Search

This paper introduces a novel shallow 3D self-supervised tensor neural network for volumetric segmentation of medical images with merits of obviating training and supervision. The proposed network is referred to as 3D Quantum-inspired Self-supervised Tensor Neural Network (3D-QNet). The underlying architecture of 3D-QNet is composed of a trinity of volumetric layers viz. input, intermediate and output layers inter-connected using an S-connected third-order neighborhood-based topology for voxel-wise processing of 3D medical image data suitable for semantic segmentation. Each of the volumetric layers contains quantum neurons designated by qubits or quantum bits. The incorporation of tensor decomposition in quantum formalism leads to faster convergence of the network operations to preclude the inherent slow convergence problems faced by the classical supervised and self-supervised networks. The segmented volumes are obtained once the network converges. The suggested 3D-QNet is tailored and tested on the BRATS 2019 Brain MR image data set and Liver Tumor Segmentation Challenge (LiTS17) data set extensively in our experiments. 3D-QNet has achieved promising dice similarity as compared to the intensively supervised convolutional network-based models like 3D-UNet, Vox-ResNet, DRINet, and 3D-ESPNet, showing a potential advantage of our self-supervised shallow network on facilitating semantic segmentation.

Due to the lack of appropriate tailoring of the inter-connection weights, the segmentation performance of the recently suggested Quantum-inspired Self-supervised Neural Network models suffers from the slow convergence problem. As a result, using quantum-inspired meta-heuristics in Quantum-Inspired Self-supervised Neural Network models improves their hyper-parameters and inter-connection weights. The goal of this paper is to propose an improved version of a Quantum-Inspired Self-supervised Neural Network (QIS-Net) model for brain lesion segmentation. The proposed Optimized Quantum-Inspired Self-supervised Neural Network (Opti-QISNet) model is based on the QIS-Net architecture, and its operations are used to get the best segmentation results. A Quantum-Inspired Optimized Multi-Level Sigmoidal (Opti-QSig) activation is the optimized activation function used in the described model. Three quantum-inspired meta-heuristics improve the Opti-QSig activation function, with fitness evaluated using Otsu’s multi-level thresholding. Experiments were carried out using brain MR images from the Cancer Imaging Archive (TCIA) in the Nature data repository. The results show that the proposed self-supervised Opti-QISNet model outperforms our recently established QIBDS Net and QIS-Net models in brain lesion segmentation, and it is a potential candidate to extensively supervised neural network based architectures (UNet and FCNNs).

A hybrid approach based on multirate signal processing and sensory data fusion is proposed for the condition monitoring and identification of fault signal signatures used in the Flight ECS (Engine Control System) unit. Though motor current signature analysis (MCSA) is widely used for fault detection now-a-days, the proposed hybrid method qualifies as one of the most powerful online/offline techniques for diagnosing the process faults. Existing approaches have some drawbacks that can degrade the performance and accuracy of a process-diagnosis system. In particular, it is very difficult to detect random stochastic noise due to the nonlinear behavior of valve controller. Using only Short Time Fourier Transform (STFT), frequency leakage and the small amplitude of the current components related to the fault can be observed, but the fault due to the controller behavior cannot be observed. Therefore, a framework of advanced multirate signal and data-processing aided with sensor fusion algorithms is proposed in this article and satisfactory results are obtained. For implementing the system, a DSP-based BLDC motor controller with three-phase inverter module (TMS 320F2812) is used and the performance of the proposed method is validated on real time data., Comment: 7 Pages, 7 figures

In recent times, non-intrusive load monitoring (NILM) has emerged as an important tool for distribution-level energy management systems owing to its potential for energy conservation and management. However, load monitoring in smart building environments is challenging due to high variability of real-time load and varied load composition. Furthermore, as the volume and dimensionality of smart meters data increases, accuracy and computational time are key concerning factors. In view of these challenges, this paper proposes an improved NILM technique using multi-head (Mh-Net) convolutional neural network (CNN) under dynamic grid voltage conditions. An attention layer is introduced into the proposed CNN model, which helps in improving estimation accuracy of appliance power consumption. The performance of the developed model has been verified on an experimental laboratory setup for multiple appliance sets with varied power consumption levels, under dynamic grid voltages. Moreover, the effectiveness of the proposed model has been verified on widely used UK-DALE data, and its performance has been compared with existing NILM techniques. Results depict that the proposed model accurately identifies appliances, power consumptions and their time-of-use even during practical dynamic grid voltage conditions.

