Automatic recognition of alertness and drowsiness from EEG by an artificial neural network

We present a novel method for classifying alert vs drowsy states from 1 s long sequences of full spectrum EEG recordings in an arbitrary subject. This novel method uses time series of interhemispheric and intrahemispheric cross spectral densities of full spectrum EEG as the input to an artificial neural network (ANN) with two discrete outputs: drowsy and alert. The experimental data were collected from 17 subjects. Two experts in EEG interpretation visually inspected the data and provided the necessary expertise for the training of an ANN. We selected the following three ANNs as potential candidates: (1) the linear network with Widrow-Hoff (WH) algorithm; (2) the non-linear ANN with the Levenberg-Marquardt (LM) rule; and (3) the Learning Vector Quantization (LVQ) neural network. We showed that the LVQ neural network gives the best classification compared with the linear network that uses WH algorithm (the worst), and the non-linear network trained with the LM rule. Classification properties of LVQ were validated using the data recorded in 12 healthy volunteer subjects, yet whose EEG recordings have not been used for the training of the ANN. The statistics were used as a measure of potential applicability of the LVQ: the t-distribution showed that matching between the human assessment and the network output was 94.37+/-1.95%. This result suggests that the automatic recognition algorithm is applicable for distinguishing between alert and drowsy state in recordings that have not been used for the training.     

Directionality of EEG synchronization in Alzheimer's disease subjects

Is directionality of electroencephalographic (EEG) synchronization abnormal in amnesic mild cognitive impairment (MCI) and Alzheimer's disease (AD)? EEG data were recorded in 64 normal elderly (Nold), 69 amnesic MCI, and 73 mild AD subjects at rest condition (closed eyes). Direction of information flux within EEG functional coupling at electrode pairs was performed by directed transfer function (DTF) at delta (2-4 Hz), theta (4-8 Hz), alpha 1 (8-10 Hz), alpha 2 (10-12 Hz), beta 1 (13-20 Hz), beta 2 (20-30 Hz), and gamma (30-40 Hz). Parietal to frontal direction of the information flux within EEG functional coupling was stronger in Nold than in MCI and/or AD subjects, namely for alpha and beta rhythms. In contrast, the directional flow within inter-hemispheric EEG functional coupling did not discriminate among the three groups. These results suggest that directionality of parieto-to-frontal EEG synchronization is abnormal not only in AD but also in amnesic MCI.     

Application of artificial neural networks for quantitative analysis of image data in chest radiographs for detection of interstitial lung disease

The authors have developed an automated computeraided diagnostic (CAD) scheme by using artificial neural networks (ANNs) on quantitative analysis of image data. Three separate ANNs were applied for detection of interstitial disease on digitized chest images. The first ANN was trained with horizontal profiles in regions of interest (ROIs) selected from normal and abnormal chest radiographs for distinguishing between normal and abnormal patterns. For training and testing of the second ANN, the vertical output patterns obtained from the 1st ANN were used for each ROI. The output value of the second ANN was used to distinguish between normal and abnormal ROIs with interstitial infiltrates. If the ratio of the number of abnormal ROIs to the total number of all ROIs in a chest image was greater than a specified threshold level, the image was classified as abnormal. In addition, the third ANN was applied to distinguish between normal and abnormal chest images. The combination of the rule-based method and the third ANN also was applied to the classification between normal and abnormal chest images. The performance of the ANNs was evaluated by means of receiver operating characteristic (ROC) analysis. The average A_z value (area under the ROC curve) for distinguishing between normal and abnormal cases was 0.976±0.012 for 100 chest radiographs that were not used in training of ANNs. The results indicate that the ANN trained with image data can learn some statistical properties associated with interstitial infiltrates in chest radiographs.     

