Improving Eye Motion Sequence Recognition Using Electrooculography Based on Context-Dependent HMM.

Eye motion-based human-machine interfaces are used to provide a means of communication for those who can move nothing but their eyes because of injury or disease. To detect eye motions, electrooculography (EOG) is used. For efficient communication, the input speed is critical. However, it is difficult for conventional EOG recognition methods to accurately recognize fast, sequentially input eye motions because adjacent eye motions influence each other. In this paper, we propose a context-dependent hidden Markov model- (HMM-) based EOG modeling approach that uses separate models for identical eye motions with different contexts. Because the influence of adjacent eye motions is explicitly modeled, higher recognition accuracy is achieved. Additionally, we propose a method of user adaptation based on a user-independent EOG model to investigate the trade-off between recognition accuracy and the amount of user-dependent data required for HMM training. Experimental results show that when the proposed context-dependent HMMs are used, the character error rate (CER) is significantly reduced compared with the conventional baseline under user-dependent conditions, from 36.0 to 1.3%. Although the CER increases again to 17.3% when the context-dependent but user-independent HMMs are used, it can be reduced to 7.3% by applying the proposed user adaptation method.

Detection of EEG-resting state independent networks by eLORETA-ICA method.

Recent functional magnetic resonance imaging (fMRI) studies have shown that functional networks can be extracted even from resting state data, the so called "Resting State independent Networks" (RS-independent-Ns) by applying independent component analysis (ICA). However, compared to fMRI, electroencephalography (EEG) and magnetoencephalography (MEG) have much higher temporal resolution and provide a direct estimation of cortical activity. To date, MEG studies have applied ICA for separate frequency bands only, disregarding cross-frequency couplings. In this study, we aimed to detect EEG-RS-independent-Ns and their interactions in all frequency bands. We applied exact low resolution brain electromagnetic tomography-ICA (eLORETA-ICA) to resting-state EEG data in 80 healthy subjects using five frequency bands (delta, theta, alpha, beta and gamma band) and found five RS-independent-Ns in alpha, beta and gamma frequency bands. Next, taking into account previous neuroimaging findings, five RS-independent-Ns were identified: (1) the visual network in alpha frequency band, (2) dual-process of visual perception network, characterized by a negative correlation between the right ventral visual pathway (VVP) in alpha and beta frequency bands and left posterior dorsal visual pathway (DVP) in alpha frequency band, (3) self-referential processing network, characterized by a negative correlation between the medial prefrontal cortex (mPFC) in beta frequency band and right temporoparietal junction (TPJ) in alpha frequency band, (4) dual-process of memory perception network, functionally related to a negative correlation between the left VVP and the precuneus in alpha frequency band; and (5) sensorimotor network in beta and gamma frequency bands. We selected eLORETA-ICA which has many advantages over the other network visualization methods and overall findings indicate that eLORETA-ICA with EEG data can identify five RS-independent-Ns in their intrinsic frequency bands, and correct correlations within RS-independent-Ns.

Noninvasive prediction of shunt operation outcome in idiopathic normal pressure hydrocephalus.

Idiopathic normal pressure hydrocephalus (iNPH) is a syndrome characterized by gait disturbance, cognitive deterioration and urinary incontinence in elderly individuals. These symptoms can be improved by shunt operation in some but not all patients. Therefore, discovering predictive factors for the surgical outcome is of great clinical importance. We used normalized power variance (NPV) of electroencephalography (EEG) waves, a sensitive measure of the instability of cortical electrical activity, and found significantly higher NPV in beta frequency band at the right fronto-temporo-occipital electrodes (Fp2, T4 and O2) in shunt responders compared to non-responders. By utilizing these differences, we were able to correctly identify responders and non-responders to shunt operation with a positive predictive value of 80% and a negative predictive value of 88%. Our findings indicate that NPV can be useful in noninvasively predicting the clinical outcome of shunt operation in patients with iNPH.

EEG and Neuronal Activity Topography analysis can predict effectiveness of shunt operation in idiopathic normal pressure hydrocephalus patients.

