Electrocardio diagnosis model system based on AI

Abstract

The invention discloses an electrocardio diagnosis model system based on AI. The electrocardiogram data uploading module is used for uploading the first electrocardiogram data to the data preprocessing module; the data preprocessing module is used for preprocessing the first electrocardiogram data; the first recognition and detection module is used for performing feature extraction and detection on the first electrocardiogram data and outputting a first anomaly detection result; the sampling frequency acquisition module is used for outputting the first electrocardio data to the signal reconstruction module when the first electrocardio data is not abnormal and the sampling frequency is not greater than the standard sampling frequency; the signal reconstruction module is used for performing signal reconstruction according to the first electrocardio data and recovering waveform details of the first electrocardio data; the second recognition and detection module is used for performing feature extraction and detection on the first electrocardiogram data after signal reconstruction and outputting a second anomaly detection result; and the conclusion uploading module is used for sending the anomaly detection reference report to the doctor client. According to the method, the recognition accuracy of the low-sampling-frequency electrocardiogram is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical artificial intelligence, and particularly to an AI-based electrocardiogram diagnosis model system. Background Art

[0002] Electrocardiogram (ECG) diagnosis is an important medical means to evaluate cardiac function and diagnose heart diseases by recording and analyzing the electrical activity of the heart. The core tool for ECG diagnosis is the electrocardiogram (ECG / EKG), which captures the weak electrical signals generated during heartbeats through body surface electrodes and converts them into visible waveform diagrams. These waveforms reflect the rhythm, frequency, and conduction path of the heart's electrical activity.

[0003] In the early stage, it relied on doctors' experience, which had problems of low efficiency and subjectivity. After the rise of artificial intelligence, researchers began to explore the application of artificial intelligence in electrocardiogram analysis. With the development of artificial intelligence, especially the application of convolutional neural networks (CNNs), AI can automatically extract features from massive data, establish a fully automatic diagnosis model, and significantly improve the diagnosis accuracy. AI electrocardiogram diagnosis can be applied to different electrocardiogram devices. However, for some devices with low sampling frequencies, the following situations may occur: 1. Loss of high-frequency features: The details of P waves and QRS waves are blurred, and the slope of the ST segment is distorted. 2. Insufficient time accuracy: The measurement errors of QT intervals and RR intervals increase. During the AI diagnosis process, the loss of high-frequency features easily affects the diagnosis of myocardial ischemia, atrial fibrillation, etc., while insufficient time accuracy easily leads to difficulties in the analysis of long QT syndrome and heart rate variability. Therefore, the recognition accuracy of AI electrocardiogram diagnosis will decrease in electrocardiograms corresponding to low sampling frequencies. Summary of the Invention

[0004] In view of the above-mentioned partial defects of the prior art, the technical problem to be solved by the present invention is to provide an AI-based electrocardiogram diagnosis model system, aiming to improve the recognition accuracy of electrocardiograms with low sampling frequencies.

