Resting-state magnetoencephalography source magnitude imaging with deep-learning neural network for classification of symptomatic combat-related mild traumatic brain injury.

Combat-related mild traumatic brain injury (cmTBI) is a leading cause of sustained physical, cognitive, emotional, and behavioral disabilities in Veterans and active-duty military personnel. Accurate diagnosis of cmTBI is challenging since the symptom spectrum is broad and conventional neuroimaging techniques are insensitive to the underlying neuropathology. The present study developed a novel deep-learning neural network method, 3D-MEGNET, and applied it to resting-state magnetoencephalography (rs-MEG) source-magnitude imaging data from 59 symptomatic cmTBI individuals and 42 combat-deployed healthy controls (HCs). Analytic models of individual frequency bands and all bands together were tested. The All-frequency model, which combined delta-theta (1-7 Hz), alpha (8-12 Hz), beta (15-30 Hz), and gamma (30-80 Hz) frequency bands, outperformed models based on individual bands. The optimized 3D-MEGNET method distinguished cmTBI individuals from HCs with excellent sensitivity (99.9 ± 0.38%) and specificity (98.9 ± 1.54%). Receiver-operator-characteristic curve analysis showed that diagnostic accuracy was 0.99. The gamma and delta-theta band models outperformed alpha and beta band models. Among cmTBI individuals, but not controls, hyper delta-theta and gamma-band activity correlated with lower performance on neuropsychological tests, whereas hypo alpha and beta-band activity also correlated with lower neuropsychological test performance. This study provides an integrated framework for condensing large source-imaging variable sets into optimal combinations of regions and frequencies with high diagnostic accuracy and cognitive relevance in cmTBI. The all-frequency model offered more discriminative power than each frequency-band model alone. This approach offers an effective path for optimal characterization of behaviorally relevant neuroimaging features in neurological and psychiatric disorders.

Marked Increases in Resting-State MEG Gamma-Band Activity in Combat-Related Mild Traumatic Brain Injury.

Combat-related mild traumatic brain injury (mTBI) is a leading cause of sustained impairments in military service members and veterans. Recent animal studies show that GABA-ergic parvalbumin-positive interneurons are susceptible to brain injury, with damage causing abnormal increases in spontaneous gamma-band (30-80 Hz) activity. We investigated spontaneous gamma activity in individuals with mTBI using high-resolution resting-state magnetoencephalography source imaging. Participants included 25 symptomatic individuals with chronic combat-related blast mTBI and 35 healthy controls with similar combat experiences. Compared with controls, gamma activity was markedly elevated in mTBI participants throughout frontal, parietal, temporal, and occipital cortices, whereas gamma activity was reduced in ventromedial prefrontal cortex. Across groups, greater gamma activity correlated with poorer performances on tests of executive functioning and visuospatial processing. Many neurocognitive associations, however, were partly driven by the higher incidence of mTBI participants with both higher gamma activity and poorer cognition, suggesting that expansive upregulation of gamma has negative repercussions for cognition particularly in mTBI. This is the first human study to demonstrate abnormal resting-state gamma activity in mTBI. These novel findings suggest the possibility that abnormal gamma activities may be a proxy for GABA-ergic interneuron dysfunction and a promising neuroimaging marker of insidious mild head injuries.

MEG Working Memory N-Back Task Reveals Functional Deficits in Combat-Related Mild Traumatic Brain Injury.

Combat-related mild traumatic brain injury (mTBI) is a leading cause of sustained cognitive impairment in military service members and Veterans. However, the mechanism of persistent cognitive deficits including working memory (WM) dysfunction is not fully understood in mTBI. Few studies of WM deficits in mTBI have taken advantage of the temporal and frequency resolution afforded by electromagnetic measurements. Using magnetoencephalography (MEG) and an N-back WM task, we investigated functional abnormalities in combat-related mTBI. Study participants included 25 symptomatic active-duty service members or Veterans with combat-related mTBI and 20 healthy controls with similar combat experiences. MEG source-magnitude images were obtained for alpha (8-12 Hz), beta (15-30 Hz), gamma (30-90 Hz), and low-frequency (1-7 Hz) bands. Compared with healthy combat controls, mTBI participants showed increased MEG signals across frequency bands in frontal pole (FP), ventromedial prefrontal cortex, orbitofrontal cortex (OFC), and anterior dorsolateral prefrontal cortex (dlPFC), but decreased MEG signals in anterior cingulate cortex. Hyperactivations in FP, OFC, and anterior dlPFC were associated with slower reaction times. MEG activations in lateral FP also negatively correlated with performance on tests of letter sequencing, verbal fluency, and digit symbol coding. The profound hyperactivations from FP suggest that FP is particularly vulnerable to combat-related mTBI.

