TW202302032A - A method of monitoring apnea and hypopnea events based on the classification of the descent rate of heartbeat intervals - Google Patents

Abstract

A method of monitoring apnea and hypopnea events based on the classification of the descent rate of heartbeat intervals includes a measurement step, a classifying step, and a monitoring and identifying step. The measurement step includes getting electrocardiogram signals of a subject by measuring. The classifying step includes calculating a descent rate of heartbeat intervals of the electrocardiogram signals according to an expression that includes an extracting/finding mode and a selecting/calculating mode to thereby be adapted to mark different intervals of the electrocardiogram signals, thereby classifying the electrocardiogram signals into an apnea and hypopnea signal group and a normal signal group. The monitoring and identifying step includes using a convolutional neural network as a learning model technology for monitoring and identifying, which is allowed to subject signals of the marked intervals to calculation of probability, thereby increasing the accuracy of monitoring the degree of severity of apnea events of the subject efficiently.

Description

Detection Method of Apnea and Insufficiency Events Based on Heartbeat Interval Decline Rate Grouping

The present invention relates to the detection of sleep-respiratory dysfunction, in particular to a method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate.

Obstructive Sleep Apnea (Obstructive Sleep Apnea; hereinafter referred to as OSA) is a common and serious sleep apnea disorder. Suspended or weakened, and according to previous studies, obstructive sleep apnea is related to the incidence of hypertension, coronary heart disease, arrhythmia, heart failure and stroke. According to the current standard method for evaluating the severity of OSA, it is through sleep multiple physiological examinations (Polysomnography; hereinafter referred to as PSG), that is, the subject must go to the sleep laboratory or sleep center to sleep for one night, and under the supervision of the nursing staff, electrode patches are attached to the neck, eyes, chin, heart and legs respectively. , and put induction belts on the chest and abdomen, blood oximeters on fingers, breathing sensors on the mouth and nose, and blood pressure monitors on the arms, so as to record the whole body through the aforementioned sensors and measuring devices. Sleep physiological data at night, including EEG, electrooculogram, electrocardiogram, jaw electromyogram, chest breathing signal, abdominal breathing signal, mouth and nose airflow, blood oxygen concentration, blood pressure changes, heart rate, and sleeping position, etc., while PSG It combines respiratory airflow, chest breathing signal, abdominal breathing signal, and blood oxygen concentration to judge and calculate the average number of apnea (Apnea) and hypopnea (Hypopnea) events per hour (that is, apnea and hypopnea) Index; Apnea and Hypopnea Index (AHI)), to assess the severity of OSA of the subject, including normal breathing (Normal; AHI <5), mild OSA (Mild; AHI between 5 and 14), moderate OSA ( Moderate; AHI between 15 and 30), and severe OSA (Severe; AHI> 30).

Continuing from the above, in view of the high cost and inconvenience of PSG examination, in recent years, some people have devoted themselves to the research of using less measured signals to develop a convenient and low-cost apnea and insufficiency event detection system. The main signal used is There are blood oxygen concentration, respiratory airflow, chest respiration, electrocardiogram, sound signal, and a combination of different signals. Please refer to Figure 1. In Figure 1, the PSG measured respiratory airflow, chest respiration signal, and abdominal respiration are shown. signal, electrocardiogram signal, and respiration notes provided by PSG (level 0 indicates the period of normal breathing, level 2 indicates the period of apnea), and the heartbeat interval signal (RR interval signal) is the adjacent ECG signal The signal composed of the interval time of the R wave, so it can be observed from Figure 1 that during the apnea, the change of the heartbeat interval signal is slow, but after the apnea is over, the heartbeat interval signal is significantly reduced and resumes after a period of time Normal, so if, after an otherwise normal, steady heartbeat interval signal, there is a sustained period of reduced heartbeat time signal followed by a return to a normal, steady heartbeat interval signal, it represents an apnea or hypopnea event, also known as apnea Heartbeat interval variation patterns of apnea and hypopnea events; however, because PSG mainly combines respiratory signals (including respiratory airflow, chest respiration, and abdominal respiration) and blood oxygen concentration to detect apnea and hypopnea events, if only respiratory Airflow, chest respiration, abdominal respiration, and blood oxygen concentration will not be able to detect all apnea insufficiency events, while the detection method based on the sound signal is limited by the sound signal is easily interfered by heart sound and environmental noise, so compared with Using respiratory airflow, chest breathing signal, blood oxygen concentration and audio signal alone, single-channel ECG is a signal that can better reflect the complete respiratory event.