[0005] To achieve the above object, the present invention discloses an AI-based electrocardiogram diagnosis model system, which includes: an electrocardiogram data uploading module, a data preprocessing module, a first recognition and detection module, a sampling frequency obtaining module, a signal reconstruction module, a second recognition and detection module, and a conclusion uploading module; The electrocardiogram data uploading module is used to upload the first electrocardiogram data to the data preprocessing module; wherein, the first electrocardiogram data at least includes a first electrocardiogram; The data preprocessing module is used to perform denoising filtering and signal segmentation on the first electrocardiogram data to eliminate baseline drift and power frequency interference, and standardize the first electrocardiogram data; The first identification and detection module is used to extract features from the preprocessed first electrocardiogram data, perform anomaly detection based on the extracted features, and output a first anomaly detection result; wherein, the first anomaly detection result includes no anomaly, various anomaly conclusions and their corresponding confidence levels; The sampling frequency acquisition module is used to obtain the first sampling frequency corresponding to the first electrocardiogram data in response to the first anomaly detection result being no anomaly; determine whether the first sampling frequency is greater than the standard sampling frequency, and if so, output that the first electrocardiogram data is normal, and if not, output the first electrocardiogram data to the signal reconstruction module; The signal reconstruction module is used to perform signal reconstruction on the first electrocardiogram data according to the first electrocardiogram data, increase the sampling frequency of the first electrocardiogram data, and restore the waveform details corresponding to the first electrocardiogram data; wherein, the signal reconstruction module is trained with multiple groups of corresponding electrocardiogram data, and the corresponding electrocardiogram data includes electrocardiogram data collected by the same user at the same time period under electrocardiogram devices with a second sampling frequency and a third sampling frequency respectively, the second sampling frequency is greater than the standard sampling frequency, and the third sampling frequency is less than the standard sampling frequency; The second identification and detection module is used to extract features from the first electrocardiogram data after signal reconstruction, perform anomaly detection based on the extracted features, and output a second anomaly detection result; The conclusion uploading module is used to send the anomaly detection reference report to the doctor client, and collect and save the doctor's opinions and treatment suggestions on the anomaly detection reference report; wherein, the anomaly detection reference report includes the first anomaly detection result or the second anomaly detection result.

[0006] Optionally, the signal reconstruction module is specifically used for: performing frequency interpolation on the first electrocardiogram data according to the first electrocardiogram data to increase the sampling frequency corresponding to the first electrocardiogram data; the training process of the signal reconstruction module includes: inputting multiple groups of the corresponding electrocardiogram data to a first training model, so that the first training model obtains the corresponding relationship between the electrocardiogram data less than the standard sampling frequency and the electrocardiogram data greater than the standard sampling frequency in the same group of the corresponding electrocardiogram data, and then obtains the signal reconstruction module, and the signal reconstruction module performs frequency interpolation on the input electrocardiogram data less than the standard sampling frequency according to the corresponding relationship, and outputs the corresponding electrocardiogram data greater than the standard sampling frequency.

[0007] Optionally, the system further includes: a report generation module; The report generation module is used to generate a comprehensive report according to the anomaly detection reference report and the doctor's opinions and treatment suggestions on the anomaly detection reference report.

[0008] Optionally, the data preprocessing module is specifically configured to: Perform denoising filtering on the first electrocardiogram data by using wavelet transform or adaptive filter to eliminate baseline drift and power frequency interference; segment heartbeats according to the R-wave peak, and standardize the first electrocardiogram data to a unified length.

[0009] Optionally, the first identification and detection module is specifically configured to: Use a convolutional neural network to extract the time-frequency domain features of the first electrocardiogram data; adopt a long short-term memory network model or a Transformer model to obtain the first anomaly detection result according to the time-frequency domain features.

[0010] Optionally, the first identification and detection module and the second identification and detection module are common modules.

[0011] Optionally, the training process of the data preprocessing module includes: Mix the noise characteristics of different electrocardiogram devices to generate first training data, and use the first training data to train the data preprocessing module to improve the robustness of the data preprocessing module for preprocessing the first electrocardiogram data.

[0012] Advantages of the present invention: 1. The signal reconstruction module of the present invention is trained by multiple groups of corresponding electrocardiogram data. According to the corresponding relationship between the low-sampling-frequency electrocardiogram data and the high-sampling-frequency data, the low-sampling-frequency electrocardiogram data can be signal-reconstructed. Therefore, when the anomaly of the low-sampling-frequency electrocardiogram data cannot be detected, the signal reconstruction module is used to perform signal reconstruction on the electrocardiogram data to increase its sampling frequency, restore its corresponding waveform details, and then perform detection. With the help of artificial intelligence, problems such as loss of high-frequency features and insufficient time accuracy caused by low sampling frequency are avoided, and the accuracy of anomaly detection is effectively improved. 2. The present invention first performs detection, and when the anomaly of the low-sampling-frequency electrocardiogram data cannot be detected, signal reconstruction is performed. Compared with performing signal reconstruction on the low-sampling-frequency electrocardiogram data at the beginning, the present invention can effectively save computing resources. 3. The present invention sends the anomaly detection reference report to the doctor client to obtain the doctor's opinions and treatment suggestions on the anomaly detection reference report, forming an effective interaction with the experts, and then analyzing the first electrocardiogram data from multiple aspects to avoid the problem of insufficient detection accuracy caused by single-level analysis.

