Search

Morphological changes related to different diseases that occur in the retina are currently extensively researched. Manual segmentation of retinal fluids is time-consuming and subject to variability, giving prominence to the demand for robust automatic segmentation methods. The standard in assessing the existence and mass of retinal fluids at present is through the optical coherence tomography (OCT) modality. In this study, semantic segmentation deep learning networks were examined in 2.5D and ensembled with 2D networks. This analysis aims to show how these networks can perform in-depth than using only a single B-scan and the effects of 2.5 patches when fitted to the deep networks. All experiments were evaluated using public data from the RETOUCH challenge as well as the OPTIMA challenge dataset and Duke dataset. The networks trained in 2.5D performed slightly better than 2D networks in all datasets. The average performance of the best network was 0.867, using the dice similarity coefficient score (DSC) metric on the RETOUCH dataset. On the DUKE dataset, Deeplabv3+Pa outperformed other networks in this study with a dice score of 0.80. Experiments showed a more robust performance when networks were ensembled. Intraretinal fluid (IRF) was recognized better than other fluids with a DSC of 0.924. Deeplabv3+Pa model outperformed all other networks with a p-value average of 0.03 on the RETOUCH challenge dataset. Methods used in this study to distinguish retinal disorders outperform human performance as well as showed competitive results to the teams who joined both challenges. Three consecutive B-scans, including partial depth information in training neural networks, were stacked as a single image built for more robust networks compared to providing only 2D information.

Motor imagery (MI)-based brain–computer interfaces have gained much attention in the last few years. They provide the ability to control external devices, such as prosthetic arms and wheelchairs, by using brain activities. Several researchers have reported the inter-communication of multiple brain regions during motor tasks, thus making it difficult to isolate one or two brain regions in which motor activities take place. Therefore, a deeper understanding of the brain’s neural patterns is important for BCI in order to provide more useful and insightful features. Thus, brain connectivity provides a promising approach to solving the stated shortcomings by considering inter-channel/region relationships during motor imagination. This study used effective connectivity in the brain in terms of the partial directed coherence (PDC) and directed transfer function (DTF) as intensively unconventional feature sets for motor imagery (MI) classification. MANOVA-based analysis was performed to identify statistically significant connectivity pairs. Furthermore, the study sought to predict MI patterns by using four classification algorithms—an SVM, KNN, decision tree, and probabilistic neural network. The study provides a comparative analysis of all of the classification methods using two-class MI data extracted from the PhysioNet EEG database. The proposed techniques based on a probabilistic neural network (PNN) as a classifier and PDC as a feature set outperformed the other classification and feature extraction techniques with a superior classification accuracy and a lower error rate. The research findings indicate that when the PDC was used as a feature set, the PNN attained the greatest overall average accuracy of 98.65%, whereas the same classifier was used to attain the greatest accuracy of 82.81% with the DTF. This study validates the activation of multiple brain regions during a motor task by achieving better classification outcomes through brain connectivity as compared to conventional features. Since the PDC outperformed the DTF as a feature set with its superior classification accuracy and low error rate, it has great potential for application in MI-based brain–computer interfaces.

Electroencephalography (EEG) signal is one of the popular approaches for analyzing the relationship between motor movement and the brain activity. This is mainly driven by the rapid development of Brain-Computer-Interface BCI devices for applications like prosthetic devices, using EEG as its input signal. The EEG is known to be highly affected by artefact and with more motor events, this may result in low classification accuracy. In this paper, classification of 3-class hand motor EEG signals, performing grasping, lifting and holding using Common Spatial Pattern (CSP) and pre-trained CNN is investigated. Thirteen electrodes capturing signals related to motor movement, C3, Cz, C4, T3, T4, F7, F3, Fz, F4, F8, P3, Pz and P4 are utilized and signal from \(\alpha \), \(\beta \), \(\varDelta \) and \(\theta \) bands are selected in the pre-processing stage. CSP filters utilizing the scheme of pair-wise are used to increase the discriminative power between two classes whereby the signals extracted by the CSP filter are converted into scalograms by utilizing Continuous Wavelet Transform (CWT). The accuracy of the proposed multi-band and CSP based classification algorithm tested using DenseNet giving average accuracy values of 97.3%, 93.8% and 100%, for GS, LT and HD movements, respectively. These results indicate that the classification framework using CSP filters and pre-trained CNN can provide a good solution in decoding hand motor movement from EEG signals.

