Neural Network Pattern Recognition Research Articles

Detecting Prostate Cancer Using Pattern Recognition Neural Networks With Flow Cytometry-Based Immunophenotyping in At-Risk Men.

Current screening methods for prostate cancer (PCa) result in a large number of false positives making it difficult for clinicians to assess disease status, thus warranting advancements in screening and early detection methods. The goal of this study was to design a liquid biopsy test that uses flow cytometry–based immunophenotyping and artificial neural network (ANN) analysis to detect PCa. Numerous myeloid and lymphoid cell populations, including myeloid-derived suppressor cells, were measured from 156 patients with PCa, 123 with benign prostatic hyperplasia (BPH), and 99 male healthy donor (HD) controls. Using pattern recognition neural network (PRNN) analysis, a type of ANN, PCa detection compared against HD resulted in 96.6% sensitivity, 87.5% specificity, and an area under the curve (AUC) value of 0.97. Detecting patients with higher risk disease (⩾Gleason 7) against lower risk disease (BPH/Gleason 6) resulted in 92.0% sensitivity, 42.7% specificity, and an AUC of 0.72. This study suggests that analyzing flow cytometry immunophenotyping data with PRNNs may prove to be a useful tool to improve PCa detection and reduce the number of unnecessary prostate biopsies performed each year.

Designing architecture of an artificial neural network for classification using robotic complexes

This article describes the basic principles of constructing artificial neural networks for pattern recognition on raster images using robotic complexes. The basic parts of convolutional neural networks are considered. Examples of evaluating the accuracy of preliminary testing results developed by the authors of a convolutional neural network model are given.

Solving Overlapping Pattern Issues in On-Chip Learning of Bio-Inspired Neuromorphic System with Synaptic Transistors

Recently, bio-inspired neuromorphic systems have been attracting widespread interest thanks to their energy-efficiency compared to conventional von Neumann architecture computing systems. Previously, we reported a silicon synaptic transistor with an asymmetric dual-gate structure for the direct connection between synaptic devices and neuron circuits. In this study, we study a hardware-based spiking neural network for pattern recognition using a binary modified National Institute of Standards and Technology (MNIST) dataset with a device model. A total of three systems were compared with regard to learning methods, and it was confirmed that the feature extraction of each pattern is the most crucial factor to avoiding overlapping pattern issues and obtaining a high pattern classification ability.

Hybrid Pattern Recognition and Multi-resolution Analysis (MRA) Based Fault Location in Power Transmission Lines

This paper proposes a fault (line-to-line) location on Ikeja West – Benin 330kV electric power transmission lines using wavelet multi-resolution analysis and neural networks pattern recognition abilities. Three-phase line-to-line current and voltage waveforms measured during the occurrence of a fault in the power transmission-line were pre-processed first and then decomposed using wavelet multi-resolution analysis to obtain the high-frequency details and low-frequency approximations. The patterns formed based on high-frequency signal components were arranged as inputs of the neural network, whose task is to indicate the occurrence of a fault on the lines. The patterns formed using low-frequency approximations were arranged as inputs of the second neural network, whose task is to indicate the exact fault type. The new method uses both low and high-frequency information of the fault signal to achieve an exact location of the fault. The neural network was trained to recognize patterns, classify data and forecast future events. Feed forward networks have been employed along with back propagation algorithm for each of the three phases in the Fault location process. An analysis of the learning and generalization characteristics of elements in power system was carried using Neural Network toolbox in MATLAB/SIMULINK environment. Simulation results obtained demonstrate that neural network pattern recognition and wavelet multi-resolution analysis approach are efficient in identifying and locating faults on transmission lines as the average percentage error in fault location was just 0.1386%. This showed that satisfactory performance was achieved especially when compared to the conventional methods such as impedance and travelling wave methods.

DESIGN AND DEVELOPMENT OF HUMAN COMPUTER INTERFACE USING ELECTROOCULOGRAM WITH DEEP LEARNING.

