CNN-based framework using spatial dropping for enhanced interpretation of neural activity in motor imagery classification

Interpretation of brain activity responses using motor imagery (MI) paradigms is vital for medical diagnosis and monitoring. Assessed by machine learning techniques, identification of imagined actions is hindered by substantial intra- and inter-subject variability. Here, we develop an architecture of Convolutional Neural Networks (CNN) with an enhanced interpretation of the spatial brain neural patterns that mainly contribute to the classification of MI tasks. Two methods of 2D-feature extraction from EEG data are contrasted: Power Spectral Density and Continuous Wavelet Transform. For preserving the spatial interpretation of extracting EEG patterns, we project the multi-channel data using a topographic interpolation. Besides, we include a spatial dropping algorithm to remove the learned weights that reflect the localities not engaged with the elicited brain response. We evaluate two labeled scenarios of MI tasks: bi-class and three-class. Obtained results in an MI database show that the thresholding strategy combined with Continuous Wavelet Transform improves the accuracy and enhances the interpretability of CNN architecture, showing that the highest contribution clusters over the sensorimotor cortex with a differentiated behavior of rhythms [Formula: see text] and [Formula: see text].

Enhanced Data Covariance Estimation Using Weighted Combination of Multiple Gaussian Kernels for Improved M/EEG Source Localization

In the recent past, estimating brain activity with magneto/electroencephalography (M/EEG) has been increasingly employed as a noninvasive technique for understanding the brain functions and neural dynamics. However, one of the main open problems when dealing with M/EEG data is its non-Gaussian and nonstationary structure. In this paper, we introduce a methodology for enhancing the data covariance estimation using a weighted combination of multiple Gaussian kernels, termed WM-MK, that relies on the Kullback-Leibler divergence for associating each kernel weight to its relevance. From the obtained results of validation on nonstationary and non-Gaussian brain activity (simulated and real-world EEG data), WM-MK proves that the accuracy of the source estimation raises by more effectively exploiting the measured nonlinear structures with high time and space complexity.

Instance-Based Representation Using Multiple Kernel Learning for Predicting Conversion to Alzheimer Disease

The early detection of Alzheimer’s disease and quantification of its progression poses multiple difficulties for machine learning algorithms. Two of the most relevant issues are related to missing data and results interpretability. To deal with both issues, we introduce a methodology to predict conversion of mild cognitive impairment patients to Alzheimer’s from structural brain MRI volumes. First, we use morphological measures of each brain structure to build an instance-based feature mapping that copes with missed follow-up visits. Then, the extracted multiple feature mappings are combined into a single representation through the convex combination of reproducing kernels. The weighting parameters per structure are tuned based on the maximization of the centered-kernel alignment criterion. We evaluate the proposed methodology on a couple of well-known classification machines employing the ADNI database devoted to assessing the combined prognostic value of several AD biomarkers. The obtained experimental results show that our proposed method of Instance-based representation using multiple kernel learning enables detecting mild cognitive impairment as well as predicting conversion to Alzheimers disease within three years from the initial screening. Besides, the brain structures with larger combination weights are directly related to memory and cognitive functions.

Fractionated electrograms and rotors detection in chronic atrial fibrillation using model-based clustering

The identification of atrial fibrillation (AF) substrates is needed to improve ablation therapy guided by electrograms, although mechanisms that sustain AF are not fully understood. Detection of complex fractionated atrial electrograms (CFAE) is used for this purpose. Nonetheless, efficacy of this method is inadequate in the case of chronic AF. Recent hypothesis proposes the rotors as fibrillatory substrate. Novel approaches seek to relate CFAE with rotor; nevertheless, such methods are not able to identify the associated substrate. Furthermore, the patterns that characterize CFAE generated by rotors remain unknown. Thus, tracking of rotors is an unsolved issue. In this paper, we propose a non-supervised method to find patterns associated with fibrillatory substrates in chronic AF. We extracted two features based on local activation wave detection and one feature based on non-linear dynamics. Gaussian mixture model-based clustering was used to discriminate CFAE patterns. Resulting clusters are visualized in an electroanatomic map. We assessed the proposed method in a real database labeled according to the level of fractionation and in a simulated episode of chronic AF in which a rotor was detected. Our results indicate that the method proposed can separate different levels of fractionation in CFAE, and provide evidence that clustering can be used to locate the vortex of the rotors. Provided approach can support ablation therapy procedures by means of CFAE patterns discrimination.