A microstrip‐fed hybrid Fractal antenna for WLAN application is presented in this article. The proposed antenna is composed of meander and Koch elements combined to form a longer curve that extends the electrical length. This approach simultaneously reduces the operating frequency and antenna size. The Green's functions and segmentation method are used to calculate the input impedance of the hybrid Fractal antenna. For analysis, the amalgamated antenna shape is decomposed into simple regular segments. The impedance parameters of the newly arranged shapes are calculated by the corresponding Green's functions. The impedance matrix equations of all the segments are presented so as to display in a single matrix whose solution determines the input impedance of the complete antenna geometry. To validate the theory, a prototype of the proposed antenna has been numerically simulated and experimentally measured.

The sensing algorithm of many high-precision sensors is implemented in Field Programmable Gate Arrays (FPGAs) because such devices meet most of the algorithm requirements. But to enhance sensors’ performance, it is often required to compensate for their output against environmental variations. This compensation requires implementing polynomial functions in some variables corresponding to the environmental variations. This paper presents a low-cost design for double-precision, reusable, and flexible computational unit for implementing such polynomial functions in digital hardware. The design can be directly used for any FPGA and even for Application-Specific Integrated Circuit (ASIC) development. In the case of ASIC design, it also features the required flexibility to handle sensor to sensor polynomial variations. The design has been verified through simulation for different FPGAs and 350nm technology node ASIC. The design has also been implemented in Spartan-6 FPGA to compensate sensors’ output in real-time.

Retinal diseases are among the significant reasons for vision loss worldwide. Age-related macular degeneration (AMD) influences older people, and 170 million individuals are diagnosed with AMD on the global level. This number is expected to increase to 288 million people by 2040. Optical coherence tomography (OCT) is the most effective and noninvasive modality to view the retinal layers. The frequent visit of patients affected with retinal diseases raised the need for developing automatic algorithms to localize and quantity the morphological changes occurring in the retina. Deep learning networks show excellent performance in classifying 2D scans at the image level and pixel level. Medical data is usually obtained with depth information, and using only 2D information could lead to lower accuracy in localizing the fluid volume size. Mimicking human performance in manually locating the diseases over medical images is the main target of automatic methods to exceed. In this study, we have used the RETOUCH challenge dataset to segment various retinal fluids. Human performance reported in the challenge scored 0.71 in the dice similarity coefficient (DSC) metric. Encoder-decoder network is demonstrated in a 3D manner for the retinal disease segmentation, and the average performance score is 0.73 in the dice metric from the Cirrus scanner data. The dataset released three different fluids, and intraretinal fluid (IRF) is more identified with 0.79 in the DSC metric.

The deterioration of the retina center could be the main reason for vision loss. Older people usually ranging from 50 years and above are exposed to age-related macular degeneration (AMD) disease that strikes the retina. The lack of human expertise to interpret the complexity in diagnosing diseases leads to the importance of developing an accurate method to detect and localize the targeted infection. Approaching the performance of ophthalmologists is the consistent main challenge in retinal disease segmentation. Artificial intelligence techniques have shown enormous achievement in various tasks in computer vision. This paper depicts an automated end-to-end deep neural network for retinal disease segmentation on optical coherence tomography (OCT) scans. The work proposed in this study shows the performance difference between convolution operations and atrous convolution operations. Three deep semantic segmentation architectures, namely U-net, Segnet, and Deeplabv3+, have been considered to evaluate the performance of varying convolution operations. Empirical outcomes show a competitive performance to the human level, with an average dice score of 0.73 for retinal diseases.

