Boreom Lee

Detection of Abnormal Living Patterns for Elderly Living Alone Using Support Vector Data Description

In this study, we developed an automated behavior analysis system using infrared (IR) motion sensors to assist the independent living of the elderly who live alone and to improve the efficiency of their healthcare. An IR motion-sensor-based activity-monitoring system was installed in the houses of the elderly subjects to collect motion signals and three different feature values, activity level, mobility level, and nonresponse interval (NRI). These factors were calculated from the measured motion signals. The support vector data description (SVDD) method was used to classify normal behavior patterns and to detect abnormal behavioral patterns based on the aforementioned three feature values. The simulation data and real data were used to verify the proposed method in the individual analysis. A robust scheme is presented in this paper for optimally selecting the values of different parameters especially that of the scale parameter of the Gaussian kernel function involving in the training of the SVDD window length, T of the circadian rhythmic approach with the aim of applying the SVDD to the daily behavior patterns calculated over 24 h. Accuracies by positive predictive value (PPV) were 95.8% and 90.5% for the simulation and real data, respectively. The results suggest that the monitoring system utilizing the IR motion sensors and abnormal-behavior-pattern detection with SVDD are effective methods for home healthcare of elderly people living alone.

Improved elimination of motion artifacts from a photoplethysmographic signal using a Kalman smoother with simultaneous accelerometry

A photoplethysmography (PPG) signal provides very useful information about a subjects hemodynamic status in a hospital or ubiquitous environment. However, PPG is very vulnerable to motion artifacts, which can significantly distort the information belonging to the PPG signal itself. Thus, the reduction of the effects of motion artifacts is an important issue when monitoring the cardiovascular system by PPG. There have been many adaptive techniques to reduce motion artifacts from PPG signals. In the present study, we compared a method based on the fixed-interval Kalman smoother with the usual adaptive filtering algorithms, e.g. the normalized least mean squares, recursive least squares and the conventional Kalman filter. We found that the fixed-interval Kalman smoother reduced motion artifacts from the PPG signal most effectively. Therefore, the use of the fixed-interval Kalman smoother can reduce motion artifacts in PPG, thus providing the most reliable information that can be deduced from the reconstructed PPG signals.

White matter neuroplastic changes in long-term trained players of the game of “Baduk” (GO): A voxel-based diffusion-tensor imaging study

Currently, one of the most challenging issues in modern neuroscience is learning-induced neural plasticity. Many researchers have identified activation-dependent structural brain plasticity in gray and white matter. The game of Baduk is known to require many cognitive processes, and long-term training in such processes would be expected to cause structural changes in related brain areas. We conducted voxel-based analyses of diffusion-tensor imaging (DTI) data and found that, compared to inexperienced controls, long-term trained Baduk players developed larger regions of white matter with increased fractional anisotropy (FA) values in the frontal, cingulum, and striato-thalamic areas that are related to attentional control, working memory, executive regulation, and problem-solving. In addition, inferior temporal regions with increased FA indicate that Baduk experts tend to develop a task-specific template for the game, as compared to controls. In contrast, decreased FA found in dorsolateral premotor and parietal areas indicate that Baduk experts were less likely than were controls to use structures related to load-dependent memory capacity. Right-side dominance in Baduk experts suggests that the tasks involved are mainly spatial processes. Altogether, long-term Baduk training appears to cause structural brain changes associated with many of the cognitive aspects necessary for game play, and investigation of the mechanism underpinning such changes might be helpful for improving higher-order cognitive capacities, such as learning, abstract reasoning, and self-control, which can facilitate education and cognitive therapies.

RRAM-based synapse for neuromorphic system with pattern recognition function

Feasibility of a high speed pattern recognition system using 1k-bit cross-point synaptic RRAM array and CMOS-based neuron chip has been experimentally demonstrated. Learning capability of a neuromorphic system comprising RRAM synapses and CMOS neurons has been confirmed experimentally, for the first time.

Electronic system with memristive synapses for pattern recognition

Memristive synapses, the most promising passive devices for synaptic interconnections in artificial neural networks, are the driving force behind recent research on hardware neural networks. Despite significant efforts to utilize memristive synapses, progress to date has only shown the possibility of building a neural network system that can classify simple image patterns. In this article, we report a high-density cross-point memristive synapse array with improved synaptic characteristics. The proposed PCMO-based memristive synapse exhibits the necessary gradual and symmetrical conductance changes, and has been successfully adapted to a neural network system. The system learns, and later recognizes, the human thought pattern corresponding to three vowels, i.e. /a /, /i /, and /u/, using electroencephalography signals generated while a subject imagines speaking vowels. Our successful demonstration of a neural network system for EEG pattern recognition is likely to intrigue many researchers and stimulate a new research direction.

LORETA imaging of P300 in schizophrenia with individual MRI and 128-channel EEG.

