KR102861010B1 - Method And Apparatus for Correcting Electrocardiogram based on User Feedback - Google Patents

Abstract

A method and device for correcting electrocardiogram reading based on user feedback are disclosed.
The present embodiment provides a method and device for correcting electrocardiogram readings based on user feedback, which provides a separate tool for re-reading each of the results of noise waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms, because there is a possibility that the automatic reading function may misread the electrocardiogram of a patient when using an electrocardiogram measuring device, so that the waveform reader can reflect the feedback results of the waveform reader into groups grouped by noise waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms.

Description

Method and Apparatus for Correcting Electrocardiogram Based on User Feedback

One embodiment of the present invention relates to a method and device for correcting electrocardiogram readings based on user feedback.

The content described below merely provides background information related to the present embodiment and does not constitute prior art.

An electrocardiogram (ECG) interpretation system has been developed to assist medical professionals in analyzing electrocardiograms. Conventional ECG interpretation systems detect R, P, and T peaks in the waveform and detect and classify arrhythmias based on rules.

Conventional electrocardiogram (ECG) interpretation systems receive and analyze a patient's entire ECG signal data, outputting the results. Deep learning technology, with its higher accuracy compared to existing methods, has recently been widely studied as an ECG interpretation algorithm.

Diagnosing arrhythmias using electrocardiograms (ECGs) is a task that only qualified medical professionals can perform, but the workforce is inadequate compared to the demand. Interpreting an ECG requires analyzing the ECG signal from multiple perspectives, including calculating the shape and time intervals of the P, QRS, and T waveforms, and analyzing the ECG rhythm. This requires significant time. While medical staff must monitor patients' ECGs in real time to monitor their condition, ongoing monitoring is hindered by a lack of personnel. Because ECG analysis directly impacts a patient's life, it must be accurate and, in the event of an emergency, must be performed quickly.

Conventional electrocardiogram analysis sometimes utilizes the endpoints and starting points of the P, QRS, and T waveforms. However, these techniques only detect peaks, limiting their utility. Rule-based algorithms for arrhythmia detection and classification often suffer from low accuracy due to waveform diversity, and new rule-based algorithms must be designed when additional arrhythmias are identified.

Conventional electrocardiogram (ECG) interpretation systems are not suitable for applications requiring real-time interpretation, such as bedside patient monitoring. Deep learning algorithms for ECG analysis vary in speed depending on the implementation model, and cannot operate hourly for real-time operation. When visualizing ECG waveforms, the output is one-dimensional data, which reduces readability during real-time interpretation.

Recently, there's a growing trend in the medical field to utilize artificial intelligence (AI) to diagnose heart conditions such as arrhythmias. However, ECG interpretation is complex, requiring the application of various algorithms. Furthermore, its accuracy is low, requiring skilled professionals to manually review and correct each automatically interpreted data. ECG interpretation requires not only the interpretation itself but also the entire electrocardiogram waveform, making it a time-consuming and cumbersome process.

Current rereading programs operate by running additional correction algorithms after all algorithms have been run. However, because interpretation criteria can vary from one ECG interpreter to another and from one patient to another, and because the results of each algorithm are interrelated, correcting the reading after all algorithms have been run takes a long time.

Korean Patent No. 10-1234567, "Deep Learning-Based Electrocardiogram Signal Analysis System."

The purpose of this embodiment is to provide a user feedback-based electrocardiogram interpretation correction method and device that provides a separate tool for re-reading each of the results of noise waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms, so that the waveform interpreter can reflect the feedback results of the waveform interpreter to groups grouped by noise waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms, since there is a possibility that the automatic interpretation function may misinterpret the electrocardiogram of a patient when using an electrocardiogram measuring device.

According to one aspect of the present embodiment, there is provided an electrocardiogram waveform acquisition unit that acquires electrocardiogram waveforms for a plurality of people; a classification unit that applies segmentation to the electrocardiogram waveform using a deep learning model to confirm a characteristic index value of each section of the electrocardiogram waveform, and generates a classification result for each section based on the characteristic index value; a beat detection unit that generates a beat detection result by detecting a heartbeat using a machine learning algorithm based on the classification result; a waveform grouping unit that groups a noise waveform, a normal beat waveform, a PVC (Premature Ventricular Contraction) waveform, a PAC (Premature Atrial Contraction) waveform, and an arrhythmia waveform based on the beat detection result; a feedback reflection unit that receives and reflects a feedback result for each of the noise waveform, the normal beat waveform, the PVC waveform, the PAC waveform, and the arrhythmia waveform; And the present invention provides an electrocardiogram reading device characterized by including a histogram unit that stores histograms for the noise waveform, the normal heartbeat waveform, the PVC waveform, the PAC waveform, and the arrhythmia waveform in which the feedback result is reflected.

As described above, according to this embodiment, since there is a possibility that the automatic reading function may misread the patient's electrocardiogram when using an electrocardiogram measuring device, a separate tool is provided for rereading the results of noise waveform, PVC waveform, PAC waveform, and arrhythmia waveform, so that the waveform reader can reflect the feedback results of the waveform reader to the groups grouped by noise waveform, PVC waveform, PAC waveform, and arrhythmia waveform.

According to this embodiment, in addition to the wearable electrocardiogram patch, it has an effect that can be utilized in various medical devices and wellness/fitness devices that measure electrocardiograms.

According to this embodiment, health checkup centers, hospitals, and medical institutions requiring electrocardiogram readings can reread noise waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms more easily, quickly, and accurately using dedicated tools.

