Search

Introduction: Modeling multi-channel electroencephalographic (EEG) time-series is a challenging tasks, even for the most recent deep learning approaches. Particularly, in this work, we targeted our efforts to the high-fidelity reconstruction of this type of data, as this is of key relevance for several applications such as classification, anomaly detection, automatic labeling, and brain-computer interfaces. Methods: We analyzed the most recent works finding that high-fidelity reconstruction is seriously challenged by the complex dynamics of the EEG signals and the large inter-subject variability. So far, previous works provided good results in either high-fidelity reconstruction of single-channel signals, or poor-quality reconstruction of multi-channel datasets. Therefore, in this paper, we present a novel deep learning model, called hvEEGNet, designed as a hierarchical variational autoencoder and trained with a new loss function. We tested it on the benchmark Dataset 2a (including 22-channel EEG data from 9 subjects). Results: We show that it is able to reconstruct all EEG channels with high-fidelity, fastly (in a few tens of epochs), and with high consistency across different subjects. We also investigated the relationship between reconstruction fidelity and the training duration and, using hvEEGNet as an anomaly detector, we spotted some data in the benchmark dataset that are corrupted and never highlighted before. Discussion: Thus, hvEEGNet could be very useful in several applications where automatic labeling of large EEG dataset is needed and time-consuming. At the same time, this work opens new fundamental research questions about (1) the effectiveness of deep learning models training (for EEG data) and (2) the need for a systematic characterization of the input EEG data to ensure robust modeling. [ABSTRACT FROM AUTHOR]

Electroencephalography (EEG) is a medical engineering technique aimed at recording the electric activity of the human brain. Brain signals derived from an EEG device can be processed and analyzed through computers by using digital signal processing, computational statistics, and machine learning techniques, that can lead to scientifically-relevant results and outcomes about how the brain works. In the last decades, the spread of EEG devices and the higher availability of EEG data, of computational resources, and of software packages for electroencephalography analysis has made EEG signal processing easier and faster to perform for any researcher worldwide. This increased ease to carry out computational analyses of EEG data, however, has made it easier to make mistakes, as well. And these mistakes, if unnoticed or treated wrongly, can in turn lead to wrong results or misleading outcomes, with worrisome consequences for patients and for the advancements of the knowledge about human brain. To tackle this problem, we present here our ten quick tips to perform electroencephalography signal processing analyses avoiding common mistakes: a short list of guidelines designed for beginners on what to do, how to do it, and what not to do when analyzing EEG data with a computer. We believe that following our quick recommendations can lead to better, more reliable and more robust results and outcome in clinical neuroscientific research. [ABSTRACT FROM AUTHOR]

Recent technological advancements such as the Internet of Things (IoT) and machine learning (ML) can lead to a massive data generation in smart environments, where multiple sensors can be used to monitor a large number of processes through a wireless sensor network (WSN). This poses new challenges for the extraction and interpretation of meaningful data. In this spirit, age of information (AoI) represents an important metric to quantify the freshness of the data monitored to check for anomalies and operate adaptive control. However, AoI typically assumes a binary representation of the information, which is actually multi-structured. Thus, deep semantic aspects may be lost. In addition, the ambient correlation of multiple sensors may not be taken into account and exploited. To analyze these issues, we study how correlation affects AoI for multiple sensors under two scenarios of (i) concurrent and (ii) time-division multiple access. We show that correlation among sensors improves AoI if concurrent transmissions are allowed, whereas the benefits are much more limited in a time-division scenario. Furthermore, we discuss how ML can be applied to extract relevant information from data and show how it can further optimize the transmission policy with savings of resources. Specifically, we demonstrate, through simulations, that ML techniques can be used to reduce the number of transmissions and that classification errors have no influence on the AoI of the system. [ABSTRACT FROM AUTHOR]

Pervasive healthcare is a promising assisted-living solution for chronic patients. However, current cutting-edge communication technologies are not able to strictly meet the requirements of these applications, especially in the case of life-threatening events. To bridge this gap, this article proposes a new architecture to support indoor healthcare monitoring, with a focus on epileptic patients. Several novel elements are introduced. The first element is the cascading of a WLAN and a cellular network, where IEEE 802.11ax is used for the wireless local area network to collect physiological and environmental data in-home and 5G-enabled Fixed Wireless Access links transfer them to a remote hospital. The second element is the extension of the network slicing concept to the WLAN, and the introduction of two new slice types to support both regular monitoring and emergency handling. Moreover, the inclusion of local computing capabilities at the WLAN router, together with a mobile edge computing resource, represents a further architectural enhancement. Local computation is required to trigger not only health-related alarms but also the network slicing change in case of emergency; in fact, proper radio resource scheduling is necessary for the cascaded networks to handle healthcare traffic together with other promiscuous everyday communication services. Numerical results demonstrate the effectiveness of the proposed approach while highlighting the performance gain achieved with respect to baseline solutions. [ABSTRACT FROM AUTHOR]

