RESUMO

Epilepsy is a neurological disease with one of the highest rates of incidence worldwide. Although EEG is a crucial tool for its diagnosis, the manual detection of epileptic seizures is time consuming. Automated methods are needed to streamline this process; although there are already several works that have achieved this, the process by which it is executed remains a black box that prevents understanding of the ways in which machine learning algorithms make their decisions. A state-of-the-art deep learning model for seizure detection and three EEG databases were chosen for this study. The developed models were trained and evaluated under different conditions (i.e., three distinct levels of overlap among the chosen EEG data windows). The classifiers with the best performance were selected, then Shapley Additive Explanations (SHAPs) and Local Interpretable Model-Agnostic Explanations (LIMEs) were employed to estimate the importance value of each EEG channel and the Spearman's rank correlation coefficient was computed between the EEG features of epileptic signals and the importance values. The results show that the database and training conditions may affect a classifier's performance. The most significant accuracy rates were 0.84, 0.73, and 0.64 for the CHB-MIT, Siena, and TUSZ EEG datasets, respectively. In addition, most EEG features displayed negligible or low correlation with the importance values. Finally, it was concluded that a correlation between the EEG features and the importance values (generated by SHAP and LIME) may have been absent even for the high-performance models.

RESUMO

Abstract: Tinnitus detection and characterization requires a carefully elaborated diagnosis mainly owing to its heterogeneity nature. The present investigation aims to find features in Electroencephalographic (EEG) signals from time and frequency domain analysis that could distinguish between healthy and tinnitus sufferers with different levels of hearing loss. For this purpose, 24 volunteers were recruited and equally divided into four groups: 1) controls, 2) slow tinnitus, 3) middle tinnitus and 4) high tinnitus. EEG signals were registered in two states, with eyes closed and opened for 60 seconds. EEG analysis was focused on two bandwidths: delta and alpha band. For time domain, the EEG features estimated were mean, standard deviation, kurtosis, maximum peak, skewness and shape. For frequency domain, the EEG features obtained were mean, skewness, power spectral density. Normality of EEG data was evaluated by the Lilliefors test, and as a result, the nonparametric technique Kruskal-Wallis H statistic to test significance was applied. Results show that EEG features are more differentiable between tinnitus sufferers and controls in frequency domain than in time domain. EEG features from tinnitus patients with high HL are significantly different from the rest of the groups in alpha frequency band activity when shape and skewness are computed.

---

Resumen: La detección y caracterización del acúfeno requiere un diagnóstico cuidadosamente elaborado debido principalmente a su naturaleza heterogénea. La presente investigación tiene como objetivo encontrar características en las señales electroencefalográficas (EEG) a partir del análisis del dominio del tiempo y frecuencia que podrían distinguir entre pacientes sanos y con acúfeno con diferentes niveles de pérdida auditiva. Para ello, se reclutaron 24 voluntarios y se dividieron por igual en cuatro grupos: 1) controles, 2) acúfeno bajo, 3) acúfeno medio y 4) acufeno alto. La actividad EEG se registró en reposo en dos condiciones: ojos cerrados y abiertos durante un minuto. El análisis de EEG se centró en anchos de banda delta y alfa. Para el dominio del tiempo, las características del EEG estimadas fueron la media, la desviación estándar, la curtosis, el pico máximo, la asimetría y la forma. Para el dominio de la frecuencia, las características de EEG obtenidas fueron media, asimetría, densidad espectral de potencia. La normalidad de los datos del EEG se evaluó mediante la prueba de Lilliefors y, como resultado, se aplicó la técnica no paramétrica del estadístico H de Kruskal-Wallis para probar la significación. Los resultados muestran que las características del EEG son más diferenciables entre los pacientes con acúfeno y los controles en el dominio de la frecuencia que en el dominio del tiempo. Las características del EEG de los pacientes con acúfeno con alta pérdida de audición son significativamente diferentes del resto de los grupos en la actividad de la banda de alfa cuando se calculan la forma y la asimetría.