FIG. 1 is a diagram illustrating biosignal data processing in the field of biosignal processing according to the present embodiment.
Figure 2 is a diagram showing the P, Q, R, S, and T waves (P wave, QRS complex, T wave) and characteristic indicators of an electrocardiogram according to the present embodiment.
Fig. 3 is a drawing showing an electrocardiogram reading device according to the present embodiment.
FIG. 4 is a diagram illustrating a method for correcting an electrocardiogram reading based on user feedback according to the present embodiment.
Fig. 5 is a drawing specifically illustrating an electrocardiogram noise stage detection process according to the present embodiment.
Fig. 6 is a diagram showing an example of feedback for noise steps by noise waveform using a dedicated tool according to the present embodiment.
FIG. 7 is a diagram showing an example of feedback for a normal heartbeat waveform and a PVC waveform using a dedicated tool according to the present embodiment.
FIG. 8 is a diagram showing an example of feedback for a normal heartbeat waveform and a PAC waveform using a dedicated tool according to the present embodiment.
FIG. 9 is a diagram showing an example of feedback for an arrhythmia detection waveform using a dedicated tool according to the present embodiment.

Hereinafter, the present embodiment will be described in detail with reference to the attached drawings.

FIG. 1 is a diagram showing the application of semantic segmentation to biosignal data processing in the field of biosignal processing according to the present embodiment.

An electrocardiogram reading device (300) is a test that records the electrical signals generated during a heartbeat as a waveform of current, and performs an electrocardiogram test to diagnose heart diseases such as arrhythmia, angina pectoris, myocardial infarction, and cardiac hypertrophy.

The electrocardiogram reading device (300) according to the present embodiment can be applied to one-dimensional (1D) biosignal data processing in the field of biosignal processing. The electrocardiogram reading device (300) divides the P wave, QRS complex (N, S, V), T wave, and noise wave included in the electrocardiogram waveform into waveform units.

The electrocardiogram reading device (300) classifies the electrocardiogram waveform into one of N (Normal beat), S (Supraventricular ectopic beat), V (Ventricular ectopic beat), F (Fusion beat), and Q (Unknown beat).

Figure 2 is a diagram showing the P, Q, R, S, and T waves (P wave, QRS complex, T wave) and characteristic indicators of an electrocardiogram according to the present embodiment.

The electrocardiogram waveform appears as a series of beats, and the beats can be broadly divided into normal beat (N), supraventricular beat (S), and ventricular beat (V). One beat of the electrocardiogram waveform basically includes a P wave, QRS wave, and T wave.

An electrocardiogram reading device (300) detects P waves, Q waves, R waves, S waves, and T waves included in an electrocardiogram waveform, and classifies normal beats (N), supraventricular beats (S), and ventricular beats (V). The electrocardiogram reading device (300) performs localization of the P waves, Q waves, R waves, S waves, and T waves to output various characteristic indices of the electrocardiogram.

The electrocardiogram reading device (300) distinguishes the PR interval, QRS interval, QT interval, ST segment, and RR interval for the input electrocardiogram waveform.

The electrocardiogram reading device (300) can classify the beat based on the characteristic information of the P wave, Q wave, R wave, S wave, and T wave based on the PR interval, QRS interval, QT interval, ST segment, and RR interval.

The electrocardiogram reading device (300) classifies the heartbeat into a normal beat (N), a supraventricular beat (S), and a ventricular beat (V). When reading an electrocardiogram waveform, the electrocardiogram reading device (300) can detect an abnormality based on localization and classification information.

The electrocardiogram reading device (300) detects abnormal conditions (arrhythmia, abnormal beat (S, V), ST, QTc, etc.) using values obtained by performing localization on the electrocardiogram waveform.

The electrocardiogram reading device (300) performs classification on the electrocardiogram waveform to classify the heartbeat into a normal beat (N), a supraventricular beat (S), and a ventricular beat (V), and detects arrhythmia.

An electrocardiogram reading device (300) applies a segmentation technique to an electrocardiogram waveform to identify each section of a P wave, a Q wave, an R wave, an S wave, a T wave, and a fibrillation waveform included in the electrocardiogram waveform. Based on each section, the electrocardiogram reading device (300) classifies the heartbeat into one of N (Normal beat), S (Supraventricular ectopic beat), V (Ventricular ectopic beat), F (Fusion beat), and Q (Unknown beat).

Fig. 3 is a drawing showing an electrocardiogram reading device according to the present embodiment.

Recently, medical devices with automatic electrocardiogram reading functions, such as smart watches and patch-type electrocardiogram monitors, have been released. However, since the automatic reading function of medical devices with automatic electrocardiogram reading functions may misinterpret a patient's electrocardiogram, it is essential to have a clinical pathologist (or doctor) re-interpret the results of the automatic reading device.

Therefore, the electrocardiogram reading device (300) provides a dedicated tool to help easily correct and confirm the automatic reading results when rereading the electrocardiogram.

The electrocardiogram reading device (300) has different standards for each reader (e.g., clinical pathologist) when reading an electrocardiogram, and since different standards are often applied to each person, it is difficult to read using a uniform algorithm. Therefore, during the execution of the reading algorithm, a dedicated tool is used to additionally reflect feedback from the reader (e.g., clinical pathologist) so as to obtain reading speed and optimized results for each reader.

The electrocardiogram reading device (300) according to the present embodiment includes an electrocardiogram waveform acquisition unit (310), a classification unit (320), a heartbeat detection unit (330), a waveform grouping unit (340), a feedback reflection unit (350), and a histogram unit (360). The components included in the electrocardiogram reading device (300) are not necessarily limited thereto.

Each component included in the electrocardiogram reading device (300) is connected to a communication path connecting software modules or hardware modules within the device and can operate organically with each other. These components communicate using one or more communication buses or signal lines.

Each component of the electrocardiogram reading device (300) illustrated in FIG. 3 means a unit that processes at least one function or operation, and can be implemented as a software module, a hardware module, or a combination of software and hardware.

