The intelligent recognition of epileptic electroencephalogram (EEG) signals is a valuable tool for the epileptic seizure detection. Recent deep learning models fail to fully consider both spectral and temporal domain representations simultaneously, which may lead to omitting the nonstationary or nonlinear property in epileptic EEGs and further produce a suboptimal recognition performance consequently. In this paper, an end-to-end EEG seizure detection framework is proposed by using a novel channel-embedding spectral-temporal squeeze-andexcitation network (CE-stSENet) with a maximum mean discrepancy-based information maximizing loss. Specifically, the CE-stSENet firstly integrates both multi-level spectral and multiscale temporal analysis simultaneously. Hierarchical multidomain representations are then captured in a unified manner with a variant of squeeze-and-excitation block. The classification net is finally implemented for epileptic EEG recognition based on features extracted in previous subnetworks. Particularly, to address the fact that the scarcity of seizure events results in finite data distribution and the severe overfitting problem in seizure detection, the CE-stSENet is coordinated with a maximum mean discrepancy-based information maximizing loss for mitigating the overfitting problem. Competitive experimental results on three EEG datasets against the state-of-the-art methods demonstrate the effectiveness of the proposed framework in recognizing epileptic EEGs, indicating its powerful capability in the automatic seizure detection.

In this study, a new predictive framework is proposed by integrating an improved grey wolf optimization (IGWO) and kernel extreme learning machine (KELM), termed as IGWO-KELM, for medical diagnosis. The proposed IGWO feature selection approach is used for the purpose of finding the optimal feature subset for medical data. In the proposed approach, genetic algorithm (GA) was firstly adopted to generate the diversified initial positions, and then grey wolf optimization (GWO) was used to update the current positions of population in the discrete searching space, thus getting the optimal feature subset for the better classification purpose based on KELM. The proposed approach is compared against the original GA and GWO on the two common disease diagnosis problems in terms of a set of performance metrics, including classification accuracy, sensitivity, specificity, precision, G-mean, F-measure, and the size of selected features. The simulation results have proven the superiority of the proposed method over the other two competitive counterparts.

Parkinson's disease (PD) is a common neurodegenerative disease, which has attracted more and more attention. Many artificial intelligence methods have been used for the diagnosis of PD. In this study, an enhanced fuzzy k-nearest neighbor (FKNN) method for the early detection of PD based upon vocal measurements was developed. The proposed method, an evolutionary instance-based learning approach termed CBFO-FKNN, was developed by coupling the chaotic bacterial foraging optimization with Gauss mutation (CBFO) approach with FKNN. The integration of the CBFO technique efficiently resolved the parameter tuning issues of the FKNN. The effectiveness of the proposed CBFO-FKNN was rigorously compared to those of the PD datasets in terms of classification accuracy, sensitivity, specificity, and AUC (area under the receiver operating characteristic curve). The simulation results indicated the proposed approach outperformed the other five FKNN models based on BFO, particle swarm optimization, Genetic algorithms, fruit fly optimization, and firefly algorithm, as well as three advanced machine learning methods including support vector machine (SVM), SVM with local learning-based feature selection, and kernel extreme learning machine in a 10-fold cross-validation scheme. The method presented in this paper has a very good prospect, which will bring great convenience to the clinicians to make a better decision in the clinical diagnosis.

As 4-sensor line scan camera technology has matured, red (R), green (G), blue (B), and near-infrared (RGB-NIR) datasets have begun to appear in large numbers. The RGB-NIR data contain the rich color features of the RGB image and the sharp edge features of the NIR image. At present, in many studies, the RGB-NIR data are input directly into the processing algorithms for calculation of the 4D data; in these cases, redundant information is included, and the high correlation between the bands results in an inability to fully exploit the characteristics of the RGB-NIR data. In this paper, we propose a double-channel convolutional neural network (CNN) algorithm that takes into account the strong correlation between the R, G, and B bands in aerial images and the weaker correlation between the NIR band and the R, G, and B bands. First, the features of the RGB and NIR bands are calculated in two different CNN networks, and subsequently, feature fusion is performed in the fully connected layer. This is followed by the classification. By combining the two neural networks of RGB-CNN and NIR-CNN, the respective characteristics of the RGB-NIR data are fully exploited.

Even with the advantages of the sine cosine algorithm (SCA) in solving multimodal problems, there are some shortcomings for this method. We observe that the random patterns utilized in SCA cause an increasing attraction toward local optima. This study developed a rationalized version of this technique to deal with several representative benchmark cases with different dimensions. The improved algorithm combines the chaotic local search mechanism and Lévy flight operator with the core trends of SCA. The new variant is named as CLSCA. The Lévy flight with long jumps is adopted to boost the exploratory tendencies of the algorithm, while the chaotic local search mechanism is used as a local search for the destination point, which helps to further enhance the exploitation capability of SCA. Therefore, a suitable equilibrium between the exploration and exploitation can be kept in the CLSCA by two embedded patterns. To investigate the effectiveness and strength of the developed method, the CLSCA was tested on many benchmark functions, including different types of tasks such as single modal, multi-modal, hybrid, and composition functions. We compare the CLSCA with well-known optimizers, like particle swarm optimization (PSO) algorithm, grey wolf optimizer (GWO), SCA with differential evolution (SCADE), opposition-based SCA (OBSCA), fuzzy self-tuning PSO (FST_PSO), chaotic salp swarm algorithm (CSSA), and Chaotic whale optimizer (CWOA). Numerical experimental results demonstrate that the exploratory and exploitative properties of the classical SCA are clearly improved. The experimental results also show that our improved CLSCA is a better technique for different kinds of optimization tasks.