Homocysteine and electroencephalographic rhythms in Alzheimer disease: a multicentric study

High plasma concentration of homocysteine is an independent risk factor for Alzheimer’s disease (AD), due to microvascular impairment and consequent neural loss [Seshadri S, Beiser A, Selhub J, Jacques PF, Rosenberg IH, D’Agostino RB, Wilson PW, Wolf PA (2002) Plasma homocysteine as a risk factor for dementia and Alzheimer’s disease. N Engl J Med 346(7):476–483]. Is high plasma homocysteine level related to slow electroencephalographic (EEG) rhythms in awake resting AD subjects, as a reflection of known relationships between cortical neural loss and these rhythms? To test this hypothesis, we enrolled 34 mild AD patients and 34 subjects with mild cognitive impairment (MCI). Enrolled people were then subdivided into four sub-groups of 17 persons: MCI and AD subjects with low homocysteine level (MCI− and AD−, homocysteine level <11 μmol/l); MCI and AD subjects with high homocysteine level (MCI+ and AD+, homocysteine level ≥11 μmol/l). Resting eyes-closed EEG data were recorded. EEG rhythms of interest were delta (2–4 Hz), theta (4–8 Hz), alpha 1 (8–10.5 Hz), alpha 2 (10.5–13 Hz), beta 1 (13–20 Hz), and beta 2 (20–30 Hz). EEG cortical sources were estimated by low-resolution brain electromagnetic tomography (LORETA). Results showed that delta (frontal and temporal), theta (central, frontal, parietal, occipital, and temporal), alpha 1 (parietal, occipital, and temporal), and alpha 2 (parietal and occipital) sources were stronger in magnitude in AD+ than AD− group. Instead, no difference was found between MCI− and MCI+ groups. In conclusion, high plasma homocysteine level is related to unselective increment of cortical delta, theta, and alpha rhythms in mild AD, thus unveiling possible relationships among that level, microvascular concomitants of advanced neurodegenerative processes, and synchronization mechanisms generating EEG rhythms.     

Deficits of spatial and task-related attentional selection in mild cognitive impairment and Alzheimer's disease

Visual selective attention was assessed with a partial-report task in patients with probable Alzheimer's disease (AD), amnestic mild cognitive impairment (MCI), and healthy elderly controls. Based on Bundesen's “theory of visual attention” (TVA), two parameters were derived: top-down control of attentional selection, representing task-related attentional weighting for prioritizing relevant visual objects, and spatial distribution of attentional weights across the left and the right hemifield.Compared with controls, MCI patients showed significantly reduced top-down controlled selection, which was further deteriorated in AD subjects. Moreover, attentional weighting was significantly unbalanced across hemifields in MCI and tended to be more lateralized in AD. Across MCI and AD patients, carriers of the apolipoprotein E ε4 allele (ApoE4) displayed a leftward spatial bias, which was the more pronounced the younger the ApoE4-positive patients and the earlier disease onset.These results indicate that impaired top-down control may be linked to early dysfunction of fronto-parietal networks. An early temporo-parietal interhemispheric asymmetry might cause a pathological spatial bias which is associated with ApoE4 genotype and may therefore function as early cognitive marker of upcoming AD.     

Methodology for hypoglycaemia detection based on the processing, analysis and classification of the electroencephalogram

Hypoglycaemia (blood glucose level below 3.8 mmol l⁻¹) is the most common complication in the treatment of diabetes with insulin and can cause a number of problems. Previous works have shown that hypoglycaemia causes changes in the electroencephalogram (EEG) signal. In this investigation, portable apparatus was developed to record the EEG, and a methodology was implemented, using digital signal processing and artificial neural networks (ANNs), to detect hypoglycaemia. Sixteen EEG recordings were made on eight subjects with diabetes (five male, three female), aged 35±13.5 years (mean ± SD), during the day, over periods of 5.7±2 min. Ten of these recordings (in seven subjects) included periods of normoglycaemia and spontaneous hypoglycaemia. The result of the off-line ANN classification for each of these ten recordings was an overall accuracy rate of 71.3%, sensitivity of 71.1% and specificity of 71.5%. In the classification using four recordings from a single subject, the accuracy was 80.6%, with a sensitivity of 77.8% and a specificity of 83.9%. In the classification using recordings from five different subjects to train the ANN, the obtained accuracy rate was 49.2%, with a sensitivity of 76% and a specificity of 32.5%. The result of the classification in real time, for one subject, was an accuracy rate of 85.2%, with a sensitivity of 60% and a specificity of 100%. In conclusion, the methodology proposed and implemented justifies further studies with the objective of constructing a hypoglycaemia detector system based on the processing and classification of the EEG.     

