KR20220097724A - Cardiovascular disease prediction system using A.I.-based cardiovascular disease prediction model for patients with SDB - Google Patents

Abstract

The present invention relates to a cardiovascular disease prediction system that predicts a future cardiovascular disease of a sleep breathing disorder patient by applying an electrocardiogram and a cardiovascular disease risk factor of the sleep breathing disorder patient to an artificial intelligence-based cardiovascular disease prediction model. In the cardiovascular disease prediction system, an operation processing part comprises: a signal-processed electrocardiogram feature extraction part; an artificial intelligence-based electrocardiogram feature extraction part; a CVD risk factor loading part; and a cardiovascular disease (CVD) prediction part.

Description

Cardiovascular disease prediction system using A.I.-based cardiovascular disease prediction model for patients with SDB}

The present invention provides a cardiovascular disease prediction model, which is an artificial intelligence (A.I.)-based algorithm using electrocardiogram and CVD risk factors, for predicting future cardiovascular disease in patients with sleep breathing disorder, and a cardiovascular disease prediction model using the same It relates to a disease prediction system.

Sleep-disordered breathing (SDB) is a partial or complete obstruction of the upper airway during sleep, resulting in apnea or hypoventilation of ventilation. Sleep breathing disorder is known to be 2-4% of the total population, and is a very common disease known to be 3.2-4.5% in Korea. Sleep apnea-hypoxia that occurs repeatedly during sleep causes repeated hypoxemia, reoxygenation, rapid changes in thoracic pressure, and arousal of the central nervous system, which acts as an acute stress factor in the cardiovascular system. Therefore, if sleep and breathing disorders are left untreated for a long time, the occurrence of hypertension, heart failure, myocardial infarction, arrhythmia, and stroke increases, which increases the risk of death from cardiovascular disease.

To diagnose sleep breathing disorder (SDB), polysomnography is usually performed. Polysomnography is a test that records and analyzes sleep states using various instruments to diagnose various abnormal conditions that occur during sleep. Polysomnography uses a variety of test equipment to diagnose diseases during sleep. Electroencephalography (EEG) to determine the state of brain function, safety test (EOG) to check eye movements, and electromyography to determine muscle status (EMG), an electrocardiogram (ECG) to check the heart rhythm, and video recording to see the overall condition, it is a method to test while sleeping, usually overnight.

The National Heart Lung & Blood Institute (NHLBI) conducted a multicenter cohort study (the Sleep Heart Health Study, SHHS) to determine the association between sleep breathing disorders and cardiovascular disease. Gottlieb et al. reported a prospective study examining the relationship between sleep and breathing disorders, coronary artery disease, and heart failure through SHHS reported to be In addition, Redline et al. investigated the relationship between sleep breathing disorder and stroke through a prospective study of SHHS, and reported that patients with mild-to-severe sleep breathing disorder showed a high correlation with ischemic stroke. In addition, studies on predictors of cardiovascular diseases such as cholesterol, blood pressure, obesity, smoking, and electrocardiogram have been actively conducted. Auer et al. reported that the ECG waveform is related to cardiovascular disease, and that ECG abnormality can be used as a predictor of coronary artery disease. However, in previous studies, only one disease was analyzed or group-based analysis such as disease morbidity or mortality was performed. Also, the usefulness of the cardiovascular disease predictor (CVD predictor) has not been evaluated whether it can actually predict future cardiovascular disease.

Therefore, there is a need for a cardiovascular disease prediction system that predicts cardiovascular diseases that will actually occur in the future as a cardiovascular disease predictor (CVD predictor).

Recently, interest in personalized medicine is increasing, and technology such as artificial intelligence and big data is applied in various fields to suggest the possibility of personalized disease diagnosis and prediction.

In the present invention, electrocardiogram and electrocardiogram for predicting future (for example, within the next 10 years) cardiovascular disease, for example, coronary heart disease, heart failure, and stroke in patients with sleep and breathing disorders We propose a cardiovascular disease prediction model, which is an artificial intelligence-based algorithm using CVD risk factors, and a cardiovascular disease prediction system to which it is applied.

As a prior art, Korean Patent Registration No. 10-1839910 discloses a subject's age, diabetes, hypertension, smoking, family history of coronary artery disease, blood pressure, cholesterol level, white blood cell level, creatine level, glycated hemoglobin level, and atrial fibrillation factors. obtaining each measurement value; assigning a risk prediction score corresponding to each acquired measurement value to each factor; And it relates to a method of predicting cardiovascular disease risk using cardiovascular disease risk factors comprising the; matching the total score by adding the risk prediction scores of each factor to the probability of developing cardiovascular disease within the prediction period. In the case of Korean Patent Registration No. 10-1839910, it only predicts the cardiovascular disease risk by adding up the predicted scores according to the cardiovascular disease risk factors, and there is some problem in accuracy.

