Tracking of Mental Workload with a Mobile EEG Sensor.

The aim of the present investigation was to assess if a mobile electroencephalography (EEG) setup can be used to track mental workload, which is an important aspect of learning performance and motivation and may thus represent a valuable source of information in the evaluation of cognitive training approaches. Twenty five healthy subjects performed a three-level N-back test using a fully mobile setup including tablet-based presentation of the task and EEG data collection with a self-mounted mobile EEG device at two assessment time points. A two-fold analysis approach was chosen including a standard analysis of variance and an artificial neural network to distinguish the levels of cognitive load. Our findings indicate that the setup is feasible for detecting changes in cognitive load, as reflected by alterations across lobes in different frequency bands. In particular, we observed a decrease of occipital alpha and an increase in frontal, parietal and occipital theta with increasing cognitive load. The most distinct levels of cognitive load could be discriminated by the integrated machine learning models with an accuracy of 86%.

Can commercially available wearable EEG devices be used for diagnostic purposes? An explorative pilot study.

Electroencephalography (EEG) is a core element in the diagnosis of epilepsy syndromes and can help to monitor antiseizure treatment. Mobile EEG (mEEG) devices are increasingly available on the consumer market and may offer easier access to EEG recordings especially in rural or resource-poor areas. The usefulness of consumer-grade devices for clinical purposes is still underinvestigated. Here, we compared EEG traces of a commercially available mEEG device (Emotiv EPOC) to a simultaneously recorded clinical video EEG (vEEG). Twenty-two adult patients (11 female, mean age 40.2 years) undergoing noninvasive vEEG monitoring for clinical purposes were prospectively enrolled. The EEG recordings were evaluated by 10 independent raters with unmodifiable view settings. The individual evaluations were compared with respect to the presence of abnormal EEG findings (regional slowing, epileptiform potentials, seizure pattern). Video EEG yielded a sensitivity of 56% and specificity of 88% for abnormal EEG findings, whereas mEEG reached 39% and 85%, respectively. Interrater reliability coefficients were better in vEEG as compared to mEEG (ϰ = 0.50 vs. 0.30), corresponding to a moderate and fair agreement. Intrarater reliability between mEEG and vEEG evaluations of simultaneous recordings of a given participant was moderate (ϰ = 0.48). Given the limitations of our exploratory pilot study, our results suggest that vEEG is superior to mEEG, but that mEEG can be helpful for diagnostic purposes. We present the first quantitative comparison of simultaneously acquired clinical and mobile consumer-grade EEG for a clinical use-case.

Recursive neural networks in hospital bed occupancy forecasting.

BACKGROUND: Efficient planning of hospital bed usage is a necessary condition to minimize the hospital costs. In the presented work we deal with the problem of occupancy forecasting in the scale of several months, with a focus on personnel's holiday planning. METHODS: We construct a model based on a set of recursive neural networks, which performs an occupancy prediction using historical admission and release data combined with external factors such as public and school holidays. The model requires no personal information on patients or staff. It is optimized for a 60 days forecast during the summer season (May-September). RESULTS: An average mean absolute percentage error (MAPE) of 6.24% was computed on 8 validation sets. CONCLUSIONS: The proposed machine learning model has shown to be competitive to standard time-series forecasting models and can be recommended for incorporation in medium-size hospitals automatized scheduling and decision making.

Correction: Kutafina, E.; Laukamp, D.; Bettermann, R.; Schroeder, U.; Jonas, S.M. Wearable Sensors for eLearning of Manual Tasks: Using Forearm EMG in Hand Hygiene Training. Sensors 2016, 16, 1221.

The authors wish to make the following erratum to this paper [...].

Wearable Sensors for eLearning of Manual Tasks: Using Forearm EMG in Hand Hygiene Training.

