Image segmentation is a fundamental task in computer vision and computational social systems, with semantic segmentation aiming to assign each pixel to a corresponding label. Due to the inherent richness of categories and contextual information in images, image segmentation remains a challenging problem. Currently, semantic segmentation models based on the segment anything model have demonstrated promising results. However, they continue to encounter challenges related to training strategies and prompt information generation. To address these issues, we propose a straightforward and efficient design method for semantic segmentation based on a prompt-free model, named SAM-SS. First, we introduce the class prompt encoder, which generates category prompts for the mask decoder to extract category-specific semantic information. Second, we incorporate the deep fusion module to bridge the semantic gap for achieving robust representation. Additionally, we observe that fine-tuning the image encoder via low-rank adaptation often leads to suboptimal convergence. To mitigate this, we propose a learning rate modulation strategy to stabilize training and boost model performance. Finally, we validate our model’s performance on three publicly available datasets. Specifically, on the Cityscapes validation set for natural images, our model achieves a mean intersection over union (mIoU) of 85.28%. Moreover, our model demonstrates strong adaptability to remote sensing imagery, achieving mIoU scores of 54.46% on LoveDA and 80.8% on the ISPRS Potsdam validation sets. These results underscore its potential utility across diverse applications.

Drug addicts are characterized by difficulty neglecting monetary reward, but its underlying neural mechanisms remain unclear. The current study aimed to investigate the behavioral and electrophysiological signatures of abnormal attentional bias based on different amounts of reward in abstinent heroin addicts (AHAs). We used a modified attentional capture task while recording EEG in 18 AHAs and 18 age-, gander-, and education-matched healthy controls (HCs). We analyzed the attentional distribution of the relative positional changes in space of the target and reward-related stimulus. When targets integrated reward-related colors, participants were more responsive and deployed more attention to targets, especially those with high-value colors. When targets and reward-related distractors were spatially separated, high-value distractors captured the AHA's attention and slowed their responses. Moreover, AHAs had weaker attentional control than HCs, exhibiting an inability to suppress the attentional bias driven by high-value stimuli. Overall, these results demonstrated that AHAs was hypersensitive to task-irrelevant and previous reward-related stimuli, possibly due to damage to brain reward circuits caused by chronic heroin abuse. Our work provides novel behavioral and neurophysiological evidence that are closely associated with the maintenance and relapse of addiction.

Objective: Diagnosis of the severity of heroin addiction with electroencephalography (EEG) signals is a challenging problem. It has been shown that brain microstates are associated with brain status and healthy condition. However, there is no study on how heroin addiction affects brain microstates. Approach: We propose a hybrid classifier based on the microstate features, extracting from resting state EEGs, to objectively and effectively identify abstinent heroin-addicted individuals (AHAIs) and healthy controls (HCs). In addition to the commonly used features such as duration, occurrence, and transition, we calculated three new features. Main Results: The results showed that the support vector machine (SVM), which allows classification of the AHAIs and HCs with a 73% accuracy rate, was an optimal classifier. Moreover, the weight setting-based genetic algorithm (GA) further improved the accuracy rate to 81%. The hybrid classification not only provides direct evidence showing the differences in EEG microstate features between AHAIs and HCs, but also offers a method to distinguish the heroin brain states of people addicted to heroin and healthy individuals and demonstrates that microstate features could serve as potential bio-markers for identifying AHAIs. Significance: our methods and the selected features may provide electrophysiological insights for the assessment of the heroin withdrawal treatment effects.

Automatic seizure onset detection in electroencephalogram (EEG) signals is vital for the monitoring and diagnosis of epilepsy. In this paper, we develop a novel detection framework employing the common spatial pattern (CSP) and discriminative log-Euclidean kernel-based Gaussian process (DLEK-GP) for distinguishing epileptic EEG signals. In the framework, the CSP is utilized as a feature extractor to reduce the dimension of multi-channel data representation and obtain distinguishing features. Afterwards, the DLEK defined on the Riemannian manifold of symmetric positive definite (SPD) matrices is combined with the GP classifier to categorize the EEG signals. The assessment results obtained from the CHB-MIT EEG dataset demonstrate that the proposed framework can achieve an average segment-based sensitivity of 97.57%, specificity of 97.26%, accuracy of 97.42%, and an average event-based sensitivity of 98.57%, false detection rate (FDR) of 0.54/h, latency of 1.02 s. The satisfactory results show that the proposed seizure detection framework holds the promising potential for clinical practice.