Current-fed half bridge converter with bidirectional switches on ac side and a full bridge converter on dc side of a high frequency transformer is an optimal topology for single stage galvanically isolated ac-dc converter for onboard vehicle charging application. AC side switches are actively commutated to achieve zero current switching (ZCS) using single phase shift modulation (SPSM) and discontinuous current phase shift modulation (DCPSM). Furthermore, zero voltage turn-on (ZVS) is achieved for dc side switches. Compared to SPSM, DCPSM maintains a constant peak current in the converter throughout the grid cycle of ac mains voltage. However, constant peak current contributes to a high circulating current near the zero crossings of ac mains voltage and also at light load conditions. This paper proposes an improved discontinuous current phase shift modulation (IDCPSM) to increase the efficiency of the converter across different loading conditions. A dual control variable is adopted to actively reduce the circulating current while maintaining soft switching of both ac and dc side switches across the grid cycle of ac mains voltage. A 1.5 kW laboratory prototype has been developed to experimentally validate the analysis, design and improvement in performance for different loading conditions., Comment: This article has been submitted to IEEE Transactions on Power Electronics for review

Background and Objective: Tuberculosis (TB) is a significant public health issue and a leading cause of death worldwide. Millions of deaths can be averted by early diagnosis and successful treatment of TB patients. Automated diagnosis of TB holds vast potential to assist medical experts in expediting and improving its diagnosis, especially in developing countries like India, where there is a shortage of trained medical experts and radiologists. To date, several deep learning based methods for automated detection of TB from chest radiographs have been proposed. However, the performance of a few of these methods on the Indian chest radiograph data set has been suboptimal, possibly due to different texture of the lungs on chest radiographs of Indian subjects compared to other countries. Thus deep learning for accurate and automated diagnosis of TB on Indian datasets remains an important subject of research. Methods: The proposed work explores the performance of convolutional neural networks (CNNs) for the diagnosis of TB in Indian chest x-ray images. Three different pre-trained neural network models, AlexNet, GoogLenet, and ResNet are used to classify chest x-ray images into healthy or TB infected. The proposed approach does not require any pre-processing technique. Also, other works use pre-trained NNs as a tool for crafting features and then apply standard classification techniques. However, we attempt an end to end NN model based diagnosis of TB from chest x-rays. The proposed visualization tool can also be used by radiologists in the screening of large datasets. Results: The proposed method achieved 93.40% accuracy with 98.60% sensitivity to diagnose TB for the Indian population. Conclusions: The performance of the proposed method is also tested against techniques described in the literature. The proposed method outperforms the state of art on Indian and Shenzhen datasets.

A chest radiograph, commonly called chest x-ray (CxR), plays a vital role in the diagnosis of various lung diseases, such as lung cancer, tuberculosis, pneumonia, and many more. Automated segmentation of the lungs is an important step to design a computer-aided diagnostic tool for examination of a CxR. Precise lung segmentation is considered extremely challenging because of variance in the shape of the lung caused by health issues, age, and gender. The proposed work investigates the use of an efficient deep convolutional neural network for accurate segmentation of lungs from CxR. We attempt an end to end DeepLabv3+ network which integrates DeepLab architecture, encoder-decoder, and dilated convolution for semantic lung segmentation with fast training and high accuracy. We experimented with the different pre-trained base networks: Resnet18 and Mobilenetv2, associated with the Deeplabv3+ model for performance analysis. The proposed approach does not require any pre-processing technique on chest x-ray images before being fed to a neural network. Morphological operations were used to remove false positives that occurred during semantic segmentation. We construct a CxR dataset of the Indian population that contain healthy and unhealthy CxRs of clinically confirmed patients of tuberculosis, chronic obstructive pulmonary disease, interstitial lung disease, pleural effusion, and lung cancer. The proposed method is tested on 688 images of our Indian CxR dataset including images with severe abnormal findings to validate its robustness. We also experimented on commonly used benchmark datasets such as Japanese Society of Radiological Technology; Montgomery County, USA; and Shenzhen, China for state-of-the-art comparison. The performance of our method is tested against techniques described in the literature and achieved the highest accuracy for lung segmentation on Indian and public datasets.

Wavelet transform is proposed in this paper for detection of islanding and fault disturbances distributed generation (DG) based power system. An IEEE 14-bus system with DG penetration is considered for the detection of disturbances under different operating conditions. The power system is a hybrid combination of photovoltaic, and wind energy system connected to different buses with different level of penetration. The voltage signal is retrieved at the point of common coupling (PCC) and processed through wavelet transform to detect the disturbances. Further, energy and standard deviation (STD) as performance indices are evaluated and compared with a suitable threshold in order to analyze a disturbance condition. Again, a comparative analysis between the existing and proposed detection is studied to prove the better performance of wavelet transform., Comment: 5 pages, 4 figures, 1 table