Using EEG signals to assess workload during memory retrieval in a real-world scenario.

Objective. The electroencephalogram (EEG) is gaining popularity as a physiological measure for neuroergonomics in human factor studies because it is objective, less prone to bias, and capable of assessing the dynamics of cognitive states. This study investigated the associations between memory workload and EEG during participants' typical office tasks on a single-monitor and dual-monitor arrangement. We expect a higher memory workload for the single-monitor arrangement.Approach. We designed an experiment that mimics the scenario of a subject performing some office work and examined whether the subjects experienced various levels of memory workload in two different office setups: (1) a single-monitor setup and (2) a dual-monitor setup. We used EEG band power, mutual information, and coherence as features to train machine learning models to classify high versus low memory workload states.Main results. The study results showed that these characteristics exhibited significant differences that were consistent across all participants. We also verified the robustness and consistency of these EEG signatures in a different data set collected during a Sternberg task in a prior study.Significance. The study found the EEG correlates of memory workload across individuals, demonstrating the effectiveness of using EEG analysis in conducting real-world neuroergonomic studies.

Assessing Pediatric Mild Traumatic Brain Injury and Its Recovery Using Resting-State Magnetoencephalography Source Magnitude Imaging and Machine Learning.

The objectives of this machine-learning (ML) resting-state magnetoencephalography (rs-MEG) study involving children with mild traumatic brain injury (mTBI) and orthopedic injury (OI) controls were to define a neural injury signature of mTBI and to delineate the pattern(s) of neural injury that determine behavioral recovery. Children ages 8-15 years with mTBI (n = 59) and OI (n = 39) from consecutive admissions to an emergency department were studied prospectively for parent-rated post-concussion symptoms (PCS) at: 1) baseline (average of 3 weeks post-injury) to measure pre-injury symptoms and also concurrent symptoms; and 2) at 3-months post-injury. rs-MEG was conducted at the baseline assessment. The ML algorithm predicted cases of mTBI versus OI with sensitivity of 95.5 ± 1.6% and specificity of 90.2 ± 2.7% at 3-weeks post-injury for the combined delta-gamma frequencies. The sensitivity and specificity were significantly better (p < 0.0001) for the combined delta-gamma frequencies compared with the delta-only and gamma-only frequencies. There were also spatial differences in rs-MEG activity between mTBI and OI groups in both delta and gamma bands in frontal and temporal lobe, as well as more widespread differences in the brain. The ML algorithm accounted for 84.5% of the variance in predicting recovery measured by PCS changes between 3 weeks and 3 months post-injury in the mTBI group, and this was significantly lower (p < 10-4) in the OI group (65.6%). Frontal lobe pole (higher) gamma activity was significantly (p < 0.001) associated with (worse) PCS recovery exclusively in the mTBI group. These findings demonstrate a neural injury signature of pediatric mTBI and patterns of mTBI-induced neural injury related to behavioral recovery.

Detection of Chronic Blast-Related Mild Traumatic Brain Injury with Diffusion Tensor Imaging and Support Vector Machines.

Blast-related mild traumatic brain injury (bmTBI) often leads to long-term sequalae, but diagnostic approaches are lacking due to insufficient knowledge about the predominant pathophysiology. This study aimed to build a diagnostic model for future verification by applying machine-learning based support vector machine (SVM) modeling to diffusion tensor imaging (DTI) datasets to elucidate white-matter features that distinguish bmTBI from healthy controls (HC). Twenty subacute/chronic bmTBI and 19 HC combat-deployed personnel underwent DTI. Clinically relevant features for modeling were selected using tract-based analyses that identified group differences throughout white-matter tracts in five DTI metrics to elucidate the pathogenesis of injury. These features were then analyzed using SVM modeling with cross validation. Tract-based analyses revealed abnormally decreased radial diffusivity (RD), increased fractional anisotropy (FA) and axial/radial diffusivity ratio (AD/RD) in the bmTBI group, mostly in anterior tracts (29 features). SVM models showed that FA of the anterior/superior corona radiata and AD/RD of the corpus callosum and anterior limbs of the internal capsule (5 features) best distinguished bmTBI from HCs with 89% accuracy. This is the first application of SVM to identify prominent features of bmTBI solely based on DTI metrics in well-defined tracts, which if successfully validated could promote targeted treatment interventions.