Idiopathic normal pressure hydrocephalus (iNPH) is a neuropsychiatric syndrome characterized by gait disturbance, cognitive impairment and urinary incontinence that affect elderly individuals. These symptoms can potentially be reversed by cerebrospinal fluid (CSF) drainage or shunt operation. Prior to shunt operation, drainage of a small amount of CSF or "CSF tapping" is usually performed to ascertain the effect of the operation. Unfortunately, conventional neuroimaging methods such as single photon emission computed tomography (SPECT) and functional magnetic resonance imaging (fMRI), as well as electroencephalogram (EEG) power analysis seem to have failed to detect the effect of CSF tapping on brain function. In this work, we propose the use of Neuronal Activity Topography (NAT) analysis, which calculates normalized power variance (NPV) of EEG waves, to detect cortical functional changes induced by CSF tapping in iNPH. Based on clinical improvement by CSF tapping and shunt operation, we classified 24 iNPH patients into responders (N = 11) and nonresponders (N = 13), and performed both EEG power analysis and NAT analysis. We also assessed correlations between changes in NPV and changes in functional scores on gait and cognition scales before and after CSF tapping. NAT analysis showed that after CSF tapping there was a significant decrease in alpha NPV at the medial frontal cortex (FC) (Fz) in responders, while nonresponders exhibited an increase in alpha NPV at the right dorsolateral prefrontal cortex (DLPFC) (F8). Furthermore, we found correlations between cortical functional changes and clinical symptoms. In particular, delta and alpha NPV changes in the left-dorsal FC (F3) correlated with changes in gait status, while alpha and beta NPV changes in the right anterior prefrontal cortex (PFC) (Fp2) and left DLPFC (F7) as well as alpha NPV changes in the medial FC (Fz) correlated with changes in gait velocity. In addition, alpha NPV changes in the right DLPFC (F8) correlated with changes in WMS-R Mental Control scores in iNPH patients. An additional analysis combining the changes in values of alpha NPV over the left-dorsal FC (∆alpha-F3-NPV) and the medial FC (∆alpha-Fz-NPV) induced by CSF tapping (cut-off value of ∆alpha-F3-NPV + ∆alpha-Fz-NPV = 0), could correctly identified "shunt responders" and "shunt nonresponders" with a positive predictive value of 100% (10/10) and a negative predictive value of 66% (2/3). In contrast, EEG power spectral analysis showed no function related changes in cortical activity at the frontal cortex before and after CSF tapping. These results indicate that the clinical changes in gait and response suppression induced by CSF tapping in iNPH patients manifest as NPV changes, particularly in the alpha band, rather than as EEG power changes. Our findings suggest that NAT analysis can detect CSF tapping-induced functional changes in cortical activity, in a way that no other neuroimaging methods have been able to do so far, and can predict clinical response to shunt operation in patients with iNPH.

Normalized power variance change between pre-ictal and ictal phase of an epilepsy patient using NAT analysis: a case study.

Variance of state variables shifts due to phase-instability and may serve as an early-warning signal of phase transition of complex systems such as an epileptic seizure of brain cortical activity. Neuronal Activity Topology (NAT) analysis calculates a normalized-power-variance (NPV) of electroencephalogram (EEG) data in each frequency band to obtain relative values comparable among different power states.

Diagnosis method of mild cognitive impairment based on power variance of EEG.

Mild cognitive impairment (MCI) patients and healthy people were classified by using a "power variance function (PVF)", namely, an index of electroencephalography (EEG) proposed in a previous report. PVF is defined by calculating variance of the power variability of an EEG signal at each frequency of the signal using wavelet transform. After confirming that the distribution of PVFs of the subjects was a normal distribution at each frequency, the distributions of PVFs of 25 MCI patients and those of 57 healthy people were compared in terms of Z-score. The comparison results indicate that for the MCI patients, the PVFs in the θ band are significantly higher in left parieto-occipital area and that those in the ß band are lower in the bitemporal area. Multidimensional discriminant analysis using the PVF in the θ-ß band recorded only on four electrodes on the left parieto-occipital area could be used to classify MCI patients from healthy people with leave-one-out accuracy of 87.5%. This indicates the possibility of diagnosing MCI by using EEG signals recorded only on a few electrodes.

The relationship between the diagnosis method of neuronal dysfunction (DIMENSION) and brain pathology in the early stages of Alzheimer's disease.