In this paper, to explore the application of depression EEG data in semi-supervised classification, we designed an improved semi-supervised graph convolutional neural network model for depression identification. Fifty-three volunteers, including 24 patients with depression and 29 healthy controls, participated in the study. Electroencephalogram (EEG) data from 128 channels were recorded in the resting state for 5 minutes. The differential entropy feature of EEG is obtained and its Pearson matrix is used to construct the node and adjacency matrix of the graph. For the classifier, we combined self-organizing incremental neural network (SOINN) and graph convolutional neural network (GCN) self-training to expand the training set, improve the effect of classification, and adopted 10-fold cross validation to verify the classification results. Compared with convolutional neural networks (CNN), long short-term memory (LSTM) neural network and classical fully supervised algorithms, such as support vector machine (SVM), the classification accuracy of the proposed model is 70.53% and 92.23% under the condition of label data of 50 and 600, respectively. Compared with the original GCN model, our method has significantly improved the performance indicators of Accuracy (Acc), Recall (Rec), Specificity (Spec) and Precision (Pre) and improved the ability of GCN label propagation. This study provides a semi-supervised learning model for detecting depression and a new method for diagnosing depression based on EEG signals.

Action recognition is essential for many human-centered applications in the Internet of Things (IoT). Especially, in the Internet of Medical Things (IoMT), action recognition shows great importance in surgical assistance, patient monitoring, etc. Recently, 3-D skeleton sequence-based action recognition draws broad attention. It is a challenging task that needs effective modeling on intraframe skeleton representations and interframe temporal dynamics. Standard long short-term memory (LSTM)-based models are widely used for sequence modeling due to its long-term memory, yet they are unable to fully model the relationship between different body joints or persons to extract crucial co-occurrence features from different levels. To handle this shortcoming, we propose an attention-based multilevel co-occurrence graph convolutional LSTM (AMCGC-LSTM). By integrating graph convolutional networks (GCNs) into LSTM, the proposed model is capable of leveraging body structural information from skeletons and strengthening the multilevel co-occurrence (MC) feature learning. Specifically, we first design the spatial attention module for feature enhancement of key joints from skeleton inputs. Second, we design MC memory units coupled with GCN to automatically model the spatial relationship between joints, and simultaneously capture the co-occurrence features from different joints, persons, and frames. Finally, we construct aggregated features of MCs (AFMCs) from MC memory units to better represent the intraframe action context encoding, and leverage a concurrent LSTM (Co-LSTM) to further model their temporal dynamics for action recognition. Our model significantly outperforms mainstream methods on NTU RGB+D 60/120 data set, mutual action subset of NTU RGB+D 60/120 data set, and Northewestern-UCLA data set.

Tens of millions of people suffer from depression worldwide. It is urgent to explore an effective method for diagnosing depression. This study developed a novel of multimodal feature fusion depression recognition method based on functional near-infrared spectroscopy (fNIRS). Sixty volunteers, including thirty patients with depression and thirty healthy controls, participated in the study. The 22-channel fNIRS device recorded the participants' brain oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) concentration changes in the positive, neutral and negative affective words' stimulation. K-nearest neighbors (KNN) and support vector machine (SVM) classifiers were used to recognize depressed patients from normal people, and 10-fold cross-validation was used to verify the classification result. Under the three single-mode features, the accuracy rates were 85.69%, 88.32% and 86.77%, corresponding to the positive condition, neutral condition and negative condition. Then, we used concatenation and linear combination for feature fusion. For the concatenation fusion method, the principal component analysis (PCA) was used to reduce the dimension. The result showed that feature fusion can relatively improve the recognition rate of people with depression, compared with single-model features. The optimal feature fusion method is to concatenate the neutral features and negative features, and the best accuracy reaches 94.45%. The study may provide a more accurate and convenient method for depression detection.