[0013] In summary, the present invention can effectively improve the recognition accuracy of low-sampling-frequency electrocardiograms. Description of the Drawings

[0014] Figure 1 is a schematic structural diagram of an AI-based electrocardiogram diagnosis model system provided by a specific embodiment of the present invention Figure 2It is a business process flowchart in the application process of an AI-based electrocardiogram diagnosis model system provided by a specific embodiment of the present invention. Detailed implementation manners

[0015] The present invention discloses an AI-based electrocardiogram diagnosis model system. Those skilled in the art can draw on the content of this article and appropriately improve technical details for implementation. It should be particularly noted that all similar replacements and modifications are obvious to those skilled in the art, and they are all considered to be included in the present invention. The device and application of the present invention have been described through preferred embodiments. Relevant personnel can obviously make changes or appropriate alterations and combinations to the devices and applications described herein without departing from the content, spirit, and scope of the present invention to implement and apply the technology of the present invention.

[0016] The applicant's research found that: Traditional electrocardiogram diagnosis mainly relies on doctors' experience, and there are problems such as low efficiency, strong subjective judgment, and easy misdiagnosis in complex cases. Most current artificial intelligence-assisted systems are limited to single-level analysis and lack effective interaction with expert consultations. In addition, AI electrocardiogram diagnosis can be applied to different electrocardiogram devices, but some devices have low sampling frequencies, resulting in the following situations: 1. Loss of high-frequency features: The details of P waves and QRS waves are blurred, and the slope of the ST segment is distorted. 2. Insufficient time accuracy: The measurement errors of QT intervals and RR intervals increase. In the process of AI diagnosis, the loss of high-frequency features easily affects the diagnosis of myocardial ischemia, atrial fibrillation, etc., while insufficient time accuracy easily leads to difficulties in the analysis of long QT syndrome and heart rate variability. Therefore, the recognition accuracy of AI electrocardiogram diagnosis will decrease in electrocardiograms corresponding to low sampling frequencies.

[0017] Therefore, the embodiments of the present invention provide an AI-based electrocardiogram diagnosis model system, as Figure 1 shown, this system's electrocardiogram diagnosis model system includes: an electrocardiogram data upload module 101, a data preprocessing module 102, a first recognition and detection module 103, a sampling frequency acquisition module 104, a signal reconstruction module 105, a second recognition and detection module 106, and a conclusion upload module 107; The electrocardiogram data upload module 101 is used to upload the first electrocardiogram data to the data preprocessing module 102; wherein, the first electrocardiogram data includes at least the first electrocardiogram; The data preprocessing module 102 is used to perform denoising filtering and signal segmentation on the first electrocardiogram data to eliminate baseline drift and power frequency interference, and standardize the first electrocardiogram data; The first recognition and detection module 103 is used to extract features from the preprocessed first electrocardiogram data, and perform anomaly detection based on the extracted features, and output a first anomaly detection result; wherein, the first anomaly detection result includes no anomaly and various anomaly conclusions and their corresponding confidence levels; A sampling frequency acquisition module 104, configured to obtain a first sampling frequency corresponding to first electrocardiogram data in response to the first anomaly detection result being normal; determine whether the first sampling frequency is greater than a standard sampling frequency, if so, output the first electrocardiogram data as normal, if not, output the first electrocardiogram data to a signal reconstruction module 105; A signal reconstruction module 105, configured to perform signal reconstruction on the first electrocardiogram data according to the first electrocardiogram data, increase the sampling frequency of the first electrocardiogram data, and restore waveform details corresponding to the first electrocardiogram data; wherein, the signal reconstruction module 105 is trained with multiple groups of corresponding electrocardiogram data, and the corresponding electrocardiogram data includes electrocardiogram data collected by the same user under electrocardiogram devices with a second sampling frequency and a third sampling frequency respectively in the same time period, the second sampling frequency is greater than the standard sampling frequency, and the third sampling frequency is less than the standard sampling frequency; A second identification and detection module 106, configured to extract features from the first electrocardiogram data after signal reconstruction, and perform anomaly detection according to the extracted features, and output a second anomaly detection result; A conclusion upload module 107, configured to send an anomaly detection reference report to a doctor client, and collect and save the doctor's opinions and treatment suggestions on the anomaly detection reference report; wherein, the anomaly detection reference report includes the first anomaly detection result or the second anomaly detection result.