Uninterruptible power supply is the main motive of power utility companies that motivate them for identifying and locating the different types of faults as quickly as possible to protect the power system prevent complete power black outs using intelligent techniques. Thus, the present research work presents a novel method for detection of fault disturbances based on Wavelet Transform (WT) and Independent Component Analysis (ICA). The voltage signal is taken offline under fault conditions and is being processed through wavelet and ICA for detection. The time-frequency resolution from WT transform detects the fault initiation instant in the signal. Again, a performance index is calculated from independent component analysis under fault condition which is used to detect the fault disturbance in the voltage signal. The proposed approach is tested to be robust enough under various operating scenarios like without noise, with 20-dB noise and variation in frequency. Further, the detection study is carried out using a performance index, energy content, by applying the existing Fourier transform (FT), short time Fourier transform (STFT) and the proposed wavelet transform. Fault disturbances are detected if the energy calculated in each scenario is greater than the corresponding threshold value. The fault detection study is simulated in MATLAB/Simulink for a typical power system., Comment: 5 pages, 6 figures, Table 1

Extensive first-principle calculations on embedded clusters containing few O, Y, Ti, and Cr atoms as well as vacancies are performed to obtain interaction parameters to be applied in Metropolis Monte Carlo simulations, within the framework of a rigid lattice model. A novel description using both pair and triple parameters is shown to be more precise than the commonly used pair parameterization. Simulated annealing provides comprehensive data on the energetics, structure and stoichiometry of nm-size clusters at . Additionally, Metropolis Monte Carlo simulations are carried out at high temperature in order to investigate the dependence of nanocluster composition on temperature. The absolute value of the binding energy per O atom unit increases with cluster size and approaches a constant for large clusters. The presence of Ti and/or vacancies increases the value of this quantity. In alloys without vacancies clusters show a planar structure, whereas the presence of vacancies leads to three-dimensional configurations. Cr is not part of the nanoclusters, except for alloys without Ti but with vacancies. In the latter case clusters consist of a core containing O, vacancies, as well as Y and a Cr shell, which was also observed experimentally. A good agreement between the existing experimental data on the ratios (Y+Ti):O, Y:Ti, (Y+Cr):O, and Y:Cr, and the simulation results is found. The comparison of experimental data with those obtained by simulations demonstrates that the assumption of nanoclusters consisting of nonstoichiometric oxides that are essentially coherent with the bcc lattice of the Fe-Cr matrix leads to reasonable results.

We report the solid state phase diagram of V-Nb, V-Ta and Nb-Ta alloys computed by combining the density functional theory total energies with the cluster expansion and Monte Carlo techniques. From the computed phase diagrams, we find that V-Nb and Nb-Ta alloys form continuous series of solid solutions and the solid solution phase is stable down to ambient temperatures, consistent with experiments. The calculated bcc V-Ta phase diagram exhibits complete miscibility. Since the current cluster expansion ignore V2 Ta phase, the chemical interaction due to relatively large electronegativity difference, which cause the ordering of V2 Ta phase from the bcc solid solution, appears to manifest by making the solid solution phase remain stable for the complete concentration range, down to ambient temperatures, perhaps with some short-range-order. This work further demonstrates the dominant role of constituent strains in the accurate calculation of phase diagram of alloys of constituents with significant size mismatches.

We propose the particle swarm optimization (PSO) trained auto associative neural network (AANN) as a single class classifier (PSOAANN). The proposed architecture consists of three layers namely input layer, hidden layer and output layer unlike that of the traditional AANN. The efficacy of the proposed single class classifier is evaluated on bankruptcy prediction datasets namely Spanish banks, Turkish banks, US banks and UK banks; UK credit dataset and the benchmark WBC dataset. PSOAANN achieved better results when compared to Modified Great Deluge Algorithm trained auto associative neural network (MGDAAANN) [1]. It is concluded that PSOAANN as a single class classifier can be used as an effective tool in classifying datasets, where the class of interest (usually the positive class) is either totally missing or disproportionately present in the training data, which is the case in many real life problems for e.g. financial fraud detection.

We propose the particle swarm optimization (PSO) trained auto associative neural network (AANN) as a single class classifier (PSOAANN). The proposed architecture consists of three layers namely input layer, hidden layer and output layer unlike that of the traditional AANN. The efficacy of the proposed single class classifier is evaluated on bankruptcy prediction datasets namely Spanish banks, Turkish banks, US banks and UK banks; UK credit dataset and the benchmark WBC dataset. PSOAANN achieved better results when compared to Modified Great Deluge Algorithm trained auto associative neural network (MGDAAANN) [1]. It is concluded that PSOAANN as a single class classifier can be used as an effective tool in classifying datasets, where the class of interest (usually the positive class) is either totally missing or disproportionately present in the training data, which is the case in many real life problems for e.g. financial fraud detection.

