A novel feature vector using an ensemble of MFCC features classified via Artificial Neural Network achieved 95.8% accuracy for predicting the likelihood of cardiac arrests from ECG signals.
Does a novel deep learning framework using an ensemble of MFCC features improve the accuracy of cardiac arrest prediction from ECG signals compared to existing techniques?
A novel deep learning framework using an ensemble of MFCC features and an Artificial Neural Network achieved 95.8% accuracy in predicting cardiac arrest from ECG signals.
Cardiovascular diseases are a major health issue that calls for prompt medical attention. In order to determine the most advantageous methods in this field, numerous techniques and studies have been carried out over the past few years. It should be mentioned that the majority of these cardiovascular illnesses are treatable with earlier detection and prediction. This paper suggests using an automated methodology to predict and classify the likelihood of cardiac arrests in a patient. Due to their collection from various sources, the Electrocardiogram (ECG) signals are initially preprocessed by normalizing them to the 0 1 range. The significant features from these preprocessed signals are then extracted using Mel-Frequency Cepstrum (MFCC), Melspectrogram, MFCC Delta 1 & 2, MFCC Merge File, etc. We also propose a novel feature vector using ensemble of MFCC features. The features are then classified using Artificial Neural Network (ANN), Support Vector Machine (SVM), TPOT and K-Nearest Neighbor (KNN). The proposed vector outperformed all other feature vectors and obtained highest accuracy of 95.8% via ANN. The study is conducted using publicly accessible dataset composed of approximately 52 thousand ECG signals. The suggested strategy is compared with existing techniques and the results indicate the robustness and effectiveness of our approach. Therefore, the proposed methodology can be effectively deployed in a clinical setting to classify ECG data and identify the risks likelihood of cardiac arrests in a patient.
Fatima et al. (Fri,) conducted a other in Cardiac arrest (n=52,000). Ensemble of MFCC features classified via Artificial Neural Network (ANN) vs. Other classifiers (SVM, TPOT, KNN) and existing techniques was evaluated on Classification accuracy. A novel feature vector using an ensemble of MFCC features classified via Artificial Neural Network achieved 95.8% accuracy for predicting the likelihood of cardiac arrests from ECG signals.