Seif Eldawlatly
Ain Shams University, Computer and Systems Engineering, Faculty Member
Objective. By means of electrical stimulation of the visual system, visual prostheses provide promising solution for blind patients through partial restoration of their vision. Despite the great success achieved so far in this field, the... more
Objective. By means of electrical stimulation of the visual system, visual prostheses provide promising solution for blind patients through partial restoration of their vision. Despite the great success achieved so far in this field, the limited resolution of the perceived vision using these devices hinders the ability of visual prostheses users to correctly recognize viewed objects. Accordingly, we propose a deep learning approach based on generative adversarial networks (GANs), termed prosthetic vision GAN (PVGAN), to enhance object recognition for the implanted patients by representing objects in the field of view based on a corresponding simplified clip art version. Approach. To assess the performance, an axon map model was used to simulate prosthetic vision in experiments involving normally-sighted participants. In these experiments, four types of image representation were examined. The first and second types comprised presenting phosphene simulation of real images containing t...
Research Interests:
Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disease that affects the nervous system causing muscle weakness, paralysis, leading to death. Given that abnormalities in spinal motoneuron (MN) excitability begin long... more
Amyotrophic Lateral Sclerosis (ALS) is a fatal neurodegenerative disease that affects the nervous system causing muscle weakness, paralysis, leading to death. Given that abnormalities in spinal motoneuron (MN) excitability begin long before symptoms manifest, developing an approach that could recognize fluctuations in MN firing could help in early diagnosis of ALS. This paper introduces a machine learning approach to discriminate between ALS and normal MN firing. The approach is based on two electrophysiological markers; namely, spiking latency and the spike-triggered average signal. The method is examined using data generated from a computational model under systematic variation of MN properties. Such variations mimic the differential dynamic changes in cellular properties that different MN types experience during ALS progression. Our results demonstrate the ability of the approach to accurately recognize ALS firing patterns across the spectrum of examined variations in MN properties.Clinical Relevance— These results represent a proof of concept for using the proposed machine-learning approach in early diagnosis of ALS.
Research Interests:
Brain-Computer Interface (BCI) is an interface that allows controlling computers using brain activity. BCI is capable of providing control commands that are hands-free and voice-free to increase the social inclusion of disabled people and... more
Brain-Computer Interface (BCI) is an interface that allows controlling computers using brain activity. BCI is capable of providing control commands that are hands-free and voice-free to increase the social inclusion of disabled people and to substitute typical devices. One common BCI paradigm is the Steady-state Visually Evoked Potentials (SSVEPs), which represent the response of the brain to flickering visual stimuli. This study proposes an ensemble classification algorithm to optimize the recognition of visual stimuli corresponding to the recorded SSVEPs. The algorithm considers the predictions of multiple classification and pre-processing approaches by taking their votes. Brain signals vary, not only from subject to subject, but also from one day to another for the same subject. Thus, most research work considers subject-dependent analysis. In this study, we evaluate the proposed approach both individually within each subject and across subjects. The algorithm is also tested for different stimulation times. The results demonstrate that for subject-dependent testing, the ensemble classification technique outperforms other approaches reaching a mean accuracy of 96.73% and 93.45% when using 5 sec and 2.5 sec stimulation times, respectively. For subject-independent testing, the ensemble classification technique achieved a mean accuracy of 92.18 % and 76.91% when using 5 sec and 2.5 sec stimulation times, respectively. These results indicate the utility of the proposed approach in enhancing the performance of SSVEP-based BCIs.
Research Interests:
Myoelectric control of prosthetic arms provides a new hope for providing naturalistic movements to amputees. Extensive work has been made in recent years to use Electromyography (EMG) signals to enhance the operation of prosthetic arms.... more
Myoelectric control of prosthetic arms provides a new hope for providing naturalistic movements to amputees. Extensive work has been made in recent years to use Electromyography (EMG) signals to enhance the operation of prosthetic arms. In this paper, we propose an EMG Kalman filter-based model, where we identify the relationship between the joint angles and recorded EMG signals. EMG signals were recorded from biceps and triceps muscles and used to train a Kalman filter decoder. We assessed the performance of the decoder by computing the correlation and the normalized root mean-square error (NRMSE) between the decoded and actual joint angles. When decoding using biceps EMG only, an average correlation of 0.61 was obtained with a NRMSE of 0.35. For triceps EMG only, an average correlation of 0.5 was obtained with a NRMSE of 0.5. Finally, when decoding the EMG of both the biceps and triceps muscles, the average correlation increased to 0.87 while the average NRMSE decreased to 0.18. These results outperform recent studies in similar applications which indicates the efficacy of the proposed decoder in decoding joint angles from recorded EMG. This could help in enhancing the control of prosthetic arms.