The present invention provides a cardiovascular disease prediction system that predicts future cardiovascular disease of sleep breathing disorder patients by applying electrocardiogram and CVD risk factors of patients with sleep breathing disorder to an artificial intelligence-based cardiovascular disease prediction model will do

In order to solve the above problems, the present invention provides a cardiovascular disease prediction system, including an arithmetic processing unit for predicting the occurrence of cardiovascular disease using an electrocardiogram signal and CVD risk factor detected during polysomnography. In the calculation processing unit, from the electrocardiogram (ECG) signal, STTc (ST-segment and T-wave (ST-T) changes) to detect a time domain electrocardiogram parameters including segments, and low-frequency to high-frequency band intensity ratio (PLF) /PHF ratio), detects a frequency domain ECG parameter, and stores the average and standard deviation of the detected time domain ECG parameter and the frequency domain ECG parameter as 'signal processed ECG characteristics' in the memory unit. ECG feature extraction unit; An electrocardiogram signal for 30 seconds is input to the pre-trained CNN-based artificial intelligence model, and the average and standard deviation of the results of each node of the flatten layer in the CNN-based artificial intelligence model are obtained, an artificial intelligence-based electrocardiogram feature extraction unit, which is stored in the memory unit as 'artificial intelligence-based electrocardiogram features'; a CVD risk factor loading unit that reads the CVD risk factor stored in the memory unit; Including a cardiovascular disease (CVD) prediction model including servo vector machines (SVM), signal-processed ECG characteristics, AI-based ECG characteristics, and ECG characteristics including CVD risk factors are added to the cardiovascular disease prediction model From the cardiovascular disease prediction model, whether or not cardiovascular disease will occur in the future It is characterized in that it comprises a; cardiovascular disease (CVD) prediction unit for outputting the information as a cardiovascular disease prediction result.

The cardiovascular disease (CVD) prediction unit consists of a pre-learned first servot vector machine, and determines whether or not a cardiovascular disease will occur in the future. a cardiovascular disease (CVD) predictive value, the cardiovascular disease occurrence determination unit; If cardiovascular disease is determined to occur in the future by the cardiovascular disease identification unit, using a one-to-one (OvO, One-Vs-One) multi-class classifier, coronary heart disease (CHD) as a future cardiovascular disease ), heart failure (HF), and stroke (stroke) to select one of the cardiovascular disease (CVD) predictive value output, a detailed disease discrimination unit; includes a.

The detailed disease discrimination unit is a servo vector machine that has already been learned to discriminate one of coronary heart disease and heart failure, and selects one of coronary heart disease and heart failure and outputs it as a result of the second servo vector machine. boat vector machine; A third servo vector machine that is pre-trained to discriminate between coronary heart disease and stroke, and selects one of coronary heart disease and stroke and outputs it as a result of the third servo vector machine machine; a fourth-servo vector machine that is pre-learned to discriminate one of heart failure and stroke, and selects one of heart failure and stroke and outputs it as a result of the fourth-servo vector machine; A voting unit that counts and outputs the selected number of times each of coronary heart disease, heart failure, and stroke from the results of the second-servo vector machine, the third-servo vector machine, and the fourth-servo vector machine; include

The voting unit is configured to output a value for a disease having the highest number of selected coronary heart disease, heart failure, and stroke as a cardiovascular disease (CVD) predicted value.

The input vector of the first servo vector machine is the average of the results of the 15th node of the flatten layer, the average of the STTc segment, the BMI index, the systolic blood pressure, the AHI index, and 11 of the flatten layer. The standard deviation of the results of the second node, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the low back pain dysfunction index (ODI), the smoking level, the results of the 12th node of the flatten layer Average, the average of the results of the first node of the flatten layer, the standard deviation of the results of the 7th node of the flatten layer, the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio) Among the standard deviation, the standard deviation of the STTc segment, the mean of the low frequency band intensity (P _LF ), and the mean of the result of the third node of the flatten layer, 11 ECG features are included.

The input vector of the second servo vector machine is the average of the results of the 4th node of the flatten layer, the HDL cholesterol index, the average of the low frequency band intensity (P _LF ), and 10 of the flatten layer The average of the results of the th node, the standard deviation of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the standard deviation of the results of the 12th node of the flatten layer, and 6 electrocardiographic characteristics of the AHI index include

The third servo vector machine is the average of the results of the 10th node of the flatten layer, the average of the STTc segment, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), diastolic blood pressure, HDL cholesterol Includes three of the index, the average of the low frequency band strength (P _LF ), and the BMI index.

The input vector of the fourth servo vector machine is the HDL cholesterol index, the average of the results of the third node of the flatten layer, the standard deviation of the low frequency band intensity (P _LF ), the cholesterol index, the AHI index, the platen Mean of the results of the 2nd node of the flatten layer, systolic blood pressure, standard deviation of high frequency band intensity (P _HF ), the average of the results of the 4th node of the flatten layer, smoking degree, platen layer It contains 7 of the standard deviations of the result of the 6th node of the (flatten layer).

In addition, the present invention provides a method of driving a cardiovascular disease prediction system, comprising a calculation processing unit for predicting the occurrence of cardiovascular disease using an electrocardiogram signal and CVD risk factor detected during polysomnography, The operation processing unit detects, from the electrocardiogram (ECG) signal, a time domain electrocardiogram parameter including a STTc (ST-segment and T-wave (ST-T) changes) segment, and a low-frequency to high-frequency band intensity ratio (PLF/PHF ratio) ), and storing the average and standard deviation of the detected time domain ECG parameter and the frequency domain ECG parameter as 'signal processed ECG characteristics' in the memory unit, extracting signal-processed ECG features step; The operation processing unit inputs an electrocardiogram signal for 30 seconds to the previously learned CNN-based artificial intelligence model, and calculates the average and standard results of each node of the flatten layer in the CNN-based artificial intelligence model. An artificial intelligence-based ECG feature extraction step of obtaining the deviation and storing it in the memory unit as 'artificial intelligence-based ECG features'; The arithmetic processing unit, the CVD risk factor loading step of reading the CVD risk factor stored in the memory unit; The arithmetic processing unit uses a cardiovascular disease (CVD) prediction model including servo vector machines (SVM), and obtains the signal-processed ECG characteristics obtained in the ECG feature extraction step and the AI-based ECG characteristic extraction step. The ECG characteristics based on artificial intelligence and the ECG characteristics including the CVD risk factors read in the CVD risk factor loading stage are input into the cardiovascular disease prediction model, and the results output from the cardiovascular disease prediction model Cardiovascular disease ( CVD) predicting step; characterized in that it includes.