Tens of millions of people suffer from depression worldwide. It is urgent to explore an effective method for diagnosing depression. This study developed a novel of multimodal feature fusion depression recognition method based on functional near-infrared spectroscopy (fNIRS). Sixty volunteers, including thirty patients with depression and thirty healthy controls, participated in the study. The 22-channel fNIRS device recorded the participants’ brain oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) concentration changes in the positive, neutral and negative affective words’ stimulation. K-nearest neighbors (KNN) and support vector machine (SVM) classifiers were used to recognize depressed patients from normal people, and 10-fold cross-validation was used to verify the classification result. Under the three single-mode features, the accuracy rates were 85.69%, 88.32% and 86.77%, corresponding to the positive condition, neutral condition and negative condition. Then, we used concatenation and linear combination for feature fusion. For the concatenation fusion method, the principal component analysis (PCA) was used to reduce the dimension. The result showed that feature fusion can relatively improve the recognition rate of people with depression, compared with single-model features. The optimal feature fusion method is to concatenate the neutral features and negative features, and the best accuracy reaches 94.45%. The study may provide a more accurate and convenient method for depression detection.

As the occurrence of depression in society becomes increasingly more common, it is an urgent task to find more objective and effective tools for real-time depression assessment. Gait analysis offers a new low-cost and contactless method for depression diagnosis. Therefore, interest in gait-based depression detection using depth sensors, such as Kinect, has grown rapidly in recent years. In this paper, a pseudo-velocity model is built to analyze the abnormal gait related to the depression by combining the velocity and angular velocity of the joints. Subsequently, we extract some features in time and frequency domain from our model to establish the classification model for depression detection. Experimental results on depression gait data recordings from 43 scored-depressed and 52 non-depressed individuals show that the proposed method achieves a good classification accuracy of 92.35 % and is superior to other existing methods. The outstanding classification performance suggests that the proposed method has notential clinical value in depression detection.

Emotion, as an advanced function of the human brain, affects kinds of human behaviors. Electroencephalographs (EEG) are widely used in the field of emotion classification owing to their low cost and portability. In this work, we study the effects of a non-linear EEG feature and a channel selection method on emotion recognition. First, the fractal dimension(FD) which could reflect the state of the brain is extracted with a sliding window. The top seven channels are screened out by calculating the F-score from the whole samples. Then, based on the signals from forehead channels, filtered channels and associated channels, emotions on valence and arousal are classified by Support Vector Machine(SVM) and K Nearest Neighbours(KNN). The result shows that the forehead channels Fp2, AF8, Fpz play an important role in valence classification. When combining the forehead channels with other channels that have higher F-score, the SVM classifier has a better accuracy on the whole set with 89.37% on valence and 87.07% on arousal. Besides, the overall accuracy calculated on each participants with associated channels get significant improvement. Especially, the KNN classifier has a much better result on every subject. This phenomenon indicates that by combining the higher F-score channels with the forehead channels, the associated channels can not only take advantage of the forehead channels' ability to categorize emotions but also consider individual differences.

As a promising non-invasive technique, functional near-infrared spectroscopy(fNIRS) can easily detect the hemodynamic responses of cortical brain activities. This paper investigated the multiclass classification of motor imagery(MI)based on fNIRS. 10 healthy individuals were recruited to move an object using their imagination. A multi-channel continuous-wave fNIRS equipment was applied to obtain the signals from the prefrontal cortex(PFC). The combined Ensemble Empirical Mode Decomposition (EEMD) and Independent Component Analysis(ICA) method was used to solve the signal-noise frequency spectrum aliasing issues caused by Mayer wave(0.1Hz), then the signal means(SM) features were extracted as an input of Support Vector Machine(SVM) classifier. The average accuracies of 4 directions, up-down and left-right were 40.55%, 73.05%, 70.7% respectively using Hbo2(8-21s). This study demonstrated that Brodmann area 4 was activated, which is consistent with previous conclusions. Furthermore, we found that the orbitofrontal cortex is also involved in MI and O2sat can also serve as a classified index.

In recent years, emotion classification from electroencephalogram (EEG) has attracted more and more attention. Because fewer channels were required in the real-time emotion classification system, many channel selection methods were proposed. In this paper, we investigated the stability of the optimal channels selected by channel selection algorithm, Mean-ReliefF-channel-selection (MRCS). EEG signals from 4 male students were recorded using 64 electrodes. With the comparisons of three effective channel sets obtained from the features (Shannon entropy (SE), differential entropy (DE) and 1st difference (1ST)) respectively, we explored the similarity of the first 10 optimal channels from the different sets. The experimental results indicated that there were average 75% common channels for the individual subject. But between subjects, the mean value of all combinations of subjects for different features can only reach 16%. This phenomenon indicates that the selected optimal channels from fewer effective features are also suitable for many other valid features. But the strong instability of channels selected by MRCS between subjects makes it a big challenge to guide an effective design of real-time emotion classification system.