The electrocardiogram waveform acquisition unit (310) acquires electrocardiogram waveforms for multiple people.

The classification unit (320) applies segmentation to the electrocardiogram waveform using a deep learning model, checks the characteristic index values of each section of the electrocardiogram waveform, and generates a classification result for each section based on the characteristic index values.

The heartbeat detection unit (330) generates a heartbeat detection result by detecting a heartbeat using a machine learning algorithm based on the classification result.

The waveform grouping unit (340) groups each of the noise waveform, normal beat waveform, PVC (Premature Ventricular Contraction) waveform, PAC (Premature Atrial Contraction) waveform, and arrhythmia waveform based on the beat detection results.

The waveform grouping unit (340) includes a noise step segmentation unit (342), a waveform shape-based grouping unit (344), a shape and interval-based grouping unit (346), and an arrhythmia detection unit (348).

The noise step segmentation unit (342) detects noise steps by detecting noise characteristics for each step of the electrocardiogram waveform based on the classification results.

The waveform shape-based grouping unit (344) generates a first normal beat waveform classification result by grouping only normal beat waveforms based on the waveform shape based on the beat detection result. The waveform shape-based grouping unit (344) generates a PVC waveform classification result by grouping only PVC (Premature Ventricular Contraction) waveforms based on the waveform shape based on the beat detection result.

The shape and interval-based grouping unit (346) generates a second normal beat waveform classification result by grouping only normal beat waveforms based on shape and interval based on the first normal beat waveform classification result. The shape and interval-based grouping unit (346) generates a PAC waveform classification result by grouping only PAC (Premature Atrial Contraction) waveforms based on shape and interval based on the first normal beat waveform classification result.

The arrhythmia detection unit (348) detects an arrhythmia waveform based on the first normal beat waveform, the second normal beat waveform, the PVC waveform, and the PAC waveform, and sets an arrhythmia standard.

The feedback reflection unit (350) receives and reflects feedback results for each of the noise waveform, normal heartbeat waveform, PVC waveform, PAC waveform, and arrhythmia waveform.

The feedback reflection unit (350) includes a noise feedback reflection unit (352), a waveform shape-based feedback reflection unit (354), a shape and interval-based feedback reflection unit (356), and an arrhythmia-based feedback reflection unit (358).

The noise feedback reflection unit (352) uses a dedicated tool to transmit the noise level for the noise waveform to the electrocardiogram waveform interpretation terminal. The noise feedback reflection unit (352) receives the noise feedback result from the electrocardiogram waveform interpretation terminal and reflects it in the noise level for each noise waveform.

The waveform shape-based feedback reflection unit (354) uses a dedicated tool to transmit the first normal beat waveform classification result and the PVC waveform classification result to the electrocardiogram waveform interpretation terminal. The waveform shape-based feedback reflection unit (354) receives the first normal beat feedback result for the first normal beat waveform classification result from the electrocardiogram waveform interpretation terminal and reflects it in the first normal beat waveform classification result. The waveform shape-based feedback reflection unit (354) receives the PVC waveform feedback result for the PVC waveform classification result and reflects it in the PVC waveform classification result.

The shape and interval-based feedback reflection unit (356) uses a dedicated tool to transmit the second normal beat waveform classification result and the PAC waveform classification result to the electrocardiogram waveform interpretation terminal. The shape and interval-based feedback reflection unit (356) receives the second normal beat feedback result for the second normal beat waveform classification result from the electrocardiogram waveform interpretation terminal and reflects it in the second normal beat waveform classification result. The shape and interval-based feedback reflection unit (356) receives the PAC waveform feedback result for the PAC waveform classification result and reflects it in the PAC waveform classification result.

The arrhythmia reference feedback reflection unit (358) transmits the arrhythmia waveform to the electrocardiogram waveform interpretation terminal using a dedicated tool. The arrhythmia reference feedback reflection unit (358) receives the arrhythmia waveform reference feedback result for the arrhythmia waveform from the electrocardiogram waveform interpretation terminal and reflects it in the arrhythmia waveform.

The histogram section (360) stores histograms for noise waveforms, normal heartbeat waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms in which feedback results are reflected.

FIG. 4 is a diagram illustrating a method for correcting an electrocardiogram reading based on user feedback according to the present embodiment.

An electrocardiogram reading device (300) obtains electrocardiogram waveforms for multiple people (S410). In step S410, the electrocardiogram reading device (300) receives raw files of electrocardiogram waveforms for multiple people from a server.

The electrocardiogram reading device (300) performs preprocessing on each electrocardiogram waveform of a plurality of people to generate preprocessed data (S412). In step S412, the electrocardiogram reading device (300) performs preprocessing to filter preset data on each electrocardiogram waveform of a plurality of people to generate preprocessed data.

The electrocardiogram reading device (300) performs segmentation on preprocessed data using a deep learning model, and generates a classification result (P, QRS [Normal, PVC], T, Noise) by classifying each section based on the characteristic index values of each section of the electrocardiogram waveform (S420).

The electrocardiogram reading device (300) classifies the noise level of the noise waveform in detail based on the classification result (P, QRS [Normal, PVC], T, Noise) (S430). In step S430, the electrocardiogram reading device (300) classifies the noise level (e.g., from 0 to 10) by using a noise segmentation algorithm based on baseline wandering, classification result, presence or absence of P wave, T wave, etc.

The electrocardiogram reading device (300) transmits the noise level for the noise waveform to the electrocardiogram waveform reading terminal using a dedicated tool. The electrocardiogram waveform reading terminal receives a feedback result corresponding to the noise level for the received noise waveform using a dedicated tool, and generates a feedback result that determines whether to use only a specific noise level (e.g., less than level 5) among the noise levels (e.g., levels 0 to 10) for analysis and to recognize the remaining noise levels (e.g., levels 5 or more) as noise and exclude them from the analysis.

