Raheel Zafar

Decoding of visual information from human brain activity: A review of fMRI and EEG studies.

Brain is the command center for the body and contains a lot of information which can be extracted by using different non-invasive techniques. Electroencephalography (EEG), Magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) are the most common neuroimaging techniques to elicit brain behavior. By using these techniques different activity patterns can be measured within the brain to decode the content of mental processes especially the visual and auditory content. This paper discusses the models and imaging techniques used in visual decoding to investigate the different conditions of brain along with recent advancements in brain decoding. This paper concludes that its not possible to extract all the information from the brain, however careful experimentation, interpretation and powerful statistical tools can be used with the neuroimaging techniques for better results.

Decoding of visual activity patterns from fMRI responses using multivariate pattern analyses and convolutional neural network

Decoding of human brain activity has always been a primary goal in neuroscience especially with functional magnetic resonance imaging (fMRI) data. In recent years, Convolutional neural network (CNN) has become a popular method for the extraction of features due to its higher accuracy, however it needs a lot of computation and training data. In this study, an algorithm is developed using Multivariate pattern analysis (MVPA) and modified CNN to decode the behavior of brain for different images with limited data set. Selection of significant features is an important part of fMRI data analysis, since it reduces the computational burden and improves the prediction performance; significant features are selected using t-test. MVPA uses machine learning algorithms to classify different brain states and helps in prediction during the task. General linear model (GLM) is used to find the unknown parameters of every individual voxel and the classification is done using multi-class support vector machine (SVM). MVPA-CNN based proposed algorithm is compared with region of interest (ROI) based method and MVPA based estimated values. The proposed method showed better overall accuracy (68.6%) compared to ROI (61.88%) and estimation values (64.17%).

EEG spectral analysis during complex cognitive task at occipital

Normal oscillations in different frequency bands have an important role in cognitive processing in the frontal region. The key issue is whether frequency oscillations of Electroencephalography (EEG) are related to cognitive task or not in occipital region. All frequency bands delta (δ), theta (θ), alpha (α), beta (β) and gamma (γ) are involved in brain tasks. This study is conducted to investigate the functional relationship between EEG frequency bands and the cognitive task. In various studies theta, alpha and beta bands are discussed for cognitive tasks; however there are few studies which have been focused on delta band for cognitive tasks. In this paper, behavior of θ, α, β and γ is described but the primary focus is on delta (δ) band during cognitive task in occipital region as it had been ignored in the literature. In conclusion, this study explains how different frequencies change during cognitive task as compared to base line (eyes open) in occipital region and the results show that there is an increase in power during the task in delta and theta band.

Electroencephalogram-based decoding cognitive states using convolutional neural network and likelihood ratio based score fusion

Electroencephalogram (EEG)-based decoding human brain activity is challenging, owing to the low spatial resolution of EEG. However, EEG is an important technique, especially for brain–computer interface applications. In this study, a novel algorithm is proposed to decode brain activity associated with different types of images. In this hybrid algorithm, convolutional neural network is modified for the extraction of features, a t-test is used for the selection of significant features and likelihood ratio-based score fusion is used for the prediction of brain activity. The proposed algorithm takes input data from multichannel EEG time-series, which is also known as multivariate pattern analysis. Comprehensive analysis was conducted using data from 30 participants. The results from the proposed method are compared with current recognized feature extraction and classification/prediction techniques. The wavelet transform-support vector machine method is the most popular currently used feature extraction and prediction method. This method showed an accuracy of 65.7%. However, the proposed method predicts the novel data with improved accuracy of 79.9%. In conclusion, the proposed algorithm outperformed the current feature extraction and prediction method.

Discriminating the different human brain states with EEG signals using Fractal dimension: A nonlinear approach

EEG signals are measured on scalp of the human brain and are widely used to address the clinical as well as in modern application like brain computer interfacing (BCI) and gaming. Feature extraction plays a fundamental role for good classification purposes. EEG features commonly extracted are linear as well as nonlinear. Nonlinear approaches are used when the complexity of EEG signals increases. Nonlinear features like correlation dimension (CD), Lyapunov exponents, approximate entropy requires higher computational complexity. On other hand Fractal dimension (FD) requires less computations. Therefore, Fractal dimension are widely used in engineering and biological sciences. In our paper, Fractal dimension has been selected to discriminate the different brain states. EEG data from 08 healthy participants have been acquired during eyes open, eyes close and during IQ task. Fractal dimensions have been computed on the EEG data acquired. Using Fractal dimension, we have successfully discriminated the different brain conditions/states like eyes open, eyes close and IQ task. Results have shown better discrimination between mental task and active brain conditions with 91.66 % accuracy using SVM classifier as compared to other classifiers. This approach can be used for fast decision making and pattern matching based on the selected epoch of the EEG signal using nonlinear approach.