Reconstruction of Neural Activity from EEG Data Using Dynamic Spatiotemporal Constraints

We present a novel iterative regularized algorithm (IRA) for neural activity reconstruction that explicitly includes spatiotemporal constraints, performing a trade-off between space and time resolutions. For improving the spatial accuracy provided by electroencephalography (EEG) signals, we explore a basis set that describes the smooth, localized areas of potentially active brain regions. In turn, we enhance the time resolution by adding the Markovian assumption for brain activity estimation at each time period. Moreover, to deal with applications that have either distributed or localized neural activity, the spatiotemporal constraints are expressed through [Formula: see text] and [Formula: see text] norms, respectively. For the purpose of validation, we estimate the neural reconstruction performance in time and space separately. Experimental testing is carried out on artificial data, simulating stationary and non-stationary EEG signals. Also, validation is accomplished on two real-world databases, one holding Evoked Potentials and another with EEG data of focal epilepsy. Moreover, responses of functional magnetic resonance imaging for the former EEG data have been measured in advance, allowing to contrast our findings. Obtained results show that the [Formula: see text]-based IRA produces a spatial resolution that is comparable to the one achieved by some widely used sparse-based estimators of brain activity. At the same time, the [Formula: see text]-based IRA outperforms other similar smooth solutions, providing a spatial resolution that is lower than the sparse [Formula: see text]-based solution. As a result, the proposed IRA is a promising method for improving the accuracy of brain activity reconstruction.

Tensor-product kernel-based representation encoding joint MRI view similarity

To support 3D magnetic resonance image (MRI) analysis, a marginal image similarity (MIS) matrix holding MR inter-slice relationship along every axis view (Axial, Coronal, and Sagittal) can be estimated. However, mutual inference from MIS view information poses a difficult task since relationships between axes are nonlinear. To overcome this issue, we introduce a Tensor-Product Kernel-based Representation (TKR) that allows encoding brain structure patterns due to patient differences, gathering all MIS matrices into a single joint image similarity framework. The TKR training strategy is carried out into a low dimensional projected space to get less influence of voxel-derived noise. Obtained results for classifying the considered patient categories (gender and age) on real MRI database shows that the proposed TKR training approach outperforms the conventional voxel-wise sum of squared differences. The proposed approach may be useful to support MRI clustering and similarity inference tasks, which are required on template-based image segmentation and atlas construction.

Very long-term ECG monitoring patch with improved functionality and wearability

Heart activity monitoring is an important task for prevention and treatment of cardiovascular diseases. However, development of long-term, wearable electrodes remains an open issue. In fact, adhesion ability and energy consumption while preserving aesthetics is one of the major concerns to implement a minimally invasive monitoring system that measures and transmits electrocardiographic signals (ECG) during long-term periods of time. Based on the concepts of functionality, wearability, and resources, we develop a new long-term ECG monitoring system under the wear-and-forget principle. We also propose a system model of the electrode-skin interface that performs real-time measurements with a minimally invasive effect when compared with another competitive and implantable systems. As a result, testing of designed prototype shows that the developed very long-term ECG monitoring patch improves energy consumption and adhesion time up to 40 days.

Local binary fitting energy solution by graph cuts for MRI segmentation

This paper proposes a new solution for local binary fitting energy minimization based on graph cuts for automatic brain structure segmentation on magnetic resonance images. The approach establishes an effective way to embed the energy formulation into a directed graph, such that the energy is minimized by maximizing the graph flow. Proposed and conventional solutions are compared by segmenting the well-known BrainWeb synthetic brain Magnetic Resonance Imaging database. Achieved results show an improvement on the computational cost (about 10 times shorter) while maintaining the segmentation accuracy (96%).