[0018] It should be noted that, in the embodiment of the present invention, through a mode of first identification - then reconstruction - then identification, the electrocardiogram data with a low sampling frequency is effectively identified, which can save computing power while ensuring accurate identification. Compared with the prior art, in the embodiment of the present invention, more cardiac problems of the electrocardiogram data with a low sampling frequency can be identified through signal reconstruction, the detection accuracy is improved, and the occurrence of missed diagnosis and misdiagnosis is avoided.

[0019] In this specific embodiment, multi - device compatibility is supported (such as a 12 - lead electrocardiograph). The system of the embodiment of the present invention is connected to each electrocardiogram device, and the electrocardiogram device automatically uploads the collected data to the cloud platform to ensure data security.

[0020] In this specific embodiment, the signal reconstruction module 105 is specifically configured to: perform frequency interpolation on the first electrocardiogram data according to the first electrocardiogram data to increase the sampling frequency corresponding to the first electrocardiogram data; the training process of the signal reconstruction module 105 includes: inputting multiple groups of corresponding electrocardiogram data to a first training model, so that the first training model obtains the corresponding relationship between the electrocardiogram data less than the standard sampling frequency and the electrocardiogram data greater than the standard sampling frequency in the same group of corresponding electrocardiogram data, and then obtains the signal reconstruction module 105. The signal reconstruction module 105 performs frequency interpolation on the input electrocardiogram data less than the standard sampling frequency according to the corresponding relationship, and outputs the corresponding electrocardiogram data greater than the standard sampling frequency.

[0021] It should be noted that by means of machine learning, the relationship between electrocardiogram (ECG) data with high sampling frequency and low sampling frequency under the same heart condition can be effectively obtained, so that the obtained model can reconstruct the signal of low-sampling-frequency ECG data into high-sampling-frequency ECG data. In addition, the ECG data collected by the same user at the second sampling frequency and the third sampling frequency of the ECG device respectively in the same time period is used to reduce variables, avoid interference from other factors, and restore the relationship as much as possible.

[0022] In this specific embodiment, the system further includes: a report generation module; The report generation module is used to generate a comprehensive report according to the abnormal detection reference report and the doctor's opinions and treatment suggestions on the abnormal detection reference report.

[0023] It should be noted that the integration of AI diagnosis, expert opinions and treatment suggestions supports the PDF format.

[0024] In this specific embodiment, the data preprocessing module 102 is specifically used for: Denoise and filter the first ECG data by using wavelet transform or adaptive filter to eliminate baseline drift and power frequency interference; segment heartbeats according to the R-wave peak, and standardize the first ECG data to a unified length.

[0025] Furthermore, in the embodiment of the present invention, the type of ECG device from which the ECG data comes is first identified, and then according to the type of ECG device, the corresponding method is rotated to perform data preprocessing on it to ensure the accuracy and reliability of the preprocessed data.

[0026] It should be noted that the just-collected ECG data generally contains noise. In order to avoid interference from noise to subsequent detections, data preprocessing effectively solves this problem.

[0027] In this specific embodiment, the first identification and detection module is specifically used for: Use a convolutional neural network to extract the time-frequency domain features of the first ECG data; use a long short-term memory network model or a Transformer model to obtain the first abnormal detection result according to the time-frequency domain features.

[0028] It should be noted that the AI diagnosis and analysis corresponding to the first identification and detection module includes: Feature extraction: Use a convolutional neural network (CNN) to extract time-frequency domain features.

