Emotion detection using electroencephalogram (EEG) signals is a rapidly evolving field with significant applications in mental health diagnostics, affective computing, and human–computer interaction. However, existing approaches often face challenges related to accuracy, interpretability, and real-time feasibility. This study leverages the DEAP dataset to explore and evaluate various machine learning and deep learning techniques for emotion recognition, aiming to address these challenges. To ensure reproducibility, we have made our code publicly available. Extensive experimentation was conducted using K-Nearest Neighbors (KNN), Support Vector Machines (SVMs), Decision Tree (DT), Random Forest (RF), Bidirectional Long Short-Term Memory (BiLSTM), Gated Recurrent Units (GRUs), Convolutional Neural Networks (CNNs), autoencoders, and transformers. Our hybrid approach achieved a peak accuracy of 85–95%, demonstrating the potential of advanced neural architectures in decoding emotional states from EEG signals. While this accuracy is slightly lower than some state-of-the-art methods, our approach offers advantages in computational efficiency and real-time applicability, making it suitable for practical deployment. Furthermore, we employed SHapley Additive exPlanations (SHAP) to enhance model interpretability, offering deeper insights into the contribution of individual features to classification decisions. A comparative analysis with existing methods highlights the novelty and advantages of our approach, particularly in terms of accuracy, interpretability, and computational efficiency. A key contribution of this study is the development of a real-time emotion detection system, which enables instantaneous classification of emotional states from EEG signals. We provide a detailed analysis of its computational efficiency and compare it with existing methods, demonstrating its feasibility for real-world applications. Our findings highlight the effectiveness of hybrid deep learning models in improving accuracy, interpretability, and real-time processing capabilities. These contributions have significant implications for applications in neurofeedback, mental health monitoring, and affective computing. Future work will focus on expanding the dataset, testing the system on a larger and more diverse participant pool, and further optimizing the system for broader clinical and industrial applications.

This study explores the use of deep learning techniques for classifying EEG signals in the context of Alzheimer’s disease (AD) and frontotemporal dementia (FTD). We propose a novel classification pipeline that combines autoencoders for feature extraction and bidirectional long short-term memory (Bi-LSTM) networks for analyzing patterns over time in EEG data. Given the complexity and high dimensionality of EEG signals, we employed an autoencoder to reduce data dimensionality while preserving key diagnostic features. The Bi-LSTM model effectively identified subtle temporal patterns in brain activity that are indicative of AD and FTD. To enhance interpretability, we applied SHapley Additive exPlanations (SHAP), providing insights into how individual features contribute to the model’s predictions. We evaluated our approach on a publicly available EEG dataset from OpenNeuro, which includes resting-state EEG recordings from 88 elderly participants—36 with AD, 23 with FTD, and 29 cognitively normal controls. EEG provides a non-invasive, cost-effective tool for brain monitoring, but presents challenges such as noise sensitivity and inter-subject variability. Despite these challenges, our approach achieved 98% accuracy while maintaining transparency, making it a promising tool for clinical applications in the diagnosis of neurodegenerative diseases.

The convergence of electroencephalography (EEG) and artificial intelligence (AI) has transformed brain-computer interfaces (BCI) from experimental medical devices into potential consumer technologies, raising unprecedented ethical challenges. This chapter analyzes ethical and transparency issues in EEG-based BCIs, examining implications for privacy, autonomy, equity, and human dignity. We explore three dimensions: technological foundations, emerging ethical risks, and practical solutions. Key challenges include “neuroprivacy” concerns, threats to mental autonomy, algorithmic bias, informed consent complexities, and security vulnerabilities. Neural data’s inferential potential through AI creates new categories of personal information exposure. We propose a framework integrating explainable AI, ethics-by-design principles, adaptive governance, and legal recognition of “neuro-rights.” Technical transparency and ethical integration are the essential conditions for responsible BCI development. While EEG-BCIs offer therapeutic potential, preserving human dignity requires ethics-centered innovation. This work provides theoretical frameworks and practical guidelines for navigating brain-machine interface ethics.

This study explores the use of deep learning techniques for classifying EEG signals in the context of Alzheimer’s disease (AD) and frontotemporal dementia (FTD). We propose a novel classification pipeline that combines autoencoders for feature extraction and bidirectional long short-term memory (Bi-LSTM) networks for temporal analysis of EEG data. Given the challenges of processing high dimensional EEG signals, we employed an Autoencoder to reduce the data’s dimensionality while preserving key features relevant for diagnosis. The Bi-LSTM model effectively captured the temporal dependencies in the EEG signals, essen- tial for detecting subtle neural changes indicative of AD and FTD. Furthermore, we leveraged SHapley Additive exPlanations (SHAP) to interpret the model’s predictions, providing transparency and ensuring that the contributions of individual features in the decision-making process are understood. The results demonstrate the efficacy of this approach, achieving an accuracy of 98%, while maintaining interpretability, making it a promising tool for clinical applications in the diagnosis of neurodegenerative diseases.

Cognitive flexibility enables individuals to adapt to changing rules, goals, or uncertainty. This study evaluates the discriminative power of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) signals from 42 healthy young adults during reversal learning tasks, using a comprehensive artificial intelligence (AI) framework. We compared classical machine learning models (e.g., Logistic Regression, Random Forest, XGBoost) and deep neural networks (MLP, GRU, BiLSTM), applied to EEG and fNIRS separately and in combination through multimodal fusion strategies. Interpretable machine learning played a significant role in our framework, with SHapley Additive exPlanations (SHAP) methodologies enabling transparent understanding of model predictions and feature importance across neurophysiological modalities. The interpretability fostered by SHAP enhances decision-making in cognitive assessments while cultivating deeper understanding of neurophysiological markers indicative of cognitive flexibility. Results show that EEG outperformed fNIRS in distinguishing "with objective (AO)" from "without objective (SO)" states. Logistic Regression achieved the best EEG performance (Accuracy = 0.810 ± 0.098, F1-score = 0.833 ± 0.089), while Random Forest was optimal for fNIRS (Accuracy = 0.647 ± 0.174, F1-score = 0.653 ± 0.228). Multimodal fusion strategies demonstrated varied effectiveness: early fusion with SVM achieved 74.17% ± 16.0% accuracy, intermediate fusion with XGBoost reached 75.3% ± 11.5%, while dynamic gating with instance-adaptive neural fusion provided dramatically superior performance at 95.00% ± 6.1% accuracy (95% Wilson CI: [0.8386, 0.9865])-representing a +14.00% improvement over best unimodal performance and +19.70% over static fusion methods. This exceptional accuracy was accompanied by 69.4% variance reduction compared to baseline approaches, demonstrating both superior classification capability and enhanced reliability. SHAP analysis revealed complementary modality roles: EEG demonstrated substantially higher gating importance (≈0.045) compared to fNIRS (≈0.028), with root mean square (RMS) features and spectral power emerging as most discriminative for EEG, while signal-to-noise ratio (SNR) features dominated fNIRS contributions. Train-test performance gap analysis (mean gap: 23.83%, range: 0.38%-40.91%) indicated generalization considerations requiring attention through expanded datasets and regularization strategies. These findings highlight the value of instance-adaptive multimodal AI integration for transparent identification of neurophysiological markers of cognitive flexibility. The dramatic performance advantage of dynamic gating over static fusion methods demonstrates that learned contextual modality weighting fundamentally surpasses fixed combination rules. However, the substantial train-test gaps underscore that these results establish methodological proof-of-concept in healthy young adults under controlled conditions, requiring extensive validation in heterogeneous populations, clinical cohorts, and naturalistic settings before clinical translation.