Feature selection for discrimination of fractionation levels in atrial electrograms

Radiofrequency catheter ablation of atrial fibrillation (AF) guided by complex fractionated atrial electrograms (CFAE) is associated with a high AF termination rate in paroxysmal AF, but not in persistent. CFAE does not always identify favorable sites for persistent AF ablation. Studies suggest that only high fractionation level should be used as a target site for ablation. Nonetheless, there are not a standardized criterion to defined fractionation levels. Therefore, a better characterization of the signal is required providing a set of more powerful features that should be extracted from CFAE. Due to the apparent difference among fractionation classes in terms of their stochastic variability, we test time-domain and time-frequency based feature extraction approaches. Also, we carried out the symmetrical uncertainty-based feature selection to determine the most relevant features which improve discrimination of fractionation levels. Obtained results on a tested real electrogram database show that most relevant features in time-domain are related with time intervals and not with amplitudes. Nonetheless, time-frequency features obtained more information from the signal and this representation is likely a better suitable discriminating approach, particularly to detect high fractionated electrograms with a sensitivity and specificity of 83.0% and 93.6%, respectively.

Three-layer-isotropic skull conductivity representation in the EEG forward problem using spherical head models

We study the influence of different conductivity models within the framework of electroencephalogram (EEG) source localization on the white matter and skull areas. Particularly, we investigate five different spherical models having either isotropic or anisotropic conductivity for both considered areas. To this end, the anisotropic finite difference reciprocity method is used for solving the EEG forward problem. We evaluate a model of a numeric skull conductivity in terms of the minimum dipole localization/orientation error. As a result, both considered models of the skull reach the lowest dipole localization error (less than 6 mm), namely: i) single anisotropic layer and ii) three isotropic layers (hard bone/spongy bone/hard bone). Additionally, two different electrode configurations (10-20 and 10-10 systems) are tested showing that the error decreases almost as much as twice for the latter one though the computational burden significantly increases.

Multiobjective optimization-based design of wearable electrocardiogram monitoring systems

Nowadays, the use of Wearable User Interfaces has been extensively growing in medical monitoring applications. However, production and manufacture of prototypes without automation tools may lead to non viable results since it is often common to find an optimization problem where several variables are in conflict with each other. Thus, it is necessary to design a strategy for balancing the variables and constraints, systematizing the design in order to reduce the risks that are present when it is exclusively guided by the intuition of the developer. This paper proposes a framework for designing wearable ECG monitoring systems using multi-objective optimization. The main contributions of this work are the model to automate the design process, including a mathematical expression relating the principal variables that make up the criteria of functionality and wearability. We also introduce a novel yardstick for deciding the location of electrodes, based on reducing interference from ECG by maximizing the electrode-skin contact.

Influence of anisotropic white matter modeling on EEG source localization

We study the influence of the anisotropic white matter within the ElectroEncephaloGraphy source localization problem. To this end, we consider three cases of the anisotropic white matter modeled in two concrete cases: by fixed or variable ratio. We extract information about highly anisotropic areas of the white matter from real Diffusion Weighted Imaging data. To validate the compared anisotropic models, we introduce the localization dipole and orientation errors. Obtained results show that the white matter model with a fixed anisotropic ratio leads to values of dipole localization error close to 1cm and may be enough in those cases avoiding localized analysis of neural brain activity. In contrast, modeling based on the anisotropic variable rate assumption becomes important in tasks regarding analysis and localization of deep sources neighboring the white matter tissue.

Identification of brain networks using time-varying spatial constraints of neural activity reconstruction

Electroencephalographic (EEG) data give a direct non-invasive measurement of neural brain activity. Nevertheless, the common assumption about EEG stationarity (time-invariant process) is a strong limitation for understanding real behavior of underlying neural networks. Here, we propose an approach for finding networks of brain regions connected by functional associations (functional connectivity) that vary along the time. To this end, we compute a set of a priori spatial dictionaries that represent brain areas with similar temporal stochastic dynamics, and then, we model relationship between areas as a time-varying process. We test our approach in both simulated and real EEG data where results show that inherent interpretability provided by the time-varying process can be useful to describe underlying neural networks.

Kernel-based atlas image selection for brain tissue segmentation

We propose a new Kernel-based Atlas Image Selection computed in the Embedding Representation space (termed KAISER) aiming to support labeling of brain tissue on 3D magnetic resonance (MR) images. KAISER approach provides efficient feature extraction from MR volumes based on an introduced inter-slice kernel (ISK). Thus, using the ISK matrix eigendecomposition, the inherent structure of data distribution is accentuated through estimation of low dimensional compact space where every pair-wise image similarity can be better measured. We compare our proposal against the whole-population atlas, randomly and demographically selected multiatlas approaches in a four-tissue image labeling task. Obtained results show that the KAISER approach outperforms other alternative techniques (98% Dice index similarity against 94%), while exhibiting better repeatability.

