A hybrid sensory feedback system for thermal nociceptive warning and protection in prosthetic hand.

Background: Advanced prosthetic hands may embed nanosensors and microelectronics in their cosmetic skin. Heat influx may cause damage to these delicate structures. Protecting the integrity of the prosthetic hand becomes critical and necessary to ensure sustainable function. This study aims to mimic the sensorimotor control strategy of the human hand in perceiving nociceptive stimuli and triggering self-protective mechanisms and to investigate how similar neuromorphic mechanisms implemented in prosthetic hand can allow amputees to both volitionally release a hot object upon a nociceptive warning and achieve reinforced release via a bionic withdrawal reflex. Methods: A steady-state temperature prediction algorithm was proposed to shorten the long response time of a thermosensitive temperature sensor. A hybrid sensory strategy for transmitting force and a nociceptive temperature warning using transcutaneous electrical nerve stimulation based on evoked tactile sensations was designed to reconstruct the nociceptive sensory loop for amputees. A bionic withdrawal reflex using neuromorphic muscle control technology was used so that the prosthetic hand reflexively opened when a harmful temperature was detected. Four able-bodied subjects and two forearm amputees randomly grasped a tube at the different temperatures based on these strategies. Results: The average prediction error of temperature prediction algorithm was 8.30 ± 6.00%. The average success rate of six subjects in perceiving force and nociceptive temperature warnings was 86.90 and 94.30%, respectively. Under the reinforcement control mode in Test 2, the median reaction time of all subjects was 1.39 s, which was significantly faster than the median reaction time of 1.93 s in Test 1, in which two able-bodied subjects and two amputees participated. Results demonstrated the effectiveness of the integration of nociceptive sensory strategy and withdrawal reflex control strategy in a closed loop and also showed that amputees restored the warning of nociceptive sensation while also being able to withdraw from thermal danger through both voluntary and reflexive protection. Conclusion: This study demonstrated that it is feasible to restore the sensorimotor ability of amputees to warn and react against thermal nociceptive stimuli. Results further showed that the voluntary release and withdrawal reflex can work together to reinforce heat protection. Nevertheless, fusing voluntary and reflex functions for prosthetic performance in activities of daily living awaits a more cogent strategy in sensorimotor control.

EMG-based Multi-User Hand Gesture Classification via Unsupervised Transfer Learning Using Unknown Calibration Gestures.

The poor generalization performance and heavy training burden of the gesture classification model contribute as two main barriers that hinder the commercialization of sEMG-based human-machine interaction (HMI) systems. To overcome these challenges, eight unsupervised transfer learning (TL) algorithms developed on the basis of convolutional neural networks (CNNs) were explored and compared on a dataset consisting of 10 gestures from 35 subjects. The highest classification accuracy obtained by CORrelation Alignment (CORAL) reaches more than 90%, which is 10% higher than the methods without using TL. In addition, the proposed model outperforms 4 common traditional classifiers (KNN, LDA, SVM, and Random Forest) using the minimal calibration data (two repeated trials for each gesture). The results also demonstrate the model has a great transfer robustness/flexibility for cross-gesture and cross-day scenarios, with an accuracy of 87.94% achieved using calibration gestures that are different with model training, and an accuracy of 84.26% achieved using calibration data collected on a different day, respectively. As the outcomes confirm, the proposed CNN TL method provides a practical solution for freeing new users from the complicated acquisition paradigm in the calibration process before using sEMG-based HMI systems.

A Robust and Real-Time Framework of Cross-Subject Myoelectric Control Model Calibration via Multi-Source Domain Adaptation.

Surface electromyogram (sEMG) has been widely used in hand gesture recognition. However, most previous studies focused on user-personalized models, which require a great amount of data from each new target user to learn the user-specific EMG patterns. In this work, we present a novel real-time gesture recognition framework based on multi-source domain adaptation, which learns extra knowledge from the data of other users, thereby reducing the data collection burdens on the target user. Additionally, compared with conventional domain adaptation methods which treat data from all users in the source domain as a whole, the proposed multi-source method treat data from different users as multiple separate source domains. Therefore, more detailed statistical information on the data distribution from each user can be learned effectively. High-density sEMG (256 channels) from 20 subjects was used to validate the proposed method. Importantly, we evaluated our method with a simulated real-time processing pipeline on continuous sEMG data stream, rather than well-segmented data. The false alarm rate during rest periods in an EMG data stream, which is typically neglected by previous studies performing offline analyses, was also considered. Our results showed that, with only 1 s sEMG data per gesture from the new user, the 10-gesture classification accuracy reached 87.66 % but the false alarm rate was reduced to 1.95 %. Our method can reduce the frustratingly heavy data collection burdens on each new user.