In the cardiovascular disease (CVD) prediction step, the arithmetic processing unit inputs an electrocardiogram characteristic including the average of STTc segments to the previously learned first servo vector machine as an input vector, and outputs the output value of the first servo vector machine, The value that determines whether or not cardiovascular disease will occur in the future is output. If the output value of the first servot vector machine determines that cardiovascular disease does not occur, the value indicating that cardiovascular disease does not occur is the cardiovascular disease (CVD) predicted value. outputting as, cardiovascular disease occurrence determination step; When the calculation processing unit determines that a cardiovascular disease will occur in the future, the cardiovascular disease occurrence determination unit uses a one-to-one (OvO, One-Vs-One) multi-class classifier, heart disease, CHD), heart failure (HF), and stroke (stroke) selected one of the cardiovascular disease (CVD) predictive value output, detailed disease identification step; includes.

In the detailed disease identification step, the operation processing unit inputs the electrocardiogram characteristics including the average of the low frequency band intensity (P _LF ) into the pre-learned second servo vector machine, and from the second servo vector machine, the second a second servo vector machine calculation step in which one of coronary heart disease and heart failure is output as a result of the servo vector machine; The arithmetic processing unit inputs the electrocardiogram characteristics including the average of the low frequency band intensities (P _LF ) into the pre-learned third servo vector machine, and the result of the third servo vector machine from the third servo vector machine a third servo vector machine calculation step in which one of coronary heart disease and stroke is output as a value; The arithmetic processing unit inputs the electrocardiogram characteristics including the standard deviation of the low frequency band intensity (P _LF ) to the pre-learned fourth servo vector machine, and from the fourth servo vector machine to the fourth servo vector machine. a fourth-servo vector machine in which one of heart failure and stroke is output as a result value; The arithmetic processing unit counts the number of times each of coronary heart disease, heart failure and stroke is selected from the result value of the second serve vector machine, the result value of the third serve vector machine, and the result value of the fourth serve vector machine and a voting step of outputting the counted values for each coronary heart disease, heart failure, and stroke.

In the voting stage, the calculation processing unit outputs the disease value with the most count values for coronary heart disease, heart failure, and stroke as cardiovascular disease (CVD) prediction values.

In addition, the present invention features a recording medium storing a computer program source for a method of driving a cardiovascular disease prediction system.

The cardiovascular disease prediction system of the present invention predicts the future cardiovascular disease of the sleep breathing disorder patient by applying the electrocardiogram and CVD risk factor of the sleep breathing disorder patient to an artificial intelligence-based cardiovascular disease prediction model.

Through this, it is possible to predict the occurrence of cardiovascular diseases, ie, coronary heart disease, heart failure, and stroke, within the next 10 years using the electrocardiogram and CVD risk factors of patients with sleep breathing disorder. It is possible to enable personalized medical services by predicting the accompanying diseases of

1 is a block diagram schematically illustrating the configuration of a cardiovascular disease prediction system using an artificial intelligence-based cardiovascular disease prediction model of a patient with sleep breathing disorder of the present invention;
2 shows a cardiovascular disease (CVD) prediction model of the present invention.

Hereinafter, the cardiovascular disease prediction system using the artificial intelligence-based cardiovascular disease prediction model of the sleep respiratory disorder patient of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram schematically illustrating the configuration of a cardiovascular disease prediction system using an artificial intelligence-based cardiovascular disease prediction model of a patient with sleep breathing disorder of the present invention;

The data input unit 100 includes an electrocardiogram detector 110 and a cardiovascular disease risk factor data input unit 160 .

The electrocardiogram detector 110 includes an electrocardiogram sensor unit 120 , a signal preprocessor 130 , and an A/D converter 150 .

The electrocardiogram sensor unit 120 detects an electrocardiogram signal of a patient with sleep breathing disorder, and the signal preprocessor 130 performs signal preprocessing to remove noise and amplify the electrocardiogram signal detected from the electrocardiogram sensor unit 120, and A / The D converter 150 converts the electrocardiogram signal preprocessed by the signal preprocessor 130 into a digital signal.

The cardiovascular disease risk factor data input unit 160 is a means for inputting cardiovascular disease risk factor data.

The electrocardiogram detector 110 may be an electrocardiogram detector and a cardiovascular disease risk factor data input device provided in the polysomnography equipment.

Here, the electrocardiogram signal may be an electrocardiogram signal detected during the night for polysomnography.

The calculation processing unit 200 stores the electrocardiogram data and cardiovascular disease risk factor data input from the data input unit 100 in the memory unit 310 . The arithmetic processing unit is made up of a computer.

Signal-processed electrocardiogram (ECG) feature extraction unit 210, AI-based ECG features extraction unit 220, CVD risk factor loading unit 230, cardiovascular disease (CVD) prediction unit ( 250).

