Showing papers by "Goutam Saha published in 2020"

FT-SDN: A Fault-Tolerant Distributed Architecture for Software Defined Network

Abstract: The traditional Software Defined Network (SDN) architecture is based on single controller in the Control Plane. Therefore, network functioning become highly dependent on the performance of the single controller in the Control Plane, which is undesirable for any reliable application. Despite many advantages of SDN, its deployment in the practical field is restricted since reliability and fault-tolerance capabilities of the system are not satisfactory. To overcome these difficulties of SDN, (FT-SDN) architecture has been proposed. The proposed architecture consists of a simple and effective distributed Control Plane with multiple controllers. FT-SDN uses a synchronized mechanism to periodically update the controller’s state within themselves. In case of failure, FT-SDN has the ability to select another working controller based on the distance and delays among different network entities. The performance of the FT-SDN architecture was examined with respect to different specifications in the presence of faults. Experimentation was done in simulation where results were found to be satisfactory.

6LE-SDN: An Edge-Based Software-Defined Network for Internet of Things

Abstract: IPv6 over low-power wireless personal area network (6LoWPAN) has been widely used for large-scale sensing and actuating purposes in the Internet of Things (IoT). Though promising, many challenges, such as high latency, heterogeneity, and packet loss persist. To mitigate these challenges, the software-defined network (SDN) technique can be hybridized with existing IoT structures that can address many of them. In this article, we propose an approach—edge-based 6LoWPAN-SDN (6LE-SDN) architecture, which can improve the system limitations mentioned. It uses an edge-based computational capability to improve the network performance over 6LoWPAN. To reduce heterogeneity, we develop a hybrid-edge switch that helps to enable communication among 6LoWPAN and SDN entities. For efficient communication between different devices, a new protocol—edge-based 6LoWPAN-SDN protocol (6LE-SDNP) is proposed, which is capable of ensuring optimal routing of the packet for efficient communication among the devices. We use the SDN-based edge controller for reducing the latency of the network apart from improving the interoperability feature. The testbed evaluation of the proposed solution indicates satisfactory performance in terms of reducing latency by 60% and network overhead by 91%. The 6LE-SDN network also succeeded in reducing the average round trip time (RTT) by 31% and the packet loss by 70% as compared to that of the traditional 6LoWPAN-based IoT.

A Fusion Based Classification of Normal, Arrhythmia and Congestive Heart Failure in ECG

Abstract: In healthcare, Electrocardiogram (ECG) signal is considered important to study life-threatening heart diseases that include arrhythmia (ARR), congestive heart failure (CHF). Mostly, atrial arrhythmia leads to CHF. Previous studies on ARR and CHF are focused on the binary classification of each category against normal sinus rhythm (NSR). So, there is a requirement to study the above disease cases together to detect the severity of the situation and take remedial action accordingly. The goal of this study is to analyse and classify these three different classes of ECG (namely ARR, CHF, and NSR) in an efficient way. We used 30 ECG recordings for each of the classes from the publicly available Physionet database. Since the temporal and spectral features by themselves may be insufficient to distinguish the classes, we sought to combine information across both. Accordingly, we considered feature representations from heart rate variability (HRV) of the ECG signal and wavelet-based features together with auto-regressive coefficients. To leverage complementary information across feature types, we employed feature-level fusion. We examined the performance of individual and fused feature types with multiple classifiers. The highest accuracy of 93.33% for three-class classification was obtained after feature fusion using Support Vector Machine (SVM). Although the performance of HRV features is relatively poor compared to wavelet-based features, their fusion improved the classification accuracy.

Windkessel Model-Based Cuffless Blood Pressure Estimation Using Continuous Wave Doppler Ultrasound System

Abstract: Blood pressure (BP) measurement plays an essential role in the prevention of cardiovascular diseases. Studies have demonstrated Ultrasound (US) based BP analysis method combining with peripheral components of the circulatory system. In this paper, a cuffless BP measurement technique has been proposed using a portable continuous wave Doppler US system which consumes < 4 Watt of power. The US blood flow signal acquired utilizing 8 MHz pencil transducer probe from the brachial artery is denoised using soft thresholding method. The spectrogram envelope of maximum frequency is obtained by an adaptive signal noise slope intersection (SNSI) method to extract hemodynamic features. In the proposed method, 2-element Windkessel (WK) model consisting of peripheral resistance and arterial compliance is employed for BP estimation. Based on the extracted features, a machine learning algorithm determines the WK model parameters. From the experiments conducted on 85 subjects, it has been observed that both systolic and diastolic BP achieve Grade B and Grade C for British Hypertension Society (BHS) and IEEE Std. 1708 protocols, respectively. Regarding Association for the Advancement of Medical Instrumentation (AAMI) standard, diastolic BP estimation error is within an acceptable limit. The robustness of the approach is examined using pre-exercise and post-exercise performance of 10 subjects. Moreover, the effect on arterial compliance for increased BP and aging is observed to characterize the dynamic property of arterial system. The proposed method is non-invasive, non-occlusive, independent of any additional interfaces, operates with a small dataset and worthy of implementation as a portable system for point-of-care application.