Evaluation of generic EMG-Torque models across two Upper-Limb joints.

Advanced single-use dynamic EMG-torque models require burdensome subject-specific calibration contractions and have historically been assumed to produce lower error than generic models (i.e., models that are identical across subjects and muscles). To investigate this assumption, we studied generic one degree of freedom (DoF) models derived from the ensemble median of subject-specific models, evaluated across subject, DoF and joint. We used elbow (N = 64) and hand-wrist (N = 9) datasets. Subject-specific elbow models performed statistically better [5.79 ± 1.89 %MVT (maximum voluntary torque) error] than generic elbow models (6.21 ± 1.85 %MVT error). However, there were no statistical differences between subject-specific vs. generic models within each hand-wrist DoF. Next, we evaluated generic models across joints. The best hand-wrist generic model had errors of 6.29 ± 1.85 %MVT when applied to the elbow. The elbow generic model had errors of 7.04 ± 2.29 %MVT when applied to the hand-wrist. The generic elbow model was statistically better in both joints, compared to the generic hand-wrist model. Finally, we tested Butterworth filter models (a simpler generic model), finding no statistical differences between optimum Butterworth and subject-specific models. Overall, generic models simplified EMG-torque training without substantive performance degradation and provided the possibility of transfer learning between joints.

sEMG-Based Inter-Session Hand Gesture Recognition via Domain Adaptation with Locality Preserving and Maximum Margin.

Surface electromyography (sEMG)-based gesture recognition can achieve high intra-session performance. However, the inter-session performance of gesture recognition decreases sharply due to the shift in data distribution. Therefore, developing a robust model to minimize the data distribution difference is crucial to improving the user experience. In this work, based on the inter-session gesture recognition task, we propose a novel algorithm called locality preserving and maximum margin criterion (LPMM). The LPMM algorithm integrates three main modules, including domain alignment, pseudo-label selection, and iteration result selection. Domain alignment is designed to preserve the neighborhood structure of the feature and minimize the overlap of different classes. The pseudo-label selection and iteration result selection can avoid the decrease in accuracy caused by mislabeled samples. The proposed algorithm was evaluated on two of the most widely used EMG databases. It achieves a mean accuracy of 98.46% and 71.64%, respectively, which is superior to state-of-the-art domain adaptation methods.

Functional force stimulation alters motor neuron discharge patterns.

Introduction: Beneficial effects have been observed for mechanical vibration stimulation (MVS), which are mainly attributed to tonic vibration reflex (TVR). TVR is reported to elicit synchronized motor unit activation during locally applied vibration. Similar effects are also observed in a novel vibration system referred to as functional force stimulation (FFS). However, the manifestation of TVR in FFS is doubted due to the use of global electromyography (EMG) features in previous analysis. Our study aims to investigate the effects of FFS on motor unit discharge patterns of the human biceps brachii by analyzing the motor unit spike trains decoded from the high-density surface EMG. Methods: Eighteen healthy subjects volunteered in FFS training with different amplitudes and frequencies. One hundred and twenty-eight channel surface EMG was recorded from the biceps brachii and then decoded after motion-artifact removal. The discharge timings were extracted and the coherence between different motor unit spike trains was calculated to quantify synchronized activation. Results and discussion: Significant synchronization within the vibration cycle and/or its integer multiples is observed for all FFS trials, which increases with increased FFS amplitude. Our results reveal the basic physiological mechanism involved in FFS, providing a theoretical foundation for analyzing and introducing FFS into clinical rehabilitation programs.

A Review on the Usability, Flexibility, Affinity, and Affordability of Virtual Technology for Rehabilitation Training of Upper Limb Amputees.