Entropy-based multichannel measure of stationarity for characterization of motor imagery patterns

We propose a novel approach for measuring the stationarity level of multichannel time-series. This measure is based on stationarity definition over time-varying spectra and aims to quantify the relationship between local (single-channel dynamics) and global (multichannel dynamics) stationarity. With the purpose of separate among several motor/imagery tasks, we asssume that movement imagination implies an increase on the EEG variability, consequently, as discriminant features, we first compute the non-stationary components of input signals, and we further obtain its stationary level throughout the proposed measure. To assess the separability level of the proposed features, we employ the t-student test. Obtained results evidence that our measure is able to accurately detect brain areas projected on the scalp where motor tasks are performed.

Feature relevance analysis supporting automatic motor imagery discrimination in EEG based BCI systems

Recently, there have been many efforts to develop Brain Computer Interface (BCI) systems, allowing identifying and discriminating brain activity, as well as, support the control of external devices, and to understand cognitive behaviors. In this work, a feature relevance analysis approach based on an eigen decomposition method is proposed to support automatic Motor Imagery (MI) discrimination in electroencephalography signals for BCI systems. We select a set of features representing the best as possible the studied process. For such purpose, a variability study is performed based on traditional Principal Component Analysis. EEG signals modelling is carried out by feature estimation of three frequency-based and one time-based. Our approach provides testing over a well-known MI dataset. Attained results show that presented algorithm can be used as tool to support discrimination of MI brain activity, obtaining acceptable results in comparison to state of the art approaches.

Biomedical data analysis by supervised manifold learning

Biomedical data analysis is usually carried out by assuming that the information structure embedded into the biomedical recordings is linear, but that statement actually does not corresponds to the real behavior of the extracted features. In order to improve the accuracy of an automatic system to diagnostic support, and to reduce the computational complexity of the employed classifiers, we propose a nonlinear dimensionality reduction methodology based on manifold learning with multiple kernel representations, which learns the underlying data structure of biomedical information. Moreover, our approach can be used as a tool that allows the specialist to do a visual analysis and interpretation about the studied variables describing the health condition. Obtained results show how our approach maps the original high dimensional features into an embedding space where simple and straightforward classification strategies achieve a suitable system performance.

Extraction of stationary components in biosignal discrimination

Biosignal recordings are widely used in the medical environment to support the evaluation and the diagnosis of pathologies. Nevertheless, the main difficulty lies in the non-stationary behavior of the biosignals, difficulting the obtention of patterns characterizing the changes in physiological or pathological states. Thus, the obtention of the stationary and non-stationary components of a biosignal is still an open issue. This work proposes a methodology to detect time-homogeneities based on time-frequency analysis with aim to extract the non-stationary behavior of the biosignal. Results show an increase in the stationarity and in the distance between classes of the reconstructions from the enhanced time-frequency representations. The stationary components extracted with the proposed approach can be used to solve biosignal classification problems.

Improvement of ECG signal quality measurement using correlation and diversity-based approaches

A large proportion of cardiovascular diseases might be preventable, however, majority of this diseases occurs in rural areas where there is a poor presence of cardiologists. To overcome this issue, the use of wearable devices within the telemedicine framework would be of benefit. However, implementation of processing algorithms in smart-phones at mobile environments imposes restrictions ensuring high measurement quality of acquired ECG data, while maintaining low computation burden. This work presents an algorithm for scoring the quality of measured ECG recordings is developed. Particularly, a quality score is provided that takes into account the proportional correlation observed in acceptable signals based on a diversity scheme, and their inverse relation with standard deviation. Testing of proposed algorithm is carried out upon two different databases, the first one is of own production, while the second one is obtained from Physionet. As a result, high values of sensitivity and specificity are achieved.

