A scalogram tensor decomposition based ECG quality assessment.

ECG quality assessment is crucial for reducing false alarms and physician strain in automated diagnosis of cardiovascular diseases. Recent researches have focused on constructing an automatic noisy ECG record rejection mechanism. This work develops a noisy ECG record rejection system using scalogram and Tucker tensor decomposition. The system can reject ECG records, which cannot be analyzed or diagnosed. Scalogram of all 12­lead ECG signals per subject are stacked to form a 3-way tensor. Tucker tensor decomposition is applied with empirical settings to obtain the core tensor. The core tensor is reshaped to form the latent features set. When tested using the PhysioNet challenge 2011 dataset in five-fold cross validation settings, the RusBoost ensemble classifier proved to be a very reliable option, producing an accuracy of 92.4% along with sensitivity of 87.1% and specificity of 93.5%. According to the experimental findings, combining the scalogram with Tucker tensor decomposition yields competitive performance and has the potential to be used in actual evaluation of ECG quality.

Early Prediction of Sepsis From Clinical Data Using Ratio and Power-Based Features.

OBJECTIVES: Early prediction of sepsis is of utmost importance to provide optimal care at an early stage. This work aims to deploy soft-computing and machine learning techniques for early prediction of sepsis. DESIGN: An algorithm for early identification of sepsis using ratio and power-based feature transformation of easily obtainable clinical data. SETTING: PhysioNet Challenge 2019 provided ICU data from three separate hospital systems. Publicly shared data from two hospital systems are used for training and validation purposes, whereas sequestered data from all the three systems is used for testing. PATIENTS: Over 60,000 ICU patients with up to 40 clinical variables are sourced for each hour of their ICU stay. The Sepsis-3 criterion is applied for annotation. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The clinical feature exploration for early prediction of sepsis is achieved using the proposed framework named genetic algorithm optimized ratio and power-based expert algorithm. An optimal feature set containing 46 ratio and power-based features is computed from the given patient covariates using genetic algorithm optimized ratio and power-based expert and grouped with identified 17 raw features and 55 statistical features to form a final feature set of 118 clinical features to predict the onset of sepsis in the proceeding 6 hours. The obtained features are fed to a hybrid Random Under-Sampling-Boosting algorithm, called RUSBoost for alleviating the involved class imbalance. The optimal RUSBoost model has achieved a normalized utility score of 0.318 on full test data. CONCLUSIONS: The proposed study supports the realization of a hospital-specific customized solution in the form of an early-warning system for sepsis. However, an extended analysis is necessary to apply this framework for hospital-independent diagnosis of sepsis in general. Nevertheless, the clinical utility of hospital-specific customized solutions based on the proposed method across a wide range of hospital systems needs to be studied.

Application of Fourier-Bessel expansion and LSTM on multi-lead ECG for cardiac abnormalities identification.

Objective. The availability of online electrocardiogram (ECG) repositories can aid researchers in developing automated cardiac abnormality diagnostic systems. Using such ECG repositories, this study aims to develop an algorithm that can assist physicians in diagnosing cardiac abnormalities.Approach. The PhysioNet/CinC 2021 Challenge has opened the venues for creating benchmark algorithms using standard and relatively diverse 12-lead ECG datasets. This work attempts to create a new machine learning approach for identifying common cardiac abnormalities using an ensemble-based classification with two models resulting from two different feature sets. The first feature set extracts RR variability based information by deploying Fourier-Bessel (FB) expansion. The second feature set is composed of time- and frequency-domains-based hand-crafted features. Two long short-term memory (LSTM)-based classifiers are trained using these two feature sets as input to categorize ECG signals. Predictions from these two models are fused to arrive at a final medical decision that improves the multi-label classification of the given ECG signals into twenty-six categories.Main results. We participated in the George B. Moody Physionet Challenge 2021 as team 'Medics', and the proposed methodology was evaluated for all five lead combinations. The challenge scoring metrics obtained on the test data for twelve-, six-, four-, three-, and two-leads combinations are 0.360, 0.368, 0.376, 0.323, and 0.381, respectively. The proposed methodology was ranked 11th among all the follow-up entries of the Challenge.Significance. The obtained results of the proposed method justify the use of an ensemble classifier developed using the extracted feature sets for devising a diagnostic system for detecting and identifying common cardiac problems.

Automated diagnosis of coronary artery disease using scalogram-based tensor decomposition with heart rate signals.

Early identification of coronary artery disease (CAD) can facilitate timely clinical intervention and save lives. This study aims to develop a machine learning framework that uses tensor analysis on heart rate (HR) signals to automate the CAD detection task. A third-order tensor representing a time-frequency relationship is constructed by fusing scalograms as vertical slices of the tensor. Each scalogram is computed from the considered time frame of a given HR signal. The derived scalogram represents the heterogeneity of data as a two-dimensional map. These two-dimensional maps are stacked one after the other horizontally along the z-axis to form a 3-way tensor for each HR signal. Each two-dimensional map is represented as a vertical slice in the xy - plane. Tensor factorization of such a fused tensor for every HR signal is performed using canonical polyadic (CP) decomposition. Only the core factor is retained later, excluding the three unitary matrices to provide the latent feature set for the detection task. The resultant latent features are then fed to machine learning classifiers for binary classification. Bayesian optimization is performed in a five-fold cross-validation strategy in search of the optimal machine learning classifier. The experimental results yielded the accuracy, sensitivity, and specificity of 96.62%, 96.53%, and 96.67%, respectively, with the bagged trees ensemble method. The proposed tensor decomposition deciphered higher-order interrelations among the considered time-frequency representations of HR signals.