Today's life assistive devices were playing significant role in our life to communicate with others. In that modality Human Computer Interface (HCI) based Electrooculogram (EOG) playing vital part. By using this method we can able to overcome the conventional methods in terms of performance and accuracy. To overcome such problem we analyze the EOG signal from twenty subjects to design nine states EOG based HCI using five electrodes system to measure the horizontal and vertical eye movements. Signals were preprocessed to remove the artifacts and extract the valuable information from collected data by using band power and Hilbert Huang Transform (HHT) and trained with Pattern Recognition Neural Network (PRNN) to classify the tasks. The classification results of 92.17% and 91.85% were shown for band power and HHT features using PRNN architecture. Recognition accuracy was analyzed in offline to identify the possibilities of designing HCI. We compare the two feature extraction techniques with PRNN to analyze the best method for classifying the tasks and recognizing single trail tasks to design the HCI. Our experimental result confirms that for classifying as well as recognizing accuracy of the collected signals using band power with PRNN shows better accuracy compared to other network used in this study. We compared the male subjects performance with female subjects to identify the performance. Finally we compared the male as well as female subjects in age group wise to identify the performance of the system. From that we concluded that male performance was appreciable compared with female subjects as well as age group between 26 to 32 performance and recognizing accuracy were high compared with other age groups used in this study.

Identifying User Suitability in sEMG Based Hand Prosthesis Using Neural Networks

Background: Identifying user suitability plays a vital role in various modalities like neuromuscular system research, rehabilitation engineering and movement biomechanics. This paper analysis the user suitability based on neural networks (NN), subjects, age groups and gender for surface electromyogram (sEMG) pattern recognition system to control the myoelectric hand. Six parametric feature extraction algorithms are used to extract the features from sEMG signals such as AR (Autoregressive) Burg, AR Yule Walker, AR Covariance, AR Modified Covariance, Levinson Durbin Recursion and Linear Prediction Coefficient. The sEMG signals are modeled using Cascade Forward Back propagation Neural Network (CFBNN) and Pattern Recognition Neural Network. Methods: sEMG signals generated from forearm muscles of the participants are collected through an sEMG acquisition system. Based on the sEMG signals, the type of movement attempted by the user is identified in the sEMG recognition module using signal processing, feature extraction and machine learning techniques. The information about the identified movement is passed to microcontroller wherein a control is developed to command the prosthetic hand to emulate the identified movement. Results: From the six feature extraction algorithms and two neural network models used in the study, the maximum classification accuracy of 95.13% was obtained using AR Burg with Pattern Recognition Neural Network. This justifies that the Pattern Recognition Neural Network is best suited for this study as the neural network model is specially designed for pattern matching problem. Moreover, it has simple architecture and low computational complexity. AR Burg is found to be the best feature extraction technique in this study due to its high resolution for short data records and its ability to always produce a stable model. In all the neural network models, the maximum classification accuracy is obtained for subject 10 as a result of his better muscle fitness and his maximum involvement in training sessions. Subjects in the age group of 26-30 years are best suited for the study due to their better muscle contractions. Better muscle fatigue resistance has contributed for better performance of female subjects as compared to male subjects. From the single trial analysis, it can be observed that the hand close movement has achieved best recognition rate for all neural network models. Conclusion: In this paper a study was conducted to identify user suitability for designing hand prosthesis. Data were collected from ten subjects for twelve tasks related to finger movements. The suitability of the user was identified using two neural networks with six parametric features. From the result, it was concluded thatfit women doing regular physical exercises aged between 26-30 years are best suitable for developing HMI for designing a prosthetic hand. Pattern Recognition Neural Network with AR Burg extraction features using extension movements will be a better way to design the HMI. However, Signal acquisition based on wireless method is worth considering for the future.