Novel heuristic search for ventricular arrhythmia detection using normalized cut clustering

Processing of the long-term ECG Holter recordings for accurate arrhythmia detection is a problem that has been addressed in several approaches. However, there is not an outright method for heartbeat classification able to handle problems such as the large amount of data and highly unbalanced classes. This work introduces a heuristic-search-based clustering to discriminate among ventricular cardiac arrhythmias in Holter recordings. The proposed method is posed under the normalized cut criterion, which iteratively seeks for the nodes to be grouped into the same cluster. Searching procedure is carried out in accordance to the introduced maximum similarity value. Since our approach is unsupervised, a procedure for setting the initial algorithm parameters is proposed by fixing the initial nodes using a kernel density estimator. Results are obtained from MIT/BIH arrhythmia database providing heartbeat labelling. As a result, proposed heuristic-search-based clustering shows an adequate performance, even in the presence of strong unbalanced classes.

Wearable and superhydrophobic hardware for ambulatory biopotential acquisition

Wearable monitoring devices are a promising trend for ambulatory and real time biosignal processing, because they improve access and coverage by means of comfortable sensors, with real-time communication via mobile networks. In this paper, we present a garment for ambulatory electrocardiogram monitoring, a smart t-shirt with a textile electrode that conducts electricity and has a coating designed to preserve the user's hygiene, allowing long-term mobile measurements. Silicon dioxide nanoparticles were applied on the surface of the textile electrodes to preserve conductivity and impart superhydrophobic properties. A model to explain these results is proposed. The best result of this study is obtained when the contact angles between the fluid and the fabric exceeded 150°, while the electrical resistivity remained below 5 Ω·cm, allowing an acquisition of high quality electrocardiograms in moving patients. Thus, this tool represents an interesting alternative for medium and long-term measurements, preserving the textile feeling of clothing and working under motion conditions.

Stochastic relevance analysis of epileptic EEG signals for channel selection and classification

Time-frequency decompositions (TFDs) are well known techniques that permit to extract useful information or features from EEG signals, being necessary to distinguish between irrelevant information and the features effectively representing the subjacent physiological phenomena, according to some evaluation measure. This work introduces a new method to obtain relevant features extracted from time-frequency plane for epileptic EEG signals. Particularly, EEG features are extracted by common spectral methods such as short time Fourier transform (STFT), wavelets transform and Empirical Mode Decomposition (EMD). Then, each method is evaluated by Stochastic Relevance Analysis (SRA) that is further used for EEG classification and channel selection. The classification measures are carried out based on the performance of the k-NN classifier, while the channels selected are validated by visual inspection and topographic scalp map. The study uses real and multi-channel EEG data and all the experiments have been supervised by an expert neurologist. Results obtained in this paper show that SRA is a good alternative for automatic seizure detection and also opens the possibility of formulating new criteria to select, classify or analyze abnormal EEG channels.

Complexity-based analysis for the detection of heart murmurs

While a healthy human heart produce a rhythmic pattern of sounds, some heart disorder induce deviations perceived as abnormal sounds called murmurs. Despite many murmurs can be considered harmless, other constitute the first basis of a heart disorder. In this sense, a correct diagnosis remains essential; however, due to the subjectivity on using human ear to make diagnosis, automatic detection systems appear as useful tools for helping medical specialists on improving diagnosis accuracy. Complexity analysis has become one important tool for the study of physiological signals, because tracking sudden alteration on the inherent complexity on biological processes might be useful for detecting pathologies. The present paper presents a complexity-based analysis methodology, which uses regularity features for the detection of heart murmurs, including Approximate Entropy, Sample Entropy, Gaussian Kernel Approximate Entropy, and Fuzzy Entropy. The results show the high discriminative power, up to 90%, of the Gaussian Kernel Approximate Entropy and Fuzzy Entropy for the proposed labour.

On determining available stochastic features by spectral splitting in obstructive sleep apnea detection

Heart rate variability (HRV) is one of the promising directions for a simple and noninvasive way for obstructive sleep apnea syndrome detection. The time-frequency representations has been proposed before to investigate the non-stationary properties of the HRV during either transient physiological or pathological episodes. Within the framework of the filter-banked feature extraction, estimation of the spectral splitting for stochastic features extraction is an open issue. Usually, this splitting is fixed empirically without taking into account the actual informative distribution of time-frequency representations. In the present work, a relevance-based approach that aims to find a priori a boundaries in the frequency domain for the spectral splitting upon t-f planes is proposed. Results show that the approach is able to find the most informative frequency bands, achieving accuracy rate over 75%.