In this paper, we propose a novel approach to eLearning that makes use of smart wearable sensors. Traditional eLearning supports the remote and mobile learning of mostly theoretical knowledge. Here we discuss the possibilities of eLearning to support the training of manual skills. We employ forearm armbands with inertial measurement units and surface electromyography sensors to detect and analyse the user's hand motions and evaluate their performance. Hand hygiene is chosen as the example activity, as it is a highly standardized manual task that is often not properly executed. The World Health Organization guidelines on hand hygiene are taken as a model of the optimal hygiene procedure, due to their algorithmic structure. Gesture recognition procedures based on artificial neural networks and hidden Markov modeling were developed, achieving recognition rates of 98 . 30 % ( ± 1 . 26 % ) for individual gestures. Our approach is shown to be promising for further research and application in the mobile eLearning of manual skills.

Measuring pain and nociception: Through the glasses of a computational scientist. Transdisciplinary overview of methods.

In a healthy state, pain plays an important role in natural biofeedback loops and helps to detect and prevent potentially harmful stimuli and situations. However, pain can become chronic and as such a pathological condition, losing its informative and adaptive function. Efficient pain treatment remains a largely unmet clinical need. One promising route to improve the characterization of pain, and with that the potential for more effective pain therapies, is the integration of different data modalities through cutting edge computational methods. Using these methods, multiscale, complex, and network models of pain signaling can be created and utilized for the benefit of patients. Such models require collaborative work of experts from different research domains such as medicine, biology, physiology, psychology as well as mathematics and data science. Efficient work of collaborative teams requires developing of a common language and common level of understanding as a prerequisite. One of ways to meet this need is to provide easy to comprehend overviews of certain topics within the pain research domain. Here, we propose such an overview on the topic of pain assessment in humans for computational researchers. Quantifications related to pain are necessary for building computational models. However, as defined by the International Association of the Study of Pain (IASP), pain is a sensory and emotional experience and thus, it cannot be measured and quantified objectively. This results in a need for clear distinctions between nociception, pain and correlates of pain. Therefore, here we review methods to assess pain as a percept and nociception as a biological basis for this percept in humans, with the goal of creating a roadmap of modelling options.

odML-Tables as a Metadata Standard in Microneurography.

Metadata standards are well-established for many types of electrophysiological methods but are still lacking for microneurographic recordings of peripheral sensory nerve fibers in humans. Finding a solution for daily work in the laboratory is a complex process. We have designed templates based on odML and odML-tables to structure and capture metadata and provided an extension to the existing GUI to enable database searching.

Corrigendum: Decoding neuropathic pain: Can we predict fluctuations of propagation speed in stimulated peripheral nerve?

[This corrects the article DOI: 10.3389/fncom.2022.899584.].

PyDapsys: an open-source library for accessing electrophysiology data recorded with DAPSYS.

In the field of neuroscience, a considerable number of commercial data acquisition and processing solutions rely on proprietary formats for data storage. This often leads to data being locked up in formats that are only accessible by using the original software, which may lead to interoperability problems. In fact, even the loss of data access is possible if the software becomes unsupported, changed, or otherwise unavailable. To ensure FAIR data management, strategies should be established to enable long-term, independent, and unified access to data in proprietary formats. In this work, we demonstrate PyDapsys, a solution to gain open access to data that was acquired using the proprietary recording system DAPSYS. PyDapsys enables us to open the recorded files directly in Python and saves them as NIX files, commonly used for open research in the electrophysiology domain. Thus, PyDapsys secures efficient and open access to existing and prospective data. The manuscript demonstrates the complete process of reverse engineering a proprietary electrophysiological format on the example of microneurography data collected for studies on pain and itch signaling in peripheral neural fibers.

odML-Tables as a Metadata Standard in Microneurography.

Metadata is essential for handling medical data according to FAIR principles. Standards are well-established for many types of electrophysiological methods but are still lacking for microneurographic recordings of peripheral sensory nerve fibers in humans. Developing a new concept to enhance laboratory workflows is a complex process. We propose a standard for structuring and storing microneurography metadata based on odML and odML-tables. Further, we present an extension to the odML-tables GUI that enables user-friendly search functionality of the database. With our open-source repository, we encourage other microneurography labs to incorporate odML-based metadata into their experimental routines.

A modelling study to dissect the potential role of voltage-gated ion channels in activity-dependent conduction velocity changes as identified in small fiber neuropathy patients.