Tensor learning of pointwise mutual information from EHR data for early prediction of sepsis.

Early detection of sepsis can facilitate early clinical intervention with effective treatment and may reduce sepsis mortality rates. In view of this, machine learning-based automated diagnosis of sepsis using easily recordable physiological data can be more promising as compared to the gold standard rule-based clinical criteria in current practice. This study aims to develop such a machine learning framework that demonstrates the quantification of heterogeneity within the tabular electronic health records (EHR) data of clinical covariates to capture both linear relationships and nonlinear correlation for the early prediction of sepsis. Here, the statistics of pairwise association for each hour-covariate pair within the EHR data for every 6-hours window-duration with selected 24 covariates is described using pointwise mutual information (PMI) matrix. This matrix gives the heterogeneity of data as a two-dimensional map. Such matrices are fused horizontally along the z-axis as vertical slices in the xy plane to form a 3-way tensor for each record with the corresponding Length of Stay (L). Tensor factorization of such fused tensor for every record is performed using Tucker decomposition, and only the core tensors are retained later, excluding the 3 unitary matrices to provide the latent feature set for the prediction of sepsis onset. A five-fold cross-validation scheme is employed wherein the obtained 120 latent features from the reshaped core tensor, are fed to Light Gradient Boosting Machine Learning models (LightGBM) for binary classification, further alleviating the involved class imbalance. The machine-learning framework is designed via Bayesian optimization, yielding an average normalized utility score of 0.4519 as defined by challenge organizers and area under the receiver operating characteristic curve (AUROC) of 0.8621 on publicly available PhysioNet/Computing in Cardiology Challenge 2019 training data. The proposed tensor decomposition of 3-way fused tensor formulated using PMI matrices leverages higher-order temporal interactions between the pairwise associations among the clinical values for early prediction of sepsis. This is validated with improved risk prediction power for every hour of admission to the ICU in terms of utility score, AUROC, and F1 score. The results obtained show a significant improvement particularly in terms of utility score of ~1.5-2% under a 5-fold cross-validation scheme on entire training data as compared to a top entrant research study that participated in the challenge.

Automated pre-screening of arrhythmia using hybrid combination of Fourier-Bessel expansion and LSTM.

Health care in developing countries demands systems-based screening solutions. In view of this, we present a new rhythm-based methodology for the point-of-care diagnosis of cardiac arrhythmia at a primary level. Such a system will reduce the workload of cardiologists significantly. The method begins by computing the RR-interval sequences from the electrocardiogram(ECG) signals. Then, the Fourier-Bessel (FB) expansion is used to obtain the intelligent series by converting the RR-interval sequences into more meaningful sequences that can characterize the underlying pathology of cardiac arrhythmia with a unique pattern. Ultimately, the obtained intelligent series are used as input to train the long short-term memory (LSTM) model for ECG classification. We have obtained an accuracy of 90.07% in classifying normal and the arrhythmia classes using MIT-BIH database. The results demonstrate that the proposed intelligent series can reveal remarkable differences between the normal and arrhythmia ECG signals. Thus, the proposed algorithm can be used as a primary screening tool for detecting cardiac arrhythmia. Potentially, the developed system can be used by paramedics in rural outreach programs with limited funding and expertise. Moreover, the use of single-lead and short-length ECG signals in the proposed system makes it a suitable candidate for applications that are intended for mobile and other hand-held or wearable devices.

Epileptic seizure classification in EEG signals using second-order difference plot of intrinsic mode functions.

Epilepsy is a neurological disorder which is characterized by transient and unexpected electrical disturbance of the brain. The electroencephalogram (EEG) is a commonly used signal for detection of epileptic seizures. This paper presents a new method for classification of ictal and seizure-free EEG signals. The proposed method is based on the empirical mode decomposition (EMD) and the second-order difference plot (SODP). The EMD method decomposes an EEG signal into a set of symmetric and band-limited signals termed as intrinsic mode functions (IMFs). The SODP of IMFs provides elliptical structure. The 95% confidence ellipse area measured from the SODP of IMFs has been used as a feature in order to discriminate seizure-free EEG signals from the epileptic seizure EEG signals. The feature space obtained from the ellipse area parameters of two IMFs has been used for classification of ictal and seizure-free EEG signals using the artificial neural network (ANN) classifier. It has been shown that the feature space formed using ellipse area parameters of first and second IMFs has given good classification performance. Experimental results on EEG database available by the University of Bonn, Germany, are included to illustrate the effectiveness of the proposed method.