Research Interests:
Motor imagery represents one Brain-Computer Interface (BCI) paradigm that has been utilized in developing applications to assist subjects with motor disability. Such paradigm relies on analyzing brain electroencephalography (EEG) activity... more
Motor imagery represents one Brain-Computer Interface (BCI) paradigm that has been utilized in developing applications to assist subjects with motor disability. Such paradigm relies on analyzing brain electroencephalography (EEG) activity to identify the intended movement direction. Existing motor imagery feature extraction techniques are focused on utilizing traditional signal processing and machine learning techniques. Recent advances in the deep learning field has inspired the development of few methods for motor imagery classification that achieved further performance improvement. This paper proposes a deep neural network approach for motor imagery classification using Long Short-Term Memory (LSTM) combined with Autoencoders based on a sequence-to-sequence architecture. The proposed network extracts features from the frequency-domain representation of EEG signals. This network is trained to obtain low-dimensional representation of EEG features that are then fed into multilayer perceptron of 3 layers for classification. Systematic and extensive examinations have been carried out by applying the approach to public benchmark EEG datasets. The obtained results outperformed classical state-of-the-art methods employing standard frequency-domain features and common spatial patterns, and comparative results to methods such as filter bank common spatial pattern and its variants. Our results indicate the efficacy of the proposed LSTM autoencoder approach in EEG motor imagery classification.
Research Interests:
Emotional distress during driving can greatly affect the safety and comfort of the driver. Being able to detect and react to the emotions of the driver would greatly improve in-car safety. It could also be utilized in a variety of... more
Emotional distress during driving can greatly affect the safety and comfort of the driver. Being able to detect and react to the emotions of the driver would greatly improve in-car safety. It could also be utilized in a variety of different applications to improve the driving experience. In this paper, we introduce a brain signal-operated emotion recognition system that is specifically tailored for the Automotive Open System Architecture (AUTOSAR) framework. The proposed system acquires brain electroencephalography (EEG) signals of the driver, identifies the underlying emotion using machine learning techniques, and feeds that emotion into the car system where different car components can react to that input. Our results demonstrate the ability of the system to recognize two emotions, namely sadness versus happiness, from the recorded EEG with a mean accuracy of 89.7% across three examined subjects using subject-dependent data. Moreover, when training the system using data recorded from multiple subjects, a mean accuracy of 91.7% is achieved. Taken together, these results indicate the ability of the proposed approach to discriminate between sadness and happiness whose extreme expression could have a significant impact on driving behavior.
Research Interests:
Utilizing brain activity to interact with the external environment is no longer impossible thanks to recent advances in developing Brain-Computer Interfaces (BCIs). This paper proposes a novel web browser application that utilizes... more
Utilizing brain activity to interact with the external environment is no longer impossible thanks to recent advances in developing Brain-Computer Interfaces (BCIs). This paper proposes a novel web browser application that utilizes user's Steady-State Visual Evoked Potentials (SSVEPs) from electroencephalography (EEG) to surf the internet. Development of a BCI web browsing application is challenging as it has to provide a comfortable web navigation using an easy-to-use interface, with the highest accuracy of command detection and correct execution. In addition, the response time needs to be as small as possible. The developed web browser was examined on three healthy subjects using the Emotiv EPOC headset to record EEG signals, and demonstrated to be fully functional in web navigation tasks with an average accuracy across subjects of 86.08 ± 15.46%.
Research Interests:
Research Interests:
Visual prostheses open the door of hope to restore functional vision for the blind. One of the main challenges facing their development is the limited number of electrodes used in the stimulation process which limits the resolution of the... more
Visual prostheses open the door of hope to restore functional vision for the blind. One of the main challenges facing their development is the limited number of electrodes used in the stimulation process which limits the resolution of the perceived images. To improve the perception, the useful features in the scene need to be enhanced while the other features should be suppressed to achieve better resolution. This paper introduces an image processing method to enhance three main features detectable by the natural visual pathway; namely the contrast, the motion and the edges. It then reduces the size of the image into an activity matrix used to generate the electric stimulation for the electrodes array. We compared the proposed method to four other image processing strategies in terms of the quality of the resulting image in addition to the perceived image using a simulation of prosthetic vision. Results demonstrate that the proposed method outperforms the other techniques in both aspects.
Research Interests:
Predicting the occurrence of epileptic seizures can provide an enormous aid to epileptic patients. This paper introduces a novel patient-specific method for seizure prediction applied to scalp Electroencephalography (EEG) signals. The... more
Predicting the occurrence of epileptic seizures can provide an enormous aid to epileptic patients. This paper introduces a novel patient-specific method for seizure prediction applied to scalp Electroencephalography (EEG) signals. The proposed method relies on the count of zero-crossings of wavelet detail coefficients of EEG signals as the major feature. This is followed by a binary classifier that discriminates between preictal and interictal states. The proposed method is practical for real-time applications given its computational efficiency as it uses an adaptive algorithm for channel selection to identify the optimum number of needed channels. Moreover, this method is robust against the variability across seizures for the same patient. Applied to data from 8 patients, the proposed method achieved high accuracy and sensitivity with an average accuracy of 94% and an average sensitivity of 96%. These results were obtained using only 10 minutes of training data as opposed to using hours of recordings typically used in traditional approaches.
