Information Theoretic-Based Interpretation of a Deep Neural Network Approach in Diagnosing Psychogenic Non-Epileptic Seizures

Abstract

Abstract:The use of a deep neural network scheme is proposed to help clinicians solve a difficult diagnosis problem in neurology. The proposed multilayer architecture includes a feature engineering step (from time-frequency transformation), a double compressing stage trained by unsupervised learning, and a classification stage trained by supervised learning. After fine-tuning, the deep network is able to discriminate well the class of patients from controls with around 90% sensitivity and specificity. This deep model gives better classification performance than some other standard discriminative learning algorithms. As in clinical problems there is a need for explaining decisions, an effort has been carried out to qualitatively justify the classification results. The main novelty of this paper is indeed to give an entropic interpretation of how the deep scheme works and reach the final decision.

A Comprehensive Machine-Learning-Based Software Pipeline to Classify EEG Signals: A Case Study on PNES vs. Control Subjects

“…Among these classifiers, linear discriminant analysis (LDA), support vector machine (SVM), and Bayesian network (BN) have been popular [13,14]. In [15], Morabito et al extracted different statistical features, such as mean (µ), standard deviation (σ), and skewness (v), from nontraditional sub-bands in the time-frequency maps of EEG signals. In many studies [15][16][17][18], the statistical features µ, σ, and v provided very robust classification scores.…”

“…In [15], Morabito et al extracted different statistical features, such as mean (µ), standard deviation (σ), and skewness (v), from nontraditional sub-bands in the time-frequency maps of EEG signals. In many studies [15][16][17][18], the statistical features µ, σ, and v provided very robust classification scores. Using artificial intelligence (AI) algorithms, Gasparini et al [15] were able to discriminate EEG time series of PNES from healthy controls.…”

“…In many studies [15][16][17][18], the statistical features µ, σ, and v provided very robust classification scores. Using artificial intelligence (AI) algorithms, Gasparini et al [15] were able to discriminate EEG time series of PNES from healthy controls. Based on previous results [15], in this paper we propose a ML pipeline to classify PNES from healthy controls (CNT) via statistical features extracted from a spectral EEG map.…”

Exploiting Deep Learning for Persian Sentiment Analysis

“…Similarly, CNNs have been proven to be very effective in sentiment analysis. However, little work has been carried out to exploit deep learning based feature representation for Persian sentiment analysis [16] [10]. In this paper, we present two deep learning models (deep autoencoders and CNNs) for Persian sentiment analysis.…”

Applications of Deep Learning in Financial Intermediation: A Systematic Literature Review

“…Existen varias áreas de conocimiento de la ingeniería computacional que emplean DL como reconocimiento de dígitos, procesamiento del lenguaje, reconocimiento de voz y técnicas de visión artificial en imágenes y en vídeo. En el campo de estudio de la salud el DL esta extensamente explotado para reconocimiento de patrones en imágenes o datos que determinen la presencia de enfermedades o mutaciones que afectan a las personas [2][3][4]. En la investigación científica se emplea para solventar problemas en los que se involucran grandes cantidades de datos como procesos de captura de CO2 [5], detectar ecuaciones o algoritmos para facilitar el desarrollo de ecuaciones [6,7].…”