Griet Goovaerts

Inferior and anterior QRS fragmentation have different prognostic value in patients who received an implantable defibrillator in primary prevention of sudden cardiac death

AIMS QRS fragmentation (fQRS) has been proposed as a predictor of sudden cardiac death (SCD) and all-cause mortality in ischemic (ICM) and non-ischemic cardiomyopathy patients. However the value of fQRS in patients with a LVEF <35% is a matter of debate. METHODS All consecutive patients with an indication for an ICD in primary prevention of SCD were included in a retrospective registry from 1996 until 2013. Twelve lead electrocardiograms before implant were analyzed for the presence of fQRS in different regions. Adjusted Cox regression analysis for first appropriate ICD shock (AS) and all-cause mortality was performed. RESULTS In total 407 patients were included with a mean follow-up of 4.2±3.3y (age 60.6±11.9y, 15.7% female and 52.8% ICM). fQRS was present in 46.7% of patients, predominantly inferior (30.7%) followed by anterior (21.4%) and lateral (11.1%) coronary artery territories. fQRS was significantly more prevalent in ICM (p=0.004). Inferior fQRS was an independent predictor of a first AS within 1y (HR 2.55, 95%CI 1.28-5.07) and 3y (HR 1.90, 95%CI 1.14-3.18) after implantation. Whereas, anterior fQRS was an independent predictor of all-cause mortality within 1y (HR 4.58, 95%CI 1.29-16.19), 3y (HR 3.92, 95%CI 1.77-8.65) and the complete follow-up (HR 2.22, 95%CI 1.33-3.69). Lateral fQRS was only a predictor of late (>3y of follow-up) all-cause mortality (HR 2.04, 95%CI 1.09-3.81). CONCLUSIONS fQRS in a specific coronary artery territory might be promising to discriminate arrhythmic from mortality risk. Inferior fQRS was a predictor of early arrhythmia, while anterior fQRS was related to mortality.

Heart beat detection in multimodal data using automatic relevant signal detection

Accurate R peak detection in the electrocardiogram (ECG) is a well-known and highly explored problem in biomedical signal processing. Although a lot of progress has been made in this area, current methods are still insufficient in the presence of extreme noise and/or artifacts such as loose electrodes. Often, however, not only the ECG is recorded, but multiple signals are simultaneously acquired from the patient. Several of these signals, such as blood pressure, can help to improve the heart beat detection. These signals of interest can be detected automatically by analyzing their power spectral density or by using the available signal type identifiers. Individual peaks from the signals of interest are combined using majority voting, heart beat location estimation and Hjorths mobility of the resulting RR intervals. Both multimodal algorithms showed significant increases in performance of up to 8.65% for noisy multimodal datasets compared to when only the ECG signal is used. A maximal performance of 90.02% was obtained on the hidden test set of the Physionet/Computing in Cardiology Challenge 2014: Robust Detection of Heart Beats in Multimodal Data.

Neural network approach for T-wave end detection: A comparison of architectures

In this paper, a new approach to the problem of detecting the end of the T wave (Te) on the electrocardiogram (ECG) using Multilayer Perceptron (MLP) neural networks is proposed and evaluated. The approach consists of a neural network acting as a regression function that estimates the Te location using the samples between two consecutive R peaks. The input vectors were taken using three dimensional reduction methods (Discrete Cosine Transform, DCT, Principal Component Analysis, PCA and resampling, RES) over a window of 100 samples. For training, Bayesian regularization has been used. A total of 1536 neural networks were trained. The results show that PCA and DCT are more feasible than RES as dimension reduction methods. Finally, a brief comparison with other algorithms proposed in the literature is included.

Automatic detection of T wave alternans using tensor decompositions in multilead ECG signals

OBJECTIVE T wave alternans (TWA) is a promising non-invasive risk stratification tool for sudden cardiac death which can be detected from surface ECG. This paper proposes a novel method to automatically detect TWA based on tensor decomposition methods. APPROACH Two different tensor decomposition approaches are examined and compared, namely canonical polyadic decomposition and the more generalized variation PARAFAC2 which allows the T waves to shift in time. RESULTS AND SIGNIFICANCE Results on different artificial and clinical signals show that the presented methods are a robust and reliable way for TWA detection, and show the potential benefit of tensors in ECG signal processing.

Influence of food intake on the QT and QT/RR relation

BACKGROUND There are conflicting data on the influence of meal intake on the QT interval. METHODS Ten healthy subjects were studied before and after a standardized breakfast and lunch with a sequence of supine resting, standing and exercise. Data collection was performed using a 12-lead Holter with semi-automated analysis. QT correction was performed using Fridericia (QTcF) correction formula and a subject-specific method based on individual QT/RR-regression (QTcI). RESULTS Meal intake induced significant changes in HR (p<0.001), but not in QTcF (p=0.512) or QTcI (p=0.739). Postural analysis showed only significant differences in supine position for HR (p=0.010), not when standing or during exercise. CONCLUSION Food intake induced an increase in heart rate limited to supine position. Using QTcF and QTcI no QTc changes were found.

