Dynamic ECG signal quality evaluation based on persistent homology and GoogLeNet method

Abstract

Cardiovascular disease is a serious health problem. Continuous Electrocardiograph (ECG) monitoring plays a vital role in the early detection of cardiovascular disease. As the Internet of Things technology continues to mature, wearable ECG signal monitors have been widely used. However, dynamic ECG signals are extremely susceptible to contamination. Therefore, it is necessary to evaluate the quality of wearable dynamic ECG signals. The topological data analysis method (TDA) with persistent homology, which can effectively capture the topological information of high-dimensional data space, has been widely studied. In this study, a brand-new quality assessment method of wearable dynamic ECG signals was proposed based on the TDA with persistent homology method. The point cloud of an ECG signal was constructed, and then the complex sequence was generated and displayed as a persistent barcode. Finally, GoogLeNet based on the transfer learning model with a 10-fold cross-validation method was used to train the classification model. A total of 12-leads ECGs Dataset and single-lead ECGs Dataset, established based on the 2011 PhysioNet/CinC challenge dataset, were both used to verify the performance of this method. In the study, 773 “acceptable” and 225 “unacceptable” signals were used as 12-leads ECGs Dataset. We relabeled 12,000 ECG signals in the challenge dataset, and treated them as single-lead ECGs Dataset after empty lead detection and balance datasets. Compared with the traditional ECG signal quality assessment method mainly based on waveform characteristics and time-frequency characteristics, the performance of the quality assessment method proposed. In this study, the classification performance of the proposed method are fairly great, mAcc = 98.04%, F1 = 98.40%, Se = 97.15%, Sp = 98.93% for 12-leads ECGs Dataset and mAcc = 98.55%, F1 = 98.62%, Se = 98.37%, Sp = 98.85% for single-lead ECGs Dataset.

Electrocardiogram arrhythmia detection with novel signal processing and persistent homology-derived predictors

“…Future directions include multiclass rhythm classification, other methods of defining the isoelectric baseline to account for baseline wander in longer ECG signals, including statistics derived from optimal cycle representatives in other approaches such as sliding window and Fast Fourier Transform embeddings, and including an isoelectric baseline prior to embedding ECG signals in higher dimensions. Several studies have used TDA-derived statistics as input to neural networks [16,53,55]; however, to the author's knowledge, there has been no study performed which utilizes persistence images [1] as the TDA-derived input for neural networks in arrhythmia detection, yielding another direction for future work.…”

“…Additionally, the Mapper algorithm has been applied to predict the presence and severity of heart disease [2]. Computer-aided ECG rhythm classification methods which utilize TDA include neural networks with topological-based features [16,53], fractal dimension in tandem with neural networks [55], mapping ECG signals to a higher dimensional space prior to computing topological features [26,27,34,36,41], and utilizing a sliding window and Fast Fourier Transform to process the ECG signal prior to computing topological features [43]. These approaches construct topological predictor variables utilizing information directly derived from the birth and death radii statistics along with extra information such as heart rate, fractal dimension statistics, and persistent entropy.…”

Current and Future Use of Artificial Intelligence in Electrocardiography

“…Those metrics are a simple way to track or compare the performances of different Those challenges try to solve relevant clinical problems, such as arrhythmia classification using different lead sets ranging from two to twelve ECG leads amongst the vast amount of twelve-lead ECG recordings in the PhysioNet CinC Challenges 2020 and 2021. They are presently known as the largest freely available repository of standard 12-lead ECG records and consistent annotations for 30 clinical diagnoses of cardiac abnormalities [45][46][47][48][49].…”

Validation of a New and Straightforward Algorithm to Evaluate Signal Quality during ECG Monitoring with Wearable Devices Used in a Clinical Setting

“…Today, dedicated signal acquisition software automatically classifies signal quality [ 18 ] with traditional and novel algorithms (for example, by implanting artificial intelligence—AI) [ 19 , 20 , 21 , 22 , 23 , 24 ]. These methods have variable sensitivity and specificity, with an overall accuracy usually >90% [ 6 ].…”