(1) Background: Prosthetic rehabilitation is essential for upper limb amputees to regain their ability to work. However, the abandonment rate of prosthetics is higher than 50% due to the high cost of rehabilitation. Virtual technology shows potential for improving the availability and cost-effectiveness of prosthetic rehabilitation. This article systematically reviews the application of virtual technology for the prosthetic rehabilitation of upper limb amputees. (2) Methods: We followed PRISMA review guidance, STROBE, and CASP to evaluate the included articles. Finally, 17 articles were screened from 22,609 articles. (3) Results: This study reviews the possible benefits of using virtual technology from four aspects: usability, flexibility, psychological affinity, and long-term affordability. Three significant challenges are also discussed: realism, closed-loop control, and multi-modality integration. (4) Conclusions: Virtual technology allows for flexible and configurable control rehabilitation, both during hospital admissions and after discharge, at a relatively low cost. The technology shows promise in addressing the critical barrier of current prosthetic training issues, potentially improving the practical availability of prosthesis techniques for upper limb amputees.

Cybersecurity in neural interfaces: Survey and future trends.

With the joint advancement in areas such as pervasive neural data sensing, neural computing, neuromodulation and artificial intelligence, neural interface has become a promising technology facilitating both the closed-loop neurorehabilitation for neurologically impaired patients and the intelligent man-machine interactions for general application purposes. However, although neural interface has been widely studied, few previous studies focused on the cybersecurity issues in related applications. In this survey, we systematically investigated possible cybersecurity risks in neural interfaces, together with potential solutions to these problems. Importantly, our survey considers interfacing techniques on both central nervous systems (i.e., brain-computer interfaces) and peripheral nervous systems (i.e., general neural interfaces), covering diverse neural modalities such as electroencephalography, electromyography and more. Moreover, our survey is organized on three different levels: (1) the data level, which mainly focuses on the privacy leakage issue via attacking and analyzing neural database of users; (2) the permission level, which mainly focuses on the prospects and risks to directly use real time neural signals as biometrics for continuous and unobtrusive user identity verification; and (3) the model level, which mainly focuses on adversarial attacks and defenses on both the forward neural decoding models (e.g. via machine learning) and the backward feedback implementation models (e.g. via neuromodulation and stimulation). This is the first study to systematically investigate cybersecurity risks and possible solutions in neural interfaces which covers both central and peripheral nervous systems, and considers multiple different levels to provide a complete picture of this issue.

Physiological measurements for driving drowsiness: A comparative study of multi-modality feature fusion and selection.

A large number of traffic accidents were caused by drowsiness while driving. In-vehicle alert system based on physiological signals was one of the most promising solutions to monitor driving fatigue. However, different physiological modalities can be used, and many relative studies compared different modalities without considering the implementation feasibility of portable or wearable devices. Moreover, evaluations of each modality in previous studies were based on inconsistent choices of fatigue label and signal features, making it hard to compare the results of different studies. Therefore, the modality comparison and fusion for continuous drowsiness estimation while driving was still unclear. This work sought to comprehensively compare widely-used physiological modalities, including forehead electroencephalogram (EEG), electrooculogram (EOG), R-R intervals (RRI) and breath, in a hardware setting feasible for portable or wearable devices to monitor driving fatigue. Moreover, a more general conclusion on modality comparison and fusion was reached based on the regression of features or their combinations and the awake-to-drowsy transition. Finally, the feature subset of fused modalities was produced by feature selection method, to select the optimal feature combination and reduce computation consumption. Considering practical feasibility, the most effective combination with the highest correlation coefficient was using forehead EEG or EOG, along with RRI and RRI-derived breath. If more comfort and convenience was required, the combination of RRI and RRI-derived breath was also promising.

A Transfer Learning based Cross-subject Generic Model for Continuous Estimation of Finger Joint Angles from a New User.

Continuous estimation of finger joints based on surface electromyography (sEMG) has attracted much attention in the field of human-machine interface (HMI). A couple of deep learning models were proposed to estimate the finger joint angles for specific subject. When applied onto a new subject, however, the performance of the subject-specific model would degrade significantly due to the inter-subject differences. Therefore, a novel cross-subject generic (CSG) model was proposed in this study to estimate continuous kinematics of finger joints for new users. Firstly, a multi-subject model based on the LSTA-Conv network was built by using sEMG and finger joint angles data from multiple subjects. Then, the subjects adversarial knowledge (SAK) transfer learning strategy was adopted to calibrate the multi-subject model with the training data from a new user. With the updated model parameters and the testing data from the new user, multiple finger joint angles could be estimated afterwards. The overall performance of the CSG model for new users was validated on three public datasets from Ninapro. The results showed that the newly proposed CSG model significantly outperformed five subject-specific models and two transfer learning models in terms of Pearson correlation coefficient, root mean square error, and coefficient of determination. Comparison analysis showed that both the long short-term feature aggregation (LSTA) module and the SAK transfer learning strategy contributed to the CSG model. Moreover, increasing number of subjects in training set improved the generalization capability of the CSG model. The novel CSG model would facilitate the application of robotic hand control and other HMI settings.