Prediction of Human Brain Activity Using Likelihood Ratio Based Score Fusion

Human brain has a complex structure with the billions of neurons, so it is a difficult and challenging task to predict the behavior of human brain. Different methods and classifiers are used to measure and classify the brain activities with higher accuracy and reliability. In this paper, instead of using mostly used classifier (support vector machine), prediction of the brain activity is done by estimating the match score densities. This method is based on likelihood ratio test which helps in finding the optimal combination of match scores. The distributions of match scores are modeled for different classes based on density score fusion in which the densities of different classes are estimated from the training data set and match scores are found by fusing the estimated densities with the testing data. The fusion is done with the data extracted from distributed activation patterns using multivariate pattern analysis (MVPA) against a visual task. MVPA is an intense strategy which helps in better understanding of the human brain. The match score-based technique is used in different biometric systems but never been used for the prediction of brain activity. In order to test the performance of proposed method, prediction accuracy is compared with the support vector machine using two data sets of different modalities, one is electroencephalography (EEG) and the other is functional magnetic resonance imaging (fMRI). The results show that the proposed method predicts the novel data with improved accuracy of 66.1% and 69.3% compared with support vector machine which have 64.15% and 65.7% for fMRI and EEG data sets, respectively.

Role of voxel selection and ROI in fMRI data analysis

Functional magnetic resonance imaging (fMRI) is one of the most popular and reliable modality to measure brain activities. The quality of fMRI data is best among other modalities such as Electroencephalography (EEG) and Magnetoencephalography (MEG). In fMRI, normally number of features are more than the number of instances so it is necessary to select the features and do dimension reduction to remove noisy and redundant data. Many techniques and methods are used to select the significant features (voxels). In this paper, the significant voxels are selected within the anatomical region of interest (ROI) based on the absolute values. In this study, we have predicted the brain states using two machine learning algorithm, i.e, Radial basis function (RBF) network and Naïve Bayes. A visual experiment with two categories is done. In conclusion, it is shown that less number of voxels and specific brain regions can increase the accuracy of prediction.

Importance of realignment parameters in fMRI data analysis

Functional magnetic resonance imaging (fMRI) is one of the finest modality to measure brain activity. Two main steps in the analysis of fMRI data are pre-processing and the statistical analysis. Pre-processing is equally an important part because it takes raw data from the scanner and prepares it for the statistical analysis. This study first explains the realignment during preprocessing and then the importance of realignment parameters (one of nuisance parameters) in General Linear model (GLM). Nuisance regressors are used to reduce noise only and are effect of no interest. In this study, it is concluded that realignment parameters have a significant effect in the model estimation because the results are improved with these parameters especially when large head movement is found during data acquisition.

Discrimination of Brain States Using Wavelet and Power Spectral Density

Cognitive task produces activation in the brain which are different from normal state. In order to study the brain behavior during cognitive state, different techniques are available. Wavelet energy and power spectral density (PSD) are well established methods for brain signal classification. In this paper, cognitive state of the brain is compared with the baseline using EEG. Data are taken from all lobes of the brain to see the effect of cognitive task in the whole brain and analyzed using wavelet energy and PSD. Graph of wavelet energy and power spectral density are plotted separately for each subject to see the effect individually. Individual results showed that the behavior of human brain change with the cognitive task and this change occurred in most of the human brain. This change is due to the neural activity which is increased during the cognitive task (IQ) and is better measured with wavelet compared to PSD.

Multiple trials in event related fMRI for different conditions

Experiment design has a key role in the functional magnetic resonance imaging (fMRI) data analyses. Block designs are suitable to localize functional areas but are not able to measure the transient changes in the brain activity. Event related design is a better approach and saves time and resources like single trial analyses. In this study, we explored the event related design with single, and multi trials with different order. In multi trials, instead of using lot of trials, we did analyses with two trials per image. The result suggest that the combination of multiple trials, order of trials and selection of significant voxels can give better results in terms of classification accuracy. Moreover, single and two trials per image saves resources as compared to many trials.