Delaram Jarchi

Regional coherence evaluation in mild cognitive impairment and Alzheimer's disease based on adaptively extracted magnetoencephalogram rhythms

This study assesses the connectivity alterations caused by Alzheimers disease (AD) and mild cognitive impairment (MCI) in magnetoencephalogram (MEG) background activity. Moreover, a novel methodology to adaptively extract brain rhythms from the MEG is introduced. This methodology relies on the ability of empirical mode decomposition to isolate local signal oscillations and constrained blind source separation to extract the activity that jointly represents a subset of channels. Inter-regional MEG connectivity was analysed for 36 AD, 18 MCI and 26 control subjects in δ, θ, α and β bands over left and right central, anterior, lateral and posterior regions with magnitude squared coherence-c(f). For the sake of comparison, c(f) was calculated from the original MEG channels and from the adaptively extracted rhythms. The results indicated that AD and MCI cause slight alterations in the MEG connectivity. Computed from the extracted rhythms, c(f) distinguished AD and MCI subjects from controls with 69.4% and 77.3% accuracies, respectively, in a full leave-one-out cross-validation evaluation. These values were higher than those obtained without the proposed extraction methodology.

A robust approach for optimization of the measurement matrix in Compressed Sensing

In this paper we address the problem of measurement matrix optimization in Compressed Sensing (CS) framework. Although the measurement matrix is generally selected randomly, some methods have been recently proposed to optimize it. It is shown that the optimized matrices can improve the quality of reconstruction and satisfy the required conditions for an efficient sampling. We propose a new optimization method with the aim of decreasing the “Mutual Coherence”. Defining a new cost function, we suggest to use a Gradient descent algorithm for this optimization problem. The advantages are less computational complexity, which makes the method suitable for large-scale problems, more robustness, and higher incoherence between the measurement matrix and sparsifying matrix (dictionary). By conducting several experiments, we obtained promising results which confirm a considerable improvement compared to those achieved by other methods.

A New Spatiotemporal Filtering Method for Single-Trial Estimation of Correlated ERP Subcomponents

A novel spatiotemporal filtering method for single trial estimation of event-related potential (ERP) subcomponents is proposed here. Unlike some previous works in ERP estimation, the proposed method is able to estimate temporally correlated ERP subcomponents such as P3a and P3b. A new cost function is, therefore, defined which can deflate one of the correlated subcomponents. The method is applied to both simulated and real data and has shown to perform very well even in low signal-to-noise ratio situations. In addition, the method is compared to spatial principal component analysis and its superiority has been confirmed by using simulated signals. The approach can be especially useful in mental fatigue analysis where the relative variability of P300 subcomponents is the key factor in detecting the level of fatigue.

Coupled particle filtering: A new approach for P300-based analysis of mental fatigue

A new method for investigating mental fatigue based on P300 variability is presented here. In this approach a new coupled particle filtering for tracking variability of P300 subcomponents, i.e., P3a and P3b, across trials is developed. The latency, amplitude, and width of each subcomponent, as the main varying parameters, are modelled using state space system. In this model the observation is modelled as a linear function of amplitude and a nonlinear function of latency and width. Two Rao-blackwellised particle filters are then coupled and employed for recursive estimation of the state of the system across trials. By including some physiological based constraints, the proposed technique prevents generation of invalid particles during estimation of the state of the system. The main advantage of the algorithm compared with other single trial based methods is its robustness in the low signal-to-noise ratio situations. The method is applied to both simulated data and real mental fatigue data. The results demonstrate potential use of the method in event-related potential (ERP) based applications.

Simultaneous localization and separation of biomedical signals by tensor factorization

In this paper, we introduce mathematical models based on multi-way data construction and analysis with a goal of simultaneously separating and localizing the sources in the brain by analysis of scalp electroencephalogram (EEG) data. we address the problem of EEG source separation and localization through a 3-way tensor analysis. We represent multi-channel EEG data using a third-order tensor with modes: space (channels), time samples and number of segments. Then we demonstrate that multi-way analysis techniques, in particular PARAFAC2, can successfully separate and localize disjoint sources within the brain. Also we used this method for separation of maternal and fetal ECG signals.