OBJECTIVES: To examine whether the diagnosis method of neuronal dysfunction (DIMENSION), a new electroencephalogram (EEG) analysis method, reflected pathological changes in the early stages of Alzheimer's disease (AD), we conducted a comparative study of cerebrospinal fluid markers and single-photon emission computed tomography. METHODS: Subjects cincluded 32 patients in the early stages of AD with a Mini-Mental State Examination score ≥24 (14 men, 18 women; mean age, 77.3 ± 9.2 years). Cerebrospinal fluid samples were collected from AD patients, and cerebrospinal fluid levels of phosphorylated tau protein (p-tau) 181 and amyloid ß (Aß) 42 were measured with sandwich ELISA. EEG recordings were performed for 5 min with the subjects awake in a resting state with their eyes closed. Then, the mean value of the EEG alpha dipolarity (Dα) and the standard deviation of the EEG alpha dipolarity (Dσ) were calculated with DIMENSION. Single-photon emission computed tomography analyses were also performed for comparison with DIMENSION measures. RESULTS: Patients with parietal hypoperfusion had significantly increasing p-tau181, decreasing Dα, and increasing Dσ. In addition, there was a negative correlation between Dα and p-tau181, p-tau181/Aß42, and a positive correlation between Dσ and p-tau181/Aß42. CONCLUSION: Dα and Dσ were related to cerebral hypoperfusion and p-tau181/Aß42. DIMENSION was able to detect changes in the early-stage Alzheimer's brain, suggesting that it is possibility as a useful examination for early-stage AD with a difficult discrimination in clinical conditions. Moreover, EEG measurement is a quick and easy diagnostic test and is useful for repeated examinations.

EEG markers for characterizing anomalous activities of cerebral neurons in NAT (neuronal activity topography) method.

A pair of markers, sNAT and vNAT, is derived from the electroencephalogram (EEG) power spectra (PS) recorded for 5 min with 21 electrodes (4-20 Hz) arranged according to the 10-20 standard. These markers form a new diagnosis tool "NAT" aiming at characterizing various brain disorders. Each signal sequence is divided into segments of 0.64 s and its discrete PS consists of eleven frequency components from 4.68 (3 × 1.56) Hz through 20.34 (13 × 1.56) Hz. PS is normalized to its mean and the bias of PS components on each frequency component across the 21 signal channels is reset to zero. The marker sNAT consists of ten frequency components on 21 channels, characterizing neuronal hyperactivity or hypoactivity as compared with NLc (normal controls). The marker vNAT consists of ten ratios between adjacent PS components denoting the over- or undersynchrony of collective neuronal activities as compared with NLc. The likelihood of a test subject to a specified brain disease is defined in terms of the normalized distance to the template NAT state of the disease in the NAT space. Separation of MCI-AD patients (developing AD in 12-18 months) from NLc is made with a false alarm rate of 15%. Locations with neuronal hypoactivity and undersynchrony of AD patients agree with locations of rCBF reduction measured by SPECT. The 2-D diagram composed of the binary likelihoods between ADc and NLc in the two representations of sNAT and vNAT enables tracing the NAT state of a test subject approaching the AD area, and the follow-up of the treatment effects.

Quantifying statistical interdependence, part III: N > 2 point processes.

Stochastic event synchrony (SES) is a recently proposed family of similarity measures. First, "events" are extracted from the given signals; next, one tries to align events across the different time series. The better the alignment, the more similar the N time series are considered to be. The similarity measures quantify the reliability of the events (the fraction of "nonaligned" events) and the timing precision. So far, SES has been developed for pairs of one-dimensional (Part I) and multidimensional (Part II) point processes. In this letter (Part III), SES is extended from pairs of signals to N > 2 signals. The alignment and SES parameters are again determined through statistical inference, more specifically, by alternating two steps: (1) estimating the SES parameters from a given alignment and (2), with the resulting estimates, refining the alignment. The SES parameters are computed by maximum a posteriori (MAP) estimation (step 1), in analogy to the pairwise case. The alignment (step 2) is solved by linear integer programming. In order to test the robustness and reliability of the proposed N-variate SES method, it is first applied to synthetic data. We show that N-variate SES results in more reliable estimates than bivariate SES. Next N-variate SES is applied to two problems in neuroscience: to quantify the firing reliability of Morris-Lecar neurons and to detect anomalies in EEG synchrony of patients with mild cognitive impairment. Those problems were also considered in Parts I and II, respectively. In both cases, the N-variate SES approach yields a more detailed analysis.

Audio representations of multi-channel EEG: a new tool for diagnosis of brain disorders.