The differential expression of the salivary peptidome in radiotherapy-induced Xerostomia

Abstract Background Head and neck cancer patients always suffer from xerostomia after radiotherapy. The parasympathomimetic is an effective drug to relieve the radiotherapy-induced xerostomia. However, the unwanted side effect annoys patient a lot. The saliva substitute is commonly used replacement of the natural saliva for relieving the xerostomia symptom and improve the quality of life. The peptides in the saliva are the main compounds to keep the moisture of the oral cavity. In this study, we would like to analyze the differential expression of the peptide in the saliva of the subjects who had xerostomia and underwent head and neck radiotherapy and compared with normal subjects. We aim to find a specific peptide as a new ingredient for developing new saliva substitute which is more compatible with oral microenvironment of the patient to maintain the oral health and quality of life like the real saliva. Methods We collected the stimulated saliva from 10 head and neck cancer patients who underwent radiotherapy and had xerostomia symptom and from 10 age- and sex-matched controls. The xerostomia was evaluated with xerostomia inventory short version. The average xerostomia score of the patient group is 11.5, and the normal group is 5.5. We separated the peptide from saliva specimens with a 10K cut off filter. The differential expression of the peptide was performed with MALDI-TOF/TOF Ultraflex mass spectrometer and plausible neural network pattern recognition software. Results We found patients who receive radiotherapy lacks two major peaks of m/z 555.19 and 754.35. Those indicated the small fragment of the several possible peptides that lacking in the patient group and contribute to xerostomia symptom. Conclusion In the future work, we will add the candidate peptides into artificial saliva and perform the in vivo study to evaluate the potential of the peptide as a new ingredient of saliva substitute.

Advanced geological prediction technology of tunnel based on image recognition

The purpose of this study is to solve the problem that the image obtained by the advanced geological prediction technology of tunnel in the tunnel engineering construction project is difficult to identify, the identification result of the original image is too dependent on the personal habits and experience of the forecaster, and the identification result is not accurate enough. In this study, the geological radar image is taken as the research object, and the method of the advanced geological prediction technology of the tunnel is studied and summarized. The image pattern recognition technology is introduced into the advanced geological prediction of the tunnel, and the image recognition of the geological radar image is carried out. After preprocessing of filtering, binarization, and refinement, it is found that the image can be more accurate, and the impact of noise on the image is also alleviated. In addition, neural network pattern recognition technology is adopted after image preprocessing to establish a neural network model. After recognition, curve targets can be successfully detected, indicating the feasibility of image recognition technology. Therefore, the image recognition technology can make up for the deficiency of the observation personnel’s understanding of the geological radar image due to personal habits, and it provides a new way to identify the advanced geological prediction technology of the tunnel.

Molybdenum Disulfide Nanosheet/Quantum Dot Dynamic Memristive Structure Driven by Photoinduced Phase Transition.

MoS2 2D nanosheets (NS) with intercalated 0D quantum dots (QDs) represent promising structures for creating low-dimensional (LD) resistive memory devices. Nonvolatile memristors based 2D materials demonstrate low power consumption and ultrahigh density. Here, the observation of a photoinduced phase transition in the 2D NS/0D QDs MoS2 structure providing dynamic resistive memory is reported. The resistive switching of the MoS2 NS/QD structure is observed in an electric field and can be controlled through local QD excitations. Photoexcitation of the LD structure at different laser power densities leads to a reversible MoS2 2H-1T phase transition and demonstrates the potential of the LD structure for implementing a new dynamic ultrafast photoresistive memory. The dynamic LD photomemristive structure is attractive for real-time pattern recognition and photoconfiguration of artificial neural networks in a wide spectral range of sensitivity provided by QDs.

Efficient IntVec: High recognition rate with reduced computational cost

In many deep neural networks for pattern recognition, the input pattern is classified in the deepest layer based on features extracted through intermediate layers. IntVec (interpolating-vector) is known to be a powerful method for this process of classification. Although the recognition error can be made much smaller by IntVec than by WTA (winner-take-all) or even by SVM (support vector machines), IntVec requires a large computational cost. This paper proposes a new method, by which the computational cost by IntVec can be reduced drastically without increasing the recognition error. Although we basically use IntVec for recognition, we substitute it with WTA, which requires much smaller computational cost, under a certain condition. To be more specific, we first try to classify the input vector using WTA. If a class is a complete loser by WTA, we judge it also a loser by IntVec and omit the calculation of IntVec for that class. If a class is an unrivaled winner by WTA, calculation of IntVec itself can be omitted for all classes.

