Neural Network Classification Algorithm Research Articles

Breath Analysis Using Quartz Tuning Forks for Predicting Blood Glucose Levels Using Artificial Neural Networks.

Diabetes Mellitus (DM), a widespread metabolic disorder, poses lifelong health implications, demanding timely diagnosis and cautious monitoring for effective disease management. Traditional blood glucose tests are invasive and require medical expertise for intermittent checking, motivating the investigation of alternative, noninvasive methods. This study introduces an approach employing breath analysis through a set of 12 quartz tuning fork-based sensors enhanced using nanomaterials and dedicated artificial neural network (ANN) algorithms for data interpretation. The breath analysis methodology involves capturing unique breath signatures using the frequency-based sensor array. The accompanying neural network classification algorithm, customized for the sensor data, enables precise classification of data from 245 individuals as diabetic, prediabetic, or healthy. A neural network regression algorithm predicted blood glucose values and was compared with the actual values obtained from medical blood glucose measurement. The clinical relevance of the predicted blood glucose has been examined using error grids. The sensor array coupled with the ANN algorithm can identify diabetic, prediabetic, and control samples with 97% test accuracy. Blood glucose was predicted using neural network regression with a correlation coefficient of 0.89 and a mean square error of 0.13.

Detection of Alcoholic EEG signal using LASSO regression with metaheuristics algorithms based LSTM and enhanced artificial neural network classification algorithms

The world has a higher count of death rates as a result of Alcohol consumption. Identification is possible because Alcoholic EEG waves have a certain behavior that is totally different compared to the non-alcoholic individual. The available approaches take longer to provide the feedback because they analyze the data manually. For this reason, in the present paper we propose a novel approach applied to detect alcoholic EEG signals automatically by using deep learning methods. Our strategy has advantages as far as fast detection is concerned; hence people can help immediately when there is a need. The potential for a significant decrease in deaths from alcohol poisoning and improvement to public health is presented by this advancement. In order to create clusters and classify the alcoholic EEG signals, this research uses a cascaded process. To begin with, an initial clustering and feature extraction is done by LASSO regression. After that, a variety of meta-heuristics algorithms like Particle Swarm Optimization (PSO), Binary Coding Harmony Search (BCHS) as well as Binary Dragonfly Algorithm (BDA) are employed for feature minimization. When this method is used, normal and alcoholic EEG signals may be differentiated using non-linear features. PSO, BCHS, and BDA features allow for estimation of statistical parameters through t-test, Friedman statistic test, Mann-Whitney U test, and Z-Score with corresponding p-values for alcoholic EEG signals. Lastly, classification is done by the use of support vector machines (SVM) (including linear, polynomial, and Gaussian kernels), random forests, artificial neural networks (ANN), enhanced artificial neural networks (EANN), and LSTM models. Results showed that LASSO regression with BDA-based EANN proposed classifier have a classification accuracy of 99.59%, indicating that our method is highly accurate at classifying alcoholic EEG signals.

Neural Network Classification Algorithm Based on Self-attention Mechanism and Ensemble Learning for MASLD Ultrasound Images

BackgroundUltrasound image examination has become the preferred choice for diagnosing metabolic dysfunction-associated steatotic liver disease (MASLD) due to its non-invasive nature. Computer-aided diagnosis (CAD) technology can assist doctors in avoiding deviations in the detection and classification of MASLD. MethodWe propose a hybrid model that integrates the pre-trained VGG16 network with an attention mechanism and a stacking ensemble learning model, which is capable of multi-scale feature aggregation based on the self-attention mechanism and multi-classification model fusion (Logistic regression, random forest, support vector machine) based on stacking ensemble learning. The proposed hybrid method achieves four classifications of normal, mild, moderate, and severe fatty liver based on ultrasound images. Result and conclusionOur proposed hybrid model reaches an accuracy of 91.34% and exhibits superior robustness against interference, which is better than traditional neural network algorithms. Experimental results show that, compared with the pre-trained VGG16 model, adding the self-attention mechanism improves the accuracy by 3.02%. Using the stacking ensemble learning model as a classifier further increases the accuracy to 91.34%, exceeding any single classifier such as LR (89.86%) and SVM (90.34%) and RF (90.73%). The proposed hybrid method can effectively improve the efficiency and accuracy of MASLD ultrasound image detection.