The signal-processed electrocardiogram (ECG) feature extraction unit 210 obtains QTc, STTc segment, SDNN, RMSSD, ultra-low-frequency band intensity (PVLF), low-frequency band intensity (PLF), high-frequency band intensity (PHF), and low-frequency high-frequency The average and standard deviation of each of the band intensity ratios (PLF/PHF ratio) are obtained and stored in the memory unit 310 as signal processing-based ECG features (referred to as 'SP-based ECG features'). do.

The AI-based ECG features extraction unit 220 is an ECG signal for 30 seconds as an input signal, and the CNN-based AI model (that is, for extracting AI-based ECG features) artificial intelligence model), and the CNN-based artificial intelligence model (that is, an artificial intelligence model for extracting electrocardiogram features) obtains the result of each node of the flatten layer (flatten layer, smoothing layer), and this result The average and standard deviation of each are obtained and stored in the memory unit 310 as AI-based ECG features.

The CVD risk factor loading unit 230 reads the CVD risk factor stored in the memory unit 310 .

The cardiovascular disease (CVD) prediction unit 250 includes a cardiovascular disease (CVD) prediction model.

The cardiovascular disease (CVD) predictive model, after undergoing a process to extract features, uses a support vector machine (SVM) model to predict coronary heart disease (CHD), heart failure (heart) within 10 years. Failure, HF) or stroke occurrence is predicted, and the prediction result is output to the output unit 320 .

Hereinafter, the operation of the arithmetic processing unit of the present invention will be described in detail.

In the present invention, future cardiovascular disease is predicted by using the electrocardiogram and CVD risk factor detected accompanying polysomnography.

To develop a predictive model for cardiovascular disease, signal processing-based ECG characteristics, AI-based ECG characteristics, and CVD risk factors were extracted.

During the experiment, polysomnography data from the baseline study were used for the electrocardiogram signal, and the electrocardiogram during sleep from sleep onset to sleep end was analyzed.

After removing baseline fluctuations and power source noise using a bandpass filter of 1.5 to 20 Hz, QRS complex and T wave were detected using the adaptive threshold algorithm and morphological method.

In general, the QRS complex (i.e., R wave) in the ECG signal exceeds a predetermined threshold (adaptive threshold) after the previous R peak. Let it be a peak (R point, R point), the minimum value immediately before the R peak is detected as a Q point (Q point), and the minimum value immediately after the R peak is detected as an S point (S point). Further, the peak following the S point is detected as the T point (T point). Since the QRS complex and the T wave detection method in the ECG signal are widely known, a detailed description thereof will be omitted.

After detecting the characteristic points of P, Q, R, S, T of ECG (electrocardiogram), QTc (corrected QT interval) and STTc segment (STTc segment, ST-segment) which are ECG characteristic parameters in the time domain and T-wave (ST-T) changes.

QTc is obtained by Equation (1).

Here, Ti, Qi, and Ri represent the time points of the i-th (ie, the i-th period) of the T point (T wave), the Q point (Q wave), and the R point, respectively, and Ri+1 is the i+1th R Indicates the point of time.

The STTc segment is obtained by Equation (2).

Here, the J point refers to the point where the QRS complex meets the ST segment.

The RR interval is expressed as Equation (3).

In this case, Ri-1 is the R point time point of the i-1th (ie, the i-1th cycle) ECG, and fs is the sampling frequency of the ECG signal.

To calculate the heart rate variability, the ectopic bit of RR was removed, and this signal was defined as NN (normal-to normal RR).

That is, the time interval between the R peak (ie, the normal R peak) and the consecutive previous R peaks (ie, the consecutive normal R peaks) is calculated, and this is the NN interval.

In addition, as an electrocardiogram characteristic parameter in the time domain, the square root of the mean (RMSSD) of the squared difference between the standard deviation (SDNN) of the entire RR interval and the adjacent RR interval may be further obtained.

The standard deviation (SDNN) of the entire RR interval is obtained as in Equation (4).

Here, N is the total heart rate, meanRRI is the average RR interval, and I(i) is the i-th RR interval.

The square root of the mean (RMSSD) of the squared difference between adjacent RR intervals is obtained as in Equation 5.

Here, I(i+1) refers to the i+1th RR interval.

For heart rate variability analysis, calculated NNs were interpolated at equal intervals and then resampled at 4 Hz. That is, when the number of samples in each RR interval is smaller than a predetermined number, interpolation is performed to fill in zeros, and the electrocardiogram interpolated in this way is re-sampling (re-sampling) at 4 Hz.

The resampled signal was subjected to a fast Fourier transform (FFT) every 30 seconds, and then the power spectrum density (PSD) was calculated by taking the square of the FFT. That is, for frequency analysis, fast Fourier transform (FFT) is performed on the interpolated and resampled ECG.

Each frequency band used to calculate the frequency domain features is very low frequency (VLF: 0~0.04 Hz), low frequency (LF: 0.04~0.15 Hz), high frequency (HF: 0.15~) 0.4 Hz).

In order to detect the ECG characteristic parameter in the frequency domain, the ultra-low frequency band intensity (P _VLF ), the low-frequency band intensity (P _LF ), and the high-frequency band intensity (P _HF ), which are the ECG characteristic parameters in the frequency domain, are obtained, and the low-frequency high-frequency band intensity is obtained. Detect the ratio (P _LF /P _HF ratio). The method of obtaining the frequency band intensity (P _LF ), the high frequency band intensity (P _HF ), and the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio) is widely known in Korean Patent Registration No. 10-0493714 and the like, so detailed description is omitted.