Objective: Patients with small fiber neuropathy (SFN) suffer from neuropathic pain, which is still a therapeutic problem. Changed activation patterns of mechano-insensitive peripheral nerve fibers (CMi) could cause neuropathic pain. However, there is sparse knowledge about mechanisms leading to CMi dysfunction since it is difficult to dissect specific molecular mechanisms in humans. We used an in-silico model to elucidate molecular causes of CMi dysfunction as observed in single nerve fiber recordings (microneurography) of SFN patients. Approach: We analyzed microneurography data from 97 CMi-fibers from healthy individuals and 34 of SFN patients to identify activity-dependent changes in conduction velocity. Using the NEURON environment, we adapted a biophysical realistic preexisting CMi-fiber model with ion channels described by Hodgkin-Huxley dynamics for identifying molecular mechanisms leading to those changes. Via a grid search optimization, we assessed the interplay between different ion channels, Na-K-pump, and resting membrane potential. Main results: Changing a single ion channel conductance, Na-K-pump or membrane potential individually is not sufficient to reproduce in-silico CMi-fiber dysfunction of unchanged activity-dependent conduction velocity slowing and quicker normalization of conduction velocity after stimulation as observed in microneurography. We identified the best combination of mechanisms: increased conductance of potassium delayed-rectifier and decreased conductance of Na-K-pump and depolarized membrane potential. When the membrane potential is unchanged, opposite changes in Na-K-pump and ion channels generate the same effect. Significance: Our study suggests that not one single mechanism accounts for pain-relevant changes in CMi-fibers, but a combination of mechanisms. A depolarized membrane potential, as previously observed in patients with neuropathic pain, leads to changes in the contribution of ion channels and the Na-K-pump. Thus, when searching for targets for the treatment of neuropathic pain, combinations of several molecules in interplay with the membrane potential should be regarded.

Spectral Fusion of Heartbeat and Accelerometer Data for Estimation of Breathing Rate in Wearable Patches.

Despite developments in wearable devices for detecting various bio-signals, continuous measurement of breathing rate (BR) remains a challenge. This work presents an early proof of concept that employs a wearable patch to estimate BR. We propose combining techniques for calculating BR from electrocardiogram (ECG) and accelerometer (ACC) signals, while applying decision rules based on signal-to-noise (SNR) to fuse the estimates for improved accuracy.

ECG Matching: An Approach to Synchronize ECG Datasets for Data Quality Comparisons.

Clinical assessment of newly developed sensors is important for ensuring their validity. Comparing recordings of emerging electrocardiography (ECG) systems to a reference ECG system requires accurate synchronization of data from both devices. Current methods can be inefficient and prone to errors. To address this issue, three algorithms are presented to synchronize two ECG time series from different recording systems: Binned R-peak Correlation, R-R Interval Correlation, and Average R-peak Distance. These algorithms reduce ECG data to their cyclic features, mitigating inefficiencies and minimizing discrepancies between different recording systems. We evaluate the performance of these algorithms using high-quality data and then assess their robustness after manipulating the R-peaks. Our results show that R-R Interval Correlation was the most efficient, whereas the Average R-peak Distance and Binned R-peak Correlation were more robust against noisy data.

EEG Spectral Changes Linked to Psychiatric Medications: Computational Pipeline for Data Mining and Analysis.

The presented computational pipeline is designed to analyze drug-induced changes in EEG data from the Temple University EEG Corpus. The data is cleaned from artifacts, pre-processed, the averaged absolute and relative frequency powers are calculated and compared to a control group. Thus, different research hypotheses can be tested with the intention to reuse accessible data collections.

Decoding Neuropathic Pain: Can We Predict Fluctuations of Propagation Speed in Stimulated Peripheral Nerve?