空间模型对单次运动相关脑电的分析

本研究提出一种从单次试验的多导EEG信号中提取运动相关去同步化和同步化电位特征的空间模型，区分左右手想象运动，作为一种新的通讯手段对外界设备进行控制。此模型根据各电极对分类的重要性自动获得其权值，并将EEG信号沿最适合分类的几个方向投影，沿投影方向计算一连续时间段内的方差，作为线性分类器的特征输入。对8名被试者左右手想象运动时59导EEG进行分类，正确率均在70％以上，与用多通道AR模型提取特征、神经网络做分类器的方法相比，效果好、速度快。     

Hippocampus and entorhinal cortex in mild cognitive impairment and early AD

Magnetic resonance imaging (MRI) has been suggested as a useful tool in early diagnosis of Alzheimer's disease (AD). Based on MRI-derived volumes, we studied the hippocampus and entorhinal cortex (ERC) in 59 controls, 65 individuals with mild cognitive impairment (MCI) and 48 patients with AD. The controls and individuals with MCI were derived from population-based cohorts. Volumes of the hippocampus and ERC were significantly reduced in the following order: control > MCI > AD. Stepwise discriminant function analysis showed that the most efficient overall classification between controls and individuals with MCI subjects was achieved with ERC measurements (65.9%). However, the best overall classification between controls and AD patients (90.7%), and between individuals with MCI and AD patients (82.3%) was achieved with hippocampal volumes. Our results suggest that the ERC atrophy precedes hippocampal atrophy in AD. The ERC volume loss is dominant over the hippocampal volume loss in MCI, whereas more pronounced hippocampal volume loss appears in mild AD.     

Quantitative electroencephalography in mild cognitive impairment: longitudinal changes and possible prediction of Alzheimer's disease

The present study evaluated the clinical course of patients with mild cognitive impairment (MCI), the pattern of electroencephalography (EEG) changes following cognitive deterioration, as well as the potential of neurophysiological measures in predicting dementia. Twenty-seven subjects with MCI were followed for a mean follow up period of 21 months. Fourteen subjects (52%) progressed (P MCI) to clinically manifest Alzheimer's disease (AD), and 13 (48%) remained stable (S MCI). The two MCI subgroups did not differ in baseline EEG measures between each other and the healthy controls (n = 16), but had significantly lower theta relative power at left temporal, temporo-occipital, centro-parietal, and right temporo-occipital derivation when compared to the reference AD group (n = 15). The P MCI baseline alpha band temporo-parietal coherence, alpha relative power values at left temporal and temporo-occipital derivations, theta relative power values at frontal derivations, and the mean frequency at centro-parietal and temporo-occipital derivations overlapped with those for AD and control groups. After the follow-up, the P MCI patients had significantly higher theta relative power and lower beta relative power and mean frequency at the temporal and temporo-occipital derivations. A logistic regression model of baseline EEG values adjusted for baseline Mini-Mental Test Examination showed that the important predictors were alpha and theta relative power and mean frequency from left temporo-occipital derivation (T5-O1), which classified 85% of MCI subjects correctly.     

MCI patients’ EEGs show group differences between those who progress and those who do not progress to AD

The theta/gamma and alpha3/alpha2 ratio were investigated as early markers for prognosticating of progression to dementia. 76 subjects with mild cognitive impairment (MCI) underwent EEG recording, MRI scans and neuropsychological (NPS) tests. After 3 years of follow-up, three subgroups were characterized as converters to Alzheimer's disease (AD, N = 18), converters to non-AD dementia (N = 14) and non-converters (N = 44). The theta/gamma and alpha3/alpha2 ratio, performance on cognitive tests and hippocampal volume, as evaluated at the time of initial MCI diagnosis, were studied in the three groups. As expected, MCI to AD converters had the smallest mean hippocampal volume and poorest performance on verbal learning tests, whereas MCI to non-AD converters had poorest cognitive performance in non-verbal learning tests, abstract thinking, and letter fluency. Increased theta/gamma ratio was associated with conversion to both AD and non-AD dementia; increased alpha3/alpha2 ratio was only associated with conversion to AD.Theta/gamma and alpha3/alpha2 ratio could be promising prognostic markers in MCI patients. In particular, the increase of high alpha frequency seems to be associated with conversion in AD. EEG markers allow a mean correct percentage of correct classification up to 88.3%. Future prospective studies are needed to evaluate the sensitivity and specificity of these measures for predicting an AD outcome.     