The discrimination of four simple limb motor imagery movements for brain-computer interface (BCI) applications is still challenging. This is because most of the movement imaginations have close spatial representations on the motor cortex area. Nevertheless, due to its potential applications in significant areas including BCI, solutions need to be formulated to overcome the task discrimination issues faced when a motor imagery movement approach is utilized. Feature extraction is one of the most important steps in any BCI system; as such, enhancement to the existing methods has been incorporated in this work. For this, we propose four-class movement imaginations of the right hand, left hand, right foot, and left foot, and develop feature extraction methods utilizing discrete wavelet transform (DWT) and empirical mode decomposition (EMD); in both methods, artificial neural network (ANN) was used as a classifier. Based on the processed electroencephalography (EEG) data recorded from eleven subjects, it can be seen that EMD features outperform DWT features; the average accuracy achieved by the EMD features is 90.02%, and 84.77% using the DWT features. EMD even performs better than DWT in discriminating the most challenging tasks involving the right foot and left foot imageries, whose EEG data were derived from the same Cz node of the motor cortex.

International audience; There are many types of collision warning systems to increase the safety of motorcyclists. These systems use different types of collision detection techniques with each one having some limitations, restricting the performance of the system. To find the effectiveness of collision warning system notifications, it is important to study the physiological response of drivers toward these systems. Existing studies are limited to the physiological response of car drivers and use only buzzer warnings for these systems. Unfortunately, no such work in that particular domain has been reported for motorcyclists. Since motorcycles have different maneuverability as compared with cars and other vehicles, it is important to investigate the response of motorcyclists toward these collision warning systems. Also, it is believed that providing verbal information about any potential hazard will further assist the motorcyclist to avoid it. The aim of this paper is to investigate the physiological responses of motorcyclists to the rear end collision warning system when auditory verbal warnings are utilized. To study the response of the motorcyclists, the N100, N200, P300, and N400 eventrelated potential components have been extracted from the recorded Electroencephalography data. It has been found that the rear end collision warning system with auditory verbal warnings significantly increases the alertness of the motorcyclist and can be helpful to avoid the possible rear-end collision scenarios. This system has shown positive effects at neural levels on motorcyclists and reduces their reaction time and attentional resources required for processing the target correctly.

This research studies the impact of the imagination of movements and associated feedbacks on the modulation of sensorimotor electroencephalographic (EEG) rhythms, for the online controls of a brain-computer interface (BCI). Nine subjects with no physical or mental impairments were selected. The number of sessions was five: one calibration and four feedback sessions. A computer screen’s cursor movement was controlled in one dimension using EEG-based four-class BCI involving motor imagery tasks of moving the right hand, the left hand, both hands, or both feet. Our findings reveal that the visual feedback applied during motor imagery movement modulates sensorimotor EEG rhythms clearly in the mu and beta bands. The analyses of event-related desynchronization/synchronization (ERD/ERS) suggest significant differences between brain activities in the calibration and feedback sessions; large ERDs during the online feedback sessions compared to that in the calibration session have been observed. The increasing ERDs in the online feedback session were noticed over the mu1 (8–10 Hz) and upper beta (18–24 Hz) rhythms, resulting in the cursor control success rate at 73.3%.

International audience; In this paper, an efficient and reliable feature fusion-based modulation classification (MC) algorithm for multiple-input multiple-output (MIMO) systems is developed. It uses two higher-order cumulants of the transmitted signal streams to classify a broad set of modulation types with no prior knowledge of the channel state information. We address the problem of the soft-decision fusion for the feature-fusion-basedMC algorithms forMIMO systems and introduce an optimal soft-decision fusion scheme to find the classification result. The complexity order of the proposed MC algorithmis studied in detail to demonstrate its low computation cost, and its performance is validated extensively by simulation results to show its practical effectiveness