Arrhythmia Classification using Deep Learning and Machine Learning with Features Extracted from Waveform-based Signal Processing.

Arrhythmia is a serious cardiovascular disease, and early diagnosis of arrhythmia is critical. In this study, we present a waveform-based signal processing (WBSP) method to produce state-of-the-art performance in arrhythmia classification. When performing WBSP, we first filtered ECG signals, searched local minima, and removed baseline wandering. Subsequently, we fit the processed ECG signals with Gaussians and extracted the parameters. Afterwards, we exploited the products of WBSP to accomplish arrhythmia classification with our proposed machine learning-based and deep learning-based classifiers. We utilized MIT-BIH Arrhythmia Database to validate WBSP. Our best classifier achieved 98.8% accuracy. Moreover, it reached 96.3% sensitivity in class V and 98.6% sensitivity in class Q, which both share one of the best among the related works. In addition, our machine learning-based classifier accomplished identifying four waveform components essential for automated arrhythmia classification: the similarity of QRS complex to a Gaussian curve, the sharpness of the QRS complex, the duration of and the area enclosed by P-wave.Clinical relevance- Early diagnosis and automated classification of arrhythmia is clinically essential.

R-peak Detection Using a Hybrid of Gaussian and Threshold Sensitivity.

We present an enhanced R-peak detection technique that incorporates both waveform shape recognition and threshold sensitivity enhancement. Waveform shape recognition was achieved with signal processing and Gaussian curve parameterization; threshold sensitivity was accomplished with the famous Pan-Tompkins algorithm. We tested all 48 records in MIT-BIH Arrhythmia Database to validate the proposed method. Our method achieved 97.41% sensitivity against a tolerance window of 10% averaged R-R interval, which improves the current state-of-the-art Pan-Tompkins algorithm by 1%. More importantly, we demonstrate that our approach outperforms the Pan-Tompkins' algorithm in 81% of the records in MIT-BIH Arrhythmia Database.Clinical relevance: High sensitivity R-peak detection is substantial in various cardiovascular disease diagnosis.

Cuff-Less Blood Pressure Monitoring with a 3-Axis Accelerometer.

OBJECTIVE: The aim was to propose a cuff-less, cost-efficient, and ultra-convenient blood pressure monitoring technique with a 3-axis accelerometer. METHODS: The efficacy of the proposed approach was examined in 8 young healthy volunteers undergoing different activities with a 3-axis accelerometer leveled on their upper chest. The 3-dimensional accelerations were exploited to select features for the calculation of systolic pressure (SP) and diastolic pressure (DP); the whole process involved signal processing, feature extraction, linear multivariate regression, and leave-one-out cross validations (LOOCV). RESULTS: DP and SP could be approximated with the linear combination of the extracted features: the L2 norm of lateral acceleration for both DP and SP, state variation (defined in the proposed algorithm) of vertical acceleration for SP, and I-J interval (defined in ballistocardiogram) of vertical acceleration for DP. The correlation coefficient (r) of the estimated and the measured DP was 0.97, and for SP, r = 0.96. In LOOCV, our best validated results in difference errors were -0.02±3.82 mmHg for DP and -0.59 ± 7.46 mmHg for SP. CONCLUSION: Compared to AAMI criteria, the proposed acceleration-based technique fulfilled the requirement. The accelerometer-based technique showed the potential to monitor blood pressure cuff-lessly, cost-efficiently, ultra-conveniently, and to be embedded in a long-term wearable device for clinical usage.

An Online Brain-Computer Interface Based on SSVEPs Measured From Non-Hair-Bearing Areas.