Classification of Autism in Young Children by Phase Angle Clustering in Magnetoencephalogram Signals

Abstract: Autism spectrum disorder (ASD) is a complex neu- rodevelopmental condition that appears in early childhood or infancy, causing delays or impairments in social interaction and restricted range of interests of a child. In this work, our goal is to classify autistic children from typically developing children using a machine learning framework. Here, we have used magnetoencephalography (MEG) signals of thirty age and gender matched children from each group. We perform a spectral domain analysis in which the features are extracted from both power and phase of large-scale neural oscillations. In this work, we propose a novel phase angle clustering (PAC) based feature and have compared its performance with commonly used power spectral density (PSD) based feature. It is observed that with Artificial Neural Network (ANN) classifier, PAC yields better classification accuracy (88.20±3.87%) than the PSD feature (82.13±2.11%). To investigate laterality of brain activity, we evaluate the classification performance of each feature type over all channels as well as over individual hemispheres. Using machine learning framework it is found that the discriminating PSD features are mostly from high gamma band i.e. 50–100 Hz frequency oscillations and the PSD features are dominant in right hemisphere. These findings are in line with studies carried before in other framework. However, PAC based feature in our study shows that the whole brain contains important attributes of autism. The discriminating PAC features are mostly from theta band (i.e. 4–8 Hz frequency oscillations) that signifies memory formation and navigation. In this study, it is found that impaired theta oscillations correlate with autistic symptoms. Overall, our findings show the potential of such signal processing and classification based study to aid the clinicians in diagnosis of ASD.

Modified Half-System Based Method for Reverse Engineering of Gene Regulatory Networks

Abstract: The accurate reconstruction of gene regulatory networks for proper understanding of the intricacies of complex biological mechanisms still provides motivation for researchers. Due to accessibility of various gene expression data, we can now attempt to computationally infer genetic interactions. Among the established network inference techniques, S-system is preferred because of its efficiency in replicating biological systems though it is computationally more expensive. This provides motivation for us to develop a similar system with lesser computational load. In this work, we have proposed a novel methodology for reverse engineering of gene regulatory networks based on a new technique: half-system . Half-systems use half the number of parameters compared to S-systems and thus significantly reduce the computational complexity. We have implemented our proposed technique for reconstructing four benchmark networks from their corresponding temporal expression profiles: an 8-gene, a 10-gene, and two 20-gene networks. Being a new technique, to the best of our knowledge, there are no comparable results for this in the contemporary literature. Therefore, we have compared our results with those obtained from the contemporary literature using other methodologies, including the state-of-the-art method, GENIE3 . The results obtained in this work stack favourably against the competition, even showing quantifiable improvements in some cases.

Sensor based Detection of Parkinson’s Disease Motor Symptoms

Abstract: Parkinson’s disease (PD) is consists of several symptoms categorically motor and non-motor symptoms. Motor symptoms are the main key symptoms for identifying the PD. In general Resting Tremor, Bradykinesia, Postural Instability and Rigidity are the major symptoms that are found in most PD patients. For measuring these symptoms there are several sensor based instruments and hardware can be developed. In this paper we will discuss about those sensor and instruments that can be helpful to measure these four motor symptoms.

A Hybrid Methodology for the Reverse Engineering of Gene Regulatory Networks

Abstract: In this work, a computational approach has been proposed based on the hybridisation of two modelling formalisms, recurrent neural networks and half-systems, for the reconstruction of gene regulatory networks from time-series gene expression datasets. To the best of our knowledge, the proposed hybridisation has not been attempted previously in this domain. Here, recurrent neural networks and half-systems have been hybridised to capture the underlying dynamics present in the temporal gene expression profiles. The motivation behind this work is to integrate the advantages of both the techniques in the proposed model such that the problem of reverse engineering of gene regulatory networks can be resolved more efficiently. Artificial bee colony optimisation has been used for the estimation of the model parameters. We have implemented the proposed hybrid methodology on the real-world experimental datasets (in vivo) of the SOS DNA Repair network of Escherichia coli. The obtained results are comparable to or better than that of other reverse engineering methodologies present in contemporary literature.