Design of atomically-thin-body field-effect sensors and pattern recognition neural networks for ultra-sensitive and intelligent trace explosive detection

There has been enormous demand for detecting trace concentrations of chemical molecules and in particular low-volatile explosives through electronic instrumentation, which however, still faces significant shortcomings in both detectability and selectivity to date. In this work, we propose a novel sensor strategy that incorporates arrays of atomically-thin-body field-effect sensors for highly-scalable, ultra-sensitive trace explosive sensors with fast response to ultra-low analyte concentrations. Sensor performance and functionalization are theoretically simulated through system-level considerations using various kinetic, electrostatic, quantum mechanics, and drift-diffusion models. Moreover, by implementing custom-built neural network models for pattern recognition, we successfully achieved automatic, accurate detection of four different types of analytes with concentrations down to 0.02 part per trillion. The intelligent sensors have the capability to differentiate analyte types with 100% accuracy and predict the concentration values with ~10% of relative error simultaneously. We envision the proposed sensor platform, design metrics, deep learning methods and the combination of these innovations will be a promising yet practical solution towards ultra-sensitive trace explosive detection and can be extended to a wide range of molecular sensing applications.

Multilevel HfO2-based RRAM devices for low-power neuromorphic networks

Training and recognition with neural networks generally require high throughput, high energy efficiency, and scalable circuits to enable artificial intelligence tasks to be operated at the edge, i.e., in battery-powered portable devices and other limited-energy environments. In this scenario, scalable resistive memories have been proposed as artificial synapses thanks to their scalability, reconfigurability, and high-energy efficiency, and thanks to the ability to perform analog computation by physical laws in hardware. In this work, we study the material, device, and architecture aspects of resistive switching memory (RRAM) devices for implementing a 2-layer neural network for pattern recognition. First, various RRAM processes are screened in view of the device window, analog storage, and reliability. Then, synaptic weights are stored with 5-level precision in a 4 kbit array of RRAM devices to classify the Modified National Institute of Standards and Technology (MNIST) dataset. Finally, classification performance of a 2-layer neural network is tested before and after an annealing experiment by using experimental values of conductance stored into the array, and a simulation-based analysis of inference accuracy for arrays of increasing size is presented. Our work supports material-based development of RRAM synapses for novel neural networks with high accuracy and low-power consumption.

Cardiac Health Prediction using ElectrocardiographyPratishtha Agnihotri

Amongst various physiological signals, that can be collected from the human body, Electrocardiogram (ECG) is one widely used signal that gives an overview of individual’s health non-invasively. Some prognostic tools, based on ECG, have already been introduced in the past. However, the diagnostic information contained in ECG is still under used. In the present study, we propose an algorithm that predicts the cardiac health (both present and future) by analyzing subject’s ECG. The prediction is based on diagnostic information like Blood Pressure (BP), Arrhythmia and Heart Rate Variability (HRV), where BP and Arrhythmia are used to predict the present cardiac health, and Arrhythmia and HRV are used to predict the future cardiac health associated with an individual. To verify the algorithm, we use: (1) Linear Regression Model to extract BP based on parameters extracted from ECG; (2) Neural Network Pattern Recognition Application to detect Arrhythmia- Right and Left bundle branch block beat, Atrial premature contraction beat, Premature ventricular contraction beat and Premature or ectopic supraventricular beats, in any ECG signal; (3) Self-Organized Maps for HRV analysis using ECG. These models are used on ECG of 30 subjects chosen from an existing database. Based on the outputs of these models our algorithm predicts the present as well as the future cardiac health of 30 subjects under study. Our predictions are compared with the present and future cardiac health of these subjects already documented in the database. The prediction accuracy showed that present and future cardiac health risk of an individual can be satisfactorily determined using the proposed algorithm, which, in future, can be easily incorporated in any health monitoring device which can record ECG.