Continuing from the above, when the current detection method of apnea and hypopnea based on single-lead ECG signals and machine learning establishes the grouping of ECG signals, that is, the ECG corresponding to normal breathing and the ECG corresponding to apnea and hypopnea, a large number of Part of it is based on the breathing notes provided by the PSG. The breathing notes are the ECG corresponding to the period of normal breathing as the normal breathing group, and the breathing notes are the ECG corresponding to the period of apnea and hypopnea as the apnea and hypopnea group. However, it can be found from Figure 1 that the ECG usually does not return to normal until after the apnea event is over, therefore, the electrocardiogram recorded as apnea and hypopnea may not necessarily reflect the impact of apnea and hypopnea, On the contrary, after the note of apnea and hypopnea is completed, the note of normal breathing is followed, but the ECG at this time is obviously affected by the note of apnea and hypopnea. Therefore, if the breathing note provided by PSG is used. The grouping of electrocardiographic signals is prone to wrong grouping. Therefore, reducing the accuracy of machine learning model training and testing is still the main research topic at present.

Therefore, the purpose of the present invention is to provide a method for detecting apnea and insufficiency events based on heartbeat interval drop rate grouping, which can detect and identify the subject with Severity of apnea and insufficiency events.

Therefore, the method for detecting apnea and insufficiency events based on the grouping of heartbeat interval drop rates in the present invention includes steps such as measurement steps, grouping steps, and detection and identification; wherein, the electrocardiograph equipped in the measurement step , can record the heartbeat, the rhythmic contraction of the surrounding muscles and the respiratory response of the subject respectively to form an electrocardiogram signal, and through the arithmetic processor equipped in the grouping step, a calculation formula can be used to reduce the heartbeat interval time Calculate the rate, the formula uses a take-out/find mode and a select/calculate mode respectively, through the take-out/find mode and select/calculate mode, calculate the positive, negative, and 0 values of the heartbeat interval time drop rate, and then Through the operation processor, the apnea and hypopnea events in the electrocardiogram signal and the interval marking operation of signal grouping are carried out, so as to obtain a group signal of apnea and hypopnea event and a signal of normal breathing group respectively , and then from the detection and identification step, a machine learning model is used by the computing processor, and a sliding window method that cooperates with the machine learning model to arrange calculations. And the training data set and test data set of electrocardiographic signals selected from different subjects are the data for detection and identification, so as to target the interval signals noted in the previous steps, so that these signals are subjected to normalization processing and implemented features Extract and obtain a plurality of better ECG signal feature maps, convert these feature maps into feature vectors and perform computerized detection, and finally output an identification result to judge whether the ECG signal obtained in the measurement step is There are apnea and hypopnea event patterns, so that through a simple detection method, it can effectively and quickly detect and classify the accuracy of classifying the severity of the apnea event of the subject.

The aforementioned and other technical contents, features and effects of the present invention will be clearly understood in the following detailed description of preferred embodiments with reference to the drawings.

Referring to Fig. 2, a preferred embodiment of the present invention, a method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate includes a measurement step, a grouping step and a detection and identification step, etc.; wherein, in the In the measurement step, an electrocardiogram machine is provided, and the electrocardiogram machine can measure heart rhythm and respiratory rate induction, aiming at the spontaneous beating of the heart in the chest of the subject and the rhythmic contraction of the surrounding muscles, and according to the cardiac tissue voltage Changes are recorded as electrocardiographic signals.