EKG characteristic parameters in the time domain of QTc, STTc segment, SDNN, and RMSSD, ultra-low-frequency band intensity (P _VLF ), low-frequency band intensity (P _LF ), and high-frequency band intensity (P _HF ) are obtained, and the low-frequency to high-frequency band intensity ratio is obtained. When the average and standard deviation of each ECG characteristic parameter in the frequency domain of (P _LF /P _HF ratio) are obtained, 18 ECG characteristic parameters (for convenience of explanation, 'signal processing-based ECG features, SP -based ECG features)) is detected.

Table 1 shows the signal-processed ECG characteristics used as input to the cardiovascular disease (CVD) prediction model of the present invention.

That is, signal processing-based ECG characteristics are extracted, and each characteristic (ie, QTc, STTc segment, SDNN, RMSSD, P _VLF , P _LF , P _HF , P _LF /P _HF ratio), calculates the average and standard deviation during the entire sleep period, and stores it in the memory unit as 'signal-processed ECG features (SP-based ECG features)'. This process can be referred to as the signal-processed ECG feature extraction step.

Next, a convolutional neural network (CNN)-based artificial intelligence model (ie, an artificial intelligence model for extracting ECG features) was designed to extract AI-based ECG features. The model receives an electrocardiogram signal for 30 seconds and undergoes batch normalization. Thereafter, features are extracted through a convolution layer and a pooling layer having a three-layer structure. The average and standard deviation of the AI-based features extracted from the nodes of the flatten layer (smoothing layer) of the CNN model learned in this way were calculated for the entire overnight sleep time.

That is, the ECG signal for 30 seconds inputs the input signal to the CNN-based AI model (that is, an AI model for extracting ECG features based on artificial intelligence), and the CNN-based AI model (that is, , an artificial intelligence model for extracting ECG features) outputs the extraction result of each node of the flatten layer (smoothing layer). Then, the average and standard deviation of each of the extraction results (15 pieces of data) are calculated, and the obtained data is stored in the memory unit as 'AI-based ECG features'. This process can be referred to as the AI-based ECG feature extraction step. Here, the data part of 'AI-based ECG features' is a total of 30 pieces of data.

Table 2 shows the AI-based ECG characteristics used as input to the cardiovascular disease (CVD) prediction model of the present invention.

In the present invention, in addition to the ECG characteristics of signal-processed ECG features (SP-based ECG features) and AI-based ECG features, in addition to the representative cardiovascular disease (CVD) risk factors, Use as input.

As CVD risk factors used in the present invention, smoking level, hypertension level (systolic blood pressure, diastolic blood pressure), cholesterol level (cholesterol index, HDL cholesterol index) obesity level (BMI index), age, A total of 10 data are used for gender, sleep breathing disorder (SDB) degree (AHI index, sleep apnea index), and low back pain dysfunction index (ODI).

Table 3 shows the CVD risk factors used as input to the cardiovascular disease (CVD) prediction model of the present invention.

That is, the arithmetic processing unit reads the CVD risk factor obtained during polysomnography and the like and stored in advance in the memory unit. This process can be referred to as the CVD risk factor step.

In this way, the inputs used in the cardiovascular disease (CVD) prediction model in the present invention are 18 signal-processed ECG characteristics, 30 AI-based ECG characteristics, and 10 CVD risk factors, a total of 58 input characteristic data. have

Next, these feature data were analyzed statistically, i.e., two independent sample t-tests and chi-square tests, where each feature between classes had a p-value <0.05. , select the feature data to be used as significantly different input data. This process can be referred to as the feature selection step.

In some cases, the feature selection step may be selected at the experimental stage and determined at the time of product shipment.

In the present invention, after undergoing a process for extracting features, coronary heart disease (CHD) within 10 years through a cardiovascular disease (CVD) prediction model using a support vector machine (SVM), It predicts the occurrence of heart failure (HF) and stroke.

2 shows a cardiovascular disease (CVD) prediction model of the present invention.

Before describing the cardiovascular disease (CVD) prediction model, ECG characteristic data applied as an input to each servot vector machine in the cardiovascular disease (CVD) prediction model of FIG. 2 is shown.

The cardiovascular disease occurrence determination unit 710 is made of the first servo vector machine (SVM_CVD), and inputs the ECG characteristic data selected in the characteristic selection step or the given ECG characteristic data, and determines whether a cardiovascular disease will occur in the future ( CVD), that is, whether or not (CVD-free) is determined. If the cardiovascular disease occurrence determining unit 710 determines that cardiovascular disease does not occur, it outputs a value indicating that cardiovascular disease does not occur (ie, CVD-free value), which is the final result of the cardiovascular disease (CVD) prediction model. output as a result.

The input vector of the first servo vector machine (SVM_CVD) is the average of the results of the 15th node of the flatten layer, the average of the STTc segment, the BMI index, the systolic blood pressure, the AHI index, the flatten layer ), the standard deviation of the results of the 11th node, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the low back pain dysfunction index (ODI), the smoking level, the 12th node of the flatten layer The average of the results of , the average of the results of the first node of the flatten layer, the standard deviation of the results of the 7th node of the flatten layer, and the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the standard deviation of the STTc segment, the average of the low frequency band intensity (P _LF ), and the average of the results of the third node of the flatten layer.

The mean and standard deviation of these data are the mean and standard deviation for each total night's sleep.