OBJECTIVE: The objective of this paper is to develop audio representations of electroencephalographic (EEG) multichannel signals, useful for medical practitioners and neuroscientists. The fundamental question explored in this paper is whether clinically valuable information contained in the EEG, not available from the conventional graphical EEG representation, might become apparent through audio representations. METHODS AND MATERIALS: Music scores are generated from sparse time-frequency maps of EEG signals. Specifically, EEG signals of patients with mild cognitive impairment (MCI) and (healthy) control subjects are considered. Statistical differences in the audio representations of MCI patients and control subjects are assessed through mathematical complexity indexes as well as a perception test; in the latter, participants try to distinguish between audio sequences from MCI patients and control subjects. RESULTS: Several characteristics of the audio sequences, including sample entropy, number of notes, and synchrony, are significantly different in MCI patients and control subjects (Mann-Whitney p < 0.01). Moreover, the participants of the perception test were able to accurately classify the audio sequences (89% correctly classified). CONCLUSIONS: The proposed audio representation of multi-channel EEG signals helps to understand the complex structure of EEG. Promising results were obtained on a clinical EEG data set.

Improving the specificity of EEG for diagnosing Alzheimer's disease.

Objective. EEG has great potential as a cost-effective screening tool for Alzheimer's disease (AD). However, the specificity of EEG is not yet sufficient to be used in clinical practice. In an earlier study, we presented preliminary results suggesting improved specificity of EEG to early stages of Alzheimer's disease. The key to this improvement is a new method for extracting sparse oscillatory events from EEG signals in the time-frequency domain. Here we provide a more detailed analysis, demonstrating improved EEG specificity for clinical screening of MCI (mild cognitive impairment) patients. Methods. EEG data was recorded of MCI patients and age-matched control subjects, in rest condition with eyes closed. EEG frequency bands of interest were θ (3.5-7.5 Hz), α(1) (7.5-9.5 Hz), α(2) (9.5-12.5 Hz), and ß (12.5-25 Hz). The EEG signals were transformed in the time-frequency domain using complex Morlet wavelets; the resulting time-frequency maps are represented by sparse bump models. Results. Enhanced EEG power in the θ range is more easily detected through sparse bump modeling; this phenomenon explains the improved EEG specificity obtained in our previous studies. Conclusions. Sparse bump modeling yields informative features in EEG signal. These features increase the specificity of EEG for diagnosing AD.

Slowing and Loss of Complexity in Alzheimer's EEG: Two Sides of the Same Coin?

Medical studies have shown that EEG of Alzheimer's disease (AD) patients is "slower" (i.e., contains more low-frequency power) and is less complex compared to age-matched healthy subjects. The relation between those two phenomena has not yet been studied, and they are often silently assumed to be independent. In this paper, it is shown that both phenomena are strongly related. Strong correlation between slowing and loss of complexity is observed in two independent EEG datasets: (1) EEG of predementia patients (a.k.a. Mild Cognitive Impairment; MCI) and control subjects; (2) EEG of mild AD patients and control subjects. The two data sets are from different patients, different hospitals and obtained through different recording systems. The paper also investigates the potential of EEG slowing and loss of EEG complexity as indicators of AD onset. In particular, relative power and complexity measures are used as features to classify the MCI and MiAD patients versus age-matched control subjects. When combined with two synchrony measures (Granger causality and stochastic event synchrony), classification rates of 83% (MCI) and 98% (MiAD) are obtained. By including the compression ratios as features, slightly better classification rates are obtained than with relative power and synchrony measures alone.

Research of the characteristics of Alzheimer's disease using EEG.

In this paper, we propose a new method for diagnosing Alzheimer's disease (AD) on the basis of electroencephalograms (EEG). The method, which is termed "Power Variance Function (PVF) method", indicates the variance of the power at each frequency. By using the proposed method, the power of EEG at each frequency was calculated using Wavelet transform, and the corresponding variances were defined as PVF. After the PVF histogram of 42 healthy people was approximated as a Generalized Extreme Value (GEV) distribution, we evaluated the PVF of 10 patients with AD and 10 patients with mild cognitive impairment (MCI). As a result, the values for all AD and MCI subjects were abnormal. In particular, the PVF in the theta band for MCI patients was abnormally high, and the PVF in the alpha band for AD patients was low.

Fronto-parietal coupling of brain rhythms in mild cognitive impairment: a multicentric EEG study.