Continuing from the foregoing, referring to Fig. 3, the grouping step is equipped with an arithmetic processor, and the arithmetic processor can perform an operation on the rate of decrease of the heartbeat interval time of the subject by a formula, and the formula includes a take-out/find-out respectively mode and a selection/calculation mode, to calculate the positive, negative, and 0 values of the heartbeat interval time drop rate through the fetching/finding mode and the selection/calculation mode, further aiming at the apnea and hypopnea in the electrocardiographic signal Events and normal breathing are grouped into signal groups to obtain an apnea and hypopnea event group signal (please refer to Figure 4 (a) to (c)), and a normal breathing group signal (please refer to Figure 5 (a) to (c)), and as shown in Figure 3, the thick line is the heartbeat interval time signal, the heartbeat interval time drop rate signal corresponding to the thin line, and the heartbeat interval drop rate signal at time t The formula is defined as follows:

Heartbeat interval time drop rate (t) =

t之後15秒(150個)心跳間隔時間中最小的50個心跳間隔時間的平均值 5 seconds (50) heartbeat interval average before t

The average value of the smallest 50 heartbeat intervals in the 15 seconds (150) heartbeat intervals after t

Through this formula, it can be seen that when the heartbeat interval decrease rate at time t is a positive value, it means that the next heartbeat interval will be reduced, and when the heartbeat interval decrease rate at time t is a negative value, it means that the next heartbeat interval will be reduced. The heartbeat interval time is increased. If the heartbeat interval time decrease rate is 0, it means that the heartbeat interval time has dropped to the lowest point or risen to the highest point. Therefore, when the calculation method uses the fetch/find mode to correspond to the ECG When the shape signal selects the apnea and hypopnea event group, it will execute at least three action modes, that is, when the action mode 1 is performed, the interval between the apnea and hypopnea events is further extracted through the heartbeat interval drop rate calculation, and When continuing to perform the action 2 mode, find out the maximum rate value Max Rate10 of the rate of decline of the heartbeat interval time in the interval of 10 seconds before the end of the apnea and hypopnea events (marked as * in Figure 3 is Max Rate10) and its The maximum position is Max Loc10, and if the maximum rate value Max Rate10 is found to be greater than 0.15 and less than 0.4 when performing action 3 mode, it means that there is a significant reduction in the interval between heartbeats, then select the maximum position Max Loc10 10 seconds before the maximum position Max Loc10 The electrocardiogram signal and the heartbeat interval time signal of the last 20 seconds enter the group of apnea and hypopnea (please refer to Figure 3). Therefore, the first 10 seconds of the selected 30-second electrocardiogram signal are heartbeat interval time. Change, increase, or decrease, and the heartbeat interval will decrease in the last 20 seconds. If the maximum rate value Max Rate10 is less than or equal to 0.15 or greater than or equal to 0.4, return to action 1 mode and continue to take out the next apnea and hypopnea event In addition, when using this selection/calculation mode to select the normal breathing group corresponding to the ECG signal, it will also perform at least three action modes, and when performing action 1 mode, it will pass the heartbeat interval. The calculation formula further selects the 30-second interval of the ECG signal and the heartbeat interval signal, and there is no apnea and hypopnea event note in this interval, and the rate of decrease of the heartbeat interval time in the aforementioned 30-second interval will be calculated when the action 2 mode is executed. The minimum rate value Mini Rate30, and the maximum rate value Max Rate30, and if the maximum rate value Max Rate30 is found to be greater than -0.02 and the maximum rate value Max Rate30 is less than 0.02 when the action 3 mode is continuously executed, it means that within the interval of 30 seconds, the heartbeat If the interval time does not change significantly, select the electrocardiogram signal and the heartbeat interval signal in the 30-second interval to enter the normal breathing group. If the maximum rate value Max Rate30 is found to be less than or equal to -0.02 or the maximum rate value Max Rate30 is greater than or equal to 0.02, then Go back to action 1 and select the next 30-second ECG signal and heartbeat interval signal interval.

Therefore, for the apnea and hypopnea group taken out above, it can be known that since the 10th second is the maximum value of the decrease rate of the corresponding heartbeat interval time, the heartbeat interval time before the 10th second may be increased, shortened or unchanged Yes, after the 10th second, the heartbeat interval time signal will decrease and then recover, as shown in (a) and (c) in Figure 4, and if the heartbeat interval time decreases for a longer duration, it will be as shown in Figure 4 As shown in (b) of 4, it will be too late to recover within 30 seconds. At the same time, it can be clearly seen from (a) to (c) in Figure 5 that for the normal breathing group selected above, the heartbeat interval The time signal showed no significant change.