The electrocardiogram waveform interpretation terminal transmits the feedback results to the electrocardiogram interpretation device (300) using a dedicated tool.

In step S430, the electrocardiogram reading device (300) classifies the noise phase by noise waveform using a dedicated tool. The electrocardiogram waveform is output to the interpreter terminal.

The electrocardiogram waveform interpretation terminal outputs the noise level for each noise waveform input from a dedicated tool of the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal selects a waveform according to the noise level of each noise waveform and then controls an input means (e.g., keyboard, mouse, joystick, pad, etc.) to pass the noise level of each noise waveform classified by the electrocardiogram interpretation device (300) to the next waveform to confirm it.

The electrocardiogram waveform interpretation terminal can use the dedicated tool provided from the electrocardiogram interpretation device (300) to subdivide the noise level by using the re-grouping function for the unclassified box.

The electrocardiogram waveform interpretation terminal controls the input means (e.g., keyboard, mouse, joystick, pad, etc.) to generate a feedback result that determines whether to use only a specific noise level (e.g., less than level 5) among noise levels (e.g., levels 0 to 10) for analysis and recognize the remaining noise levels (e.g., levels 5 and above) as noise and exclude them from analysis.

In step S430, the electrocardiogram reading device (300) uses a dedicated tool to receive feedback results from the electrocardiogram waveform reading terminal, which determine whether to use only a specific noise level (e.g., less than level 5) among noise levels (e.g., levels 0 to 10) for analysis and recognize the remaining noise levels (e.g., levels 5 or more) as noise and exclude them from analysis, and then reflects the feedback results in the noise level for each noise waveform.

The electrocardiogram reading device (300) generates a beat detection result by performing a beat detection (peak detection) based on the classification result (P, QRS [Normal, PVC], T, Noise) (S440). In step S440, the electrocardiogram reading device (300) generates a beat detection result by detecting the peak of the QRS based on the classification result (P, QRS [Normal, PVC], T, Noise).

The electrocardiogram reading device (300) performs waveform shape-based grouping based on the heartbeat detection results (S450). In step S450, the electrocardiogram reading device (300) performs S wave detection (S detection), and since each reader has different criteria for re-correction, the grouping is performed using only the shape.

The electrocardiogram reading device (300) generates a first normal beat waveform classification result by grouping only normal beat waveforms by clustering them in the process of performing waveform shape-based grouping (S452).

In step S452, the electrocardiogram reading device (300) performs shape-based grouping, and identifies the RR interval using the peak points of the P wave and QRS in the beat detection results. The electrocardiogram reading device (300) clusters normal beat waveforms based on the RR interval and generates a first normal beat waveform classification result by grouping them.

In step S452, the electrocardiogram reading device (300) uses a dedicated tool to output the first normal beat waveform classification result (first normal beat waveform group) clustered into the first normal beat waveform to the electrocardiogram waveform reading terminal.

The electrocardiogram waveform interpretation terminal outputs the first normal beat waveform classification result (first normal beat waveform group) clustered with the first normal beat waveform input from the dedicated tool of the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can select a representative shape of the first normal heartbeat waveform and then control an input means (e.g., keyboard, mouse, joystick, pad, etc.) to move the shape of the first normal heartbeat waveform classified by the electrocardiogram interpretation device (300) to the next waveform to confirm it.

The electrocardiogram waveform interpretation terminal can segment unclassified boxes using the re-grouping function using a dedicated tool provided from the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can control an input means (e.g., a keyboard, a mouse, a joystick, a pad, etc.) to move the representative shape of the first normal rhythm waveform from the first normal rhythm waveform to the PVC waveform. The electrocardiogram waveform interpretation terminal can control an input means (e.g., a keyboard, a mouse, a joystick, a pad, etc.) to move the representative shape of the first normal rhythm waveform from the PVC waveform to the first normal rhythm waveform.

In step S452, the electrocardiogram reading device (300) receives feedback on the first normal heartbeat waveform classification result from the electrocardiogram waveform reading terminal using a dedicated tool and then reflects it in the first normal heartbeat waveform classification result.

The electrocardiogram reading device (300) performs grouping based on waveform shape by clustering only the PVC (Premature Ventricular Contraction) waveform (S454).

In step S454, the electrocardiogram reading device (300) performs shape-based grouping, and identifies the RR interval using the peak points of the P wave and QRS from the beat detection results. The electrocardiogram reading device (300) clusters and groups only the PVC waveforms based on the RR interval.

In step S454, the electrocardiogram reading device (300) outputs a group of PVC waveforms clustered into PVC waveforms using a dedicated tool to the electrocardiogram waveform reading terminal.

The electrocardiogram waveform interpretation terminal outputs a group of PVC waveforms clustered with PVC waveforms input from a dedicated tool of the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can select a representative shape of the PVC waveform and then control an input means (e.g., keyboard, mouse, joystick, pad, etc.) to pass the shape of the PVC waveform classified by the electrocardiogram interpretation device (300) to the next waveform to confirm it.

The electrocardiogram waveform interpretation terminal can segment unclassified boxes using the re-grouping function using a dedicated tool provided from the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can control the representative shape of the PVC waveform by input means (e.g., keyboard, mouse, joystick, pad, etc.) to move from the PVC waveform to the normal heartbeat waveform.

In step S454, the electrocardiogram reading device (300) receives feedback on the result of classifying the PVC waveform from the electrocardiogram waveform reading terminal using a dedicated tool and then reflects it in the PVC waveform classification result.