In general, a support vector machine (SVM) aims to maximize its generalizing capability. Using the representative feature vectors of each class, we search for an optimized decision hyperplane with maximum margin between categories. The servo vector machine is a well-known classifier, and a detailed description thereof will be omitted.

When the cardiovascular disease occurrence determination unit 710 determines that a cardiovascular disease will occur in the future (CVD), the detailed disease determination unit 720 uses a one-to-one (OvO, One-Vs-One) multi-class classifier to , Select one of coronary heart disease (CHD), heart failure (HF), and stroke as the cardiovascular disease that will occur in the future and print it out. The detailed disease determination unit 720 includes a second serve vector machine (SVM_C-H) 721 , a third serve vector machine (SVM_C-S) 722 , and a fourth serve vector machine (SVM_H-S). It consists of 3 servot vector machines of (723).

The second servo vector machine (SVM_C-H) 721 is a means for determining one of coronary heart disease (CHD), heart failure (HF), coronary heart disease (coronary heart disease, CHD) and heart failure (HF), the result (cardiovascular disease (CVD) prediction result candidate) is output.

The input vector of the second servo vector machine (SVM_C-H) 721 is the average of the results of the fourth node of the flatten layer, the HDL cholesterol index, the average of the low frequency band intensity (P _LF ), the flat The average of the results of the 10th node of the flatten layer, the standard deviation of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the standard deviation of the results of the 12th node of the flatten layer, It may include an AHI index.

The third servo vector machine (SVM_C-S) 722 is a means for discriminating one of coronary heart disease (CHD) and stroke, coronary heart disease (CHD) and A result of selecting one of the strokes (cardiovascular disease (CVD) prediction result candidate) is output.

The input vector of the third servo vector machine (SVM_C-S) 722 is the average of the results of the 10th node of the flatten layer, the average of the STTc segment, and the low-frequency to high-frequency band intensity ratio (P _LF / The average of P _HF ratio), the diastolic blood pressure, the HDL cholesterol index, the average of the low frequency band intensity (P _LF ), and the BMI index may be included.

The fourth servo vector machine (SVM_H-S) 723 is a means for determining one of heart failure (HF) and stroke, and one of heart failure (HF) and stroke. Outputs the selected result (cardiovascular disease (CVD) prediction result candidate).

The input vector of the fourth servo vector machine (SVM_H-S) 723 is the HDL cholesterol index, the average of the results of the third node of the flatten layer, the standard deviation of the low frequency band intensity (P _LF ), Cholesterol index, AHI index, mean of the results of the second node of the flatten layer, systolic blood pressure, the standard deviation of the high frequency band intensity (P _HF ), the results of the fourth node of the flatten layer The mean, smoking level, and standard deviation of the result of the 6th node of the flatten layer may be included.

To the voting unit 770 , the result of the second serve vector machine (SVM_C-H) 721, the result of the third serve vector machine (SVM_C-S) 722, and the fourth serve vector machine (SVM_H-) S) From the result of 723, the number of times each of coronary heart disease (CHD), heart failure (HF), and stroke is selected is counted and output. And the disease with the most selected number of times is the disease that can occur within the next 10 years, and the final result is output.

In other words, the cardiovascular disease (CVD) predictive model, after undergoing a process for extracting features, uses a support vector machine (SVM) model to predict coronary heart disease (CHD), It predicts the occurrence of heart failure (HF) and stroke. To do this, first go through CVD and SVM (SVM_CVD) that predicts CVD-free, and if classified as CVD, it goes through a one-to-one (OvO, One-Vs-One) multi-class classifier to classify CHD, HF, and stroke. OvO is a well-known method for selecting the class with the most discriminant values through K(K-1)/2 binary class classification by selecting a combination of two classes among K target classes. As such, a detailed description survives.

In this specification, the detailed description of the contents that can be sufficiently recognized and inferred by those of ordinary skill in the technical field of the present invention is omitted, and the technical spirit or essential configuration of the present invention is changed in addition to the specific examples described in the present specification More various modifications are possible within the range not to do so. Therefore, the present invention may be practiced in a manner different from that specifically described and illustrated in this specification, which will be understood by those of ordinary skill in the art of the present invention.

100: data input unit 110: electrocardiogram detection unit
120: electrocardiogram sensor unit 130: signal pre-processing unit
150: A/D conversion unit 160: Cardiovascular disease risk factor data input unit
210: signal-processed electrocardiogram feature extraction unit
220: artificial intelligence-based electrocardiogram feature extraction unit
230: CVD risk factor loading unit 250: cardiovascular disease prediction unit
310: memory unit 320: output unit

Claims (19)