As for the detection and identification step, it uses the arithmetic processor to further compare the apnea and hypopnea event group signals with the normal breathing group through a machine learning model and a sliding window method that cooperates with the machine learning model. The signal is trained to generate detection and recognition results, and the machine learning model uses a convolutional neural network as a learning model technology, and uses training data recorded independently of ECG signals selected from different subjects The training data set and the test data set are data for detection/identification, and the data of the training data set and test data set used above are the sleep multiple physiological data provided by the Sleep Heart Health Study (SHHS). Check (Polysomnography; referred to as PSG) database to build, and the training data set and test data set respectively include normal breathing, and apnea and hypopnea group 30-second electrocardiographic signal and heartbeat interval time signal, in order to use the training data The pooled ECG and heartbeat interval signals are used to train an optimized machine learning model to identify whether the input ECG and heartbeat interval signals correspond to normal breathing or apnea and hypopnea events, and the test The electrocardiogram signal and heartbeat interval time signal in the data set are used to test the correctness of the machine learning model after optimization for the recognition accuracy of the electrocardiogram signal and heartbeat interval time signal outside the training data set, which can test the optimized The real performance of machine learning models.

Furthermore, please refer to FIG. 6. In this embodiment, the machine learning model is a deep learning model technology based on a convolutional neural network (CONVOLUTIONAL NEURAL NETWORKS; CNN for short), and its input signal can be a separate 30-second ECG signal, a single 30-second heartbeat interval signal, or a 30-second ECG signal and a heartbeat interval signal at the same time, and the sampling frequency is 100Hz, so the length of the input signal can be 1×3000 or 2× 3000 input, and the deep learning model technology includes at least eight feature extraction layers with the same structure, at least one flat layer connected to the eight feature extraction layers, a first classification layer connected to the flat layer, a A second classification layer connected to the first classification layer, and a third classification layer connected to the second classification layer, and each of the aforementioned feature extraction layers includes a method for obtaining at least 45 1D feature maps Convolutional layer, a batch normalization layer, an activation layer, a maximum pooling layer with a pooling size of 2, and a discarding layer with a 50% discarding rate. The electrocardiographic signal input in the grouping step is normalized, and the feature extraction is performed on the electrocardiographic signal to obtain better feature maps of the electrocardiographic signal, and the flattening layer converts 45 1D feature maps into 1D Feature vectors for subsequent classification layers, while the first classification layer includes a fully connected layer with 2000 neurons, a batch normalization layer, an activation layer, and a dropout layer with a 50% dropout rate layer, and the second classification layer includes a fully connected layer with 1000 neurons, a batch normalization layer, an activation layer and a dropout layer with a 50% dropout rate, and the third classification layer includes There is a fully connected layer with 2 neurons, and the activation function (Softmax) is used to calculate the probability of the two outputs of the classification layer. The category with high probability is the result of identification, that is, the input ECG signal corresponds to breathing Normal or apneic and hypopneic events.

Continuing from the above, the sliding window method is to complete the model training and testing of the subject's ECG signal for the machine learning model and cooperate with the arrangement calculation, which can collect before or after a certain action according to the size of the window of the sliding window method. action, and cooperate with the calculation and calculation of the specific gravity value. Specifically, take the ECG signal or heartbeat interval signal of 3 minutes (18,000 sampling points) as an example, when the window length L is 3000, each Take 3000 sampling points from the subject's ECG signal or heartbeat interval signal at a time, and input them into the optimized model to get a classification probability of apnea and hypopnea events. The window is sampled every 300 The point (3 seconds) slides to the next position, and another 3000 sample points are taken, while the aforementioned sliding, sampling pattern continues until the 3 minute length is completed, as shown in Figure 7; thus, a total of 60 lengths are taken during the 3 minute period The signals of 3000 sampling points are respectively input into the optimized model to obtain the classification probability of 60 apnea and hypopnea events, that is, the example shown in Figure 8, the classification probability of apnea and hypopnea events is greater than When it is equal to 0.5, it means that apnea and hypopnea events occur in the window of the sliding window method, and it is marked as A. Of course, when the continuous pane is classified as A, it is regarded as the same apnea and hypopnea event , that is, as shown in the middle of Figure 8, there are 6 consecutive panes classified as A, and each pane corresponds to a window interval of 300 sampling points (3 seconds), so the signal interval of 18 seconds (6×3 seconds) It is detected as an apnea and hypopnea event, and finally an identification result is output, so as to generate a better detection and identification method through simple calculation probability, which can effectively and quickly detect, identify and classify the apnea event of the subject serious accuracy.