The electrocardiogram reading device (300) performs shape and interval based grouping based on the normal heartbeat waveform classification result (S460).

In step S460, the electrocardiogram reading device (300) performs regrouping based on the interval based on the normal heartbeat waveform classification result, since the criteria for re-correction are different for each reader after performing S wave detection (S detection).

The electrocardiogram reading device (300) generates a second normal beat wave classification result by clustering only normal beat waves in the process of performing shape and interval-based grouping (S462).

In step S462, the electrocardiogram reading device (300) determines the RR interval using the peak points of the P wave and QRS in the remaining beat waveforms, excluding the waveforms determined to be PVCs, in the process of performing shape and interval-based grouping. The electrocardiogram reading device (300) clusters normal beat waveforms based on the RR interval and generates a second normal beat waveform classification result by grouping them.

In step S462, the electrocardiogram reading device (300) outputs the second normal beat waveform classification result (second normal beat waveform group) clustered into the second normal beat waveform using a dedicated tool to the electrocardiogram waveform reading terminal.

The electrocardiogram waveform interpretation terminal outputs a second normal beat waveform classification result (second normal beat waveform group) clustered with the second normal beat waveform input from a dedicated tool of the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can select a representative shape of the second normal heartbeat waveform and then control an input means (e.g., keyboard, mouse, joystick, pad, etc.) to move the shape of the second normal heartbeat waveform classified by the electrocardiogram interpretation device (300) to the next waveform to confirm it.

The electrocardiogram waveform interpretation terminal can segment unclassified boxes using the re-grouping function using a dedicated tool provided from the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can control an input means (e.g., a keyboard, a mouse, a joystick, a pad, etc.) to move the representative shape of the second normal rhythm waveform from the second normal rhythm waveform to the PAC waveform. The electrocardiogram waveform interpretation terminal can control an input means (e.g., a keyboard, a mouse, a joystick, a pad, etc.) to move the representative shape of the second normal rhythm waveform from the PAC waveform to the second normal rhythm waveform.

In step S462, the electrocardiogram reading device (300) receives feedback on the result of classifying the second normal heartbeat waveform from the electrocardiogram waveform reading terminal using a dedicated tool and then reflects it in the result of classifying the second normal heartbeat waveform.

The electrocardiogram reading device (300) performs grouping by clustering only the PAC (Premature Atrial Contraction) waveform in the process of performing shape and interval-based grouping (S464).

In step S464, the electrocardiogram reading device (300) determines the RR interval using the peak points of the P wave and QRS in the remaining beat waveforms, excluding the waveforms determined to be PVCs, in the process of performing shape and interval-based grouping. The electrocardiogram reading device (300) clusters and groups the PAC waveforms based on the RR interval.

In step S464, the electrocardiogram reading device (300) outputs a group of PAC waveforms clustered into PAC waveforms using a dedicated tool to the electrocardiogram waveform reading terminal.

The electrocardiogram waveform interpretation terminal outputs a PAC waveform group clustered with PAC waveforms input from a dedicated tool of the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can select a representative shape of the PAC waveform and then control an input means (e.g., keyboard, mouse, joystick, pad, etc.) to pass the shape of the PAC waveform classified by the electrocardiogram interpretation device (300) to the next waveform to confirm it.

The electrocardiogram waveform interpretation terminal can segment unclassified boxes using the re-grouping function using a dedicated tool provided from the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can move the representative shape of the PAC waveform from the PAC waveform to the normal heartbeat waveform by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.).

In step S464, the electrocardiogram reading device (300) receives feedback on the result of classifying the PAC waveform from the electrocardiogram waveform reading terminal using a dedicated tool and then reflects it in the PAC waveform classification result.

The electrocardiogram reading device (300) performs arrhythmia detection based on the first normal beat waveform, the second normal beat waveform, the PAC waveform, and the PVC waveform, and selects only the arrhythmia waveform to set the arrhythmia standard (S470).

In step S470, the electrocardiogram reading device (300) outputs an arrhythmia waveform to the electrocardiogram waveform reading terminal using a dedicated tool. The electrocardiogram waveform reading terminal outputs the arrhythmia waveform received from the dedicated tool of the electrocardiogram reading device (300).

The electrocardiogram waveform interpretation terminal can select a representative shape of an arrhythmia waveform and then control an input means (e.g., keyboard, mouse, joystick, pad, etc.) to move the shape of the arrhythmia waveform classified by the electrocardiogram interpretation device (300) to the next waveform to confirm it.

The electrocardiogram waveform interpretation terminal can segment unclassified boxes using the re-grouping function using a dedicated tool provided from the electrocardiogram interpretation device (300).

The electrocardiogram waveform interpretation terminal can determine the arrhythmia waveform criteria (minimum/maximum beat count, BPM, duration) by controlling the representative shape of the arrhythmia waveform using an input means (e.g., keyboard, mouse, joystick, pad, etc.).

In step S470, the electrocardiogram reading device (300) receives feedback on the confirmation result for the arrhythmia waveform from the electrocardiogram waveform reading terminal using a dedicated tool and then reflects it in the arrhythmia waveform classification result.

The electrocardiogram reading device (300) stores histograms of noise waveforms, PVC waveforms, PAC waveforms, and arrhythmia waveforms that reflect user feedback (S480).

Although FIG. 4 describes steps S410 to S480 as being executed sequentially, this is not necessarily limited to the sequential order. In other words, it is possible to modify the steps described in FIG. 4 and execute them, or to execute one or more steps in parallel, and thus FIG. 4 is not limited to a chronological order.

As described above, the method for correcting an electrocardiogram reading based on user feedback according to the present embodiment described in FIG. 4 can be implemented as a program and recorded on a computer-readable recording medium. The computer-readable recording medium on which the program for implementing the method for correcting an electrocardiogram reading based on user feedback according to the present embodiment is recorded includes any type of recording device that stores data that can be read by a computer system.