To understand neural encoding of neuropathic pain, evoked and resting activity of peripheral human C-fibers are studied via microneurography experiments. Before different spiking patterns can be analyzed, spike sorting is necessary to distinguish the activity of particular fibers of a recorded bundle. Due to single-electrode measurements and high noise contamination, standard methods based on spike shapes are insufficient and need to be enhanced with additional information. Such information can be derived from the activity-dependent slowing of the fiber propagation speed, which in turn can be assessed by introducing continuous "background" 0.125-0.25 Hz electrical stimulation and recording the corresponding responses from the fibers. Each fiber's speed propagation remains almost constant in the absence of spontaneous firing or additional stimulation. This way, the responses to the "background stimulation" can be sorted by fiber. In this article, we model the changes in the propagation speed resulting from the history of fiber activity with polynomial regression. This is done to assess the feasibility of using the developed models to enhance the spike shape-based sorting. In addition to human microneurography data, we use animal in-vitro recordings with a similar stimulation protocol as higher signal-to-noise ratio data example for the models.

Consistency of Feature Importance Algorithms for Interpretable EEG Abnormality Detection.

Recent advances in machine learning show great potential for automatic detection of abnormalities in electroencephalography (EEG). While simple and interpretable models combined with expert-comprehensible input features offer full control of the decision making process, these methods commonly lag behind complex deep learning and feature extraction methods in terms of performance. Here we study a feasibility of a bridging solution, where deep learning is combined with interpretable input and an algorithm computing the importance of particular EEG features in the decision process. We built a convolutional neural network with multi-channel EEG frequency bands as input and investigated four different methods for feature importance attribution: Layer-wise Relevance Propagation (LRP), DeepLIFT, Integrated Gradients (IG) and Guided GradCAM. Our analysis showed consistency between the first three methods, and deviating attributions of the fourth method, suggesting the importance of using a package of methods together to ensure the robustness of medical interpretation.

Towards Interpretable Machine Learning in EEG Analysis.

In this paper a machine learning model for automatic detection of abnormalities in electroencephalography (EEG) is dissected into parts, so that the influence of each part on the classification accuracy score can be examined. The most successful setup of several shallow artificial neural networks aggregated via voting results in accuracy of 81%. Stepwise simplification of the model shows the expected decrease in accuracy, but a naive model with thresholding of a single extracted feature (relative wavelet energy) is still able to achieve 75%, which remains strongly above the random guess baseline of 54%. These results suggest the feasibility of building a simple classification model ensuring accuracy scores close to the state-of-the-art research but remaining fully interpretable.

Modeling of Activity-Induced Changes in Signal Propagation Speed of Mechano-Electrically Stimulated Nerve Fiber.

One of the important questions in the research on neural coding is how the preceding axonal activity affects the signal propagation speed of the following one. We present an approach to solving this problem by introducing a multi-level spike count for activity quantification and fitting a family of linear regression models to the data. The best-achieved score is R2=0.89 and the comparison of different models indicates the importance of long and very short nerve fiber memory. Further studies are required to understand the complex axonal mechanisms responsible for the discovered phenomena.

openMNGlab: Data Analysis Framework for Microneurography - A Technical Report.

openMNGlab is an open-source software framework for data analysis, tailored for the specific needs of microneurography - a type of electrophysiological technique particularly important for research on peripheral neural fibers coding. Currently, openMNGlab loads data from Spike2 and Dapsys, which are two major data acquisition solutions. By building on top of the Neo software, openMNGlab can be easily extended to handle the most common electrophysiological data formats. Furthermore, it provides methods for data visualization, fiber tracking, and a modular feature database to extract features for data analysis and machine learning.

Comparison of mobile and clinical EEG sensors through resting state simultaneous data collection.

Development of mobile sensors brings new opportunities to medical research. In particular, mobile electroencephalography (EEG) devices can be potentially used in low cost screening for epilepsy and other neurological and psychiatric disorders. The necessary condition for such applications is thoughtful validation in the specific medical context. As part of validation and quality assurance, we developed a computer-based analysis pipeline, which aims to compare the EEG signal acquired by a mobile EEG device to the one collected by a medically approved clinical-grade EEG device. Both signals are recorded simultaneously during 30 min long sessions in resting state. The data are collected from 22 patients with epileptiform abnormalities in EEG. In order to compare two multichannel EEG signals with differently placed references and electrodes, a novel data processing pipeline is proposed. It allows deriving matching pairs of time series which are suitable for similarity assessment through Pearson correlation. The average correlation of 0.64 is achieved on a test dataset, which can be considered a promising result, taking the positions shift due to the simultaneous electrode placement into account.