Noise removal in electroencephalogram signals using an artificial neural network based on the simultaneous perturbation method

Auto-Denoising for EEG Signals Using Generative Adversarial Network

The brain-computer interface (BCI) has many applications in various fields. In EEG-based research, an essential step is signal denoising. In this paper, a generative adversarial network (GAN)-based denoising method is proposed to denoise the multichannel EEG signal automatically. A new loss function is defined to ensure that the filtered signal can retain as much effective original information and energy as possible. This model can imitate and integrate artificial denoising methods, which reduces processing time; hence it can be used for a large amount of data processing. Compared to other neural network denoising models, the proposed model has one more discriminator, which always judges whether the noise is filtered out. The generator is constantly changing the denoising way. To ensure the GAN model generates EEG signals stably, a new normalization method called sample entropy threshold and energy threshold-based (SETET) normalization is proposed to check the abnormal signals and limit the range of EEG signals. After the denoising system is established, although the denoising model uses the different subjects' data for training, it can still apply to the new subjects' data denoising. The experiments discussed in this paper employ the HaLT public dataset. Correlation and root mean square error (RMSE) are used as evaluation criteria. Results reveal that the proposed automatic GAN denoising network achieves the same performance as the manual hybrid artificial denoising method. Moreover, the GAN network makes the denoising process automatic, representing a significant reduction in time.

High-Precise Bipolar Disorder Detection by Using Radial Basis Functions Based Neural Network

Presently, several million people suffer from major depressive and bipolar disorders. Thus, the modelling, characterization, classification, diagnosis, and analysis of such mental disorders bears great significance in medical research. Electroencephalogram records provide important information to improve clinical diagnosis and are very useful in the scientific community. In this work, electroencephalogram records and patient data from the Hospital Virgen de la Luz in Cuenca (Spain) were processed for a correct classification of bipolar disorders. This work implemented an innovative radial basis function-based neural network employing a fuzzy means algorithm. The results show that the proposed method is an effective approach for discrimination of two kinds of classes, i.e., bipolar disorder patients and healthy persons. The proposed algorithm achieved the best performance compared with other machine learning techniques such as Bayesian linear discriminant analysis, Gaussian naive Bayes, decision trees, K-nearest neighbour, or support vector machine, showing a very high accuracy close to 97%. Therefore, the neural network technique presented could be used as a new tool for the diagnosis of bipolar disorder, considering the possibility of integrating this method into medical software.

Wavelet skeletons in sleep EEG-monitoring as biomarkers of early diagnostics of mild cognitive impairment

Many neuro-degenerative diseases are difficult to diagnose in their early stages. For example, early diagnosis of Mild Cognitive Impairment (MCI) requires a wide variety of tests to distinguish MCI symptoms and normal consequences of aging. In this article, we use the wavelet-skeleton approach to find some characteristic patterns in the electroencephalograms (EEGs) of healthy adult patients and patients with cognitive dysfunctions. We analyze the EEG activity recorded during natural sleep of 11 elderly patients aged between 60 and 75, six of whom have mild cognitive impairment, and apply a nonlinear analysis method based on continuous wavelet transformskeletons. Our studies show that a comprehensive analysis of EEG signals of the entire sleep state allows us to identify a significant decrease in the average duration of oscillatory patterns in the frequency band [12; 14] Hz in the presence of mild cognitive impairment. Thus, the changes in this frequency range can be interpreted as related to the activity in the motor cortex, as a candidate for developing the criteria for early objective MCI.

A new automated multi-stage system of non-local means and multi-kernel adaptive filtering techniques for EEG noise and artifacts suppression

Context.Electroencephalography (EEG) signals are contaminated with diverse types of noises and artifacts, which greatly distort EEG recording and increase the difficulty in obtaining accurate diagnosis.Objective.This paper investigates, for the first time, multi-kernel normalized least mean square with coherence-based sparsification (MKNLMS-CS) algorithm for suppressing different artifact components, and the 1D patch-based non-local means (NLM) algorithm for eliminating white and colored noises.Approach.A novel multi-stage system based on combining the NLM algorithm with the MKNLMS-CS algorithm is proposed for eliminating different noise and artifact sources by targeting each noise or artifact component in a single stage.Main Results.The proposed approach is applied to clinical real EEG data, and the results reveal the superior performance of the proposed system in removing white and colored noises, suppressing different artifact components, preserving the important and tiny features of the original EEG signal, and keeping the morphology of EEG frequency components.Significance.The proposed multi-stage design succeeds not only to suppress different artifact components and noise sources under low and high noise conditions, but also to achieve accurate sleep spindle detection from the filtered high-quality EEG signals. This demonstrates the usefulness of the proposed approach for obtaining high-resolution EEG signal from noisy and contaminated EEG recordings.

An Efficient Teager Energy Operator-Based Automated QRS Complex Detection

Database

The efficiency and robustness of the proposed method has been tested on Fantasia Database (FTD), MIT-BIH Arrhythmia Database (MIT-AD), and MIT-BIH Normal Sinus Rhythm Database (MIT-NSD).

Aim

Because of the importance of QRS complex in the diagnosis of cardiovascular diseases, improvement in accuracy of its measurement has been set as a target. The present study provides an algorithm for automatic detection of QRS complex on the ECG signal, with the benefit of energy and reduced impact of noise on the ECG signal.

Method

The method is basically based on the Teager energy operator (TEO), which facilitates the detection of the baseline threshold and extracts QRS complex from the ECG signal.

Results

The testing of the undertaken method on the Fanatasia Database showed the following results: sensitivity (Se) = 99.971%, positive prediction (P+) = 99.973%, detection error rate (DER) = 0.056%, and accuracy (Acc) = 99.944%. On MIT-AD involvement, Se = 99.74%, P+ = 99.97%, DER = 0.291%, and Acc = 99.71%. On MIT-NSD involvement, Se = 99.878%, P+ = 99.989%, DER = 0.134%, and Acc = 99.867%.

Conclusion

Despite the closeness of the recorded peaks which inflicts a constraint in detection of the two consecutive QRS complexes, the proposed method, by applying 4 simple and quick steps, has effectively and reliably detected the QRS complexes which make it suitable for practical purposes and applications.

Shannon's Energy Based Algorithm in ECG Signal Processing

Physikalisch-Technische Bundesanstalt (PTB) database is electrocardiograms (ECGs) set from healthy volunteers and patients with different heart diseases. PTB is provided for research and teaching purposes by National Metrology Institute of Germany. The analysis method of complex QRS in ECG signals for diagnosis of heart disease is extremely important. In this article, a method on Shannon energy (SE) in order to detect QRS complex in 12 leads of ECG signal is provided. At first, this algorithm computes the Shannon energy (SE) and then makes an envelope of Shannon energy (SE) by using the defined threshold. Then, the signal peaks are determined. The efficiency of the algorithm is tested on 70 cases. Of all 12 standard leads, ECG signals include 840 leads of the PTB Diagnostic ECG Database (PTBDB). The algorithm shows that the Shannon energy (SE) sensitivity is equal to 99.924%, the detection error rate (DER) is equal to 0.155%, Positive Predictivity (+P) is equal to 99.922%, and Classification Accuracy (Acc) is equal to 99.846%.