Systematic and Symptomatic Review for Parkinson’s Disease

Abstract: After two hundred years of the Shaking Palsy by Dr. James Parkinson, we have revealed many factors and causes behind Parkinson’s disease (PD). Before Shaking Palsy the symptoms were known as some disorders. 5000 years ago in Ayurveda, the Indian medical manuscript and 2500 years ago in Nei Ping, the first Chinese medical manuscript some disorders were mentioned those are similar to PD and also several treatment procedures were also described. This proves that PD is not a disease that evolved only in modern industrial age. But it is true that, after Shaking Palsy, this PD comes into spotlight of the modern medical practices. To understand this disease we need to go through the complete etiology of PD including Genetical and Environmental factor that may lead us to the different causative factors of PD. Each factor has its unique signification and outburst as a multiple combination of different Parkinsonian symptoms. In this review we will discuss about all possible etiological factors related to PD with different Parkinsonian symptoms categorized with motor and non-motor symptoms. Before going to the brief review, we will also discuss about several milestone researches which actually open a new window in PD research on its time.

Biomarker Gene Identification Using a Quantum Inspired Clustering Approach

Abstract: In this paper, we have implemented an unsupervised approach for finding out the significant genes from microarray gene expression datasets. The proposed method is based on implements a quantum clustering approach to represent gene-expression data as equations and uses the procedure to search for the most probable set of clusters given the available data. The main contribution of this approach lies in the ability to take into account the essential features or genes using clustering. Here, we present a novel clustering approach that extends ideas from scale-space clustering and support-vector clustering. This clustering method is used as a feature selection method. Our approach is fundamentally based on the representation of datapoints or features in the Hilbert space, which is then represented by the Schrodinger equation, of which the probability function is a solution. This Schrodinger equation contains a potential function that is extended from the initial probability function.The minima of the potential values are then treated as cluster centres. The cluster centres thus stand out as representative genes. These genes are evaluated using classifiers, and their performance is recorded over various indices of classification. From the experiments, it is found that the classification performance of the reduced set is much better than the entire dataset.The only free-scale parameter, sigma, is then altered to obtain the highest accuracy, and the corresponding biological significance of the genes is noted.

Effect of Familiarity on Recognition of Pleasant and Unpleasant Emotional States Induced by Hindi Music Videos

Abstract: Valence is an important dimension representing the hedonic value of emotion labeled as positive or negative. Inducing these emotions and making an understanding from brain responses have huge practical significance. Music is a powerful tool to induce emotions, and the induced emotions are generally getting influenced by factors external to music such as familiarity. This work presents a novel study on the effect of familiarity on recognition of pleasant (positive) and unpleasant (negative) emotional states induced by Hindi music videos. For this, we recorded 32-channel EEG from six healthy subjects while they watched Hindi music videos and self-reported ratings of felt emotions on valence and familiarity scale. We used a machine learning framework for emotion classification from power spectral and functional connectivity features. The framework consists of SVD-QRcp and F-ratio based feature selection and an SVM classifier. The classification was performed under three cases of familiarity, namely, familiar, unfamiliar, and regardless of familiarity of the music videos. We found that for the familiar case, the classification performance was higher than unfamiliar and regardless of familiarity cases for all considered features. The best performing features were from the individual electrodes and these features were from the frontal and left parietal regions which indicate the lateralized processing of valence. In addition to classification, we analyzed the feature and electrode usage for all the cases of familiarity. It was found that the features from theta, alpha, and gamma band covering the frontal and parietal brain regions were dominantly involved.

Linking flexibility of brain networks to cognitive development in preschool children

Abstract: Recent studies of functional connectivity networks (FCNs) suggest that the reconfiguration of brain network across time, both at rest and during task, is linked with cognition in human adults. In this study, we tested this prediction, i.e. cognitive ability is associated with a flexible brain network in preschool children of 3-4 years - a critical age, representing a ‘blossoming period’ for brain development. We recorded magnetoen-cephalogram (MEG) data from 88 preschoolers, and assessed their cognitive ability by a battery of cognitive tests. We estimated FCNs obtained from the source reconstructed MEG recordings, and characterized the temporal variability at each node using a novel path-based measure of temporal variability; the latter captures reconfiguration of the node’s interactions to the rest of the network across time. Using connectome predictive modeling, we demonstrated that the temporal variability of fronto-temporal nodes in the dynamic FCN can reliably predict out-of-scanner performance of short-term memory and attention distractability in novel participants. Further, we observed that the network-level temporal variability increased with age, while individual nodes exhibited an inverse relationship between temporal variability and node centrality. These results demonstrate that functional brain networks, and especially their reconfiguration ability, are important to cognition at an early but a critical stage of human brain development.