Training using short-time features for OSA discrimination

Heart rate variability (HRV) is one of the promising directions for a simple and noninvasive way for obstructive sleep apnea syndrome detection (OSA). The interaction between the sympathetic and parasympathetic systems on the HRV recordings, gives rise to several non-stationary components added to the signal. Aiming to improve the classifier accuracy for obstructive sleep apnoea detection, the use of more appropriated techniques for leading with non-stationarity and mixed dynamics, are needed. This work aims at searching a convenient training strategy of combining the feature set to be further fed in to the classifier, which should take into consideration the different dynamics in the HRV signal. Therefore, a set of the short-time features, extracted from a given HRV time-varying decomposition, and selected by spectral splitting is considered. Additionally, three methods of projection are used: none, simple, and multivariate. Finally, the different approaches are tested and compared, using k-nn and support vector machines (SVM) classifiers. Attained results show that using continuous wavelet transform with short-time features and multivariate projection, followed by a SVM classifier, allow to obtain a suitable OSA detection.

Effect of latency on clustering of P300 recordings for ADHD discrimination

This paper is focused on testing the latency contribution as regards the quality of formed groups for discriminating between healthy and attention deficit hyperactivity disorder children. To this end, two different cases are considered: nonaligned original recordings and aligned signals according to P300 position. For latter case, a novel approach to conduct time location of P300 component is introduced, which is based on derivative of event-related potential signals. The used database holds event-related potentials registered in auditory and visual oddball paradigm. Several experiments are carried out testing both configurations of considered data matrix. For grouping input data matrices, the k-means clustering technique is employed. To assess the quality of formed clusters and the relevance for clustering of latency-based features, relative values of distances between centroids and data points are computed in order to apprise separability and compactness of estimated clusters. Experimental results show that time localization of P300 component is not a decisive feature in formation of compact and well-defined groups within a discrimination framework for two considered data classes under certain conditions.

Novel signal-dependent filter bank method for identification of multiple basal ganglia nuclei in Parkinsonian patients

Microelectrode recordings are a valuable tool for assisting localization targets during deep brain stimulation procedures in Parkinson's disease neurosurgery. Attempts to automate and standardize this process have been limited by variability in patient neurophysiology and strong dynamics of microelectrode recordings. In this paper, a methodology for the identification of basal ganglia nuclei is presented that is based on a signal-dependent filter bank method using microelectrode recordings. The method is a customized realization of the discrete wavelet transform via the lifting scheme that is optimally tuned by genetic algorithms. Using this method, unique mother wavelet functions that exhibit an adaptable spectrum to the microelectrode recording dynamic are generated. Additionally, by extracting morphological features from the space-transformed microelectrode recording, it is possible to integrate them into three-dimensional (3D) feature spaces with maximum class separability. Finally, high discriminant feature spaces are fed into basic classifiers to recognize up to four basal nuclei. Comparison with several existing wavelets highlights the characteristics of new mother wavelets. Additionally, classification results show that identification of addressed nuclei in the basal ganglia can be performed with 95% confidence.

Evaluation of feature extraction techniques on event-related potentials for detection of attention-deficit/hyperactivity disorder

Event-related potentials (ERPs) are one of the most informative and dynamic methods of monitoring cognitive processes, which are widely used in clinical research to deal a variety of psychiatric and neurological disorders as attention-deficit/hyperactivity disorder (ADHD). This work proposes an extraction and selection methodology for discriminating between normal and pathological patients with ADHD by using ERPs. Three different sets of features (morphological, wavelets, and nonlinear based) are analyzed, looking for the best classification accuracy. The results show that the wavelet features provided a good discriminative capability, but it improved by combining all the set of features and applying a feature selection algorithm, reaching a maximum accuracy rate of 91.3%.