Research Interests:
Research Interests:
Advances in Brain-Computer Interfaces (BCIs) have made BCIs come in use mainly for the disabled to communicate. Practical usage of BCIs requires that the whole BCI system be portable so that disabled subjects can use them anywhere. The... more
Advances in Brain-Computer Interfaces (BCIs) have made BCIs come in use mainly for the disabled to communicate. Practical usage of BCIs requires that the whole BCI system be portable so that disabled subjects can use them anywhere. The key aspect in mobility is to use mobile devices for processing by developing software applications with low-computational complexity. In this thesis, a low-computational P300 speller application is developed for Android using an Emotiv wireless EEG neuroheadset. Given the limited resources of mobile devices, a novel ensemble classifier approach is proposed that uses Principal Component Analysis (PCA) features to identify evoked P300 signals from EEG recordings. The performance of the method is demonstrated on benchmark data and on our own data. Results demonstrate the capability of the PCA ensemble classifier to classify P300 data recorded using the Emotiv neuroheadset with an average online classification accuracy of 97.22%.
Research Interests:
Dynamic Bayesian Networks (DBNs) are efficient graphical tools that could be used to detect causal relationships in multivariate systems. Here, we utilize DBNs to infer causality among electroencephalography (EEG) electrodes during a... more
Dynamic Bayesian Networks (DBNs) are efficient graphical tools that could be used to detect causal relationships in multivariate systems. Here, we utilize DBNs to infer causality among electroencephalography (EEG) electrodes during a motor imagery task. We inferred the causal relationships between EEG electrodes during each of right and left hands imagery movements from 9 different subjects. We demonstrate how using the inferred connectivity as a feature enhances the discrimination among right and left hands imagery movements compared to using traditional band power features. Our analysis reveals a distinctive connectivity pattern manifested by an increase in the number of incoming connections to the right hemisphere motor area compared to the left hemisphere during right hand imagery movements. This pattern is reversed during left hand imagery movements.
Research Interests:
Restoring vision is no longer impossible as a result of recent advances in neural interfaces. Successful demonstrations of retinal implants motivate the development of more effective visual prostheses. The thalamic Lateral Geniculate... more
Restoring vision is no longer impossible as a result of recent advances in neural interfaces. Successful demonstrations of retinal implants motivate the development of more effective visual prostheses. The thalamic Lateral Geniculate Nucleus (LGN) is one potential deep-brain interfacing site for visual prostheses. A main challenge in developing thalamic as well as other visual prostheses is optimizing the parameters of electrical stimulation. This paper introduces a Kalman-based optimal encoder whose function is to determine the optimal electrical stimulation parameters required to induce a certain visual sensation. The performance of the proposed approach is demonstrated using a probabilistic model of LGN neurons. Results demonstrate a significant similarity between neuronal responses obtained using electrical stimulation and the responses obtained using the corresponding visual stimuli with a mean correlation of 0.62 (P <; 0.01, n = 54). These results indicate the efficacy of the proposed optimal encoder in driving LGN neurons to induce visual sensations.
Research Interests:
Visual prosthesis opens new perspectives in the field of restoring vision for blind people. It aims to bypass the defective stages of the natural visual pathway and to provide proper inputs to the later stages. The starting point of any... more
Visual prosthesis opens new perspectives in the field of restoring vision for blind people. It aims to bypass the defective stages of the natural visual pathway and to provide proper inputs to the later stages. The starting point of any visual prosthesis is an image processing model which is typically performed in two phases: Feature extraction to highlight the important information of visual scene, followed by a visual pathway stimulation which replaces the functionality of the defective stages of the visual pathway. This paper introduces a MATLAB toolbox for visual prostheses to handle all the required image processing. This toolbox supports the most common image processing techniques needed for both static and dynamic scenes recognition. In addition, it supports different kinds of visual prostheses like retinal and thalamic visual prostheses depending on how deep the visual stimulation phase will go. The paper demonstrates the value of the toolbox through a case study in which it...
Visual prostheses have recently shown success in partially restoring vision to the blind. While retinal implants are considered the most successful type of visual prostheses, other techniques are needed for patients with completely... more
Visual prostheses have recently shown success in partially restoring vision to the blind. While retinal implants are considered the most successful type of visual prostheses, other techniques are needed for patients with completely degenerated retina or injured optic nerve. Thalamic visual prostheses that target the Lateral Geniculate Nucleus (LGN) represent one promising type. However, one challenge in tuning thalamic visual prostheses is to understand how visual information is encoded in the firing of LGN neurons. In this paper, we introduce an artificial neural network visual encoding model that incorporates visual stimulation history to predict the firing of LGN neurons in response to visual stimulation. To assess the performance of the model, we recorded stimulus-driven activity from the LGN in three anesthetized rats. Multielectrode arrays with 32 channels were used to simultaneously record the extracellular activity of LGN neurons in response to single-pixel flashing stimulat...
