A pyramid-like model for heartbeat classification from ECG recordings

A lightweight network based on multi-feature pseudo-color mapping for arrhythmia recognition

Heartbeats classification is a crucial tool for arrhythmia diagnosis. In this study, a multi-feature pseudo-color mapping (MfPc Mapping) was proposed, and a lightweight FlexShuffleNet was designed to classify heartbeats. MfPc Mapping converts one-dimensional (1-D) electrocardiogram (ECG) recordings into corresponding two-dimensional (2-D) multi-feature RGB graphs, and it offers good excellent interpretability and data visualization. FlexShuffleNet is a lightweight network that can be adapted to classification tasks of varying complexity by tuning hyperparameters. The method has three steps. The first step is data preprocessing, which includes de-noising the raw ECG recordings, removing baseline drift, extracting heartbeats, and performing data balancing, the second step is transforming the heartbeats using MfPc Mapping. Finally, the FlexShuffleNet is employed to classify heartbeats into 14 categories. This study was evaluated on the test set of the MIT-BIH arrhythmia database (MIT/BIH DB), and it yielded the results i.e., accuracy of 99.77%, sensitivity of 94.60%, precision of 89.83% and specificity of 99.85% and F1-score of 0.9125 in 14-category classification task. Additionally, validation on Shandong Province Hospital database (SPH DB) yielded the results i.e., accuracy of 92.08%, sensitivity of 93.63%, precision of 91.25% and specificity of 99.85% and F1-score of 0.9315. The results show the satisfied performance of the proposed method.

Relationship between Circle of Willis Variations and Cerebral or Cervical Arteries Stenosis Investigated by Computer Tomography Angiography and Multitask Convolutional Neural Network

Circle of Willis (CoW) is the most critical collateral pathway that supports the redistribution of blood supply in the brain. The variation of CoW is closely correlated with cerebral hemodynamic and cerebral vessel-related diseases. But what is responsible for CoW variation remains unclear. Moreover, the visual evaluation for CoW variation is highly time-consuming. In the present study, based on the computer tomography angiography (CTA) dataset from 255 patients, the correlation between the CoW variations with age, gender, and cerebral or cervical artery stenosis was investigated. A multitask convolutional neural network (CNN) was used to segment cerebral arteries automatically. The results showed the prevalence of variation of the anterior communicating artery (Aco) was higher in the normal senior group than in the normal young group and in females than in males. The changes in the prevalence of variations of individual segments were not demonstrated in the population with stenosis of the afferent and efferent arteries, so the critical factors for variation are related to genetic or physiological factors rather than pathological lesions. Using the multitask CNN model, complete cerebral and cervical arteries could be segmented and reconstructed in 120 seconds, and an average Dice coefficient of 78.2% was achieved. The segmentation accuracy for precommunicating part of anterior cerebral artery and posterior cerebral artery, the posterior communicating arteries, and Aco in CoW was 100%, 99.2%, 94%, and 69%, respectively. Artificial intelligence (AI) can be considered as an adjunct tool for detecting the CoW, particularly related to reducing workload and improving the accuracy of the visual evaluation. The study will serve as a basis for the following research to determine an individual's risk of stroke with the aid of AI.

Research on Data Analysis of Traditional Chinese Medicine with Improved Differential Evolution Clustering Algorithm

Medical data analysis is an important part of intelligent medicine, and clustering analysis is a commonly used method for data analysis of Traditional Chinese Medicine (TCM); however, the classical K-Means algorithm is greatly affected by the selection of initial clustering center, which is easy to fall into the local optimal solution. To avoid this problem, an improved differential evolution clustering algorithm is proposed in this paper. The proposed algorithm selects the initial clustering center randomly, optimizes and locates the clustering center in the process of evolution iteration, and improves the mutation mode of differential evolution to enhance the overall optimization ability, so that the clustering effect can reach the global optimization as far as possible. Three University of California, Irvine (UCI), data sets are selected to compare the clustering effect of the classical K-Means algorithm, the standard DE-K-Means algorithm, the K-Means++ algorithm, and the proposed algorithm. The experimental results show that, in terms of global optimization, the proposed algorithm is obviously superior to the other three algorithms, and in terms of convergence speed, the proposed algorithm is better than DE-K-Means algorithm. Finally, the proposed algorithm is applied to analyze the drug data of Traditional Chinese Medicine in the treatment of pulmonary diseases, and the analysis results are consistent with the theory of Traditional Chinese Medicine.

Heartbeats Classification Using Hybrid Time-Frequency Analysis and Transfer Learning Based on ResNet

The classification of heartbeats is an important method for cardiac arrhythmia analysis. This study proposes a novel heartbeat classification method using hybrid time-frequency analysis and transfer learning based on ResNet-101. The proposed method has the following major advantages over the afore-mentioned methods: it avoids the need for manual features extraction in the traditional machine learning method, and it utilizes 2-D time-frequency diagrams which provide not only frequency and energy information but also preserve the morphological characteristic within the ECG recordings, and it owns enough deep to make better use of performance of CNN. The method deploys a hybrid time-frequency analysis of the Hilbert transform (HT) and the Wigner-Ville distribution (WVD) to transform 1-D ECG recordings into 2-D time-frequency diagrams which were then fed into a transfer learning classifier based on ResNet-101 for two classification tasks (i.e., 5 heartbeat categories assigned by the ANSI/AAMI standard (i.e., N, V, S, Q and F) and 14 original beat kinds of the MIT/BIH arrhythmia database). For 5 heartbeat categories classification, the results show the F1-score of N, V, S, Q and F categories are FN 0.9899, FV 0.9845, FS 0.9376, FQ 0.9968, FF 0.8889, respectively, and the overall F1-score is 0.9595 using the combination data balancing. The results show the average values for accuracy, sensitivity, specificity, predictive value and F1-score on test set for 14 beat kinds the MIT-BIH arrhythmia database are 99.75%, 91.36%, 99.85%, 90.81% and 0.9016, respectively. Compared with other methods, the proposed method can yield more accurate results.

Extended Segmented Beat Modulation Method for Cardiac Beat Classification and Electrocardiogram Denoising

Beat classification and denoising are two challenging and fundamental operations when processing digital electrocardiograms (ECG). This paper proposes the extended segmented beat modulation method (ESBMM) as a tool for automatic beat classification and ECG denoising. ESBMM includes four main steps: (1) beat identification and segmentation into PQRS and TU segments; (2) wavelet-based time-frequency feature extraction; (3) convolutional neural network-based classification to discriminate among normal (N), supraventricular (S), and ventricular (V) beats; and (4) a template-based denoising procedure. ESBMM was tested using the MIT–BIH arrhythmia database available at Physionet. Overall, the classification accuracy was 91.5% while the positive predictive values were 92.8%, 95.6%, and 83.6%, for N, S, and V classes, respectively. The signal-to-noise ratio improvement after filtering was between 0.15 dB and 2.66 dB, with a median value equal to 0.99 dB, which is significantly higher than 0 (p < 0.05). Thus, ESBMM proved to be a reliable tool to classify cardiac beats into N, S, and V classes and to denoise ECG tracings.