Stereoscopic 3D (S3D) displays provide a very realistic visualization experience to the viewers through a sensation of depth perception. Despite the rapid growth of S3D technology (S3D) in education, very little has been known about the effects of S3D on human behavior and the corresponding brain's responses over traditional 2D display during conscious learning and Long-Term Memory (LTM) recall process. The present study aims to explore the effects of S3D technology on human behavior and the brain responses during learning and memory using an electroencephalography (EEG) technique. A sample of 68 participants between 18 and 30 years of age was recruited to perform three tasks: (1) Raven's Advanced Progressive Matrices (RAPM) test to assess the fluid intelligence ability; (2) learning task to study the S3D or 2D based contents for learning; and (3) memory recall task to assess the memory recall and retention performance. The participants were randomly assigned to two groups―2D group and S3D group, in such a way that their fluid intelligence ability, age, and background knowledge about the learning material are controlled. The analysis of behavioral data suggested that responses of both groups in terms of reaction time after two months of retention were statistically significantly different, F(2.66,125.09) = 4.47, p = .007, η p 2 = 0.087. The differences between the S3D and 2D groups were identified by EEG analysis, i.e., the classifier correctly discriminated the EEG signals of both groups with more than 90% accuracy rate. The EEG source analysis showed statistically significant differences between the groups in the brain regions of BA 7, BA 10, BA 11, and BA 25, reflecting widespread neuronal networks involved in the S3D group as compared to the 2D group in the LTM recollection. In conclusion, it was experimentally showed that the human brain processed the S3D contents differently by utilizing more cortical regions and neuronal networks than the traditional 2D contents, which modulates the behavior of the participants by recollecting the LTM faster in the recall task.

Blind algorithms for multiple-inputmultiple-output (MIMO) signals interception have recently received considerable attention because of their important applications in modern civil and military communication fields. One key step in the interception process is to blindly recognize the modulation type of the MIMO signals. This can be performed by employing a Modulation Classification (MC) algorithm, which can be feature-based or likelihood-based. To overcome the problems associated with the existing likelihood-based MC algorithms, a new algorithm is developed in this paper. We formulated the MC problem as maximizing a global likelihood function formed by combining the likelihood functions for the estimated transmitted signals, where Minimum Mean Square Error (MMSE) filtering is employed to separate the MIMO channel into several sub-channels. Simulation results showed that the proposed algorithm works well under various operating conditions, and performs close to the performance upper bound with reasonable complexity. © 2016 American Scientific Publishers All rights reserved.

Blind techniques and algorithms for Multiple-Input Multiple-Output (MIMO) signals interception have recently attracted a great deal of research efforts. This is due to their important applications in the military and civil telecommunications domains. One essential step in the signal interception process is to blindly recognize the modulation scheme of the MIMO signals. This process is formally called Modulation Classification (MC). This paper discusses the modulation classification for MIMO systems and presents a comprehensive and critical literature review of the existing MC algorithms for MIMO systems; where possible, gaps in the knowledge base are identified and future directions for the research work are suggested.

Background and objectives: Aged people usually are more to be diagnosed with retinal diseases in developed countries. Retinal capillaries leakage into the retina swells and causes an acute vision loss, which is called age-related macular degeneration (AMD). The disease can not be adequately diagnosed solely using fundus images as depth information is not available. The variations in retina volume assist in monitoring ophthalmological abnormalities. Therefore, high-fidelity AMD segmentation in optical coherence tomography (OCT) imaging modality has raised the attention of researchers as well as those of the medical doctors. Many methods across the years encompassing machine learning approaches and convolutional neural networks (CNN) strategies have been proposed for object detection and image segmentation. Methods: In this paper, we analyze four wide-spread deep learning models designed for the segmentation of three retinal fluids outputting dense predictions in the RETOUCH challenge data. We aim to demonstrate how a patch-based approach could push the performance for each method. Besides, we also evaluate the methods using the OPTIMA challenge dataset for generalizing network performance. The analysis is driven into two sections: the comparison between the four approaches and the significance of patching the images. Results: The performance of networks trained on the RETOUCH dataset is higher than human performance. The analysis further generalized the performance of the best network obtained by fine-tuning it and achieved a mean Dice similarity coefficient (DSC) of 0.85. Out of the three types of fluids, intraretinal fluid (IRF) is more recognized, and the highest DSC value of 0.922 is achieved using Spectralis dataset. Additionally, the highest average DSC score is 0.84, which is achieved by PaDeeplabv3+ model using Cirrus dataset. Conclusions: The proposed method segments the three fluids in the retina with high DSC value. Fine-tuning the networks trained on the RETOUCH dataset makes the network perform better and faster than training from scratch. Enriching the networks with inputting a variety of shapes by extracting patches helped to segment the fluids better than using a full image.