Research Interests:
Transfer learning is one approach that could be used to better train deep neural networks. It plays a key role in initializing a network in computer vision applications as opposed to implementing a network from scratch which could be... more
Transfer learning is one approach that could be used to better train deep neural networks. It plays a key role in initializing a network in computer vision applications as opposed to implementing a network from scratch which could be time-consuming. Natural Language Processing (NLP) shares a similar concept of transferring from large-scale data. Recent studies demonstrated that pretrained language models can be used to achieve state-of-the-art results on a multitude of NLP tasks such as sentiment analysis, machine translation and text summarization. In this paper, we demonstrate that a free RNN/CNN self-attention model used for sentiment analysis can be improved with 2.53% by using contextualized word representation learned in a language modeling task.
Research Interests:
Research Interests:
Motor imagery is currently one of the main applications of Brain-Computer Interface (BCI) which aims at providing the disabled with means to execute motor commands. One of the major stages of motor imagery systems is reducing the... more
Motor imagery is currently one of the main applications of Brain-Computer Interface (BCI) which aims at providing the disabled with means to execute motor commands. One of the major stages of motor imagery systems is reducing the dimensions of the input data and enhancing the features prior to applying a classification stage to recognize the intended movement. In this paper, we utilize autoencoders as a powerful tool to enhance the input features of the band power filtered electroencephalography (EEG) data. We compare the performance of the autoencoder-based approach to using Principal Component Analysis (PCA). Our results demonstrate that using autoencoders with non-linear activation function achieves better performance compared to using PCA. We demonstrate the effects of varying the number of hidden nodes of the autoencoder as well as the activation function on the performance. We finally examine the characteristics of the trained autoencoders to identify the features that are mos...
Research Interests:
Research Interests:
Brain-Computer Interface (BCI) is one of the most promising fields nowadays in hope to assist individuals with cognitive or sensory-motor disabilities. It facilitates human interaction with modern technologies and tools through brain... more
Brain-Computer Interface (BCI) is one of the most promising fields nowadays in hope to assist individuals with cognitive or sensory-motor disabilities. It facilitates human interaction with modern technologies and tools through brain activity. This could grant some communication capabilities back to severely disabled individuals. The P300 speller is one of the most successful BCI communication applications. A typical P300 speller is presented to the user as a 6-by-6 matrix whose cells contain the essential letters, digits, and characters that are randomly flashed. The usage of this application typically takes a significant amount of time due to the need to perform multiple trials to accurately recognize the target cell. We introduce a deep neural network approach to enhance the recognition of the target cell from a limited number of trials. We recorded the electroencephalography (EEG) signals from two subjects using Emotiv Epoc neuroheadset. A deep neural network consisting of multiple autoencoder layers and a softmax classification layer was trained with filtered EEG data. Compared to using Principal Component Analysis (PCA) and linear classifiers, the proposed model achieved an average increase in accuracy of 4.5% when applied to averaged 5 trials P300 data, and, more importantly, an increase of 8% when applied to single trial P300 data. The obtained results elucidate the significant effect of deep learning techniques in this paradigm.
Research Interests:
Spinal motoneurons (MNs) play a crucial role in movement control. Decoding the firing activity of spinal MNs could help in real-life challenges, such as enhancing the control of myoelectric prostheses and diagnosing neurodegenerative... more
Spinal motoneurons (MNs) play a crucial role in movement control. Decoding the firing activity of spinal MNs could help in real-life challenges, such as enhancing the control of myoelectric prostheses and diagnosing neurodegenerative diseases. In this paper, we propose a machine learning approach to automatically classify MNs based on their firing activity. Applying the proposed approach to data from a MN computational model, the classification accuracy of all examined datasets exceeded 95%. We extended the approach to detecting the death of a given MN type using clustering validity index. Results indicated that 86% of the examined death-detection cases were detected accurately. These results demonstrate that the proposed approach is a successful step in automating neuronal cell-type classification.
Research Interests:
National Basketball Association (NBA) All Star Game is a demonstration game played between the selected Western and Eastern conference players. The selection of players for the NBA All Star game purely depends on votes. The fans and... more
National Basketball Association (NBA) All Star Game is a demonstration game played between the selected Western and Eastern conference players. The selection of players for the NBA All Star game purely depends on votes. The fans and coaches vote for the players and decide who is going to make the All Star roster. A player who continues to receive enough votes in following years will play more All Star games. The selection of All Star players in NBA is subjective based on voting and there are no selection criteria that take out the human bias and opinion. Analyzing data from previous sports leagues can provide insight into the factors that lead to winning games and titles. This study aims to classify the players into regular or All Star players from the National Basketball Association and identify the most important characteristics that make a player an All Star player. To accomplish this, the performance per minute of play and per average of total minutes of player were analyzed usi...