How early can we predict Alzheimer's disease using computational anatomy?

Computational anatomy with magnetic resonance imaging (MRI) is well established as a noninvasive biomarker of Alzheimer's disease (AD); however, there is less certainty about its dependency on the staging of AD. We use classical group analyses and automated machine learning classification of standard structural MRI scans to investigate AD diagnostic accuracy from the preclinical phase to clinical dementia. Longitudinal data from the Alzheimer's Disease Neuroimaging Initiative were stratified into 4 groups according to the clinical status—(1) AD patients; (2) mild cognitive impairment (MCI) converters; (3) MCI nonconverters; and (4) healthy controls—and submitted to a support vector machine. The obtained classifier was significantly above the chance level (62%) for detecting AD already 4 years before conversion from MCI. Voxel-based univariate tests confirmed the plausibility of our findings detecting a distributed network of hippocampal-temporoparietal atrophy in AD patients. We also identified a subgroup of control subjects with brain structure and cognitive changes highly similar to those observed in AD. Our results indicate that computational anatomy can detect AD substantially earlier than suggested by current models. The demonstrated differential spatial pattern of atrophy between correctly and incorrectly classified AD patients challenges the assumption of a uniform pathophysiological process underlying clinically identified AD.     

Resting EEG Discrimination of Early Stage Alzheimer’s Disease from Normal Aging Using Inter-Channel Coherence Network Graphs

Amnestic mild cognitive impairment (MCI) is a degenerative neurological disorder at the early stage of Alzheimer’s disease (AD). This work is a pilot study aimed at developing a simple scalp-EEG-based method for screening and monitoring MCI and AD. Specifically, the use of graphical analysis of inter-channel coherence of resting EEG for the detection of MCI and AD at early stages is explored. Resting EEG records from 48 age-matched subjects (mean age 75.7 years)—15 normal controls (NC), 16 with early-stage MCI, and 17 with early-stage AD—are examined. Network graphs are constructed using pairwise inter-channel coherence measures for delta–theta, alpha, beta, and gamma band frequencies. Network features are computed and used in a support vector machine model to discriminate among the three groups. Leave-one-out cross-validation discrimination accuracies of 93.6% for MCI vs. NC (p < 0.0003), 93.8% for AD vs. NC (p < 0.0003), and 97.0% for MCI vs. AD (p < 0.0003) are achieved. These results suggest the potential for graphical analysis of resting EEG inter-channel coherence as an efficacious method for noninvasive screening for MCI and early AD.     

Classification of diffusion tensor images for the early detection of Alzheimer's disease

Early detection of Alzheimer's disease (AD) is important since treatments are more efficacious when used at the beginning of the disease. Despite significant advances in diagnostic methods for AD, there is no single diagnostic method for AD with high accuracy. We developed a support vector machine (SVM) model that classifies mild cognitive impairment (MCI) and normal control subjects using probabilistic tractography and tract-based spatial statistics of diffusion tensor imaging (DTI) data. MCI is an intermediate state between normal aging and AD, so finding MCI is important for an early diagnosis of AD. The key features of DTI data we identified through extensive analysis include the fractional anisotropy (FA) values of selected voxels, their average FA value, and the volume of fiber pathways from a pre-defined seed region. In particular, the volume of the fiber pathways to thalamus is the most powerful single feature in classifying MCI and normal subjects regardless of the age of the subjects. The best performance achieved by the SVM model in a 10-fold cross validation and in independent testing was sensitivity of 100%, specificity of 100% and accuracy of 100%.     

