The timely detection and accurate localization of brain tumors is crucial in preserving people’s quality of life. Thankfully, intelligent computational systems have proven invaluable in addressing these challenges. In particular, the UNET model can extract essential pixel-level features to automatically identify the tumor’s location. However, known deep learning-based works usually directly feed the 3D volume into the model, which causes excessive computational complexity. This paper presents an approach to boost the UNET network, reducing computational workload while maintaining superior efficiency in locating brain tumors. This concept could benefit portable or embedded recognition systems with limited resources for operating in real time. This enhancement involves an automatic slice selection from the MRI T2 modality volumetric images containing the most relevant tumor information and implementing an adaptive learning rate to avoid local minima. Compared with the original model (7.7 M parameters), the proposed UNET model uses only 2 M parameters and was tested on the BraTS 2017, 2020, and 2021 datasets. Notably, the BraTS2021 dataset provided outstanding binary metric results: 0.7807 for the Intersection Over the Union (IoU), 0.860 for the Dice Similarity Coefficient (DSC), 0.656 for the Sensitivity, and 0.9964 for the Specificity compared to vanilla UNET.

Electroencephalograms provide a non-invasive and effective method for studying emotion recognition and developing Artificial Intelligence (AI) models to understand human behavior and decision-making processes. This study involved testing several machine learning classification kernels to develop an accurate emotion recognition model capable of classifying emotions stimuli such as “Boring,” “Calm,” “Happy,” and “Fear” during gameplay. An emotion classifier was assessed using the publicly available database for an emotion recognition system based on EEG signals and various computer games (GAMEEMO). The signal processing method, referred to as Regression EEG (REGEEG), involves an efficient electrode pairing selector developed for EEG signal processing using a regression algorithm, rotation matrices, director vectors, and robust statistical and polynomial feature extraction. REGEEG and feature extraction methods were evaluated with 28 machine learning kernels, resulting in five kernels with classification performance above 80%, with the K-Nearest Neighbors (k-NN) based model outperforming the rest (achieving over 95% accuracy, F1-Score, and kappa-score). REGEEG performance was further validated using 30 Cross-Validation (CV) folders and 28 in the Leave-one Subject-out (LoSo) technique without impacting the average classification performance. The classification highlights revealed low variance in the CV, while the LoSo approach helped identify outliers in the GAMEEMO dataset. Furthermore, the EEG pair channels selector demonstrates superior performance in classification, indicating a correlation between features and each processed pair of channels.

Popular comments suggest that continuous exposure of children and adolescents to video games yields a non-benefit behavior in the players’ mental health. Contrarily, several studies have proven that commercial and serious games improve mental activity; some are used in treating psychological and physical disorders. This paper presents a method based on electroencephalogram signals analysis to classify multiple emotions recorded from subjects’ gameplay seasons. In the core of this study, a self-assessed labeling method is evaluated using the Force, EEG, and Emotion Labelled Dataset (FEEL) for emotion recognition tasks. Besides, a 1-D Local Binary Pattern (LBP) method transforms the EEG temporal behavior to extract time-frequency features. Complementarily, the database artifacts were removed using a novel Conflict Learning approach for machine learning models, associating the extracted samples with the subjects’ emotion labeling. A semi-supervised clustering method was employed to show the similarity between self-assessed subjects’ labels. Finally, numerical results suggested a conflict between 23 original labels, improving the classification by over 92% in accuracy for 19 self-assessed classes.