The fifth generation (5G) of cellular networks has ambitious targets of data rate, end-toend latency, and connection reliability. With the recent spreading of information and communication technology (ICT) into healthcare, leading to an Internet of medical things, new e-health services are being offered, in particular to the elderly population who needs daily assistance. This paper aims at identifying the most representative healthcare scenarios that can benefit from 5G networks and synthesizing the communications requirements in these particular scenarios. The impact of three key 5G technologies, i.e., medical network slices, mobile edge computing, and the management of heterogeneous networks, will be discussed in the representative case study of the connected ambulance. [ABSTRACT FROM AUTHOR]

In today's society, chronic diseases are a well-known issue related to the average increasing age of a population, especially in the most developed countries. The elderly, who are often the most impaired individuals, experience a significant reduction of independence in their daily life. This, consequently, affects their psychological conditions as well as their social attitudes and relationships. Therefore, industry, academia, and government health organizations are actively investigating and testing large-scale affordable solutions to improve the overall Quality of Life (QoL) in this population. [ABSTRACT FROM PUBLISHER]

In post-stroke motor rehabilitation of the upper limb, Brain-Computer Interfaces (BCIs) are becoming widespread complementary tools of the standard clinical practice to reinforce the beneficial effects of the treatments administered during the hospitalization. This kind of systems takes advantage of the so-called neuroplasticity, an exciting - but sometimes controversial - property of the brain that allows the recovery of lost functions by substituting the use of damaged neural paths with alternative ones, usually redundant. BCIs then artificially induce patients to unconsciously modify their cerebral activity to identify and reinforce - with a specific training - such new paths. Operant-conditioning is the most employed strategy for realizing this artificial learning and its effectiveness mainly depends on the reliability of the features that drive the BCI. After a brief excursus on the standard rehabilitative methods, an example of BCI application is presented with a following discussion about strong points and criticisms to cope with yet. [ABSTRACT FROM AUTHOR]

The classification of different fine hand movements from electroencephalogram (EEG) signals represents a relevant research challenge, e.g., in BCI applications for motor rehabilitation. Here, we analyzed two different datasets where fine hand movements (touch, grasp, palmar, and lateral grasp) were performed in a self-paced modality. We trained and tested a newly proposed CNN, and we compared its classification performance with two well-established machine learning models, namely, shrinkage-linear discriminant analysis (LDA) and Random Forest (RF). Compared to previous literature, we included neuroscientific evidence, and we trained our Convolutional Neural Network (CNN) model on the so-called movement-related cortical potentials (MRCPs). They are EEG amplitude modulations at low frequencies, i.e., (0.3 , 3) Hz that have been proved to encode several properties of the movements, e.g., type of grasp, force level, and speed. We showed that CNN achieved good performance in both datasets (accuracy of 0.70 ± 0.11 and 0.64 ± 0.10 , for the two datasets, respectively), and they were similar or superior to the baseline models (accuracy of 0.68 ± 0.10 and 0.62 ± 0.07 with sLDA; accuracy of 0.70 ± 0.15 and 0.61 ± 0.07 with RF, with comparable performance in precision and recall). In addition, compared to the baseline, our CNN requires a faster pre-processing procedure, paving the way for its possible use in online BCI applications. [ABSTRACT FROM AUTHOR]

Introduction: Modeling multi-channel electroencephalographic (EEG) time-series is a challenging tasks, even for the most recent deep learning approaches. Particularly, in this work, we targeted our efforts to the high-fidelity reconstruction of this type of data, as this is of key relevance for several applications such as classification, anomaly detection, automatic labeling, and brain-computer interfaces., Methods: We analyzed the most recent works finding that high-fidelity reconstruction is seriously challenged by the complex dynamics of the EEG signals and the large inter-subject variability. So far, previous works provided good results in either high-fidelity reconstruction of single-channel signals, or poor-quality reconstruction of multi-channel datasets. Therefore, in this paper, we present a novel deep learning model, called hvEEGNet, designed as a hierarchical variational autoencoder and trained with a new loss function. We tested it on the benchmark Dataset 2a (including 22-channel EEG data from 9 subjects)., Results: We show that it is able to reconstruct all EEG channels with high-fidelity, fastly (in a few tens of epochs), and with high consistency across different subjects. We also investigated the relationship between reconstruction fidelity and the training duration and, using hvEEGNet as an anomaly detector, we spotted some data in the benchmark dataset that are corrupted and never highlighted before., Discussion: Thus, hvEEGNet could be very useful in several applications where automatic labeling of large EEG dataset is needed and time-consuming. At the same time, this work opens new fundamental research questions about (1) the effectiveness of deep learning models training (for EEG data) and (2) the need for a systematic characterization of the input EEG data to ensure robust modeling., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2024 Cisotto, Zancanaro, Zoppis and Manzoni.)