Steady state visual evoked potential (SSVEP)-based brain-computer interface (BCI) has gained a lot of attention due to its robustness and high information transfer rate (ITR). However, transitioning well-controlled laboratory-oriented BCI demonstrations to real-world applications poses severe challenges for this exciting field. For instance, conducting BCI experiments usually requires skilled technicians to abrade the area of skin underneath each electrode and apply an electrolytic gel or paste to acquire high-quality SSVEPs from hair-covered areas. Our previous proof-of-concept study has proposed an alternative approach that employed electroencephalographic signals collected from easily accessible non-hair-bearing areas including neck, behind the ears, and face to realize an SSVEP-based BCI. The study results showed that, with proper electrode placements and advanced signal-processing algorithms, the SSVEPs measured from non-hair-bearing areas in off-line SSVEP experiments could achieve comparable SNR to that obtained from the hair-bearing occipital areas. This study extended the previous work to systematically investigate the costs and benefits of non-hair SSVEPs. Furthermore, this study developed and evaluated an online BCI system based solely on non-hair EEG signals. A 12-target identification task was employed to quantitatively assess the performance of the online SSVEP-based BCI system. All subjects successfully completed the tasks using non-hair SSVEPs with 84.08 ± 15.60% averaged accuracy and 30.21 ± 10.61 bits/min averaged ITR. The empirical results of this study demonstrated the practicality of implementing an SSVEP-based BCI based on signals from non-hair-bearing areas, significantly improving the feasibility and practicality of real-world BCIs.

Developing an online steady-state visual evoked potential-based brain-computer interface system using EarEEG.

The purpose of this study is to demonstrate an online steady-state visual evoked potential (SSVEP)-based BCI system using EarEEG. EarEEG is a novel recording concept where electrodes are embedded on the surface of earpieces customized to the individual anatomical shape of users' ear. It has been shown that the EarEEG can be used to record SSVEPs in previous studies. However, a long distance between the visual cortex and the ear makes the signal-to-noise ratio (SNR) of SSVEPs acquired by the EarEEG relatively low. Recently, filter bank- and training data-based canonical correlation analysis algorithms have shown significant performance improvement in terms of accuracy of target detection and information transfer rate (ITR). This study implemented an online four-class SSVEP-based BCI system using EarEEG. Four subjects participated in offline and online BCI experiments. For the offline classification, an average accuracy of 82.71±11.83 % was obtained using 4 sec-long SSVEPs acquired from earpieces. In the online experiment, all subjects successfully completed the tasks with an average accuracy of 87.92±12.10 %, leading to an average ITR of 16.60±6.55 bits/min. The results suggest that EarEEG can be used to perform practical BCI applications. The EarEEG has the potential to be used as a portable EEG recordings platform, that could enable real-world BCI applications.

Developing an EEG-based on-line closed-loop lapse detection and mitigation system.

In America, 60% of adults reported that they have driven a motor vehicle while feeling drowsy, and at least 15-20% of fatal car accidents are fatigue-related. This study translates previous laboratory-oriented neurophysiological research to design, develop, and test an On-line Closed-loop Lapse Detection and Mitigation (OCLDM) System featuring a mobile wireless dry-sensor EEG headgear and a cell-phone based real-time EEG processing platform. Eleven subjects participated in an event-related lane-keeping task, in which they were instructed to manipulate a randomly deviated, fixed-speed cruising car on a 4-lane highway. This was simulated in a 1st person view with an 8-screen and 8-projector immersive virtual-reality environment. When the subjects experienced lapses or failed to respond to events during the experiment, auditory warning was delivered to rectify the performance decrements. However, the arousing auditory signals were not always effective. The EEG spectra exhibited statistically significant differences between effective and ineffective arousing signals, suggesting that EEG spectra could be used as a countermeasure of the efficacy of arousing signals. In this on-line pilot study, the proposed OCLDM System was able to continuously detect EEG signatures of fatigue, deliver arousing warning to subjects suffering momentary cognitive lapses, and assess the efficacy of the warning in near real-time to rectify cognitive lapses. The on-line testing results of the OCLDM System validated the efficacy of the arousing signals in improving subjects' response times to the subsequent lane-departure events. This study may lead to a practical on-line lapse detection and mitigation system in real-world environments.

Empirical mode decomposition improves detection of SSVEP.