EEG has been largely used in both clinical and research applications. Brain computer interface (BCI) system is one of the major EEG research applications which can provide a new way of communications for special users who cannot communicate via normal pathways. This paper focuses on the development of the brain controlled wheelchair which incorporates two additional control interfaces including joystick and a remote control through an android phone. All three controls are integrated in such a way that it allows the user to change the mode of control by simply changing the state of the slide switch. This work utilizes the Neurosky Mindwave Mobile headset to capture the EEG signals through a single channel placed at the FP1 position. Eye blinks and attention levels are the key features of the captured EEG that are extracted and identified through an android application. The design also assimilates ultrasonic sensors based safety system which is capable of detecting the obstacles in all four directions to ensure the safety of the user.

Epilepsy, a common neurological disorder, is generally detected by electroencephalogram (EEG) signals. Visual inspection and interpretation of EEGs is a slow, time consuming process that is vulnerable to error and subjective variability. Consequently, several efforts to develop automatic epileptic seizure detection and classification methods have been made. The present study proposes a novel computer aided diagnostic technique (CAD) based on the discrete wavelet transform (DWT) and arithmetic coding to differentiate epileptic seizure signals from normal (seizure-free) signals. The proposed CAD technique comprises three steps. The first step decomposes EEG signals into approximations and detail coefficients using DWT while discarding non-significant coefficients in view of threshold criteria; thus, limiting the number of significant wavelet coefficients. The second step converts significant wavelet coefficients to bit streams using arithmetic coding to compute the compression ratio. In the final step, the compression feature set is standardized, whereupon machine-learning classifiers detect seizure activity from seizure-free signals. We employed the widely used benchmark database from Bonn University to compare and validate the technique with results from prior approaches. The proposed method achieved a perfect classification performance (100% accuracy) for the detection of epileptic seizure activity from EEG data, using both linear and non-liner machine-learning classifiers. This CAD technique can thus be considered robust with an extraordinary detection capability that discriminates epileptic seizure activity from seizure-free and normal EEG activity with simple linear classifiers. The method has the potential for efficient application as an adjunct for the clinical diagnosis of epilepsy.

This study develops a feature-based Automatic Modulation Classification (AMC) algorithm for spatially multiplexed Multiple-Input Multiple-Output (MIMO) systems employing two Higher Order Cumulants (HOCs) of the estimated transmit signal streams as discriminating features and a multiclass Support Vector Machine (SVM) as a classification system. The algorithm under study has the capability to recognize a wide range of modulation schemes without any prior information about the channel state. The classification performance of the proposed algorithm was evaluated via extensive simulations under different operating conditions and was also compared with the one obtained with the optimal Hybrid Likelihood Ratio Test (HLRT) approach. The results show that the proposed algorithm is capable of classifying the considered modulation schemes with good classification accuracy and can achieve performance comparable to that of the HLRT approach while having a significantly lower computational complexity.

International audience; In many countries, the motorcyclist fatality rate is much higher than that of other vehicle drivers. Among many other factors, motorcycle rear-end collisions are also contributing to these biker fatalities. To increase the safety of motorcyclists and minimize their road fatalities, this paper introduces a vision-based rear-end collision detection system. The binary road detection scheme contributes significantly to reduce the negative false detections and helps to achieve reliable results even though shadows and different lane markers are present on the road. The methodology is based on Harris corner detection and Hough transform. To validate this methodology, two types of dataset are used: (1) self-recorded datasets (obtained by placing a camera at the rear end of a motorcycle) and (2) online datasets (recorded by placing a camera at the front of a car). This method achieved 95.1% accuracy for the self-recorded dataset and gives reliable results for the rear-end vehicle detections under different road scenarios. This technique also performs better for the online car datasets. The proposed technique's high detection accuracy using a monocular vision camera coupled with its low computational complexity makes it a suitable candidate for a motorbike rear-end collision detection system. (C) 2017 SPIE and IS&T.