Arrhythmia recognition and classification using combined linear and nonlinear features of ECG signals.

Arrhythmia is a cardiac condition caused by abnormal electrical activity of the heart, and an electrocardiogram (ECG) is the non-invasive method used to detect arrhythmias or heart abnormalities. Due to the presence of noise, the non-stationary nature of the ECG signal (i.e. the changing morphology of the ECG signal with respect to time) and the irregularity of the heartbeat, physicians face difficulties in the diagnosis of arrhythmias. The computer-aided analysis of ECG results assists physicians to detect cardiovascular diseases. The development of many existing arrhythmia systems has depended on the findings from linear experiments on ECG data which achieve high performance on noise-free data. However, nonlinear experiments characterize the ECG signal more effectively sense, extract hidden information in the ECG signal, and achieve good performance under noisy conditions. This paper investigates the representation ability of linear and nonlinear features and proposes a combination of such features in order to improve the classification of ECG data. In this study, five types of beat classes of arrhythmia as recommended by the Association for Advancement of Medical Instrumentation are analyzed: non-ectopic beats (N), supra-ventricular ectopic beats (S), ventricular ectopic beats (V), fusion beats (F) and unclassifiable and paced beats (U). The characterization ability of nonlinear features such as high order statistics and cumulants and nonlinear feature reduction methods such as independent component analysis are combined with linear features, namely, the principal component analysis of discrete wavelet transform coefficients. The features are tested for their ability to differentiate different classes of data using different classifiers, namely, the support vector machine and neural network methods with tenfold cross-validation. Our proposed method is able to classify the N, S, V, F and U arrhythmia classes with high accuracy (98.91%) using a combined support vector machine and radial basis function method.

Adapted filter banks for feature extraction in transcranial magnetic stimulation evoked responses.

A novel adaptive and approximate shift-invariant wavelet packet feature extraction scheme for event-related potentials (ERPs) in the electroencephalogram (EEG) is introduced in this paper. In this algorithm, the shift-invariant wavelet packed decomposition is done by integrating a cost function for decimation decision in each sub-band expansion. Additionally, a shape adaptation of the wavelet is implemented to find the best adapted wavelet shape for a given class of ERPs. This scheme is used to analyze the time course of the impact of single-pulse transcranial magnetic stimulation (TMS) to the auditory ERPs. We show that the proposed scheme is able to extract even slightest impacts of TMS, making it a promising tool for the extraction of weak ERPs components, particularly in hybrid TMS-EEG/ERP setups.

An investigation on the effects of single pulse transcranial magnetic stimulation in a modified maximum entropy auditory stimulation paradigm.

In this paper, we intend to investigate further the effects of single pulse TMS (sTMS) on auditory attention through an experimental design that combines a modified version of maximum entropy stimulation paradigm. Single pulses of TMS with 4.4s inter-stimulus interval (ISI) were applied to the left temporal lobe of subjects while three randomized auditory stimuli with constant ISI of 1.1s were delivered to the contralateral side within the TMS stimulation duration. Our main focus was to examine the time course of the auditory late responses (ALRs) due to TMS stimulation by a phase clustering on the unit circle measure and an adaptive shift- invariant feature extraction method. In the attention scheme, a significant difference in the phase stability between TMS and no-TMS was found in the range of the N1 wave of ALRs. However, the difference occurs only for the data after 1.1s. Furthermore, there is an absence of differences in the amplitude of the ALR. In addition, the effects of TMS and attention can also be discriminated very well and illuminate the effects of TMS in auditory attention. It is concluded that even sTMS might have the potential to alter the attentional states and the effects can last about 1s, at least when considering the large- scale neural correlates of attention in ALR sequences.

Detection of the uncomfortable loudness level by electroencephalographic data: a novelty detection approach using habituation correlates.

The objective fitting of hearing aids and cochlear implants in uncooperative patients still remains a challenge. Especially in determining the threshold of uncomfortable loudness which cannot be predicted from the auditory threshold. In this study, we propose a single sweeps processing method which employs a hybrid approach of adaptive frame decomposition adaptation by a tight wavelet frame and the gaussian novelty detection for the detection of large-scale electroencephalographic responses correlates of habituation in late auditory evoked potentials. For this, habituation is discerns as a novel event. It is concluded that the new approach provides a fast and reliable method in the discrimination of uncomfortable loudness level from comfortable loudness level. It can be further use in more clinically oriented studies related to an objective frequency specific fitting of hearing aids or cochlear implants.

Optimized shift-invariant wavelet packet feature extraction for electroencephalographic evoked responses.

Local discriminant bases (LDB) have a major disadvantage in their representation which is sensitive to signal translations. The discriminant features will be not consistent when the same but shifted signal is applied. Thus, to overcome this problem, an approximate shift-invariant features extraction based on local discriminant bases is introduced. This technique is based on approximate shift-invariant wavelet packed decomposition which integrate a cost function for decimation decision in each sub-band expansion. This technique gives a consistent best tree selection both in top-down and bottom-up search method. It also provides a consistent wavelet shape in a shape-adapted wavelet method to determine the best wavelet library for a particular signal. This method has an advantage especially in electroencephalographic (EEG) measurement in which there is an inter-individual shift in time for the signals. An application of this method is provided by the discrimination between signals with transcranial magnetic stimulation (TMS) and acoustic-somatosensory stimulation (ASS).

Adaptive time-scale feature extraction in electroencephalographic responses to transcranial magnetic stimulation.

Electroencephalographic responses evoked by transcranial magnetic stimulation (TMS) gain more and more interest for basic neurophysiological research and possibly diagnostic purposes. However, the separation of magnetically from non- magnetically induced brain activity still remains a challenge due to superimposed secondary effects, in particular auditory and somatosensory evoked potentials. In this study, we use optimized tight wavelet frames for the adaptive extraction of discriminant electroencephalographic time-scale features during TMS using figure-of-eight coil for focal stimulation and a combined auditory and somatosensory stimulation (ASS) paradigm. We restrict our focus to large-scale features which correspond to slow wave cortical potentials (SCPs). These potentials might reflect thalamocortical dynamics and are frequently used in biofeedback therapies. The proposed methods allows for a robust extraction of slow wave components and separated clearly the TMS from the ASS data. It is concluded that our study strongly supports recent suggestions that TMS modulates SCPs, reinforcing the theory that TMS leads to long term changes in the cortical excitability.