Sleep apnoea detection in children using PPG envelope-based dynamic features

Photopletysmography signal has been developed for monitoring of Obstructive Sleep Apnoea, in particular, whenever an apneic episode occurs, that is reflected by decreases in the photopletysmography signal amplitude fluctuation. However, other physiological events such as artifacts and deep inspiratory gasp produce sympathetic activation, being unrelated to apnea. Thus, its high sensitivity can produce misdetections and overestimate apneic episodes. In this regard, a methodology for selecting a set of relevant non-stationary features to increase the specificity of the obstructive sleep apnea detector is discussed. A time-evolving version of the standard linear multivariate decomposition is discussed to perform stochastic dimensionality reduction. As a result, performed outcomes of accuracy bring enough evidence that if using a subset of cepstral-based dynamic features, then patient classification accuracy is 83.3%. Therefore, photoplethysmography--based detection provides an adequate scheme for obstructive sleep apnea diagnosis.

EEG seizure identification by using optimized wavelet decomposition

A methodology for wavelet synthesis based on lifting scheme and genetic algorithms is presented. Often, the wavelet synthesis is addressed to solve the problem of choosing properly a wavelet function from an existing library, but which may be not specially designed to the application in hand. The task under consideration is the identification of epileptic seizures over electroencephalogram recordings. Although basic classifiers are employed, results rendered that the proposed methodology is successful in the considered study achieving similar classification rates that had been reported in literature.

TFR-based feature extraction using PCA approaches for discrimination of heart murmurs

Discrimination of murmurs in heart sounds is accomplished by means of time-frequency representations (TFR) which help to deal with non-stationarity. Nevertheless, classification with TFR is not straightforward given their large dimension and redundancy. In this paper we compare several methodologies to apply Principal Component Analysis (PCA) to TFR as a dimensional reduction scheme, which differ in the form that features are represented. Besides, we propose a method which maximizes information among TFR preserving information within TFRs. Results show that the methodologies that represent TFRs as matrices improve discrimination of heart murmurs, and that the proposed methodology shrinks variability of the results.

A new approach to discriminative HMM training for pathological voice classification

This paper presents a new approach that improves discriminative training criterion for Hidden Markov Models, and is oriented to pathological voice identification. This technique is aimed at maximizing the Area under the Curve of a receiver operating characteristic curve by adjusting the model parameters using as objective function the Mahalanobis distance and the distance between means of the underlying probability density functions associated with each class. The results show that the proposed technique significantly outperforms the accuracy in a classification system compared with other training criteria. Results are provided using the MEEIVL voice disorders database.

On improvement of detection of Obstructive Sleep Apnea by partial least square-based extraction of dynamic features

This paper presents a methodology for Obstructive Sleep Apnea (OSA) detection based on the HRV analysis, where as a measure of relevance PLS is used. Besides, two different combining approaches for the selection of the best set of contours are studied. Attained results can be oriented in research focused on finding alternative methods minimizing the HRV-derived parameters used for OSA diagnosing, with a diagnostic accuracy comparable to a polysomnogram. For two classes (normal, apnea) the results for PLS are: specificity 90%, sensibility 91% and accuracy 93.56%.

Estimation of dynamic neural activity using a Kalman filter approach based on physiological models

This paper presents a new method to estimate dynamic neural activity from EEG signals. The method is based on a Kalman filter approach, using physiological models that take both spatial and temporal dynamics into account. The filter's performance (in terms of estimation error) is analyzed for the cases of linear and nonlinear models having either time invariant or time varying parameters. The best performance is achieved with a nonlinear model with time-varying parameters.

Unsupervised feature selection in cardiac arrhythmias analysis

The problem of detecting clinical events related to cardiac arrhythmias in long term electrocardiograms is a difficult one due to the large amount of irrelevant information that hides such events. This problem has been addressed in the literature by means of clustering or classification algorithms that create data partitions according to a cost function based on heartbeat features dissimilarity measures. However, studies about the type or number of heartbeat features is lacking. Usually, the feature sets used are relevant but redundant, which degrades algorithm performance. This paper describes a method for automatic selection of heartbeat features. This method is assessed using real signals from the MIT database and common features used in previous works.

Feature extraction of speech signals in emotion identification

In this work, the acoustic and spectral characteristics and the automatic recognition of human emotional states through speech analysis have been studied. Acoustic features have been evaluated and features from time-frequency representation are proposed. The method is based in the representation of speech signal through energy distributions (Gabor transform and WVD) and discrete coefficients (DWT and linear prediction analysis). Recognition accuracy of 94.6% for emotion detection are obtained from SES database of emotional speech in spanish language.