Deep solar radiation forecasting with convolutional neural network and long short-term memory network algorithms

This paper designs a hybridized deep learning framework that integrates the Convolutional Neural Network for pattern recognition with the Long Short-Term Memory Network for half-hourly global solar radiation (GSR) forecasting. The Convolution network is applied to robustly extract data input features from predictive variables (i.e., statistically significant antecedent inputs) while Long Short-Term Memory absorbs them for prediction. Half-hourly GSR for Alice Springs (Australia: 01 January 2006 to 31 August 2018) are extracted with stationarity checks applied via unit-root and mutual information test to capture antecedent GSR values required to forecast future GSR. The proposed hybrid model is benchmarked with standalone models as well as other Deep Learning, Single Hidden Layer and Tree based models. The results show that the benchmarked models are not able to generate satisfactory GSR predictions and the proposed hybrid model outperforms all other counterparts. The hybrid model registers superior results with over 70% of predictive errors lying below ±10 Wm−2 and outperforms the benchmark model for 1-Day half-hourly GSR prediction with low Relative Root Mean Square Error (≈1.515%), Mean Absolute Percentage Error (≈4.672%) and Absolute Percentage Bias (≈1.233%). This study ascertains that a proposed hybrid model based on a convolution network framework can accurately predict GSR and enable energy availability to be regularly monitored over multi-step horizons when coupled with a low latency Long Short-Term Memory network. Furthermore, it also concludes that the proposed model can have practical implications in forecasting GSR, capitalizing its versatility as a stratagem in monitoring solar powered systems by integrating freely available solar radiation into a real power grid system.

Integration of Forearm sEMG Signals with IMU Sensors for Trajectory Planning and Control of Assistive Robotic Arm.

Patients with issues such as cerebral palsy, spinal cord injury, and multiple sclerosis have difficulties with activities of daily living (ADL). Their abilities to perform tasks can be improved through vigorous physical therapy. When that therapy is either not effective or lacking in its progression an assistive robotic device can be used to improve patients' quality of life and help them in accomplishing ADL's. This study presents implementation of an EMG controlled assistive robotic arm to aid patients with upper limb mobility limitations. Using the MYO armband, EMG signals were obtained from three healthy human test subjects and were analyzed in MATLAB® Simulink®. Post signal acquisition, signals were classified to be used as inputs for a Kinova MICO 6 DOF manipulator. Trajectories are planned based on the user EMG signals and robot position data obtained from the Polhemus 6D motion tracker, an IMU-type sensor, which automatically provide position and orientation data. An inverse dynamics controller is developed to drive the robot joints accordingly. Results have shown that the classification accuracy of the EMG signals to control commands for the robot was greater than 90%. The classification accuracy was achieved through the use of a pattern recognition neural network. This preliminary investigation demonstrates the possible future implementation of the system for its intended application.

Perceptual Influence of Non-Spectral Cues on Spectral Properties of Korean Vowel Signals in Seoul Corpus: A Neural Network Modeling Study

Though the importance of spectral characteristics of Korean vowels in the hVd syllable has been recognized in the literature, it has never been studied whether static spectral measurements at a (steady-state) central section are enough to characterize Korean vowels in spontaneous speech, or dynamic spectral measurements across the temporal dimension can better characterize vowels. Despite ample reported evidence of the perceptual influence of non-spectral cues on spectral properties of vowels in the literature, no reports have yet been released on the difference in the degree of the perceptual influence of non-spectral cues (e.g., place and manner of the preceding or following phones, F0, speaking rate, prosody, and gender) on spectral properties of vowels. Through Neural Network pattern recognition modeling in a supervised mode, it was found that dynamic spectral models with non-spectral cues better explain vowel perception than static spectral models and furthermore, flanking phone identities, and manner and place of flanking phones are perceptually the most influential while duration, F0 and speaking rate are perceptually far less contributive than argued in the literature.

Efficient Spatial Sampling for AFM-Based Cancer Diagnostics: A Comparison between Neural Networks and Conventional Data Analysis

Atomic force microscopy (AFM) in spectroscopy mode receives a lot of attention because of its potential in distinguishing between healthy and cancer tissues. However, the AFM translational process in clinical practice is hindered by the fact that it is a time-consuming technique in terms of measurement and analysis time. In this paper, we attempt to address both issues. We propose the use of neural networks for pattern recognition to automatically classify AFM force–distance (FD) curves, with the aim of avoiding curve-fitting with the Sneddon model or more complicated ones. We investigated the applicability of this method to the classification of brain cancer tissues. The performance of the classifier was evaluated with receiving operating characteristic (ROC) curves for the approach and retract curves separately and in combination with each other. Although more complex and comprehensive models are required to demonstrate the general applicability of the proposed approach, preliminary evidence is given for the accuracy of the results, and arguments are presented to support the possible applicability of neural networks to the classification of brain cancer tissues. Moreover, we propose a possible strategy to shorten measurement times based on the estimation of the minimum number of FD curves needed to classify a tissue with a confidence level of 0.005. Taken together, these results have the potential to stimulate the design of more effective protocols to reduce AFM measurement times and to get rid of curve-fitting, which is a complex and time-consuming issue that requires experienced staff with a strong data-analysis background.