Explainable and Robust Deep Forests for EMG-Force Modeling.

Machine and deep learning techniques have received increasing attentions in estimating finger forces from high-density surface electromyography (HDsEMG), especially for neural interfacing. However, most machine learning models are normally employed as block-box modules. Additionally, most previous models suffer from performance degradation when dealing with noisy signals. In this work, we propose to employ a forest ensemble model for HDsEMG-force modeling. Our model is explainable and robust against noise. Additionally, we explored the effect of increasing the depth of forest models in EMG-force modeling problems. We evaluated the performance of deep forests with a finger force estimation task. Training and testing data were acquired 3-25 days apart, approximating realistic scenarios. Results showed that deep forests significantly outperformed other models. With artificial signal distortion in 20% channels, deep forests also showed a higher robustness, with the error reduced from that of the baseline by 50% compared with all other models. We provided explanations for the proposed model using the mean decrease impurity (MDI) metric, revealing a strong correspondence between the model and physiology.

Virtual regression-based myoelectric hand-wrist prosthesis control and electrode site selection using no force feedback.

Most transradial prosthesis users with conventional "Sequential" myoelectric control have two electrode sites which control one degree of freedom (DoF) at a time. Rapid EMG co-activation toggles control between DoFs (e.g., hand and wrist), providing limited function. We implemented a regression-based EMG control method which achieved simultaneous and proportional control of two DoFs in a virtual task. We automated electrode site selection using a short-duration (90 s) calibration period, without force feedback. Backward stepwise selection located the best electrodes for either six or 12 electrodes (selected from a pool of 16). We additionally studied two, 2-DoF controllers: "Intuitive" control (hand open-close and wrist pronation-supination controlled virtual target size and rotation, respectively) and "Mapping" control (wrist flexion-extension and ulnar-radial deviation controlled virtual target left-right and up-down movement, respectively). In practice, a Mapping controller would be mapped to control prosthesis hand open-close and wrist pronation-supination. Eleven able-bodied subjects and 4 limb-absent subjects completed virtual target matching tasks (fixed target moves to a new location after being "matched," and subject immediately pursues) and fixed (static) target tasks. For all subjects, both 2-DoF controllers with 6 optimally-sited electrodes had statistically better target matching performance than Sequential control in number of matches (average of 4-7 vs. 2 matches, p< 0.001) and throughput (average of 0.75-1.25 vs. 0.4 bits/s, p< 0.001), but not overshoot rate and path efficiency. There were no statistical differences between 6 and 12 optimally-sited electrodes for both 2-DoF controllers. These results support the feasibility of 2-DoF simultaneous, proportional myoelectric control.

Sleep postures monitoring based on capacitively coupled electrodes and deep recurrent neural networks.

BACKGROUND: Capacitively coupled electrode (CC electrode), as a non-contact and unobtrusive technology for measuring physiological signals, has been widely applied in sleep monitoring scenarios. The most common implementation is capacitive electrocardiogram (cECG) that could provide useful clinical information for assessing cardiac function and detecting cardiovascular diseases. In the current study, we sought to explore another potential application of cECG in sleep monitoring, i.e., sleep postures recognition. METHODS: Two sets of experiments, the short-term experiment, and the overnight experiment, were conducted. The cECG signals were measured by a smart mattress based on flexible CC electrodes and sleep postures were recorded simultaneously. Then, a classifier model based on a deep recurrent neural network (RNN) was proposed to distinguish sleep postures (supine, left lateral and right lateral). To verify the reliability of the proposed model, leave-one-subject-out cross-validation was introduced. RESULTS: In the short-term experiment, the overall accuracy of 96.2% was achieved based on 30-s segment, while the overall accuracy was 88.8% using one heart beat segment. For the unconstrained overnight experiment, the accuracy of 91.0% was achieved based on 30-s segment, while the accuracy was 81.4% using one heart beat segment. CONCLUSIONS: The results suggest that cECG could render valuable information about sleep postures detection and potentially be helpful for sleep disorder diagnosis.