Tensor-based detection of T wave alternans using ECG

T wave alternans is defined as changes in the T wave amplitude in an ABABAB-pattern. It can be found in ECG signals of patients with heart diseases and is a possible indicator to predict the risk on sudden cardiac death. Due to its low amplitude, robust automatic T wave alternans detection is a difficult task. We present a new method to detect T wave alternans in multichannel ECG signals. The use of tensors (multidimensional matrices) permits the combination of the information present in different channels, making detection more reliable. The possibility of decomposition of incomplete tensors is exploited to deal with noisy ECG segments. Using a sliding window of 128 heartbeats, a tensor is constructed of the T waves of all channels. Canonical Polyadic Decomposition is applied to this tensor and the resulting loading vectors are examined for information about the T wave behavior in three dimensions. T wave alternans is detected using a sign change counting method that is able to extract both the T wave alternans length and magnitude. When applying this novel method to a database of patients with multiple positive T wave alternans tests using the clinically available spectral method tests, both the length and the magnitude of the detected T wave alternans is larger for these subjects than for subjects in a control group.

Detection of irregular heartbeats using tensors

Automatic classification of heartbeats in different categories is important for ECG analysis. The number of irregular heartbeats in a signal can for example be used as a risk stratifier for sudden cardiac death. Current heart-beat classification methods typically use time or frequency domain features to characterize heartbeats. We propose the use of tensors to incorporate the structural information that is present in multilead ECG channels. Since different ECG leads provide information on a particular orientation in space, more robust detection can be done if all leads are considered. After preprocessing and heartbeat detection using wavelet-based methods, the ECG signal is segmented beat-by-beat. The different heartbeats are then placed in a three-dimensional tensor with dimensions time, channels and heartbeats. Canonical Polyadic Decomposition is used to decompose the tensor. The results are three loading vectors, corresponding to the dimensions of the original tensor. Through analysis of these loading vectors, irregular heartbeats can be detected using a simple thresholding procedure. The method has been applied to a subset of the St.-Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database available on Physionet. When applying the method to the first 10 signals, we obtain a mean sensitivity and specificity of more than 90%. These results indicate that the presented method is a new and reliable way of performing irregular heartbeat detection.

Advanced EEG Processing for the Detection of Drowsiness in Drivers

Drowsiness is a serious problem for drivers which causes many accidents every day. It is estimated that drowsiness is the cause of four deaths and 100 injuries per day in the United States. In this paper two methods have been developed to detect drowsiness based on features of ocular artifacts in EEG signals. The ocular artifacts are derived from the EEG signals by using Canonical Correlation Analysis (BSS-CCA). Wavelet transforms are used to automatically select components containing eye blinks. Sixteen features are then calculated from the eye blink and used for drowsiness detection. The first method is based on linear regression, the second on fuzzy detection. For the first method, the drowsiness level is correctly detected in 72% of the epochs. The second method uses fuzzy detection and detects the drowsiness correctly in 65% of the epochs. The best results are obtained when using one single eye blink feature.

Automated T Wave End Detection Methods: Comparison of Four Different Methods for T Wave End Detection

T wave end detection is essential for electrocardiogram (ECG) processing and analysis. Several methods have been proposed and tested, but an objective comparison is lacking. In this paper, four different (semi-) automated methods are compared with the manually annotated T wave ends of the PhysioNet QT database. The first method is a semi-automatic method, based on a template matching algorithm. The second method uses the tangent of the steepest point of the descending limb of the T wave. The third and fourth method perform a maximum area search of, respectively, a trapezium and the area under the curve. In order to evaluate the accuracy and repeatability of the proposed algorithms, the mean and standard deviation (sd) of the detection errors were computed. This was performed for leads I and II separately, after selection of the best annotated T wave end per beat and after selection of the best lead. We demonstrated that the trapezium method is the least repeatable of all methods tested (sd=29.7ms), whilst the integral method scores best in terms of accuracy (mean=2.2ms). These findings were strengthened by the analysis of the generated Bland-Altman plots, where the smallest bias was observed for the integral method (-1.89ms).

Inter- and intra-observer variability of visual fragmented QRS scoring in ischemic and non-ischemic cardiomyopathy

BACKGROUND Fragmented QRS (fQRS) on a 12-lead ECG has been linked with adverse outcome. However, the visual scoring of ECGs is prone to inter- and intra-observer variability. METHODS Five observers, two experienced and three novel, assessed fQRS in 712 digital ECGs, 100 were re-evaluated to assess intra-observer variability. Fleiss and Cohens Kappa were calculated and compared between subgroups. RESULTS The inter-observer variability for assessing fQRS in all leads combined was substantial with a Kappa of 0.651. Experienced observers only had a better agreement with a Kappa of 0.823. Intra-observer variability ranged from 0.736 to 0.880. In the subgroup with ventricular pacing the inter-observer variability was even significantly larger when compared to ECGs with normal QRS duration (Kappa 0.493 vs 0.664, p<0.001). CONCLUSION The visual assessment of QRS fragmentation is prone to inter- and intra-observer variability, mainly influenced by the experience of the observers, the underlying rhythm and QRS morphology.