A Study on Abnormalities Detection Techniques from Echocardiogram

Abstract: Echocardiography is one of the most widely used tools in abnormalities detection in cardiac perspective. A person with difficulty in breathing or any symptoms that shows a weak heart is asked to follow the test. This test is vital and is done manually where a transducer is used to obtain a specific image that can visually locate the presence of abnormalities. Automated methodologies have emerged to solve the problem faced by manual treatment. This will help the physician to reduce misdiagnosis of echo images. This paper is based on the study of different existing techniques that can be used in the detection of abnormalities in cardiac system using echo images.

Mitigating the Effects of External Perturbations on a Gene Regulatory Network using Feedback Controllers

Abstract: Gene regulatory networks are generally robust in nature. However, unwanted perturbations arising out of extreme environmental conditions or external pathogen attacks may lead them to malfunction. Potentially, this can have an adverse effect on the biochemical functions of a living system. In this work, we have proposed a computational model based on negative feedback control to eliminate the effects of such unwanted perturbations. We have implemented the recurrent neural network formalism for modelling the underlying network dynamics from a given time-series gene expression dataset. The artificial bee colony optimisation technique has been employed for model parameter estimation. The controller used in this work is of the proportional-integral-derivative type. To the best of our knowledge, this is one of the first research works in this domain to consider a completely non-linear scenario. A 10-gene DREAM4 benchmark network has been considered in this work. The results obtained herein show that the proposed formalism can mitigate the unwanted effects of external disturbances effectively.

A Low-Error, Memory-Based Fast Binary Logarithmic Converter

Abstract: This brief presents a memory based fast binary logarithmic converter based on a piecewise linear approximation technique. The proposed method is simple and arithmetic operation-less, which achieves 10−4 to 10−3 maximum absolute error (MAE) while maintaining a high speed. The approach partitions the logarithmic curve of the fractional component into $2^{L}$ uniform regions and a block RAM (size $2^{L} \times $ bits) stores the approximate value of each sub-region. For any number, most significant $L$ bits of the fractional component address the memory location of the logarithmic converter. The hardware synthesis result, implemented with 26 bits fractional precision on Virtex-6 field-programmable gate array device, shows 75% improvement in MAE and 27% decrease in critical path delay compared to the current state-of-the-art techniques in the worst-case scenario. In Otsu’s image thresholding algorithm, the proposed logarithmic converter with 3.08 ( $=2^{8} \times 12$ ) kbits memory size adequately meets the accuracy requirement for improved image segmentation.

Feature Selection for High-Dimensional Data Through Instance Vote Combining

Abstract: Supervised feature selection (FS) is used to select a discriminative and non-redundant subset of features in classification problems dealing with high dimensional inputs. In this paper, feature selection is posed akin to the set-covering problem where the goal is to select a subset of features such that they cover the instances. To solve this formulation, we quantify the local relevance (i.e., votes assigned by instances) of each feature that captures the extent to which a given feature is useful to classify the individual instances correctly. In this work, we propose to combine the instance votes across features to infer their joint local relevance. The votes are combined on the basis of geometric principles underlying classification and feature spaces. Further, we show how such instance vote combining may be employed to derive a heuristic search strategy for selecting a relevant and non-redundant subset of features. We illustrate the effectiveness of our approach by evaluating the classification performance and robustness to data variations on publicly available benchmark datasets. We observed that the proposed method outperforms state-of-the-art mutual information based FS techniques and performs comparably to other heuristic approaches that solve the set-covering formulation of feature selection.

Incorporating Multiple Assay Operations and Cross Contamination Avoidance in Digital Microfluidic Biochips with Reduced Number of Pins

Abstract: Digital Microfluidic Biochips (DMFBs) are becoming more and more capable in areas of biomedical science, biochemistry, and also in microelectronics. It is popularly known as lab-on-a-chip and as the name suggests, laboratory experiments are carried out in it. Multiple assay operations can be performed here effectively and if possible, simultaneously. Thus, in this perspective, parallelism is having tremendous effect in designing biochips, while the dimension of a chip is a limit. In order to enhance the credibility of a chip, one must take care of its throughput, electrode utilization, and pin count as well. If the number of pins that can accomplish all the desired activities of multiple bioassay operations in synchronization can be minimized, the design can be treated as efficient and cost-effective. Earlier work shows that 21 pins are the minimum requirement to achieve the said goal. Whereas, in this article, we have drastically reduced the total number of pins for 15 × 15 arrays. The algorithm developed in this paper requires only 11 pins for the desired tasks. This design is able to avoid the problem of cross contamination, and it has been validated for some considerable real-life assay operations. Thus, in this paper, we are able to show the multiple assay operations in parallel.