Automatic Upper-Limb Brunnstrom Recovery Stage Evaluation via Daily Activity Monitoring.

Motor function assessment is crucial for post-stroke rehabilitation. Conventional evaluation methods are subjective, heavily depending on the experience of therapists. In light of the strong correlation between the stroke severity level and the performance of activities of daily living (ADLs), we explored the possibility of automatically evaluating the upper-limb Brunnstrom Recovery Stage (BRS) via three typical ADLs (tooth brushing, face washing and drinking). Multimodal data (acceleration, angular velocity, surface electromyography) were synchronously collected from 5 upper-limb-worn sensor modules. The performance of BRS evaluation system is known to be variable with different system parameters (e.g., number of sensor modules, feature types and classifiers). We systematically searched for the optimal parameters from different data segmentation strategies (five window lengths and four overlaps), 42 types of features, 12 feature optimization techniques and 9 classifiers with the leave-one-subject-out cross-validation. To achieve reliable and low-cost monitoring, we further explored whether it was possible to obtain a satisfactory result using a relatively small number of sensor modules. As a result, the proposed approach can correctly recognize the stages of all 27 participants using only three sensor modules with the optimized data segmentation parameters (window length: 7s, overlap: 50%), extracted features (simple square integral, slope sign change, modified mean absolute value 1 and modified mean absolute value 2), the feature optimization method (principal component analysis) and the logistic regression classifier. According to the literature, this is the first study to comprehensively optimize sensor configuration and parameters in each stage of the BRS classification framework. The proposed approach can serve as a factor-screening tool towards the automatic BRS classification and is promising to be further used at home.

The Effects of Deep Brain Stimulation on Motor Unit Activities in Parkinson's Disease based on High-Density Surface EMG Analysis.

Tremor in Parkinson's disease (PD) is caused by synchronized activation bursts in limb muscles. Deep Brain Stimulation (DBS) is an effective clinical therapy for inhibiting tremor and improving movement disorders in PD patients. However, the neural mechanism of how tremor symptom is suppressed by DBS at motor unit (MU) level remains unclear. This paper developed a data acquisition platform for collecting physiological data in PD patients. Both high-density surface Electromyography (HD-sEMG) and kinematics data were collected concurrently before and after DBS surgery. The MU behaviors were obtained via HD-sEMG decomposition algorithm to reveal the effect of DBS on PD tremor. A data set of one tremor dominant PD patient acquired in pre-operation and post-operation (DBS-on) phases was analyzed. Preliminary results showed significant changes in MU firing rate and MU synchronization. The analysis approach introduced in this paper provides a novel perspective for studying the neural mechanism of DBS as revealed by MU activities. Clinical Relevance- This study presented an approach to investigate the effect of DBS therapy on improving tremor disorder of PD patients.

Random Channel Masks for Regularization of Least Squares-Based Finger EMG-Force Modeling to Improve Cross-Day Performance.

Estimating the finger forces from surface electromyography (sEMG) is essential for diverse applications (e.g., human-machine interfacing). The performance of pre-trained sEMG-force models degenerates significantly when applied on a second day, due to the large cross-day variation of sEMG characteristics. Previous studies mainly employed transfer learning algorithms to tackle this problem. However, transfer learning algorithms normally require data collected on the second day for model calibration, increasing the inconvenience in practical use. In this work, we investigated the effect of model regularization on this issue. Specifically, 256-channel high-density sEMG (HDsEMG) signals with varying finger forces were collected on different days (3-25 days apart). We applied randomly generated channel perturbations ("masks") to feature maps of randomly selected channels in training dataset. The channel masks of the training set were generated randomly and independently in each narrow time window (~20 ms). We assumed that by learning from randomly masked feature maps (randomness is the central aspect), the model would not be biased by a small number of features but would be based on learning from a global perspective, therefore avoiding overfitting to the within-day EMG patterns. Moore-Penrose inverse model regularization was also employed as a baseline method, with results showing that cross-day EMG-force models require a higher tolerance parameter compared with within-day applications. In combination with the Moore-Penrose inverse model regularization, further applying random channel masks to the training set significantly improved model performance in cross-day validation.

fNIRS-Based Dynamic Functional Connectivity Reveals the Innate Musical Sensing Brain Networks in Preterm Infants.