Discretizing continuous attributes is one essential and important data preprocessing step in data mining. Various data mining techniques are designed to be applied to discrete attributes. There have been tremendous efforts to propose... more
Discretizing continuous attributes is one essential and important data preprocessing step in data mining. Various data mining techniques are designed to be applied to discrete attributes. There have been tremendous efforts to propose discretization techniques with different characteristics. However, a clear pathway that can guide the choice of the needed discretization technique for different types of datasets is lacking. This paper proposes a taxonomy based on the existence of class information and relationship between attributes in the analyzed dataset. We review different discretization techniques classified according to the proposed taxonomy. The proposed taxonomy emphasizes the advantages and disadvantages of each discretization technique to be able theoretically to find a suitable discretization technique for a particular dataset.
Neurodegenerative diseases (NDDs) lead to extreme locomotion disorders, which stem from the impairment of motor function caused by these diseases. They alter the gait rhythms and gait dynamics of their patients, which can be reflected in... more
Neurodegenerative diseases (NDDs) lead to extreme locomotion disorders, which stem from the impairment of motor function caused by these diseases. They alter the gait rhythms and gait dynamics of their patients, which can be reflected in their time-series recordings of footfall contact times. Studies has shown that human gait can be used to identify NDDs. In this paper, an ensemble classification-based diagnostic system that uses patients’ gait data to diagnose NDDs is introduced. The diagnostic system is an ensemble of four binary classifications. The proposed technique is applied to gait dynamics data recorded from 64 subjects representing three NDDs: Amyotrophic Lateral Sclerosis (ALS), Parkinson’s Disease (PD), and Huntington’s Disease (HD), in addition to healthy subjects. Our results demonstrate the ability of the proposed technique to recognize gait dynamics of ALS, PD and HD with an accuracy of 96.8, 86.8% and 85.5%, respectively. These results demonstrate the efficacy of th...
A Brain-Machine Interface (BMI) is an electronic device that interfaces directly with the brain to restore a lost function caused by an injury or disorder. BMIs currently provide new hope for people with disabilities, whose brain function... more
A Brain-Machine Interface (BMI) is an electronic device that interfaces directly with the brain to restore a lost function caused by an injury or disorder. BMIs currently provide new hope for people with disabilities, whose brain function is still intact. By recording brain activity, BMIs could translate recorded brain activity to actions that can interact with the external environment. There are two major types of BMIs: Non-invasive and invasive. Non-invasive BMIs use multiple electrodes to record electroencephalography (EEG) signals from the scalp. Such recording technique enables extracting several time-domain and frequency-domain features that correlate with the subjects' intended behavior. Invasive BMIs, which involve a brain surgery to implant electrodes into the brain, enable recording the activity of individual neurons in the brain in addition to the ability to electrically stimulating such neurons. This talk will give an overview of both the theoretical and practical as...
I M A G E B Y A N D R I J B O R Y S A S S O C I A T E S HEALTH CHALLENGES REPRESENT one of the longstanding issues in the Arab region that hinder its ability to develop. Prevalence of diseases such as cardiovascular diseases, liver... more
I M A G E B Y A N D R I J B O R Y S A S S O C I A T E S HEALTH CHALLENGES REPRESENT one of the longstanding issues in the Arab region that hinder its ability to develop. Prevalence of diseases such as cardiovascular diseases, liver cirrhosis and cancer among many others has contributed to the deteriorated health status across the region leading to lower life expectancy compared to other regions. For instance, the average life expectancy in the Arab world is approximately 70 years, which is at least 10 years lower than most high-income countries. Among many directions of healthcare development across the region, biomedical computing research represents one main arm of tackling health challenges. Advances in computational technologies have enabled the emergence of biomedical computing as one of the most influential research areas worldwide. In recent years, life sciences have witnessed an explosion in the volumes of biomedical data generated by high-throughput technologies and other sources. The enormity of volumes and interdependence in biomedical data pose great analytical challenges in the quest to infer deeply hidden knowledge buried under this complexity. As such, the biomedical computing research community in the Arab world has been actively contributing to the efforts that tackle long-standing biomedical challenges. Research in biomedical computing in the region dates back to mid1970s with the establishment of the Systems and Biomedical Engineering department at Cairo University in Egypt. Since then, the number of related programs has steadily increased and researchers from different disciplines have developed interest in biomedical computing applications. Despite the limited available resources, researchers from the Arab region have made over the years strong contributions to the rapid advances that occur in the field of biomedical computing. Recent successful efforts by researchers across the region have been evident in three broad areas of biomedical computing; namely, biomedical imaging, biomedical signal analysis and bioinformatics. These efforts have materialized in advancing a diverse spectrum of biomedical computing applications, as well as stimulating clear commercial interest. This article sheds light on notable research efforts in the Arab world in each of the aforementioned areas of biomedical computing. It also demonstrates how this research Biomedical Computing in the Arab World: Unlocking the Potential of a Growing Research Community DOI:10.1145/3447746