Seizure source localization using a hybrid second order blind identification and extended rival penalized competitive learning algorithm

Abstract Localization of seizure sources prior to neurosurgery is crucial. In this paper, a new method is proposed to localize the seizure sources from multi-channel electroencephalogram (EEG) signals. Blind source separation based on second order blind identification (SOBI) is primarily applied to estimate the brain source signals in each window of the EEG signals. A new clustering method based on rival penalized competitive learning (RPCL) is then developed to cluster the rows of the estimated unmixing matrices in all the windows. The algorithm also includes pre and post-processing stages. By multiplying each cluster center to the EEG signals, the brain signal sources are approximated. According to a complexity value measure, the main seizure source signal is separated from the others. This signal is projected back to the electrodes’ space and is subjected to the dipole source localization using a single dipole model. The simulation results verify the accuracy of the system. In addition, correct localization of the seizure source is consistent with the clinical tests derived using the simultaneous intracranial recordings.

Semi-blind signal separation and channel estimation in MIMO communication systems by tensor factorization

In this paper, we introduce a tensor-factorization method for signal detection in MIMO applications. We address the detection problem through a 3-way tensor analysis. We represent the 4 × 4 MIMO received signals as a third-order tensor with modes: receiver antennas, user data symbols at each packet, and finally number of packets. Then, we demonstrate that by multi-way analysis using PARAFAC2 we can successfully solve the blind MIMO signal detection problem. In order to solve the permutation and scaling ambiguities of the detected signals we used different M-Sequence training symbols are used. For evaluating the method we compared our BER results with those of MMSE-VBLAST signal detection method.

Estimation of trial to trial variability of P300 subcomponents by coupled Rao-blackwellised particle filtering

In this paper a new method based on Rao-blackwellised particle filtering for tracking variability of event related-potential (ERP) subcomponents in different trials is presented. The latency, amplitude, and width of each subcomponent is formulated in the state space model. Then, the observation is modeled as a linear function of amplitude and a nonlinear function of latency and width. The Rao-blackwellised particle filtering is then applied for recursive estimation of the state of the system in different trials. To prevent generation of some invalid particles and also to have a reliable estimation in every situation, using some prior knowledge about some ERP subcomponents, a coupled Rao-blackwellised particle filter is designed to detect variability of the desired ERP subcomponents. The method is applied to both simulated and real P300 data. The algorithm has the ability of tracking the variability of P300 subcomponents i.e. P3a and P3b, in single trials even in the low signal-to-noise ratio situations.

Separating and tracking ERP subcomponents by constrained particle filtering

In this paper a new method based on particle filtering for separating and tracking event related-potential (ERP) subcomponents in different trials is presented. The latency and amplitude of each ERP subcomponent is formulated in the state space model. Based on some knowledge about ERP subcomponents, a constraint on the state space variables is provided to prevent the generation of invalid particles and also make use of a small number of particles which are most effective especially in high dimensions. The method is applied on the simulated and real P300 data. The algorithm has the ability of tracking P300 subcomponents i.e. P3a and P3b, in single trials even in the low signal-to-noise ratio situations.

Seizure Source Selection via a Hybrid Second Order Blind Identification and Gap Statistic Algorithm

In this paper a new method for isolating main seizure source from multichannel scalp EEG is proposed. First, the number of active brain sources in each time frame is determined. Then, second order blind identification (SOBI) is applied to isolate brain sources in a number of signal segments. After applying SOBI to the signals during seizure, rows of resulted unmixing matrix arc clustered by k-means clustering method. The value of k is estimated via gap statistic method. By multiplying each cluster center to the electrode signals, the brain sources are extracted. The results showed that seizure source could be effectively separated from other brain sources.