We propose the particle swarm optimization (PSO) trained auto associative neural network (AANN) as a single class classifier (PSOAANN). The proposed architecture consists of three layers namely input layer, hidden layer and output layer unlike that of the traditional AANN. The efficacy of the proposed single class classifier is evaluated on bankruptcy prediction datasets namely Spanish banks, Turkish banks, US banks and UK banks; UK credit dataset and the benchmark WBC dataset. PSOAANN achieved better results when compared to Modified Great Deluge Algorithm trained auto associative neural network (MGDAAANN) [1]. It is concluded that PSOAANN as a single class classifier can be used as an effective tool in classifying datasets, where the class of interest (usually the positive class) is either totally missing or disproportionately present in the training data, which is the case in many real life problems for e.g. financial fraud detection.

We propose the particle swarm optimization (PSO) trained auto associative neural network (AANN) as a single class classifier (PSOAANN). The proposed architecture consists of three layers namely input layer, hidden layer and output layer unlike that of the traditional AANN. The efficacy of the proposed single class classifier is evaluated on bankruptcy prediction datasets namely Spanish banks, Turkish banks, US banks and UK banks; UK credit dataset and the benchmark WBC dataset. PSOAANN achieved better results when compared to Modified Great Deluge Algorithm trained auto associative neural network (MGDAAANN) [1]. It is concluded that PSOAANN as a single class classifier can be used as an effective tool in classifying datasets, where the class of interest (usually the positive class) is either totally missing or disproportionately present in the training data, which is the case in many real life problems for e.g. financial fraud detection.

We propose the particle swarm optimization (PSO) trained auto associative neural network (AANN) as a single class classifier (PSOAANN). The proposed architecture consists of three layers namely input layer, hidden layer and output layer unlike that of the traditional AANN. The efficacy of the proposed single class classifier is evaluated on bankruptcy prediction datasets namely Spanish banks, Turkish banks, US banks and UK banks; UK credit dataset and the benchmark WBC dataset. PSOAANN achieved better results when compared to Modified Great Deluge Algorithm trained auto associative neural network (MGDAAANN) [1]. It is concluded that PSOAANN as a single class classifier can be used as an effective tool in classifying datasets, where the class of interest (usually the positive class) is either totally missing or disproportionately present in the training data, which is the case in many real life problems for e.g. financial fraud detection.

Oxide Dispersion Strengthened (ODS) steels are candidate materials for use in nuclear environment as they offer excellent swelling resistance characteristics and enhanced high temperature mechanical properties. ODS steels derive their strength from the dispersion of oxide (Y-Ti-O) nanoclusters in the ferritic matrix. The mechanism of formation, structure, thermal & radiation stability and composition of these nanoclusters are not yet fully understood. We are studying the basic energetic and kinetic properties of solutes and their clusters along with vacancy in bcc Fe using density functional theory and evolution of these nanoclusters using kinetic Monte Carlo simulations. The diffusion coefficients of these solute atoms in bcc Fe have been calculated using Le Claire's nine frequency model. Among these, Y is believed to be rate limiting for the growth of nanoclusters and our calculated diffusion coefficient is found to be three orders of magnitude higher than the experimental one [J. Nucl. Mater. 419, 208 (2011)]. In order to further understand the diffusion behavior of Y in bcc Fe and hence its role in the kinetics of nanocluster formation, we review the Le Claire's diffusion model both in dilute and concentrated limits. Also, in continuation of our previous study of solute atom interaction with vacancy, we discuss our recent results of DFT calculations of solute atom interactions with self-interstitials.

Oxide Dispersion Strengthened (ODS) steels are candidate materials for use in nuclear environment as they offer excellent swelling resistance characteristics and enhanced high temperature mechanical properties. ODS steels derive their strength from the dispersion of oxide (Y-Ti-O) nanoclusters in the ferritic matrix. The mechanism of formation, structure, thermal & radiation stability and composition of these nanoclusters are not yet fully understood. We are studying the basic energetic and kinetic properties of solutes and their clusters along with vacancy in bcc Fe using density functional theory and evolution of these nanoclusters using kinetic Monte Carlo simulations. The diffusion coefficients of these solute atoms in bcc Fe have been calculated using Le Claire's nine frequency model. Among these, Y is believed to be rate limiting for the growth of nanoclusters and our calculated diffusion coefficient is found to be three orders of magnitude higher than the experimental one [J. Nucl. Mater. 419, 208 (2011)]. In order to further understand the diffusion behavior of Y in bcc Fe and hence its role in the kinetics of nanocluster formation, we review the Le Claire's diffusion model both in dilute and concentrated limits. Also, in continuation of our previous study of solute atom interaction with vacancy, we discuss our recent results of DFT calculations of solute atom interactions with self-interstitials.