[0029] Abnormal detection: Identify arrhythmias (such as atrial fibrillation, premature ventricular contractions) based on a long short-term memory network (LSTM) or a Transformer model.

[0030] Confidence score: Output an abnormal detection reference report and confidence (the threshold is adjustable, and the default is ≥90% for high confidence).

[0031] In this specific embodiment, the first identification and detection module 103 and the second identification and detection module 106 are common modules.

[0032] It should be noted that the same module can reduce system redundancy.

[0033] In this specific embodiment, the training process of the data preprocessing module 102 includes: Mix the noise characteristics of different electrocardiogram devices to generate first training data, and use the first training data to train the data preprocessing module 102 to improve the robustness of the data preprocessing module 102 for preprocessing the first electrocardiogram data.

[0034] It should be noted that through the above training method, the performance of data preprocessing can be improved, and more accurate electrocardiogram data can be obtained.

[0035] In specific applications, the business process diagram can be as Figure 2 shown, including: the patient side, the hospital side, the doctor side, and the AI side.

[0036] Figure 2 The flowchart of [] details all the links from data collection to report generation, ensuring that every step is accurate. First, the electrocardiogram data of the patient is collected through professional equipment, and then these data are automatically uploaded to the cloud platform for secure storage. In the cloud, the data preprocessing steps begin, including denoising filtering and signal segmentation, to optimize the data quality and lay a solid foundation for subsequent analysis. Next, the corresponding module of AI diagnosis and analysis intervenes. Through feature extraction, anomaly detection, and confidence scoring, it accurately identifies electrocardiogram abnormalities such as arrhythmia. Once the diagnosis is completed, the system automatically generates a comprehensive report containing the AI anomaly detection reference report, expert opinions, and treatment suggestions and distributes it to the patient. The entire process is scientifically designed and efficient, fully reflecting the application value of the embodiments of the present invention in the electrocardiogram consultation scenario.

[0037] The signal reconstruction module 105 of the embodiments of the present invention is trained by multiple groups of corresponding electrocardiogram data. It can perform signal reconstruction on the low-sampling-frequency electrocardiogram data according to the corresponding relationship between the low-sampling-frequency electrocardiogram data and the high-sampling-frequency data. Therefore, when the abnormality of the low-sampling-frequency electrocardiogram data cannot be detected, the signal reconstruction module 105 is used to perform signal reconstruction on the electrocardiogram data to increase its sampling frequency, restore its corresponding waveform details, and then perform detection. With the help of artificial intelligence, problems such as high-frequency feature loss and insufficient time accuracy caused by low sampling frequency are avoided, effectively improving the accuracy of anomaly detection.

[0038] In the embodiments of the present invention, first, detection is performed. When abnormalities in low-sampling-frequency electrocardiogram data cannot be detected, signal reconstruction is then carried out. Compared with performing signal reconstruction on low-sampling-frequency electrocardiogram data at the beginning, the embodiments of the present invention can effectively save computing power resources.

[0039] The embodiments of the present invention send the abnormal detection reference report to the doctor client to obtain the doctor's opinions and treatment suggestions on the abnormal detection reference report, form an effective interaction with the experts, and then analyze the first electrocardiogram data from multiple aspects to avoid the problem of insufficient detection accuracy caused by single-level analysis.

[0040] In summary, the embodiments of the present invention can effectively improve the recognition accuracy of low-sampling-frequency electrocardiograms.

[0041] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including the element.

[0042] Each embodiment in this specification is described in a related manner. For the same or similar parts between the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment.

[0043] The above are only the preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are all included in the protection scope of the present invention.