Inkjet-printing is a well-known technology that has been recently revalued for the production of flexible sensors and biosensors, thank to the use of engineered nanostructured inks. In a previous work, we developed a general-purpose biosensors printing platform that made use of a simple and low-cost consumer printer and allowed to produce customized flexible electrodes with an ultra-short concept-to-prototype time, without requiring any sintering step. In this study we show the preliminary results about the use of such a newly easily-accessible, low-cost inkjet-based platform to produce flexible and fully customizable electrodes for reliable surface electromyographic (sEMG) recordings.

Wearable technology for the automatic detection of gait events has recently gained growing interest, enabling advanced analyses that were previously limited to specialist centres and equipment (e.g., instrumented walkway). In this study, we present a novel method based on dilated convolutions for an accurate detection of gait events (initial and final foot contacts) from wearable inertial sensors. A rich dataset has been used to validate the method, featuring 71 people with Parkinson's disease (PD) and 67 healthy control subjects. Multiple sensors have been considered, one located on the fifth lumbar vertebrae and two on the ankles. The aims of this study were: (i) to apply deep learning (DL) techniques on wearable sensor data for gait segmentation and quantification in older adults and in people with PD; (ii) to validate the proposed technique for measuring gait against traditional gold standard laboratory reference and a widely used algorithm based on wavelet transforms (WT); (iii) to assess the performance of DL methods in assessing high-level gait characteristics, with focus on stride, stance and swing related features. The results showed a high reliability of the proposed approach, which achieves temporal errors considerably smaller than WT, in particular for the detection of final contacts, with an inter-quartile range below 70 ms in the worst case. This study showes encouraging results, and paves the road for further research, addressing the effectiveness and the generalization of data-driven learning systems for accurate event detection in challenging conditions.

In a proof-of-principle prototypical demonstration we describe a new type of brain-machine interface (BMI) paradigm for upper limb motor-training. The proposed technique allows a fast contingent and proportionally modulated stimulation of afferent proprioceptive and motor output neural pathways using operant learning. Continuous and immediate assisted-feedback of force proportional to rolandic rhythm oscillations during actual movements was employed and illustrated with a single case experiment. One hemiplegic patient was trained for 2 weeks coupling somatosensory brain oscillations with force-field control during a robot-mediated center-out motor-task whose execution approaches movements of everyday life. The robot facilitated actual movements adding a modulated force directed to the target, thus providing a non-delayed proprioceptive feedback. Neuro-electric, kinematic, and motor-behavioral measures were recorded in pre- and post-assessments without force assistance. Patient's healthy arm was used as control since neither a placebo control was possible nor other control conditions. We observed a generalized and significant kinematic improvement in the affected arm and a spatial accuracy improvement in both arms, together with an increase and focalization of the somatosensory rhythm changes used to provide assisted-force-feedback. The interpretation of the neurophysiological and kinematic evidences reported here is strictly related to the repetition of the motor-task and the presence of the assisted-force-feedback. Results are described as systematic observations only, without firm conclusions about the effectiveness of the methodology. In this prototypical view, the design of appropriate control conditions is discussed. This study presents a novel operant-learning-based BMI-application for motor-training coupling brain oscillations and force feedback during an actual movement.

Brain-computer interface (BCI) technology has been used for rehabilitation after stroke and there are a number of reports involving stroke patients in BCI-feedback training. Most publications have demonstrated the efficacy of BCI technology in post-stroke rehabilitation using output devices such as Functional Electrical Stimulation, robot, and orthosis. The aim of this review is to focus on the progress of BCI-based rehabilitation strategies and to underline future challenges. A brief history of clinical BCI-approaches is presented focusing on stroke motor rehabilitation. A context for three approaches of a BCI-based motor rehabilitation program is outlined: the substitutive strategy, classical conditioning and operant conditioning. Furthermore, we include an overview of a pilot study concerning a new neuro-forcefeedback strategy. This pilot study involved healthy participants. Finally we address some challenges for future BCI-based rehabilitation.