Therefore, the present invention mainly aims at whether the subject has apnea and hypopnea events, that is, first calculates the decline rate of the heartbeat interval time of the subject's electrocardiogram signal by using a formula, and uses the extraction/finding method respectively The positive, negative, and zero values of the heartbeat interval time drop rate are calculated in the output mode and the selection/calculation mode, and the apnea and hypopnea events of the electrocardiogram signal are further grouped into signal groups with normal breathing, and then through the detection In the identification step, the technique of using the ECG signal training of the training data set to obtain the optimized model allows the signals of the apnea and hypopnea event group and the normal breathing group to be normalized and executed after grouping. Feature extraction to obtain better feature maps of a plurality of electrocardiogram signals, and further convert these feature maps into feature vectors and perform calculation calculations, so that the electrocardiogram input into the test data set can test the real performance of the machine learning model, and The results show that the accuracy of both training and testing can reach more than 95%, so that a simple detection method can be used to finally output a recognition result, which can effectively and quickly detect and classify the serious apnea events of the subjects. accuracy.

To sum up the above, the present invention is based on the detection method of apnea and hypopnea events grouped by heart rate interval drop rate, which mainly aims at whether the subject has apnea and hypopnea events, and then performs detection and identification after heartbeat interval drop rate grouping , by using a convolutional neural network as a learning model technique in the detection and identification step, using data from a training data set and a testing data set of electrocardiogram signals from different subjects as a detection At the same time, the size of the sliding window method is combined with the alignment calculation, and the ECG signal obtained from the measurement step is further calculated and trained through the learning model technology, so that the signal can be normalized and executed. Extract and obtain better feature maps of a plurality of electrocardiographic signals, and convert these feature maps into feature vectors and perform calculations to output a recognition result, so as to effectively and quickly detect, identify and classify the subjects Accuracy of severe apnea events.

But the above is only to illustrate the preferred embodiments of the present invention, and should not limit the scope of the present invention, that is, all the simple equivalent changes and modifications made according to the patent scope of the present invention and the content of the description of the invention , should still fall within the scope covered by the patent of the present invention.

none

Fig. 1 is a schematic diagram of known breathing signals, breathing notes, ECG and heartbeat interval legends. Fig. 2 is a flowchart of a preferred embodiment of the present invention. FIG. 3 is a schematic diagram of the calculation of the rate of decrease in time between heartbeat intervals (RR intervals) in the present invention. Fig. 4 is a schematic illustration of the electrocardiogram signal and heartbeat interval time signal legend of the apnea and hypopnea groups of the present invention. Fig. 5 is a schematic diagram of the legend of the normal breathing group signal and the heartbeat interval time signal according to the present invention. FIG. 6 is a schematic diagram of a deep learning model based on a convolutional neural network of the preferred embodiment. Fig. 7 is a schematic diagram of sliding sampling points of the sliding window method of the preferred embodiment. Fig. 8 is a schematic diagram of the classification results of 60 apnea and hypopnea events during 3 minutes using the sliding window method of the preferred embodiment.