Steady State Visual Evoked Potentials (SSVEPs) have been used to quantify attention-related neural activity to visual targets. This study investigates how empirical mode decomposition (EMD) can improve detection accuracy and rate of SSVEPs. First, the scalp-recorded electroencephalogram (EEG) signals are decomposed into intrinsic mode functions (IMFs) by EMD. Then, IMF components accounting for SSVEPs are selected for target frequency detection. Finally, target frequency is identified by two methods: Gabor transform and Canonical Correlation Analysis (CCA). This study quantitatively explores the impact of EMD on the target frequency detection. Empirical results show that the EMD improves their recognition accuracy when Gabor transform is used, even in a shorter Gaussian window, but has little effects on the performance of the CCA. Further, this study finds that harmonic responses of the target frequency can be used to enhance the SSVEP detection both for the Gabor transform and CCA.

Developing stimulus presentation on mobile devices for a truly portable SSVEP-based BCI.

This study integrates visual stimulus presentation and near real-time data processing on a mobile device (e.g. a Tablet or a cell-phone) to implement a steady-state visual evoked potentials (SSVEP)-based brain-computer interface (BCI). The goal of this study is to increase the practicability, portability and ubiquity of an SSVEP-based BCI for daily use. The accuracy of flickering frequencies on the mobile SSVEP BCI system was tested against that on a laptop/desktop used in our previous studies. This study then analyzed the power spectrum density of the electroencephalogram signals elicited by the visual stimuli rendered on the mobile BCIs. Finally, this study performed an online test with the Tablet-based BCI system and obtained an averaged information transfer rate of 33.87 bits/min in three subjects. The current integration leads to a truly practical and ubiquitous SSVEP BCI on mobile devices for real-life applications.

Measuring steady-state visual evoked potentials from non-hair-bearing areas.

Steady-State Visual Evoked Potential (SSVEP)-based Brain-Computer Interface (BCI) applications have been widely applied in laboratories around the world in the recent years. Many studies have shown that the best locations to acquire SSVEPs were from the occipital areas of the scalp. However, for some BCI users such as quadriparetic patients lying face up during ventilation, it is difficult to access the occipital sites. Even for the healthy BCI users, acquiring good-quality EEG signals from the hair-covered occipital sites is inevitably more difficult because it requires skin preparation by a skilled technician and conductive gel usage. Therefore, finding an alternative approach to effectively extract high-quality SSVEPs for BCI practice is highly desirable. Since the non-hair-bearing scalp regions are more accessible by all different types of EEG sensors, this study systematically and quantitatively investigated the feasibility of measuring SSVEPs from non-hair-bearing regions, compared to those measured from the occipital areas. Empirical results showed that the signal quality of the SSVEPs from non-hair-bearing areas was comparable with, if not better than, that measured from hair-covered occipital areas. These results may significantly improve the practicality of a BCI system in real-life applications; especially used in conjunction with newly available dry EEG sensors.

Analyzing high-density ECG signals using ICA.

The analysis of ECG signals is of fundamental importance for cardiac diagnosis. Conventional ECG recordings, however, use a limited number of channels (12) and each records a mixture of activities generated in different parts of the heart. Therefore, direct observation of the ECG signals collected on the body surface is likely an inefficient way to study and diagnose cardiac abnormalities. This study describes new experimental and analytical methods to capture more meaningful ECG component signals, each representing more directly a physical cardiac source. This study first describes a simply applied method for collecting high-density ECG signals. The recorded signals are then separated by independent component analysis (ICA) to obtain spatially fixed and temporally independent component activations. Results from five subjects show that P-, QRS-, and T-waves can be clearly separated from the recordings, suggesting ICA might be an effective and useful tool for high-density ECG analysis, interpretation, and diagnosis.

Spatial density reduction in the study of the ECG signal using independent component analysis.

Conventional 12-lead ECG can only record limited aspects of the heart electrical signals. Previously, an approach for the analysis of the ECG signal using high spatial resolution over 101 or 75 channels on the chest and back of the subject for recording the heart signal has been proposed by Zhu et al [3]. They decomposed the original signals into distinct temporal source components by applying independent component analysis. In the current work, we reduce the number of recording channels to 24 and 12 while still able to decompose the ECG signals into similar source components. In addition, we propose a method to estimate the source components dipoles based on the Nelder-Mead simplex method. This approach facilitates the feasibility of the ICA study on the patients with moderate spatial resolution, since it requires significantly fewer leads for recording and easier diagnosis setup.