Claims

1. An AI-based electrocardiogram diagnosis model system, characterized in that, The electrocardiogram diagnosis model system includes: an electrocardiogram data upload module, a data preprocessing module, a first identification and detection module, a sampling frequency acquisition module, a signal reconstruction module, a second identification and detection module, and a conclusion upload module; The electrocardiogram data upload module is used to upload the first electrocardiogram data to the data preprocessing module; wherein, the first electrocardiogram data at least includes a first electrocardiogram; The data preprocessing module is used to perform denoising filtering and signal segmentation on the first electrocardiogram data to eliminate baseline drift and power frequency interference, and standardize the first electrocardiogram data; The first identification and detection module is used to extract features from the preprocessed first electrocardiogram data, and perform anomaly detection based on the extracted features, and output a first anomaly detection result; wherein, the first anomaly detection result includes no anomaly and various anomaly conclusions and their corresponding confidence levels; The sampling frequency acquisition module is used to obtain the first sampling frequency corresponding to the first electrocardiogram data in response to the first anomaly detection result being no anomaly; judge whether the first sampling frequency is greater than the standard sampling frequency, if so, output that the first electrocardiogram data is normal, if not, output the first electrocardiogram data to the signal reconstruction module; The signal reconstruction module is used to perform signal reconstruction on the first electrocardiogram data according to the first electrocardiogram data to increase the sampling frequency of the first electrocardiogram data and restore the waveform details corresponding to the first electrocardiogram data; wherein, the signal reconstruction module is trained with multiple groups of corresponding electrocardiogram data, and the corresponding electrocardiogram data includes electrocardiogram data collected by the same user at the same time period under electrocardiogram devices with a second sampling frequency and a third sampling frequency respectively, the second sampling frequency is greater than the standard sampling frequency, and the third sampling frequency is less than the standard sampling frequency; The second identification and detection module is used to extract features from the first electrocardiogram data after signal reconstruction, and perform anomaly detection based on the extracted features, and output a second anomaly detection result; The conclusion upload module is used to send the anomaly detection reference report to the doctor client, and collect and save the doctor's opinions and treatment suggestions on the anomaly detection reference report; wherein, the anomaly detection reference report includes the first anomaly detection result or the second anomaly detection result.

2. The AI-based electrocardiogram diagnosis model system according to claim 1, wherein The signal reconstruction module is specifically used for: performing frequency interpolation on the first electrocardiogram data according to the first electrocardiogram data to increase the sampling frequency corresponding to the first electrocardiogram data; the training process of the signal reconstruction module includes: inputting multiple groups of the corresponding electrocardiogram data to a first training model, so that the first training model obtains the corresponding relationship between the electrocardiogram data less than the standard sampling frequency and the electrocardiogram data greater than the standard sampling frequency in the same group of the corresponding electrocardiogram data, and then obtains the signal reconstruction module, and the signal reconstruction module performs frequency interpolation on the input electrocardiogram data less than the standard sampling frequency according to the corresponding relationship, and outputs the corresponding electrocardiogram data greater than the standard sampling frequency.

3. The AI-based electrocardiogram diagnosis model system according to claim 1, characterized in that, The system further includes: a report generation module; The report generation module is used to generate a comprehensive report according to the abnormal detection reference report and the doctor's opinions and treatment suggestions on the abnormal detection reference report.

4. The AI-based electrocardiogram diagnosis model system according to claim 1, wherein The data preprocessing module is specifically used for: Performing denoising filtering on the first electrocardiogram data by using wavelet transform or an adaptive filter to eliminate baseline drift and power frequency interference; segmenting heartbeats according to the R-wave peak, and normalizing the first electrocardiogram data to a unified length.

5. The AI-based electrocardiogram diagnosis model system according to claim 1, characterized in that, The first identification and detection module is specifically used for: Using a convolutional neural network to extract the time-frequency domain features of the first electrocardiogram data; obtaining the first abnormal detection result according to the time-frequency domain features by using a long short-term memory network model or a Transformer model.

6. The AI-based electrocardiogram diagnosis model system according to claim 1, characterized in that, The first identification and detection module and the second identification and detection module are common modules.

7. The AI-based electrocardiogram diagnosis model system according to claim 1, wherein The training process of the data preprocessing module includes: Mixing the noise characteristics of different electrocardiogram devices to generate first training data, and using the first training data to train the data preprocessing module to improve the robustness of the data preprocessing module for preprocessing the first electrocardiogram data.

Images