Lina Abou-Abbas

Emotions are caused by a human brain reaction to objective events. The purpose of this study is to investigate emotion identification by machine learning using electroencephalography (EEG) data. Current research in EEG-based emotion recognition faces significant challenges due to the high-dimensionality and variability of EEG signals, which complicate accurate classification. Traditional methods often struggle to extract relevant features from noisy and high-dimensional data, and they typically fail to capture the complex temporal dependencies within EEG signals. Recent progress in machine learning by deep neural networks has opened up opportunities to develop methods highly efficient and practicable as to serve useful real-world applications. The purpose of this study is to investigate a novel end-to-end deep learning method of emotion recognition using EEG data, which prefaces a combination of two-dimensional (2D) convolutional network (CNN) and Long short-term memory network (LSTM) by an autoencoder. The autoencoder layers seek a lower dimensionality encoding for optimal input signal reconstruction, and the 2D CNN/LSTM combination layers capture both spatial and temporal features that best describe the emotion classes present in the data. Experiments in four-category classification of emotions, using the public and freely available DEAP dataset, revealed that the method reached superior performance: 90.04% for the "arousal" category, 89.97% for "valence", 87.73% for "dominance," and 90.84% for liking", as measured by the accuracy metric.

Background: Imbalanced datasets pose challenges for developing accurate seizure detection systems based on electroencephalogram (EEG) data. Generative AI techniques may help augment minority class data to facilitate automatic epileptic seizure detection.New method: This study investigates the impact of various data augmentation (DA) approaches, including Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP), Vanilla GAN, Conditional GAN (CGAN), and Cramer GAN, on classification performance with Random Forest models. The best-performing GAN variant, WGAN-GP, was then integrated with a bidirectional Long Short-Term Memory (LSTM) architecture and compared against traditional and synthetic oversampling methods.Results: The evaluation of different GAN variants for data augmentation with Random Forest classifiers identified WGAN-GP as the most effective approach. The integration of WGAN-GP with bidirectional LSTM yielded substantial performance improvements, outperforming traditional oversampling methods and achieving an accuracy of 91.73% on the augmented data, compared to 86% accuracy on real data without augmentation.Comparison with existing methods: The proposed generative AI approach combining WGAN-GP and recurrent neural network models outperforms comparative synthetic oversampling methods on metrics relevant for reliable seizure detection from imbalanced EEG datasets.Conclusions: Incorporating the WGAN-GP generative AI technique for data augmentation and integrating it with bidirectional LSTM elevates seizure detection accuracy for imbalanced EEG datasets, surpassing the performance of traditional oversampling and class weight adjustment methods. This approach shows promise for improving epilepsy monitoring and management through enhanced automated detection system effectiveness.

Background:: Characterizing the condition of patients suffering from knee osteoarthritis is complex due to multiple associations between clinical, functional, and structural parameters. While significant variability exists within this population, especially in candidates for total knee arthroplasty, there is increasing interest in knee kinematics among orthopedic surgeons aiming for more personalized approaches to achieve better outcomes and satisfaction. The primary objective of this study was to identify distinct kinematic phenotypes in total knee arthroplasty candidates and to compare different methods for the identification of these phenotypes. Methods:: Three-dimensional kinematic data obtained from a Knee Kinesiography exam during treadmill walking in the clinic were used. Various aspects of the clustering process were evaluated and compared to achieve optimal clustering, including data preparation, transformation, and representation methods. Results:: A K-Means clustering algorithm, performed using Euclidean distance, combined with principal component analysis applied on data transformed by standardization, was the optimal approach. Two unique kinematic phenotypes were identified among 80 total knee arthroplasty candidates. The two distinct phenotypes divided patients who significantly differed both in terms of knee kinematic representation and clinical outcomes, including a notable variation in 63.3% of frontal plane features and 81.8% of transverse plane features across 77.33% of the gait cycle, as well as differences in the Pain Catastrophizing Scale, highlighting the impact of these kinematic variations on patient pain and function. Conclusion:: Results from this study provide valuable insights for clinicians to develop personalized treatment approaches based on patients’ phenotype affiliation, ultimately helping to improve total knee arthroplasty outcomes.

Deep learning has served pattern classification in many applications, with a performance which often well exceeds that of other machine learning paradigms. Yet, in general, deep learning has used computational architectures built, albeit partially, by ad hoc means, and its classification decisions are not necessarily interpretable in terms of knowledge relevant to the application it serves. This is often referred to as the black box problem, which in certain applications, such as epileptic seizure prediction, can be a serious impediment. The purpose of this study is to investigate an interpretable deep learning classifier for epileptic EEG-driven seizure prediction. This neural network is interpretable because its layers can be visualized and interpreted as a result of a novel architecture where the learned weights follow from signal processing computations such as frequency sub-band and spatial filters. Consequently, the extracted features are no longer abstract as they correspond to the features commonly used for decoding EEG data. In addition, the network uses layer-wise relevance propagation to reveal pertinent features which can further explain the computations leading to the decisions. In seizure prediction experiments using the CHB-MIT data set, the method produced classification results which improved on the state-of-the art, with first network layer filters corresponding to clinically relevant frequency bands, and the input channels in the brain location in which the seizure originates contributing most significantly to the network predictions.

Monitoring patients at risk of epileptic seizure is critical for optimal treatment and ensuing the reduction of seizure risk and complications. In general, seizure detection is done manually in hospitals and involves time-consuming visual inspection and interpretation by experts of electroencephalography (EEG) recordings. The purpose of this study is to investigate the pertinence of band-limited spectral power and signal complexity in order to discriminate between seizure and seizure-free EEG brain activity. The signal complexity and spectral power are evaluated in five frequency intervals, namely, the delta, theta, alpha, beta, and gamma bands, to be used as EEG signal feature representation. Classification of seizure and seizure-free data was performed by prevalent potent classifiers. Substantial comparative performance evaluation experiments were performed on a large EEG data record of 341 patients in the Temple University Hospital EEG seizure database. Based on statistically validated criteria, results show the efficiency of band-limited spectral power and signal complexity when using random forest and gradient-boosting decision tree classifiers (95% of the area under the curve (AUC) and 91% for both F-measure and accuracy). These results support the use of these automatic classification schemes to assist the practicing neurologist interpret EEG records more accurately and without tedious visual inspection.