Deep learning-based magnetocardiogram signal multi-angle feature extraction and selection method

Through deep learning technology, multi-angle features of cardiac magnetic signals are extracted, combined with the feature selection method of adaptive LASSO, the problems of strong subjectivity and low robustness of feature extraction in traditional methods are solved, and automatic classification of diseases with high accuracy and reliability are achieved.

CN119719748AActive Publication Date: 2025-03-28BEIHANG UNIV
5 Cites 0 Cited by

Patent Information

Application Number
CN202411862868.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-17
Publication Date
2025-03-28
Estimated Expiration
2044-12-17

AI Technical Summary

Technical Problem

The traditional magnetic cardiac signal classification method relies on manual feature extraction and time-frequency domain analysis, and has problems such as strong subjectivity, low characteristic robustness and difficulty in adapting to different signal characteristics.

Method used

The multi-angle feature extraction method based on deep learning is adopted, combining time domain, frequency domain and multimodal features, time domain features are extracted using learnable position coding and CNN-Transformer model, frequency domain features are extracted by HHT and CNN models, multimodal features are extracted by adaptive VMD and center frequency weighting method, and feature selection is performed through adaptive LASSO.

Benefits of technology

An automated feature extraction and classification mechanism is realized, which improves the accuracy and reliability of automatic classification of diseases, reduces the dependence on expert knowledge, and can adapt to different disease diagnosis tasks.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to a magnetocardiogram signal multi-angle feature extraction and selection method based on deep learning, and the method comprises the steps: carrying out the preprocessing of a magnetocardiogram signal; a magnetocardiogram time domain feature extraction method based on learnable position coding and a CNN-Transformer deep learning model, a magnetocardiogram frequency domain feature extraction method based on HHT and the CNN deep learning model, and a magnetocardiogram multi-modal feature extraction method based on adaptive VMD and center frequency weighting are designed. And selecting a magnetocardiogram multi-angle feature subset which has the greatest influence on disease classification through a feature selection model based on adaptive LASSO. Compared with a traditional magnetocardiogram feature extraction and selection method, the feature extraction and selection method provided by the invention can automatically and more accurately extract magnetocardiogram multi-angle features, and the extracted features are multiple in angle, good in robustness, high in reliability and wide in applicable disease classification range.
Need to check novelty before this filing date? Find Prior Art