In the cardiovascular disease prediction system, comprising an arithmetic processing unit for predicting the occurrence of cardiovascular disease using the electrocardiogram signal and CVD risk factor detected during polysomnography,
arithmetic processing unit,
From the electrocardiogram (ECG) signal, time domain electrocardiogram parameters including ST-segment and T-wave (ST-T) changes (STTc) segments are detected, and a low-frequency to high-frequency band intensity ratio (PLF/PHF ratio) is included. a signal-processed electrocardiogram feature extracting unit that detects a frequency domain electrocardiogram parameter, and stores the average and standard deviation of the detected time domain electrocardiogram parameter and the frequency domain electrocardiogram parameter as 'signal processed electrocardiogram features' in the memory unit;
An electrocardiogram signal for 30 seconds is input to the pre-trained CNN-based artificial intelligence model, and the average and standard deviation of the results of each node of the flatten layer in the CNN-based artificial intelligence model are obtained, an artificial intelligence-based electrocardiogram feature extraction unit, which is stored in the memory unit as 'artificial intelligence-based electrocardiogram features';
a CVD risk factor loading unit that reads the CVD risk factor stored in the memory unit;
Including a cardiovascular disease (CVD) prediction model including servo vector machines (SVM), signal-processed ECG characteristics, AI-based ECG characteristics, and ECG characteristics including CVD risk factors are added to the cardiovascular disease prediction model From the cardiovascular disease prediction model, whether or not cardiovascular disease will occur in the future a cardiovascular disease (CVD) prediction unit that outputs the information as a cardiovascular disease prediction result;
A cardiovascular disease prediction system comprising a. The method of claim 1, wherein the cardiovascular disease (CVD) prediction unit,
It consists of a pre-learned first servot vector machine to determine whether or not cardiovascular disease will occur in the future. Cardiovascular disease occurrence determination unit that outputs as;
If cardiovascular disease is determined to occur in the future by the cardiovascular disease identification unit, using a one-to-one (OvO, One-Vs-One) multi-class classifier, coronary heart disease (CHD) as a future cardiovascular disease ), heart failure (HF), and stroke (stroke) selected one of the cardiovascular disease (CVD) predicted value, the detailed disease discrimination unit;
A cardiovascular disease prediction system comprising a. The method of claim 2, wherein the detailed disease discrimination unit
a second servo vector machine pre-trained for discriminating one of coronary heart disease and heart failure, wherein one of the coronary heart disease and heart failure is selected and outputted as a result of the second servo vector machine;
A third servo vector machine that is pre-trained to discriminate between coronary heart disease and stroke, and selects one of coronary heart disease and stroke and outputs it as a result of the third servo vector machine machine;
a fourth-servo vector machine that is pre-learned to discriminate one of heart failure and stroke, and selects one of heart failure and stroke and outputs it as a result of the fourth-servo vector machine;
A cardiovascular disease prediction system comprising a. The method of claim 3, wherein the detailed disease discrimination unit
a voting unit that counts and outputs the number of times each of coronary heart disease, heart failure, and stroke is selected from the results of the second-servo vector machine, the third-servo vector machine, and the fourth-servo vector machine;
A cardiovascular disease prediction system, characterized in that it further comprises. 6. The method of claim 5,
A cardiovascular disease prediction system, characterized in that the voting unit outputs the values for diseases with the highest number of selected coronary heart disease, heart failure, and stroke as cardiovascular disease (CVD) prediction values. 6. The method of claim 5,
The input vector of the first servo vector machine is the average of the results of the 15th node of the flatten layer, the average of the STTc segment, the BMI index, the systolic blood pressure, the AHI index, and 11 of the flatten layer. The standard deviation of the results of the second node, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the low back pain dysfunction index (ODI), the smoking level, the results of the 12th node of the flatten layer Average, the average of the results of the first node of the flatten layer, the standard deviation of the results of the 7th node of the flatten layer, the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio) Cardiovascular disease prediction, characterized by including 11 electrocardiogram features among standard deviation, standard deviation of STTc segment, mean of low frequency band intensity (P _LF ), and mean of the result of the third node of the flatten layer system. 6. The method of claim 5,
The input vector of the second servo vector machine is the average of the results of the 4th node of the flatten layer, the HDL cholesterol index, the average of the low frequency band intensity (P _LF ), and 10 of the flatten layer The average of the results of the th node, the standard deviation of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the standard deviation of the results of the 12th node of the flatten layer, and 6 electrocardiographic characteristics of the AHI index A cardiovascular disease prediction system, characterized in that it comprises. 6. The method of claim 5,
The third servo vector machine is the average of the results of the 10th node of the flatten layer, the average of the STTc segment, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), diastolic blood pressure, HDL cholesterol A cardiovascular disease prediction system, characterized in that it includes three of the index, the average of the low frequency band intensity (P _LF ), and the BMI index. 7. The method of claim 6,
The input vector of the fourth servo vector machine is the HDL cholesterol index, the average of the results of the third node of the flatten layer, the standard deviation of the low frequency band intensity (P _LF ), the cholesterol index, the AHI index, the platen Mean of the results of the 2nd node of the flatten layer, systolic blood pressure, standard deviation of high frequency band intensity (P _HF ), the average of the results of the 4th node of the flatten layer, smoking degree, platen layer (Flatten layer) characterized in that it includes 7 of the standard deviation of the result of the 6th node, cardiovascular disease prediction system. In the driving method of a cardiovascular disease prediction system, comprising a calculation processing unit for predicting the occurrence of cardiovascular disease using an electrocardiogram signal and CVD risk factor detected during polysomnography,
The operation processing unit detects, from the electrocardiogram (ECG) signal, a time domain electrocardiogram parameter including a STTc (ST-segment and T-wave (ST-T) changes) segment, and a low-frequency to high-frequency band intensity ratio (PLF/PHF ratio) ), and storing the average and standard deviation of the detected time domain ECG parameter and the frequency domain ECG parameter in the memory unit as 'signal processed ECG characteristics', extracting signal-processed ECG features step;