Fig. 5 is a drawing specifically illustrating an electrocardiogram noise stage detection process according to the present embodiment.

An electrocardiogram noise segmentation device (500) obtains electrocardiogram waveforms for multiple people (S510).

The electrocardiogram noise segmentation device (500) executes a pre-trained deep learning model (S520). In step S520, the electrocardiogram noise segmentation device (500) performs waveform classification on electrocardiogram data for which a noise level is desired to be output using the deep learning model.

The electrocardiogram noise segmentation device (500) applies segmentation to the electrocardiogram waveform using a deep learning model to identify the characteristic index values of each section of the electrocardiogram waveform. The electrocardiogram noise segmentation device (500) labels the classification values for each section based on the characteristic index values (S530).

In step S530, the electrocardiogram noise segmentation device (500) classifies the waveform into P wave, QRS wave, T wave, and noise, derives the result point by point, and assigns a wave segmentation value corresponding to the length of the electrocardiogram signal.

The electrocardiogram noise segmentation device (500) checks the characteristic index values of each section of the P wave, Q wave, R wave, S wave, T wave, and noise waveform included in the electrocardiogram waveform. The electrocardiogram noise segmentation device (500) performs noise analysis based on the characteristic index values of each section and generates noise analysis results (whether baseline changes occur, P wave and T wave non-detection ratio, noise ratio).

The electrocardiogram noise segmentation device (500) outputs the noise level (noise level 0 to noise level 4) based on the noise analysis results (whether there is baseline fluctuation, P wave and T wave non-detection ratio, noise ratio) (S540).

In step S540, the electrocardiogram noise segmentation device (500) analyzes the noise by the presence of baseline wander, the P wave and T wave non-detection ratio, and the noise ratio to segment the noise stage. The electrocardiogram noise segmentation device (500) segments the noise stage into noise stage 0 to noise stage 10.

The electrocardiogram noise segmentation device (500) checks whether the segmented noise stage is less than noise stage 5 (S550).

As a result of the verification in step S550, if the distinguished noise level is less than noise level 5, the electrocardiogram noise segmentation device (500) uses a machine learning algorithm to time-seriesly check the classification values for each section while the values labeled as noise are removed, and classifies and detects electrocardiogram beats (if a P wave is followed by a QRS wave, but not a P wave, arrhythmia, abnormal beat (S, V), ST, QTc, etc.) (S570).

In step S570, the electrocardiogram noise segmentation device (500) determines whether to extract arrhythmia corresponding to the ST segment value, the QT interval (Corrected QT Interval) value, heart rate variability (HRV), or arrhythmia (AFIB, AVB) according to the segmented noise stage.

The electrocardiogram noise segmentation device (500) calculates ST segment, QT interval, HRV (heart rate variability), and arrhythmia when the noise level is classified as noise level 0 (S582). In step S582, the electrocardiogram noise segmentation device (500) calculates all parameters when the noise level is classified as noise level 0, there is no baseline fluctuation in the waveform, both P wave and T wave are detected normally, and there is a normal section without noise.

The electrocardiogram noise segmentation device (500) calculates HRV (heart rate variability) and arrhythmia when the noise level is classified as noise level 1 (S584). In step S584, the electrocardiogram noise segmentation device (500) calculates HRV (heart rate variability) and arrhythmia values when the baseline fluctuation is severe when the noise level is classified as noise level 1, but does not obtain ST segment and QT interval values that are affected by the baseline.

The electrocardiogram noise segmentation device (500) calculates only HRV (heart rate variability) when the noise level is divided into two stages (S586). In step S586, the electrocardiogram noise segmentation device (500) can calculate HRV (heart rate variability) values when the noise level is divided into two stages and the P wave and T wave are not detected normally, but does not obtain ST segment, QT interval, or arrhythmia values.

As a result of the verification in step S550, if the distinguished noise level is not less than noise level 5, the electrocardiogram noise segmentation device (500) outputs noise level 5 to noise level 10 with the entire section processed as noise (S588). In step S588, if the electrocardiogram noise segmentation device (500) is distinguished into noise level 5, in order to increase the processing speed, it does not calculate all parameter values (multiple parameter values such as ST segment, QT interval, arrhythmia (AFIB, AVB), HRV (heart rate variability) values of the corresponding section).

In step S588, if the electrocardiogram noise segmentation device (500) divides the noise stage into 5 noise stages, if the electrocardiogram waveform has a high noise ratio even though some QRS waves are detected normally, it processes the corresponding certain section as noise and does not calculate all parameters.

In step S588, if the electrocardiogram noise segmentation device (500) is classified into noise levels 6 or higher, it does not calculate all parameter values (multiple parameter values such as ST segment, QT interval, arrhythmia (AFIB, AVB), HRV (heart rate variability) values of the corresponding section) to increase the processing speed.

In step S588, if the electrocardiogram noise segmentation device (500) is classified as having a noise level of 6 or higher, and the electrocardiogram patch is not properly attached or connected, and is processed as noise in hardware, all corresponding sections are processed as noise and no parameters are calculated.

Although FIG. 5 describes steps S510 to S588 as being executed sequentially, this is not necessarily the case. In other words, it is possible to change the steps described in FIG. 5 and execute them, or to execute one or more steps in parallel, and thus FIG. 5 is not limited to a chronological order.

As described above, the electrocardiogram noise phase detection process according to the present embodiment described in FIG. 5 can be implemented as a program and recorded on a computer-readable recording medium. The computer-readable recording medium on which the program for implementing the electrocardiogram noise phase detection process according to the present embodiment is recorded includes all types of recording devices that store data that can be read by a computer system.

