then by extracting related electrocardiogram and electroencephalogram features. According to a first aspect of the present disclosure, there is provided a prediabetes detection system based on combination of electrocardiogram and electroencephalogram information. The system includes: a signal obtaining module, configured to synchronously obtain an electrocardiogram (ECG) signal and an electroencephalogram (EEG) signal of a user in a noninvasive manner by utilizing a wearable device;a feature extraction module, configured to: perform dimension reduction processing on a combined feature set composed of an ECG feature and an EEG feature in various ways to obtain a plurality of dimension-reduced combined feature sets, and select the ECG feature and the EEG feature meeting a preset criteria of correlation by analyzing a correlation between the plurality of dimension-reduced combined feature sets and a blood glucose concentration value to constitute an optimized combined feature set; anda multimodal fusion module, configured to respectively input the optimized combined feature set into a plurality of types of trained neural network models, to obtain a detection result indicating whether the user is a prediabetic by fusing output results of the plurality of types of neural networks. In one embodiment, the synchronously obtaining an electrocardiogram signal and an electroencephalogram signal of a user includes: placing six ECG electrodes V1 to V6 configured to monitor the ECG signal onto the user's chest, wherein the ECG electrode V1 is placed in a fourth intercostal space at a right border of a sternum, the ECG electrode V2 is placed in a fourth intercostal space at a left border of the sternum, and the ECG electrode V3 is placed in a midpoint of a connecting line between the ECG electrode V2 and the ECG electrode V4, the ECG electrode V4 is placed at an intersection between a left mid-clavicular line and a fifth intercostal space, the ECG electrode V5 is parallel to an anterior axillary line, and the ECG electrode V6 is parallel to a midaxillary line;wearing an EEG electrode cap on the user's head, six EEG electrodes configured to monitor the EEG signal being provided in the EEG electrode cap, wherein the six EEG electr