基于MR图像纹理特征的阿尔茨海默病分类模型

目的基于MR图像,提取脑部海马区域纹理特征参数建立阿尔茨海默病(Alzheimer disease,AD)的早期分类预测模型。方法研究数据来源于美国国立老年研究所ADNI数据库,收集研究对象的磁共振(magnetic resonance,MR)脑图像,分别基于左、右和双侧海马图像,通过区域增长法和Contourlet变换提取纹理特征参数,结合研究对象的基本信息作为特征变量采用高斯过程分类方法建立AD患者和健康对照的诊断模型以及轻度认知障碍(mild cognitive impairment,MCI)患者转变为AD的预测模型,并评价模型的灵敏度、特异度以及ROC曲线下面积。结果研究共纳入420例研究对象。基于AD和健康对照两组构建的分类模型,双侧海马区的灵敏度、特异度以及ROC曲线下面积分别为92.7%、87.1%和0.922,均大于基于左侧或右侧海马区图像建立的模型。基于MCI数据建立的AD早期预测模型中,灵敏度最高为82.4%,ROC曲线下面积最高为0.836。结论基于脑部海马区的Contourlet纹理特征构建预测模型,可以识别AD早期的病变情况,这将有助于早期监测MCI进展为AD,为减缓和治疗AD发病提供依据。     

Working memory load-related electroencephalographic parameters can differentiate progressive from stable mild cognitive impairment

Recent studies described several changes of endogenous event-related potentials (ERP) and brain rhythm synchronization during memory activation in patients with Alzheimer's disease (AD). To examine whether memory-related EEG parameters may predict cognitive decline in mild cognitive impairment (MCI), we assessed P200 and N200 latencies as well as beta event-related synchronization (ERS) in 16 elderly controls (EC), 29 MCI cases and 10 patients with AD during the successful performance of a pure attentional detection task as compared with a highly working memory demanding two-back task. At 1 year follow-up, 16 MCI patients showed progressive cognitive decline (PMCI) and 13 remained stable (SMCI). Both P200 and N200 latencies in the two-back task were longer in PMCI and AD cases compared with EC and SMCI cases. During the interval 1000 ms to 1700 ms after stimulus, beta ERS at parietal electrodes was of lower amplitude in PMCI and AD compared with EC and SMCI cases. Univariate models showed that P200, N200 and log% beta values were significantly related to the SMCI/PMCI distinction with areas under the receiver operating characteristic curve of 0.93, 0.78 and 0.72, respectively. The combination of all three EEG hallmarks was the stronger predictor of MCI deterioration with 90% of correctly classified MCI cases. Our data reveal that PMCI and clinically overt AD share the same pattern of working memory-related EEG activation characterized by increased P200-N200 latencies and decreased beta ERS. They also show that P200 latency during the two-back task may be a simple and promising EEG marker of rapid cognitive decline in MCI.     

Brain oscillatory responses to an auditory-verbal working memory task in mild cognitive impairment and Alzheimer's disease.

We report preliminary findings on EEG oscillatory correlates of working memory in mild cognitive impairment (MCI) and Alzheimer's disease (AD). Event-related desynchronization (ERD) and synchronization (ERS) of the 1-20 Hz EEG frequencies were studied using wavelet transforms in elderly controls, MCI patients and mild probable AD patients performing an auditory-verbal Sternberg memory task. Behaviourally, the AD patients made more errors than the controls and the MCI group. Statistically significant differences during the encoding of the memory set were found between the controls and the MCI group, such that the latter group showed ERD in the approximately 10-20 Hz frequencies. The findings may reflect different, compensatory encoding strategies in MCI. During retrieval, the most obvious differences were observed between the controls and the AD group: the ERD in the approximately 7-17 Hz frequencies was absent in the AD group particularly in anterior and left temporal electrode locations. This finding might indicate that AD is associated with deficient lexical-semantic processing during the retrieval phase in working memory tasks. Future studies with larger patient groups are needed to establish the diagnostic value of ERD/ERS patterns in MCI and AD.     

Regional Magnetic Resonance Imaging Measures for Multivariate Analysis in Alzheimer’s Disease and Mild Cognitive Impairment