The operation processing unit inputs an electrocardiogram signal for 30 seconds to the previously learned CNN-based artificial intelligence model, and calculates the average and standard results of each node of the flatten layer in the CNN-based artificial intelligence model. An artificial intelligence-based ECG feature extraction step of obtaining a deviation and storing it in a memory unit as 'artificial intelligence-based ECG features';
The arithmetic processing unit, the CVD risk factor loading step of reading the CVD risk factor stored in the memory unit;
The arithmetic processing unit uses a cardiovascular disease (CVD) prediction model including servo vector machines (SVM), and obtains the signal-processed ECG characteristics obtained in the ECG feature extraction step and the AI-based ECG characteristic extraction step. The ECG characteristics based on artificial intelligence and the ECG characteristics including the CVD risk factors read in the CVD risk factor loading stage are input into the cardiovascular disease prediction model, and the results output from the cardiovascular disease prediction model Cardiovascular disease ( CVD) prediction step;
A method of driving a cardiovascular disease prediction system, characterized in that it comprises a. 11. The method of claim 10, wherein the cardiovascular disease (CVD) prediction step,
The arithmetic processing unit inputs an electrocardiogram characteristic including the average of STTc segments to the previously learned first servo vector machine as an input vector, and determines whether or not cardiovascular disease will occur in the future by using the output value of the first servo vector machine A cardiovascular disease occurrence determination step in which a value indicating that cardiovascular disease does not occur is output as a cardiovascular disease (CVD) predicted value if the output value of the first servo vector machine determines that cardiovascular disease does not occur ;
When the calculation processing unit determines that a cardiovascular disease will occur in the future, the cardiovascular disease occurrence determination unit uses a one-to-one (OvO, One-Vs-One) multi-class classifier, heart disease (CHD), heart failure (HF), and stroke (stroke) selected one of the cardiovascular disease (CVD) predictive value, a detailed disease identification step;
A method of driving a cardiovascular disease prediction system, characterized in that it comprises a. The method of claim 11, wherein the detailed disease identification step
The arithmetic processing unit inputs the electrocardiogram characteristics including the average of the low frequency band intensity (P _LF ) into the pre-learned second serve vector machine, and the result of the second serve vector machine from the second serve vector machine a second servo vector machine calculation step, in which one of coronary heart disease and heart failure is output as a value;
The arithmetic processing unit inputs the electrocardiogram characteristics including the average of the low frequency band intensities (P _LF ) into the pre-learned third servo vector machine, and the result of the third servo vector machine from the third servo vector machine a third servo vector machine calculation step in which one of coronary heart disease and stroke is output as a value;
The arithmetic processing unit inputs the electrocardiogram characteristics including the standard deviation of the low frequency band intensity (P _LF ) to the pre-learned fourth servo vector machine, and from the fourth servo vector machine to the fourth servo vector machine. a fourth-servo vector machine in which one of heart failure and stroke is output as a result value;
A method of driving a cardiovascular disease prediction system, characterized in that it comprises a. The method of claim 12, wherein the detailed disease identification step is
The arithmetic processing unit counts the number of times each of coronary heart disease, heart failure and stroke is selected from the result value of the second serve vector machine, the result value of the third serve vector machine, and the result value of the fourth serve vector machine a voting step of outputting counted values for each coronary heart disease, heart failure, and stroke;
A method of driving a cardiovascular disease prediction system, characterized in that it further comprises. 14. The method of claim 13,
In the voting step, the calculation processing unit outputs the disease value with the highest count value for coronary heart disease, heart failure, and stroke as a cardiovascular disease (CVD) prediction value. 14. The method of claim 13,
The input vector of the first servo vector machine is the average of the results of the 15th node of the flatten layer, the average of the STTc segment, the BMI index, the systolic blood pressure, the AHI index, and 11 of the flatten layer. The standard deviation of the results of the second node, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the low back pain dysfunction index (ODI), the smoking level, the results of the 12th node of the flatten layer Average, the average of the results of the first node of the flatten layer, the standard deviation of the results of the 7th node of the flatten layer, the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio) Cardiovascular disease prediction, characterized by including 11 electrocardiogram features among standard deviation, standard deviation of STTc segment, mean of low frequency band intensity (P _LF ), and mean of the result of the third node of the flatten layer How the system works. 14. The method of claim 13,
The input vector of the second servo vector machine is the average of the results of the 4th node of the flatten layer, the HDL cholesterol index, the average of the low frequency band intensity (P _LF ), and 10 of the flatten layer The average of the results of the th node, the standard deviation of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), the standard deviation of the results of the 12th node of the flatten layer, and 6 electrocardiographic characteristics of the AHI index A method of driving a cardiovascular disease prediction system, comprising: 14. The method of claim 13,
The third servo vector machine is the average of the results of the 10th node of the flatten layer, the average of the STTc segment, the average of the low-frequency to high-frequency band intensity ratio (P _LF /P _HF ratio), diastolic blood pressure, HDL cholesterol A method of driving a cardiovascular disease prediction system, characterized in that it includes three of an index, an average of the low frequency band intensity (P _LF ), and a BMI index. 14. The method of claim 13,
The input vector of the fourth servo vector machine is the HDL cholesterol index, the average of the results of the third node of the flatten layer, the standard deviation of the low frequency band intensity (P _LF ), the cholesterol index, the AHI index, the platen Mean of the results of the 2nd node of the flatten layer, systolic blood pressure, standard deviation of high frequency band intensity (P _HF ), the average of the results of the 4th node of the flatten layer, smoking degree, platen layer A method of driving a cardiovascular disease prediction system, characterized in that it includes 7 of the standard deviation of the result of the 6th node of the (flatten layer). A recording medium for storing a computer program source for the driving method of the cardiovascular disease prediction system according to any one of claims 10 to 18.

Images