Research Interests:
The Lateral Geniculate Nucleus (LGN) represents one of the major processing sites along the visual pathway. Despite its crucial role in processing visual information and its utility as one target for recently developed visual prostheses,... more
The Lateral Geniculate Nucleus (LGN) represents one of the major processing sites along the visual pathway. Despite its crucial role in processing visual information and its utility as one target for recently developed visual prostheses, it is much less studied compared to the retina and the visual cortex. In this paper, we introduce a deep learning encoder to predict LGN neuronal firing in response to different visual stimulation patterns. The encoder comprises a deep Convolutional Neural Network (CNN) that incorporates visual stimulus spatiotemporal representation in addition to LGN neuronal firing history to predict the response of LGN neurons. Extracellular activity was recorded in vivo using multi-electrode arrays from single units in the LGN in 12 anesthetized rats with a total neuronal population of 150 units. Neural activity was recorded in response to single-pixel, checkerboard and geometrical shapes visual stimulation patterns. Extracted firing rates and the corresponding ...
Research Interests:
Research Interests:
Developing visual prostheses that target inner brain structures along the visual pathway represent a new hope for patients with completely damaged early visual pathway sites. One of the major challenges in the development of subcortical... more
Developing visual prostheses that target inner brain structures along the visual pathway represent a new hope for patients with completely damaged early visual pathway sites. One of the major challenges in the development of subcortical and cortical visual prostheses is tuning electrical stimulation that could optimally induce desired visual percepts. In this paper, we propose a Kalman filter-based strategy that could be used to identify electrical stimulation patterns that mimic a specific visual input for thalamic visual prostheses. We demonstrate the performance of the proposed strategy using a population of lateral geniculate nucleus neurons modeled using an adapted generalized non-linear model. A mean correlation of 0.69 is obtained between visually evoked and electrically evoked responses-driven using the proposed strategy-for an optimal electrode-placement setup. In addition, we demonstrate the performance for a random electrode-placement setup in which a mean correlation of 0.26 is obtained. For this latter setup, our analysis reveals an inversely proportional relationship between the obtained correlation and the distance between each neuron and the nearest electrode. The proposed strategy could be thus utilized to tune and enhance the performance of thalamic visual prostheses as well as other prosthesis systems.
Research Interests:
Visual prosthesis holds hope of vision restoration for millions with retinal degenerative diseases. Machine learning techniques such as artificial neural networks could help in improving prosthetic devices as they could learn how the... more
Visual prosthesis holds hope of vision restoration for millions with retinal degenerative diseases. Machine learning techniques such as artificial neural networks could help in improving prosthetic devices as they could learn how the brain encodes information and imitate that code. This paper introduces an autoencoder-based approach for tuning thalamic visual prostheses. The objective of the proposed approach is to estimate electrical stimuli that are equivalent to a given natural visual stimulus, in a way such that they both elicit responses that are as similar as possible when introduced to a Lateral Geniculate Nucleus (LGN) population. Applying the proposed method to a probabilistic model of LGN neurons, results demonstrate a significant similarity between both responses with a mean correlation of 0.672 for optimal electrodes placement and 0.354 for random electrodes placement. The results indicate the efficacy of the proposed approach in estimating an electrical stimulus equivalent to a specific visual stimulus.
Research Interests:
... 450-477, 2009. [16] S. Eldawlatly, Y. Zhou, R. Jin, and K. Oweiss, &amp;amp;amp;amp;amp;quot;On The Use of DynamicBayesian Networks in Reconstructing Functional Neuronal Networks from Spike TrainEnsembles, submitted to Neural... more
... 450-477, 2009. [16] S. Eldawlatly, Y. Zhou, R. Jin, and K. Oweiss, &amp;amp;amp;amp;amp;quot;On The Use of DynamicBayesian Networks in Reconstructing Functional Neuronal Networks from Spike TrainEnsembles, submitted to Neural Computation, 2009. 625
Research Interests:
ABSTRACT Recent advances in developing Brain-Computer Interfaces (BCIs) have opened up a new realm for designing efficient systems that could enable disabled people to communicate. The P300 speller is one important BCI application that... more
ABSTRACT Recent advances in developing Brain-Computer Interfaces (BCIs) have opened up a new realm for designing efficient systems that could enable disabled people to communicate. The P300 speller is one important BCI application that allows the selection of characters on a virtual keyboard by analyzing recorded electroencephalography (EEG) activity. In this work, we propose an ensemble classifier that uses Principal Component Analysis (PCA) features to identify evoked P300 signals from EEG recordings. We examine the performance of the proposed method, using different linear classifiers, on the datasets provided by the BCI competition III. Results demonstrate a classification accuracy of 91% using the proposed method. In addition, our results indicate a significant improvement in classification accuracy compared to traditional feature extraction and classification approaches. The proposed method results in low across-subjects variability compared to other methods with minimal parameter tuning required which could be useful in mobile platform P300 applications.