Fig. 6 is a diagram showing an example of feedback for noise steps by noise waveform using a dedicated tool according to the present embodiment.

As illustrated in Fig. 6, the electrocardiogram reading device (300) classifies the noise level of the noise waveform in detail based on the classification result (P, QRS [Normal, PVC], T, Noise). The electrocardiogram reading device (300) classifies the noise level (e.g., from 0 to 10) by using a noise segmentation algorithm based on baseline wandering, classification result, presence or absence of P wave, T wave, etc. The electrocardiogram reading device (300) transmits the noise level of the noise waveform to the electrocardiogram waveform reading terminal using a dedicated tool.

As illustrated in FIG. 6, the ECG waveform interpretation terminal controls an input means (e.g., a keyboard, a mouse, a joystick, a pad, etc.) to generate a feedback result that determines whether to use only a specific noise level (e.g., less than level 5) among the noise levels received (e.g., from 0 to 10 levels) for analysis and to recognize the remaining noise levels (e.g., more than level 5) as noise and exclude them from analysis. The ECG waveform interpretation terminal transmits the feedback result to the ECG interpretation device (300) using the dedicated tool.

FIG. 7 is a diagram showing an example of feedback for a normal heartbeat waveform and a PVC waveform using a dedicated tool according to the present embodiment.

As illustrated in Fig. 7, the electrocardiogram reading device (300) performs waveform shape-based grouping based on the heartbeat detection results.

The electrocardiogram reading device (300) clusters and groups only normal beat waves in the process of performing waveform shape-based grouping. The electrocardiogram reading device (300) confirms the RR interval using the peak points of the P wave and QRS in the beat detection results in the process of performing shape-based grouping. The electrocardiogram reading device (300) clusters and groups normal beat waves based on the RR interval.

The electrocardiogram reading device (300) clusters and groups only the PVC (Premature Ventricular Contraction) waveforms in the process of performing waveform shape-based grouping. The electrocardiogram reading device (300) confirms the RR interval using the peak points of the P wave and QRS from the beat detection results in the process of performing shape-based grouping. The electrocardiogram reading device (300) clusters and groups only the PVC waveforms based on the RR interval.

As illustrated in Fig. 7, the electrocardiogram reading device (300) outputs the normal heartbeat waveform and PVC waveform classified using a dedicated tool to the electrocardiogram waveform reading terminal.

As illustrated in Figure 7, the ECG waveform interpreter terminal generates a feedback result that moves all peaks in the corresponding box together when the peak points are finely adjusted by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.). Consequently, the ECG waveform interpreter terminal generates a feedback result that reflects the fine adjustment of the peak points to affect S wave detection.

The electrocardiogram waveform interpretation terminal can control the representative shape of the normal rhythm waveform by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.) to move from the normal rhythm waveform to the PVC waveform. The electrocardiogram waveform interpretation terminal can control the representative shape of the PVC waveform by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.) to move from the PVC waveform to the normal rhythm waveform.

The ECG waveform interpreter terminal can generate feedback results included in a report by selecting one of Iso, Couplet, Bigeminy, or Run by using a dedicated tool to input a representative shape box for the PVC waveform.

FIG. 8 is a diagram showing an example of feedback for a normal heartbeat waveform and a PAC waveform using a dedicated tool according to the present embodiment.

As illustrated in Fig. 8, the electrocardiogram reading device (300) performs shape and interval based grouping based on the normal heartbeat waveform classification result (S460).

The electrocardiogram reading device (300) clusters and groups only normal beat waves in the process of performing shape and interval-based grouping. The electrocardiogram reading device (300) confirms the RR interval using the peak points of the P wave and QRS in the remaining beat waves, excluding the waveforms determined to be PVC, in the process of performing shape and interval-based grouping. The electrocardiogram reading device (300) clusters and groups normal beat waves based on the RR interval.

The electrocardiogram reading device (300) clusters and groups only the PAC (Premature Atrial Contraction) waveforms in the process of performing shape and interval-based grouping. The electrocardiogram reading device (300) confirms the RR interval using the peak points of the P wave and QRS in the remaining beat waveforms excluding the waveforms determined to be PVC in the process of performing shape and interval-based grouping. The electrocardiogram reading device (300) clusters and groups the PAC waveforms based on the RR interval.

As illustrated in Fig. 8, the electrocardiogram reading device (300) outputs the normal heartbeat waveform and PAC waveform classified using a dedicated tool to the electrocardiogram waveform reading terminal.

As illustrated in Figure 8, the ECG waveform interpreter terminal generates a feedback result that moves all peaks in the corresponding box together when the peak points are finely adjusted by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.). Consequently, the ECG waveform interpreter terminal generates a feedback result that reflects the fine adjustment of the peak points to affect S wave detection.

The electrocardiogram waveform interpreter terminal can control the representative shape of the normal rhythm waveform by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.) to move from the normal rhythm waveform to the PAC waveform. The electrocardiogram waveform interpreter terminal can control the representative shape of the PAC waveform by controlling the input means (e.g., keyboard, mouse, joystick, pad, etc.) to move from the PVC waveform to the normal rhythm waveform.

The ECG waveform interpreter terminal can generate feedback results included in a report by selecting one of Iso, Couplet, Bigeminy, or Run by using a dedicated tool to input a representative shape box for the PAC waveform.

FIG. 9 is a diagram showing an example of feedback for an arrhythmia detection waveform using a dedicated tool according to this embodiment.

As illustrated in Fig. 9, the electrocardiogram reading device (300) performs arrhythmia detection based on a normal heartbeat waveform, PAC waveform, and PVC waveform, and selects only the arrhythmia waveform to set the arrhythmia criteria.