We investigated the characteristics of P300 generators in schizophrenics by using voxel-based statistical parametric mapping of current density images. P300 generators, produced by a rare target tone of 1500 Hz (15%) under a frequent nontarget tone of 1000 Hz (85%), were measured in 20 right-handed schizophrenics and 21 controls. Low-resolution electromagnetic tomography (LORETA), using a realistic head model of the boundary element method based on individual MRI, was applied to the 128-channel EEG. Three-dimensional current density images were reconstructed from the LORETA intensity maps that covered the whole cortical gray matter. Spatial normalization and intensity normalization of the smoothed current density images were used to reduce anatomical variance and subject-specific global activity and statistical parametric mapping (SPM) was applied for the statistical analysis. We found that the sources of P300 were consistently localized at the left superior parietal area in normal subjects, while those of schizophrenics were diversely distributed. Upon statistical comparison, schizophrenics, with globally reduced current densities, showed a significant P300 current density reduction in the left medial temporal area and in the left inferior parietal area, while both left prefrontal and right orbitofrontal areas were relatively activated. The left parietotemporal area was found to correlate negatively with Positive and Negative Syndrome Scale total scores of schizophrenic patients. In conclusion, the reduced and increased areas of current density in schizophrenic patients suggest that the medial temporal and frontal areas contribute to the pathophysiology of schizophrenia, the frontotemporal circuitry abnormality.

Neuromorphic speech systems using advanced ReRAM-based synapse

We demonstrate an advanced ReRAM based analog artificial synapse for neuromorphic systems. Nitrogen doped TiN/PCMO based artificial synapse is proposed to improve the performance and reliability of the neuromorphic systems by using simple identical spikes. For the first time, we develop fully unsupervised learning with proposed analog synapses which is illustrated with the help of auditory and electroencephalography (EEG) applications.

Enhancing the estimation of blood pressure using pulse arrival time and two confounding factors

A new method of blood pressure (BP) estimation using multiple regression with pulse arrival time (PAT) and two confounding factors was evaluated in clinical and unconstrained monitoring situations. For the first analysis with clinical data, electrocardiogram (ECG), photoplethysmogram (PPG) and invasive BP signals were obtained by a conventional patient monitoring device during surgery. In the second analysis, ECG, PPG and non-invasive BP were measured using systems developed to obtain data under conditions in which the subject was not constrained. To enhance the performance of BP estimation methods, heart rate (HR) and arterial stiffness were considered as confounding factors in regression analysis. The PAT and HR were easily extracted from ECG and PPG signals. For arterial stiffness, the duration from the maximum derivative point to the maximum of the dicrotic notch in the PPG signal, a parameter called TDB, was employed. In two experiments that normally cause BP variation, the correlation between measured BP and the estimated BP was investigated. Multiple-regression analysis with the two confounding factors improved correlation coefficients for diastolic blood pressure and systolic blood pressure to acceptable confidence levels, compared to existing methods that consider PAT only. In addition, reproducibility for the proposed method was determined using constructed test sets. Our results demonstrate that non-invasive, non-intrusive BP estimation can be obtained using methods that can be applied in both clinical and daily healthcare situations.

Generators of the gamma-band activities in response to rare and novel stimuli during the auditory oddball paradigm.

In this study, we report the cortical sources of the gamma-band activity emitted during the auditory oddball paradigm using the adaptive beamformer algorithm and non-parametric permutation test and then compare them with those of the P3a and P3b components. The results of the gamma-band current sources revealed that the same gamma-band activities are in process during the rare target and novelty task. In the low (30-55 Hz) gamma-band activity, the common sources were localized in the (inferior) anterior cingulate and adjacent inferior frontal cortex. In the high (65-85 Hz) gamma-band activity, the generator was represented in the superior frontal cortex. On the other hand, the P3a and P3b generators showed widespread distributions including the well-known fronto-parietal network [J. Polich, Theoretical overview of P3a and P3b, in: J. Polish (Ed.), Detection of Change: Event-Related Potential and fMRI Findings, Kluwer Academic Press, Boston, 2003, pp. 83-98]. In conclusion, the same frontal generators of gamma-band activities in the present study may be associated with the functions of attentional control for the binding of consecutive cognitive stages corresponding to earlier P3a and later P3b components, which have distinct source distributions except for some overlaps.

Multiclass Classification for the Differential Diagnosis on the ADHD Subtypes Using Recursive Feature Elimination and Hierarchical Extreme Learning Machine: Structural MRI Study.

The classification of neuroimaging data for the diagnosis of certain brain diseases is one of the main research goals of the neuroscience and clinical communities. In this study, we performed multiclass classification using a hierarchical extreme learning machine (H-ELM) classifier. We compared the performance of this classifier with that of a support vector machine (SVM) and basic extreme learning machine (ELM) for cortical MRI data from attention deficit/hyperactivity disorder (ADHD) patients. We used 159 structural MRI images of children from the publicly available ADHD-200 MRI dataset. The data consisted of three types, namely, typically developing (TDC), ADHD-inattentive (ADHD-I), and ADHD-combined (ADHD-C). We carried out feature selection by using standard SVM-based recursive feature elimination (RFE-SVM) that enabled us to achieve good classification accuracy (60.78%). In this study, we found the RFE-SVM feature selection approach in combination with H-ELM to effectively enable the acquisition of high multiclass classification accuracy rates for structural neuroimaging data. In addition, we found that the most important features for classification were the surface area of the superior frontal lobe, and the cortical thickness, volume, and mean surface area of the whole cortex.