Humans have the ability to appreciate and create music. However, why and how humans have this distinctive ability to perceive music remains unclear. Additionally, the investigation of the innate perceiving skill in humans is compounded by the fact that we have been actively and passively exposed to auditory stimuli or have systematically learnt music after birth. Therefore, to explore the innate musical perceiving ability, infants with preterm birth may be the most suitable population. In this study, the auditory brain networks were explored using dynamic functional connectivity-based reliable component analysis (RCA) in preterm infants during music listening. The brain activation was captured by portable functional near-infrared spectroscopy (fNIRS) to simulate a natural environment for preterm infants. The components with the maximum inter-subject correlation were extracted. The generated spatial filters identified the shared spatial structural features of functional brain connectivity across subjects during listening to the common music, exhibiting a functional synchronization between the right temporal region and the frontal and motor cortex, and synchronization between the bilateral temporal regions. The specific pattern is responsible for the functions involving music comprehension, emotion generation, language processing, memory, and sensory. The fluctuation of the extracted components and the phase variation demonstrates the interactions between the extracted brain networks to encode musical information. These results are critically important for our understanding of the underlying mechanisms of the innate perceiving skills at early ages of human during naturalistic music listening.

Myoelectric Control Performance of Two Degree of Freedom Hand-Wrist Prosthesis by Able-Bodied and Limb-Absent Subjects.

Recent research has advanced two degree-of-freedom (DoF), simultaneous, independent and proportional control of hand-wrist prostheses using surface electromyogram signals from remnant muscles as the control input. We evaluated two such regression-based controllers, along with conventional, sequential two-site control with co-contraction mode switching (SeqCon), in box-block, refined-clothespin and door-knob tasks, on 10 able-bodied and 4 limb-absent subjects. Subjects operated a commercial hand and wrist using a socket bypass harness. One 2-DoF controller (DirCon) related the intuitive hand actions of open-close and pronation-supination to the associated prosthesis hand-wrist actions, respectively. The other (MapCon) mapped myoelectrically more distinct, but less intuitive, actions of wrist flexion-extension and ulnar-radial deviation. Each 2-DoF controller was calibrated from separate 90 s calibration contractions. SeqCon performed better statistically than MapCon in the predominantly 1-DoF box-block task (>20 blocks/minute vs. 8-18 blocks/minute, on average). In this task, SeqCon likely benefited from an ability to easily focus on 1-DoF and not inadvertently trigger co-contraction for mode switching. The remaining two tasks require 2-DoFs, and both 2-DoF controllers each performed better (factor of 2-4) than SeqCon. We also compared the use of 12 vs. 6 optimally-selected EMG electrodes as inputs, finding no statistical difference. Overall, we provide further evidence of the benefits of regression-based EMG prosthesis control of 2-DoFs in the hand-wrist.

Epileptic Seizure Detection by Cascading Isolation Forest-Based Anomaly Screening and EasyEnsemble.

The electroencephalogram (EEG), for measuring the electrophysiological activity of the brain, has been widely applied in automatic detection of epilepsy seizures. Various EEG-based seizure detection algorithms have already yielded high sensitivity, but training those algorithms requires a large amount of labelled data. Data labelling is often done with a lot of human efforts, which is very time-consuming. In this study, we propose a hybrid system integrating an unsupervised learning (UL) module and a supervised learning (SL) module, where the UL module can significantly reduce the workload of data labelling. For preliminary seizure screening, UL synthesizes amplitude-integrated EEG (aEEG) extraction, isolation forest-based anomaly detection, adaptive segmentation, and silhouette coefficient-based anomaly detection evaluation. The UL module serves to quickly locate the determinate subjects (seizure segments and seizure-free segments) and the indeterminate subjects (potential seizure candidates). Afterwards, more robust seizure detection for the indeterminate subjects is performed by the SL using an EasyEnsemble algorithm. EasyEnsemble, as a class-imbalance learning method, can potentially decrease the generalization error of the seizure-free segments. The proposed method can significantly reduce the workload of data labelling while guaranteeing satisfactory performance. The proposed seizure detection system is evaluated using the Children's Hospital Boston-Massachusetts Institute of Technology (CHB-MIT) scalp EEG dataset, and it achieves a mean accuracy of 92.62%, a mean sensitivity of 95.55%, and a mean specificity of 92.57%. To the best of our knowledge, this is the first epilepsy seizure detection study employing the integration of both the UL and the SL modules, achieving a competitive performance superior or similar to that of the state-of-the-art methods.