Feature Selection for Recognition of Online Handwritten Bangla Characters

Feature selection through optimization techniques provides an interesting approach to minimize computational time with enhanced prediction capability, and has better cognizance of data in any pattern recognition application. This paper is an extended version of previously published work (Sen et al. in: 7th IAPR TC3 workshop on artificial neural networks in pattern recognition, Ulm, Germany, pp 246–256, 2016), where a quad-tree based image segmentation approach has been discussed to estimate some topological and shape based features (192-attributed) for the recognition of online handwritten Bangla characters. The previous work achieved a recognition accuracy of 98.5% on a database consisting of 10,000 handwritten Bangla characters. In this paper, parameters, used in the previous version during feature estimation, are tuned to improve the performance of the overall system. Thereafter, krill-herd a bio-inspired, meta-heuristic algorithm has been applied to find the optimal feature vector by reducing the dimension of the original feature vector. The reduced feature vector has been fed to Sequential Minimal Optimization classifier for the recognition of the same online handwritten Bangla character database used in previous work. It has been observed that result obtained with this optimal feature set is almost equivalent as the result produced by the entire feature vector.

Explicit Characterization of Spatial Heterogeneity Based on Water Quality, Sediment Contamination, and Ichthyofauna in a Riverine-to-Coastal Zone.

Our study aims to identify the spatial characteristics of water quality and sediment conditions in relation to fisheries resources, since the productivity of fisheries resources is closely related to the ambient conditions of the resource areas. We collected water quality samples and sediment contaminants from twenty-one sites at Gwangyang Bay, Korea, in the summer of 2018. Our study sites covered the area from the Seomjin River estuary to the inner and outer bays. To spatially characterize physicochemical features of Gwangyang Bay, we used Self-Organizing Map (SOM), which is known as a robust and powerful tool of unsupervised neural networks for pattern recognition. The present environmental conditions of Gwangyang Bay were spatially characterized according to four different attributes of water quality and sediment contamination. From the results, we put emphasis on several interesting points: (i) the SOM manifests the dominant physicochemical attributes of each geographical zone associated with the patterns of water quality and sediment contamination; (ii) fish populations appear to be closely associated with their food sources (e.g., shrimps and crabs) as well as the ambient physicochemical conditions; and (iii) in the context of public health and ecosystem services, the SOM result can potentially offer guidance for fish consumption associated with sediment heavy metal contamination. The present study may have limitations in representing general features of Gwangyang Bay, given the inability of snapshot data to characterize a complex ecosystem. In this regard, consistent sampling and investigation are needed to capture spatial variation and to delineate the temporal dynamics of water quality, sediment contamination, and fish populations. However, the SOM application is helpful and useful as a first approximation of an environmental assessment for the effective management of fisheries resources.

Breast Tumor Classification Using Fast Convergence Recurrent Wavelet Elman Neural Networks

This paper develops an intelligent classification system for breast tumors that uses fine needle aspirate image data. A recurrent wavelet Elman neural network is used to classify the breast tumor as either benign or malignant. The structure of the RWENN uses different wavelet functions for hidden layers so that the generalization and search space are significantly greater than those of a conventional neural network. In this paper, there is also a stable convergence analysis of the RWENN classifier and the optimal learning rates are derived to guarantee the fastest convergence for the classification system. The performance of the developed classifier is compared with the Matlab neural network pattern recognition toolbox and other literature that uses a tenfold cross validation on the Wisconsin breast cancer dataset. The simulation results show that the proposed RWENN classifier has better classification results than other existing methods.