Automated structural magnetic resonance imaging (MRI) processing pipelines are gaining popularity for Alzheimer’s disease (AD) research. They generate regional volumes, cortical thickness measures and other measures, which can be used as input for multivariate analysis. It is not clear which combination of measures and normalization approach are most useful for AD classification and to predict mild cognitive impairment (MCI) conversion. The current study includes MRI scans from 699 subjects [AD, MCI and controls (CTL)] from the Alzheimer’s disease Neuroimaging Initiative (ADNI). The Freesurfer pipeline was used to generate regional volume, cortical thickness, gray matter volume, surface area, mean curvature, gaussian curvature, folding index and curvature index measures. 259 variables were used for orthogonal partial least square to latent structures (OPLS) multivariate analysis. Normalisation approaches were explored and the optimal combination of measures determined. Results indicate that cortical thickness measures should not be normalized, while volumes should probably be normalized by intracranial volume (ICV). Combining regional cortical thickness measures (not normalized) with cortical and subcortical volumes (normalized with ICV) using OPLS gave a prediction accuracy of 91.5 % when distinguishing AD versus CTL. This model prospectively predicted future decline from MCI to AD with 75.9 % of converters correctly classified. Normalization strategy did not have a significant effect on the accuracies of multivariate models containing multiple MRI measures for this large dataset. The appropriate choice of input for multivariate analysis in AD and MCI is of great importance. The results support the use of un-normalised cortical thickness measures and volumes normalised by ICV.     

Mild cognitive impairment as predictor for Alzheimer's disease in clinical practice: effect of age and diagnostic criteria

BACKGROUND: We investigated whether the predictive accuracy of mild cognitive impairment (MCI) for Alzheimer-type dementia (AD) in a clinical setting is dependent on age and the definition of MCI used. METHOD: Non-demented subjects older than 40 (n=320) who attended a memory clinic of a university hospital were reassessed 5 years later for the presence of AD. MCI was diagnosed according to the criteria of amnestic MCI, mild functional impairment (MFI), ageing-associated cognitive decline (AACD), and age-associated memory impairment (AAMI). The main outcome measure was the area under the curve (AUC) of a receiver operating characteristic (ROC) curve. Analyses were conducted on the entire sample and on subgroups of subjects aged 40-54, 55-69 and 70-85 years. RESULTS: A diagnosis of AD at follow-up was made in 58 subjects. Four of them were in the 40-54 age group, 29 in the 55-69 age group and 25 in the 70-85 age group. The diagnostic accuracy in the entire sample was low to moderately high with AUCs ranging from 0.56 (AACD) to 0.75 (amnestic MCI). A good predictive accuracy with an AUC >0.80 was only observed in subjects aged 70-85 using the criteria of amnestic MCI (AUC=0.84). CONCLUSIONS: The predictive accuracy of MCI for AD is dependent on age and the definition of MCI used. The predictive accuracy is good only for amnestic MCI in subjects 70-85 years. As subjects with prodromal AD are often younger than 70, the usefulness of MCI as predictor of AD in clinical practice is limited.     

Regions with different evoked frequency band responses during early-stage visual processing distinguish mild Alzheimer dementia from mild cognitive impairment and normal aging

Although diagnostic procedures have been developed for detection of mild cognitive impairment (MCI) and mild Alzheimer dementia (AD), more valid noninvasive tools are needed. In this work, we apply a procedure based on the evidences that different evoked frequency band responses may emerge from different sources during early-stage visual processing in a mental state-specific manner, while subjects were passively viewing a visual stimulus. In this case, spatial differences should arise across mental conditions such as mild Alzheimer dementia, mild cognitive impairment, and normal aging. With the use of EEG source image we found three different neural patterns in aged individuals: (1) left hippocampus and midbrain in mild AD, (2) left lateral orbitofrontal gyrus, left nucleus accumbens, caudate nucleus, thalamus, posterior cinguli, right precuneous, right superior parietal lobe in MCI, and (3) right lateral-medial orbitofrontal gyrus, caudate nucleus, thalamus, right lateral occipitotemporal gyrus in elderly controls. Although preliminary, these results show remarkably robust differences that distinguish between an age-matched control group, a group with MCI, and a group with mild AD. Because the method applied in this work differentiates among clinical entities with varying severity of cognitive decline, it may eventually serve as an electrophysiological marker in the early detection of neurodegeneration.     

轻度认知障碍老年人脑电的非线性相互依赖脑网络分析

本研究旨在初步探究患轻度认知障碍(MCI)的老年人及正常老年人进行情绪调控时的大脑网络特性.记录了14位MCI患者及18个正常人进行认知重评的情绪调控实验时的头皮脑电信号(EEG),包括中性观看、负性观看和负性重评任务.并采用非线性相互依赖性的方法衡量不同频段的大脑区域间的连接强度,然后根据非线性相互依赖性指数构建两组...