RESUMO

Pneumonia and pulmonary thromboembolism (PTE) are both respiratory diseases; their diagnosis is difficult due to their similarity in symptoms, medical subjectivity, and the large amount of information from different sources necessary for a correct diagnosis. Analysis of such clinical data using computational tools could help medical staff reduce time, increase diagnostic certainty, and improve patient care during hospitalization. In addition, no studies have been found that analyze all clinical information on the Mexican population in the Spanish language. Therefore, this work performs automatic diagnosis of pneumonia and pulmonary thromboembolism using machine-learning tools along with clinical laboratory information (structured data) and clinical text (unstructured data) obtained from electronic health records. A cohort of 173 clinical records was obtained from the Mexican Social Security Institute. The data were preprocessed, transformed, and adjusted to be analyzed using several machine-learning algorithms. For structured data, naïve Bayes, support vector machine, decision trees, AdaBoost, random forest, and multilayer perceptron were used; for unstructured data, a BiLSTM was used. K-fold cross-validation and leave-one-out were used for evaluation of structured data, and hold-out was used for unstructured data; additionally, 1-vs.-1 and 1-vs.-rest approaches were used. Structured data results show that the highest AUC-ROC was achieved by the naïve Bayes algorithm classifying PTE vs. pneumonia (87.0%), PTE vs. control (75.1%), and pneumonia vs. control (85.2%) with the 1-vs.-1 approach; for the 1-vs.-rest approach, the best performance was reported in pneumonia vs. rest (86.3%) and PTE vs. rest (79.7%) using naïve Bayes, and control vs. diseases (79.8%) using decision trees. Regarding unstructured data, the results do not present a good AUC-ROC; however, the best F1-score were scored for control vs. disease (72.7%) in the 1-vs.-rest approach and control vs. pneumonia (63.6%) in the 1-to-1 approach. Additionally, several decision trees were obtained to identify important attributes for automatic diagnosis for structured data, particularly for PTE vs. pneumonia. Based on the experiments, the structured datasets present the highest values. Results suggest using naïve Bayes and structured data to automatically diagnose PTE vs. pneumonia. Moreover, using decision trees allows the observation of some decision criteria that the medical staff could consider for diagnosis.

RESUMO

Epilepsy is a disease that decreases the quality of life of patients; it is also among the most common neurological diseases. Several studies have approached the classification and prediction of seizures by using electroencephalographic data and machine learning techniques. A large diversity of features has been extracted from electroencephalograms to perform classification tasks; therefore, it is important to use feature selection methods to select those that leverage pattern recognition. In this study, the performance of a set of feature selection methods was compared across different classification models; the classification task consisted of the detection of ictal activity from the CHB-MIT and Siena Scalp EEG databases. The comparison was implemented for different feature sets and the number of features. Furthermore, the similarity between selected feature subsets across classification models was evaluated. The best F1-score (0.90) was reported by the K-nearest neighbor along with the CHB-MIT dataset. Results showed that none of the feature selection methods clearly outperformed the rest of the methods, as the performance was notably affected by the classifier, dataset, and feature set. Two of the combinations (classifier/feature selection method) reporting the best results were K-nearest neighbor/support vector machine and random forest/embedded random forest.

Assuntos

Epilepsia , Qualidade de Vida , Algoritmos , Eletroencefalografia/métodos , Epilepsia/diagnóstico , Humanos , Convulsões/diagnóstico , Processamento de Sinais Assistido por Computador , Máquina de Vetores de Suporte

RESUMO

Tinnitus is an auditory condition that causes humans to hear a sound anytime, anywhere. Chronic and refractory tinnitus is caused by an over synchronization of neurons. Sound has been applied as an alternative treatment to resynchronize neuronal activity. To date, various acoustic therapies have been proposed to treat tinnitus. However, the effect is not yet well understood. Therefore, the objective of this study is to establish an objective methodology using electroencephalography (EEG) signals to measure changes in attentional processes in patients with tinnitus treated with auditory discrimination therapy (ADT). To this aim, first, event-related (de-) synchronization (ERD/ERS) responses were mapped to extract the levels of synchronization related to the auditory recognition event. Second, the deep representations of the scalograms were extracted using a previously trained Convolutional Neural Network (CNN) architecture (MobileNet v2). Third, the deep spectrum features corresponding to the study datasets were analyzed to investigate performance in terms of attention and memory changes. The results proved strong evidence of the feasibility of ADT to treat tinnitus, which is possibly due to attentional redirection.

Assuntos

Zumbido , Estimulação Acústica , Atenção , Percepção Auditiva , Eletroencefalografia , Humanos , Zumbido/terapia

RESUMO

Genomic signal processing (GSP) methods which convert DNA data to numerical values have recently been proposed, which would offer the opportunity of employing existing digital signal processing methods for genomic data. One of the most used methods for exploring data is cluster analysis which refers to the unsupervised classification of patterns in data. In this paper, we propose a novel approach for performing cluster analysis of DNA sequences that is based on the use of GSP methods and the K-means algorithm. We also propose a visualization method that facilitates the easy inspection and analysis of the results and possible hidden behaviors. Our results support the feasibility of employing the proposed method to find and easily visualize interesting features of sets of DNA data.