Electroencephalogram (EEG) is a noninvasive method to record electrical activity of brain and it has been used extensively in research of brain function due to its high time resolution. However raw EEG is a mixture of signals, which contains noises such as Ocular Artifact (OA) that is irrelevant to the cognitive function of brain. To remove OAs from EEG, many methods have been proposed, such as Independent Components Analysis (ICA), Discrete Wavelet Transform (DWT), Adaptive Noise Cancellation (ANC) and Wavelet Packet Transform (WPT). In this paper, we present a novel hybrid de-noising method which uses Discrete Wavelet Transform (DWT) and Kalman Filtering to remove OAs in EEG. Firstly, we used this method on simulated data. The Mean Squared Error (MSE) of DWT-Kalman method was 0.0017, significantly lower compared to results using WPT-ICA and DWT-ANC, which were 0.0468 and 0.0052, respectively. Meanwhile, the Mean Absolute Error (MAE) using DWT-Kalman achieved an average of 0.0052, which also performed better than WPT-ICA and DWT-ANC, which were 0.0218 and 0.0115, respectively. Then we applied the proposed approach to the raw data collected by our prototype three-channel EEG collector and 64-channel Braincap from BRAIN PRODUCTS. On both data, our method achieved satisfying results. This method does not rely on any particular electrode or the number of electrodes in certain system, so it is recommended for ubiquitous applications.

Modeling and prediction of Electroencephalogram (EEG) signals is very important for Portable applications; EEG signals are however widely regarded as being chaotic in nature. An adaptive modeling technique that combines Discrete Wavelet Transformation (DWT) to predict contaminated EEG signals for removal of ocular artifacts (OAs) from EEG records is proposed as an effective a data processing tool for Interventions in Mental Illness Based on Bio-feedback. The proposed method is well suited for use in portable environments where constraints with respect to acceptable wearable sensor attachments usually dictate single channel devices. Using simulated and measured data the accuracy of the proposed model is compared to the accuracy of other pre-existing methods based on Wavelet Packet Transform (WPT) and independent component analysis (ICA) using DWT and adaptive noise cancellation (ANC) for Portable applications. The results show that the our new model not only demonstrates an improved performance with respect to the recovery of true EEG signals, achieves improved computational speed, and demonstrates better tracking performance.

A new model to remove ocular artifacts (OA) from electroencephalograms (EEGs) is presented. The model is based on discrete wavelet transformation (DWT) and adaptive noise cancellation (ANC). Using simulated and measured data, the accuracy of the model is compared with the accuracy of other existing methods based on stationary wavelet transforms and our previous work based on wavelet packet transform and independent component analysis. A particularly novel feature of the new model is the use of DWTs to construct an OA reference signal, using the three lowest frequency wavelet coefficients of the EEGs. The results show that the new model demonstrates an improved performance with respect to the recovery of true EEG signals and also has a better tracking performance. Because the new model requires only single channel sources, it is well suited for use in portable environments where constraints with respect to acceptable wearable sensor attachments usually dictate single channel devices. The model is also applied and evaluated against data recorded within the EUFP 7 Project--Online Predictive Tools for Intervention in Mental Illness (OPTIMI). The results show that the proposed model is effective in removing OAs and meets the requirements of portable systems used for patient monitoring as typified by the OPTIMI project.

With the booms of mobile communication, especially mobile smart phone, technologies to identify individuals for mobile security calls for some more strict requirements in user-friendly, real-time and ubiquitous aspects. In addition to traditional approaches (for example, password check), some advanced biometric methodologies have been applied in practice, such as fingerprint and iris based solutions, however, these solutions generally lack a true ubiquitous nature for mobile security. In this paper, we present a real time EEG based individual identification interface to support ubiquitous applications. The EEG signals are collected through a mono-polar single channel in real time via a mobile EEG device. An experiment involving about 20 subjects has been conducted to evaluate the interface. The experiment comprises three types of tests: accuracy test, time dimension test and capacity dimension test. The results of these experiments demonstrate that our approach is highly suitable to the demands of mobile security in ubiquitous environment. In addition, we integrate this interface into scenarios of ubiquitous application - Online Predictive Tools for Intervention in Mental Illness (OPTIMI).