Electroencephalographic (EEG) data were recorded in 69 normal elderly (Nold), 88 mild cognitive impairment (MCI), and 109 mild Alzheimer's disease (AD) subjects at rest condition, to test whether the fronto-parietal coupling of EEG rhythms is in line with the hypothesis that MCI can be considered as a pre-clinical stage of the disease at group level. Functional coupling was estimated by synchronization likelihood of Laplacian-transformed EEG data at electrode pairs, which accounts for linear and non-linear components of that coupling. Cortical rhythms of interest were delta (2-4Hz), theta (4-8Hz), alpha 1 (8-10.5Hz), alpha 2 (10.5-13Hz), beta 1 (13-20Hz), beta 2 (20-30Hz), and gamma (30-40Hz). Compared to the Nold subjects, the AD patients presented a marked reduction of the synchronization likelihood (delta to gamma) at both fronto-parietal and inter-hemispherical (delta to beta 2) electrodes. As a main result, alpha 1 synchronization likelihood progressively decreased across Nold, MCI, and mild AD subjects at midline (Fz-Pz) and right (F4-P4) fronto-parietal electrodes. The same was true for the delta synchronization likelihood at right fronto-parietal electrodes (F4-P4). For these EEG bands, the synchronization likelihood correlated with global cognitive status as measured by the Mini Mental State Evaluation. The present results suggest that at group level, fronto-parietal coupling of the delta and alpha rhythms progressively becomes abnormal though MCI and mild AD. Future longitudinal research should evaluate whether the present EEG approach is able to predict the cognitive decline in individual MCI subjects.

EEG filtering based on blind source separation (BSS) for early detection of Alzheimer's disease.

OBJECTIVE: Development of an EEG preprocessing technique for improvement of detection of Alzheimer's disease (AD). The technique is based on filtering of EEG data using blind source separation (BSS) and projection of components which are possibly sensitive to cortical neuronal impairment found in early stages of AD. METHODS: Artifact-free 20s intervals of raw resting EEG recordings from 22 patients with Mild Cognitive Impairment (MCI) who later proceeded to AD and 38 age-matched normal controls were decomposed into spatio-temporally decorrelated components using BSS algorithm 'AMUSE'. Filtered EEG was obtained by back projection of components with the highest linear predictability. Relative power of filtered data in delta, theta, alpha 1, alpha 2, beta 1, and beta 2 bands were processed with Linear Discriminant Analysis (LDA). RESULTS: Preprocessing improved the percentage of correctly classified patients and controls computed with jack-knifing cross-validation from 59 to 73% and from 76 to 84%, correspondingly. CONCLUSIONS: The proposed approach can significantly improve the sensitivity and specificity of EEG based diagnosis. SIGNIFICANCE: Filtering based on BSS can improve the performance of the existing EEG approaches to early diagnosis of Alzheimer's disease. It may also have potential for improvement of EEG classification in other clinical areas or fundamental research. The developed method is quite general and flexible, allowing for various extensions and improvements.

A new EEG method for estimating cortical neuronal impairment that is sensitive to early stage Alzheimer's disease.

OBJECTIVES: To test the hypothesis that elecetroencephalographic (EEG) analysis is sensitive to cortical neuronal impairment in early stage Alzheimer's disease (AD), and that this analysis correlates with corresponding changes in cerebral blood flow. METHODS: We examined an EEG measure of neuronal impairment in the cerebral cortex in terms of its ability to detect very mild AD. This measure, the mean value of the resting state EEG alpha dipolarity (D(alpha)), approaches unity without cortical sulcal lesions, whereas brains with randomly distributed cortical sulcal lesions lower D(alpha) values well below unity. D(alpha) was evaluated in 25 patients with very mild AD, 33 patients with moderately severe AD, and 56 normal age-matched subjects. These subjects also received SPECT, and strong correlation between D(alpha) and regional cerebral blood flow (rCBF) was observed. RESULTS: D(alpha) values greater than 0.977 correctly classified normal subjects, but also included 10% of very mild AD. D(alpha) values less than 0.952 correctly classified very mild AD as well as moderately severe AD, but also included 10% of normal subjects. D(alpha) values also correlated positively with bilateral temporal-parietal rCBF (a characteristic finding in AD patients); both declined with increasing dementia severity. CONCLUSIONS: Analysis of D(alpha) in this sample supports the hypothesis that early stages of AD can be discriminated from normal aging using measures of cortical neuronal impairment. Furthermore, dementia severity, as reflected by the degree of impairment, is reflected in declining D(alpha) values and increasing variance (greater spread of the D(alpha) values).