The electrocardiogram reading device (300) outputs an arrhythmia waveform to an electrocardiogram waveform reading terminal using a dedicated tool.

The ECG waveform interpretation terminal outputs the interpretation results based on the default value for each ECG rhythm, and according to the control of the input means, it passes the corresponding box, checks the waveform captured with the corresponding rhythm, and generates a feedback result confirming the presence or absence of an abnormality.

The ECG waveform interpreter terminal can generate feedback results by selecting appropriate arrhythmia criteria by creating multiple new rhythm boxes by changing the minimum beat rate, maximum beat rate, BPM, duration, etc. if the criteria need to be different.

The ECG waveform interpreter terminal can generate feedback results included in a report by selecting a desired rhythm waveform among the arrhythmia waveforms using a dedicated tool and using an input method.

The above description is merely an example of the technical idea of the present embodiment, and those skilled in the art will appreciate that various modifications and variations can be made without departing from the essential characteristics of the present embodiment. Therefore, the present embodiments are not intended to limit the technical idea of the present embodiment, but rather to explain it, and the scope of the technical idea of the present embodiment is not limited by these embodiments. The scope of protection of the present embodiment should be interpreted by the claims below, and all technical ideas within a scope equivalent thereto should be interpreted as being included in the scope of rights of the present embodiment.

300: Electrocardiogram reading device
310: ECG waveform acquisition unit
320: Classification Department
330: Pulse detection unit
340: Waveform grouping section
350: Feedback Reflection Section
360: Histogram section

Claims (9)

An electrocardiogram waveform acquisition unit for acquiring electrocardiogram waveforms for multiple people;
A classification unit that applies segmentation to the electrocardiogram waveform using a deep learning model, identifies characteristic indicator values of each section of the electrocardiogram waveform, and generates a classification result for each section based on the characteristic indicator values;
A heartbeat detection unit that generates a heartbeat detection result by detecting a heartbeat using a machine learning algorithm based on the above classification result;
A waveform grouping unit that groups each of a noise waveform, a normal beat waveform, a PVC (Premature Ventricular Contraction) waveform, a PAC (Premature Atrial Contraction) waveform, and an arrhythmia waveform based on the above beat detection results;
A feedback reflection unit that receives and reflects feedback results for each of the noise waveform, the normal heartbeat waveform, the PVC waveform, the PAC waveform, and the arrhythmia waveform; and
A histogram unit that stores histograms for the noise waveform, the normal heartbeat waveform, the PVC waveform, the PAC waveform, and the arrhythmia waveform in which the feedback results are reflected;
The waveform grouping unit further includes a noise step segmentation unit that detects noise steps by detecting noise characteristics for each step of the electrocardiogram waveform based on the classification result;
An electrocardiogram reading device characterized in that the feedback reflection unit further includes a noise feedback reflection unit that transmits the noise phase for the noise waveform to an electrocardiogram waveform reading terminal using a dedicated tool and reflects the noise feedback result received from the electrocardiogram waveform reading terminal in the noise phase. delete delete In the first paragraph,
The above waveform grouping unit,
Based on the above beat detection results, a first normal beat waveform classification result is generated by grouping and clustering only normal beat waveforms based on the waveform shape.
A waveform shape-based grouping unit that generates a PVC waveform classification result by grouping only the PVC (Premature Ventricular Contraction) waveforms based on the waveform shape based on the above pulse detection results.
An electrocardiogram reading device characterized by including a . In paragraph 4,
The above feedback reflection section is,
Using a dedicated tool, the first normal rhythm waveform classification result and the PVC waveform classification result are transmitted to the electrocardiogram waveform interpretation terminal,
After receiving the first normal beat feedback result for the first normal beat waveform classification result from the electrocardiogram waveform interpretation terminal, reflect it in the first normal beat waveform classification result,
A waveform shape-based feedback reflection unit that reflects the PVC waveform feedback result for the PVC waveform classification result after receiving the PVC waveform feedback result for the PVC waveform classification result.
An electrocardiogram reading device characterized by including a . In paragraph 4,
The above waveform grouping unit,
Based on the first normal heartbeat waveform classification result, a second normal heartbeat waveform classification result is generated by grouping and clustering only normal heartbeat waveforms based on shape and interval.
A shape and interval-based grouping unit that generates a PAC waveform classification result by clustering and grouping only the PAC (Premature Atrial Contraction) waveforms based on shape and interval based on the first normal rhythm waveform classification result above.
An electrocardiogram reading device characterized by including a . In paragraph 6,
The above feedback reflection section is,
Using a dedicated tool, the second normal heartbeat waveform classification result and the PAC waveform classification result are transmitted to the electrocardiogram waveform interpretation terminal,
After receiving the second normal beat feedback result for the second normal beat waveform classification result from the electrocardiogram waveform interpretation terminal, reflect it in the second normal beat waveform classification result,
A shape and interval-based feedback reflection unit that reflects the PAC waveform classification result after receiving the PAC waveform feedback result for the above PAC waveform classification result.
An electrocardiogram reading device characterized by including a . In the first paragraph,
The above waveform grouping unit,
An arrhythmia detection unit that detects the arrhythmia waveform based on the normal heartbeat waveform, the PVC waveform, and the PAC waveform and sets an arrhythmia standard.
An electrocardiogram reading device characterized by including a . In paragraph 4,
The above feedback reflection section is,
Using a dedicated tool, the arrhythmia waveform is transmitted to the electrocardiogram waveform interpretation terminal,
An arrhythmia reference feedback reflection unit that receives an arrhythmia waveform reference feedback result for the arrhythmia waveform from the electrocardiogram waveform interpretation terminal and then reflects it in the arrhythmia waveform.
An electrocardiogram reading device characterized by including a .