Adaptive Toeplitz Convolution- enhanced Classifier for Anomaly Detection in ECG Big Data DOI Creative Commons

Lili Wu,

Majid Khan Majahar Ali, Tao Li

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Июль 30, 2024

Abstract The anomaly detection of electrocardiogram (ECG) data is crucial for identifying deviations from normal heart rhythm patterns and providing timely interventions high-risk patients. Various autoencoder (AE) models within machine learning (ML) have been proposed this task. However, these often do not explicitly consider the specific in ECG time series, thereby impacting their efficiency. In contrast, we adopt a method based on prior knowledge series shapes, employing multi-stage preprocessing, adaptive convolution kernels, Toeplitz matrices to replace encoding part AE. This approach combines inherent features with symmetry matrices, effectively extracting signals reducing dimensionality. Our model consistently outperforms state-of-the-art detection, achieving an overall accuracy exceeding 99.6%, Precision Area Under Receiver Operating Characteristic Curve (AUC) reaching 99.8%, Recall peaking at 99.9%. Moreover, runtime significantly reduced. These results demonstrate that our technique detects anomalies through automatic feature extraction enhances performance ECG5000 dataset, benchmark collection heartbeat signals.

Язык: Английский

Adaptive Toeplitz Convolution- enhanced Classifier for Anomaly Detection in ECG Big Data DOI Creative Commons

Lili Wu,

Majid Khan Majahar Ali, Tao Li

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Июль 30, 2024

Abstract The anomaly detection of electrocardiogram (ECG) data is crucial for identifying deviations from normal heart rhythm patterns and providing timely interventions high-risk patients. Various autoencoder (AE) models within machine learning (ML) have been proposed this task. However, these often do not explicitly consider the specific in ECG time series, thereby impacting their efficiency. In contrast, we adopt a method based on prior knowledge series shapes, employing multi-stage preprocessing, adaptive convolution kernels, Toeplitz matrices to replace encoding part AE. This approach combines inherent features with symmetry matrices, effectively extracting signals reducing dimensionality. Our model consistently outperforms state-of-the-art detection, achieving an overall accuracy exceeding 99.6%, Precision Area Under Receiver Operating Characteristic Curve (AUC) reaching 99.8%, Recall peaking at 99.9%. Moreover, runtime significantly reduced. These results demonstrate that our technique detects anomalies through automatic feature extraction enhances performance ECG5000 dataset, benchmark collection heartbeat signals.

Язык: Английский

Процитировано

0