Claims (7)

A detection method for apnea and insufficiency events grouped based on heartbeat interval drop rate, which includes: a measurement step, which is equipped with an electrocardiogram machine, which can measure heart rhythm and respiratory rate sensing, so as to The spontaneous beating of the heart of the subject's chest and the rhythmic contraction of the surrounding muscles are recorded according to the changes in the voltage of the heart tissue to form an electrocardiogram signal; a grouping step, which is equipped with an arithmetic processor, which can obtain a calculation formula Calculations are performed on the rate of decrease of the heartbeat interval time, and the calculation formula includes a take-out/find mode and a selection/calculation mode respectively, so as to calculate the positive value of the heartbeat interval time decrease rate through the take-out/find mode and the selection/calculation mode , negative, and 0 values, and further perform signal grouping actions for apnea and hypopnea events and normal breathing in the electrocardiogram signal, and obtain a group signal of apnea and hypopnea event and a signal of normal breathing group respectively; Among them, the calculation formula is defined as follows: heartbeat interval time decrease rate (t) = 5 seconds (50) heartbeat interval time average value before t

The average value of the smallest 50 heartbeat intervals in the 15 seconds (150) heartbeat intervals after t, that is, when the heartbeat interval decline rate at time t is a positive value, it means that the next heartbeat interval is reduced. And when the heartbeat interval decrease rate of time t is a negative value, it means that the next heartbeat interval is increasing, and if the heartbeat interval decrease rate is 0, it means that the heartbeat interval drops to the lowest point or rises to the highest points; and a detection and identification step, the operation processor uses a machine learning model, and a sliding window method that is aligned with the machine learning model, and the machine learning model uses a convolutional neural network as a learning model technology, and wherein the training data set and the test data set recorded with electrocardiographic signals that are independent and selected from different subjects are used as the data for detection and identification, and the machine learning model includes at least eight identical structures Feature extraction layer, at least one flat layer connected to the eight feature extraction layers, a first classification layer connected to the flat layer, a second classification layer connected to the first classification layer, and a The second classification layer is connected to the third classification layer, and the aforementioned feature extraction layers can just normalize the signal input to the grouping step, perform feature extraction on the signal and obtain better multiple ECG patterns The feature map of the signal, and the flattening layer will convert the feature map into feature vectors for use by the classification layers, and the classification layers will perform calculations based on the feature vectors, and finally output a detection and identification result , so as to determine whether the electrocardiographic signal obtained in the measuring step has apnea and hypopnea event patterns. According to the method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate according to claim 1, wherein the fetching/finding mode of the grouping step is further fetched by the arithmetic processor through the heartbeat interval drop rate formula The method extracts the note interval of the apnea and the hypopnea event, and finds out the method to find out the maximum rate value Max Rate10 and the maximum position Max Loc10 of the rate of decline of the heartbeat interval time in the interval of 10 seconds before the end of the apnea and hypopnea event; In addition, the selection/calculation mode of the grouping step is that the calculation processor further uses the heartbeat interval time drop rate formula to select the heartbeat interval time signal annotation interval of the 30-second interval by selection, and there is no apnea and respiration in this interval Insufficient event notation, and the minimum rate value Mini Rate30 and the maximum rate value Max Rate30 of the heartbeat interval time drop rate within the 30-second interval are calculated by calculation. According to the method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate according to claim 1 or 2, wherein each feature extraction layer includes a convolution layer, a batch normalization layer, an activation layer, and a maximum Pooling layer and a dropout layer and the aforementioned convolution layer is a setting that can obtain at least 45 ID feature maps, the maximum pooling layer is a setting with a pooling size of 2, and the dropout layer is a setting with a 50% dropout rate setting. According to the method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate according to claim 1, wherein the first and second classification layers both include a fully connected layer, a batch normalization layer, and an activation layer and a dropout layer, and the fully connected layer of the first classification layer has 2000 neurons, and the fully connected layer of the second classification layer has 1000 neurons. According to the method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate according to claim 3, wherein the first and second classification layers both include a fully connected layer, a batch normalization layer, and an activation layer and a dropout layer, and the fully connected layer of the first classification layer has 2000 neurons, and the fully connected layer of the second classification layer has 1000 neurons. According to the method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate according to claim 1, wherein the third classification layer includes a fully connected layer with 2 neurons, and the fully connected layer uses Activation function (Softmax) to calculate rate. According to the method for detecting apnea and insufficiency events grouped based on heartbeat interval drop rate according to claim 5, wherein the third classification layer includes a fully connected layer with 2 neurons, and the fully connected layer uses Activation function (Softmax) to calculate rate.

Images