Research Interests:
The current trend to use Brain-Computer Interfaces (BCIs) with mobile devices mandates the development of efficient EEG data processing methods. In this paper, we demonstrate the performance of a Principal Component Analysis (PCA)... more
The current trend to use Brain-Computer Interfaces (BCIs) with mobile devices mandates the development of efficient EEG data processing methods. In this paper, we demonstrate the performance of a Principal Component Analysis (PCA) ensemble classifier for P300-based spellers. We recorded EEG data from multiple subjects using the Emotiv neuroheadset in the context of a classical oddball P300 speller paradigm. We compare the performance of the proposed ensemble classifier to the performance of traditional feature extraction and classifier methods. Our results demonstrate the capability of the PCA ensemble classifier to classify P300 data recorded using the Emotiv neuroheadset with an average accuracy of 86.29% on cross-validation data. In addition, offline testing of the recorded data reveals an average classification accuracy of 73.3% that is significantly higher than that achieved using traditional methods. Finally, we demonstrate the effect of the parameters of the P300 speller paradigm on the performance of the method.
Research Interests:
Research Interests:
ABSTRACT Advances in Brain-computer Interface (BCI) technology have opened the door to assisting millions of people worldwide with disabilities. In this work, we focus on assessing brain attention state that could be used to selectively... more
ABSTRACT Advances in Brain-computer Interface (BCI) technology have opened the door to assisting millions of people worldwide with disabilities. In this work, we focus on assessing brain attention state that could be used to selectively run an application on a hand-held device. We examine different classification techniques to assess brain attention state. Spectral analysis of the recorded EEG activity was performed to compute the Alpha band power for different subjects during attentive and non-attentive tasks. The estimated power values were used to train a number of classical classifiers to discriminate among the two attention states. Results demonstrate a classification accuracy of 70% using both individual- and multi-channel data. We then utilize a graphical approach to assess the causal influence among EEG electrodes for each of the two attention states. The inferred graphical representations for each state were used as signatures for state classification. A classification accuracy of 83% was obtained using the graphical approach outperforming the examined classical classifiers.
Graphical models are powerful tools to infer statistical relationships between simultaneously observed random variables. Here, we used Dynamic Bayesian Networks (DBN) to infer causal relationships between simultaneously recorded neurons... more
Graphical models are powerful tools to infer statistical relationships between simultaneously observed random variables. Here, we used Dynamic Bayesian Networks (DBN) to infer causal relationships between simultaneously recorded neurons in the rat somatosensory (barrel) cortex in response to whisker stimulation. DBNs attempt to explain the activity of the observed neurons by searching for the best network connectivity that explains the observed data. The results demonstrate that the networks inferred for the same whisker are stable across multiple repeated trials. In contrast to networks obtained using classical cross-correlograms, DBN was able to discriminate between direct and indirect connectivity. We also found strong consistency between the inferred connections and the sequence of neural firing relative to the stimulus onset.
Research Interests:
... In this paper, we investigate the applicability of applying DBN to reconstruct functional corticalnetworks from ... We demonstrated the use of DBN in identifying the structures of neural networks from observed ... [3] K. Murphy,... more
... In this paper, we investigate the applicability of applying DBN to reconstruct functional corticalnetworks from ... We demonstrated the use of DBN in identifying the structures of neural networks from observed ... [3] K. Murphy, Dynamic Bayesian Networks: Representation, Inference ...
Research Interests:
Clustering analysis is an important tool to study the functional interdependency among large ensembles of neurons from the observed spiking activity. An important question is how to determine the number of independent clusters when... more
Clustering analysis is an important tool to study the functional interdependency among large ensembles of neurons from the observed spiking activity. An important question is how to determine the number of independent clusters when neuronal ensembles are dynamically recruited to process and store information, for e.g. during learning and behavior. In this paper, we propose a new approach based on
Research Interests:
... [10] S. Eldawlatly, Y. Zhou, R. Jin, and K. Oweiss, "On The Use of DynamicBayesian Networks in Reconstructing Functional Neuronal Networks from SpikeTrain Ensembles," Submitted to Neural Computation, 2008. ...
