Method, device, equipment and storage medium for evaluating reliability of electrocardio detection result

The reliability of ECG test results is evaluated by a neural network model trained by a self-supervised method, which solves the problem of overfitting in ECG analysis by deep learning algorithms and achieves efficient and low-cost reliability evaluation, applicable to a variety of ECG analysis tasks.

CN116172571BActive Publication Date: 2026-02-27GUANGZHOU SHIYUAN ELECTRONICS CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202111433648.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-11-29
Publication Date
2026-02-27
Estimated Expiration
2041-11-29

AI Technical Summary

Technical Problem

Deep learning algorithms are prone to overfitting in electrocardiogram (ECG) analysis. Mismatched training data distributions lead to unreliable analysis results. Furthermore, data acquisition is costly, and it is difficult to determine whether the data distribution reflects the true ECG distribution.

Method used

A self-supervised method was used to train a waveform feature extraction and reconstruction model. The reliability of the ECG test results was evaluated by analyzing the error between the reconstructed waveform and the original waveform. A pre-trained neural network model was used for waveform feature extraction and reconstruction, and a reliability score was calculated.

Benefits of technology

No additional manual annotation is required, reducing data annotation costs. It is lightweight and does not affect the running time and accuracy of ECG analysis algorithms. It can be applied to a variety of ECG analysis tasks and accurately assess the reliability of ECG test results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116172571B_ABST
    Figure CN116172571B_ABST
Patent Text Reader

Abstract

The application discloses a method, device and equipment for evaluating the reliability of electrocardio detection results and a storage medium. The method comprises the following steps: inputting an electrocardio detection original waveform into a pre-trained waveform feature extraction model to obtain extracted waveform features; inputting the waveform features into a pre-trained waveform reconstruction model to obtain a reconstructed waveform; and obtaining the reliability of the electrocardio detection results according to the reconstructed waveform and the electrocardio detection original waveform. According to the method for evaluating the reliability of electrocardio detection results provided in the application, the waveform features of electrocardio detection can be reconstructed through a pre-trained neural network model, and the reliability of the electrocardio detection results can be obtained by analyzing the error between the reconstructed waveform and the original waveform. The method does not require additional manual labeling, has no influence on the original electrocardio analysis algorithm, and can accurately obtain the reliability of the electrocardio detection results.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Power system fault diagnosis method comprehensively using electricity amount and timing sequence information

    CN104297637A

  • Three-heart beat multi-model comprehensive decision electrocardiogram feature classification method fused with source end influence

    CN111297350A

  • Garbage classification model modeling method, garbage classification method, garbage classification device and medium

    CN113705729A

  • Coding architectures for automatic analysis of waveforms

    US20210304855A1