Research Interests:
Identifying functional connectivity from simultaneously recorded spike trains is important in understanding how the brain processes information and instructs the body to perform complex tasks. We investigate the applicability of dynamic... more
Identifying functional connectivity from simultaneously recorded spike trains is important in understanding how the brain processes information and instructs the body to perform complex tasks. We investigate the applicability of dynamic Bayesian networks (DBN) to infer the structure of neural circuits from observed spike trains. A probabilistic point process model was used to assess the performance. Results confirm the utility of DBNs in inferring functional connectivity as well as directions of signal flow in cortical network models. Results also demonstrate that DBN outperforms the Granger causality when applied to populations with highly non-linear synaptic integration mechanisms.
Research Interests:
Research Interests: Algorithms, Nonlinear dynamics, Multidisciplinary, Spike train analysis, Learning and Memory, and 17 moreNeuronal Network, Computer Simulation, Cerebral Cortex, Effective Connectivity, Mathematical Computing, Synaptic Transmission, Neurons, Mathematical Concepts, dynamic Bayesian network, SYNAPSES, Network structure, Neural Computation, Neural pathways, Bayes Theorem, Action Potentials, Poisson Distribution, and Time varying(Neuronal Network, Computer Simulation, Cerebral Cortex, Effective Connectivity, Mathematical Computing, Synaptic Transmission, Neurons, Mathematical Concepts, dynamic Bayesian network, SYNAPSES, Network structure, Neural Computation, Neural pathways, Bayes Theorem, Action Potentials, Poisson Distribution, and Time varying)
(Neuronal Network, Computer Simulation, Cerebral Cortex, Effective Connectivity, Mathematical Computing, Synaptic Transmission, Neurons, Mathematical Concepts, dynamic Bayesian network, SYNAPSES, Network structure, Neural Computation, Neural pathways, Bayes Theorem, Action Potentials, Poisson Distribution, and Time varying)
Brain machine interfaces (BMIs) have recently received significant attention from the neuroscience and engineering communities as a result of striking advances in monitoring, processing, and modeling brain function at multiple temporal... more
Brain machine interfaces (BMIs) have recently received significant attention from the neuroscience and engineering communities as a result of striking advances in monitoring, processing, and modeling brain function at multiple temporal and spatial resolutions. These advances, however, have also raised significant challenges to both communities that are becoming the focus of numerous ongoing research efforts. Broadly categorized based on their
Research Interests:
The problem of identifying plasticity in a recorded neural population has long been the subject of intense research. With the ability to simultaneously record large ensembles of single unit activity over extended periods of time, it is... more
The problem of identifying plasticity in a recorded neural population has long been the subject of intense research. With the ability to simultaneously record large ensembles of single unit activity over extended periods of time, it is becoming central to the ability to efficiently decode neuronal responses. In a previous study, we demonstrated that a graph theoretic approach can identify
Research Interests:
Identifying clusters of neurons that exhibit functional interdependency in a recorded population has recently emerged as a direct result of the ability to simultaneously record multiple single unit activity with high-density... more
Identifying clusters of neurons that exhibit functional interdependency in a recorded population has recently emerged as a direct result of the ability to simultaneously record multiple single unit activity with high-density microelectrode arrays. We demonstrated in a previous study that a graph theoretic approach can identify functional interdependency over multiple time scales between models of neuronal firing in response to a common input or synaptically-coupled in a multi-cluster population. In this paper, we investigate the performance of the technique in the case of neuronal interaction arising at various latencies and interval lengths. We report the capability of the approach to track these variable degrees of interactions. This feature can be very useful in decoding variable motor cortical response patterns during sensorimotor integration in Brain Machine Interface applications.
Research Interests:
Research Interests: Information Theory, Graph Theory, Coding, Entropy, Information Processing, and 14 moreStatistical Learning, Learning, Neural coding, Spike train analysis, Connectedness, Message Passing, Encoding, Joint, Statistical Model, Articulation, Electrical And Electronic Engineering, Maximum entropy, Entropie, and Conditional Independence
This paper develops a new fuzzy clustering approach that is suitable for image processing applications. The developed approach is based upon the classical fuzzy c-means (FCM) and referred to as the spatial FCM (SFCM). Its effectiveness is... more
This paper develops a new fuzzy clustering approach that is suitable for image processing applications. The developed approach is based upon the classical fuzzy c-means (FCM) and referred to as the spatial FCM (SFCM). Its effectiveness is due to two mechanisms. The first is the replacement of the Euclidean distance traditionally used to measure similarity between input images and clusters prototypes by a novel similarity measure that considers spatial relationships between image pixels and thus becomes less sensitive to image perturbations. The second SFCM mechanism for effectiveness is the addition of a similarity penalty term to FCM's objective function. The aim is to encourage clustering similar images into same clusters. The SFCM is compared to the FCM and some of its variants in the difficult application of synthetic aperture radar (SAR) target clustering. It is shown that the SFCM consistently yields better performance