Automated identification of pesticide mixtures via machine learning analysis of TLC-SERS spectra

Identification of components in pesticide mixtures has been a major challenge in spectral analysis. In this paper, we assembled monolayer Ag nanoparticles on Thin-layer chromatography (TLC) plates to prepare TLC-Ag substrates with mixture separation and surface-enhanced Raman scattering (SERS) detection. Spectral scans were performed along the longitudinal direction of the TLC-Ag substrate to generate SERS spectra of all target analytes on the TLC plate. Convolutional neural network classification and spectral angle similarity machine learning algorithms were used to identify pesticide information from the TLC-SERS spectra. It was shown that the proposed automated spectral analysis method successfully classified five categories, including four pesticides (thiram, triadimefon, benzimidazole, thiamethoxam) as well as a blank TLC-Ag data control. The location of each pesticide on the TLC plate was determined by the intersection of the information curves of the two algorithms with 100 % accuracy. Therefore, this method is expected to help regulators understand the residues of mixed pesticides in agricultural products and reduce the potential risk of agricultural products to human health and the environment.

Hydrogel-based flexible degradable triboelectric nanogenerators for human activity recognition

The development of Internet of Things (IoT) and Artificial Intelligence (AI) technologies has led to an increased demand for wearable electronic devices that can be used in a variety of contexts. Hydrogel sensors have been regarded as an ideal material for the fabrication of multifunctional flexible wearable devices. However, it is difficult to create hydrogel sensors with ideal electrical and mechanical properties. In this study, a conductive hydrogel prepared by a simple method has been introduced. The hydrogel was immersed in a binary solvent of glycerol, water and sodium citrate to form a transparent, long-term stable and highly conductive gelatin-NaCl organohydrogel (GNOH). A stretchable and durable gelatin Ecoflex triboelectric nanogenerator (GE-TENG) was created by combining an organic hydrogel with Ecoflex elastomer. It has high electrical and mechanical performance, with an open-circuit voltage of 142 V, current of 3.8 μA, power output of 19.36 μW and response time of 85 ms. Moreover, the single-electrode mode of the GE-TENG provides high electrical output for powering portable electronic devices and can capture biomechanical energy in temperatures as low as −20 °C. As a result of these capabilities, GE-TENG can effectively identify human movements with precision and transmit signals wirelessly, thereby broadening its potential applications. The GE-TENG-based self-powered intelligent sensing system (GSIS) for human-computer interaction demonstrates a wearable intelligent system for controlling the movement of a tank model and a manipulator by integrating the signal acquisition, signal processing and signal output components. The system achieves 100% signal recognition accuracy when combined with covariance matrix feature analysis and Convolutional Neural Network classification algorithms. This research demonstrates the potential of wearable hydrogel-based electronic devices for monitoring human motion, harvesting biomechanical energy and human-computer interaction.

Efficient Framework to Manipulate Data Compression and Classification of Power Quality Disturbances for Distributed Power System

There is some risk of power quality disturbances at many stages of production, transformation, distribution, and energy consumption. The cornerstone for dealing with power quality problems is the characterization of power quality disturbances (PQDs). However, past research has focused on a narrow topic: noise disruption, overfitting, and training time. A new strategy is suggested to address this problem that combines efficient one-dimensional dataset compression with the convolutional neural network (CNN) classification algorithm. First, three types of compression algorithms: wavelet transform, autoencoder, and CNN, are proposed to be evaluated. According to the IEEE-1159 standard, the synthetic dataset was built with fourteen different PQD types. Furthermore, the PQD classification procedure integrated compressed data with the CNN classification algorithm. Finally, the suggested method demonstrates that combining CNN compression and classification methods can efficiently recognize PQDs. Even in noisy environments, PQD signal processing achieved up to 98.25% accuracy and managed the overfitting.

Identification of breast cancer subtypes based on graph convolutional network

Identification of molecular subtypes of malignant tumors plays a vital role in individualized diagnosis, personalized treatment, and prognosis prediction of cancer patients. The continuous improvement of comprehensive tumor genomics database and the ongoing breakthroughs in deep learning technology have driven further advancements in computer-aided tumor classification. Although the existing classification methods based on gene expression omnibus database take the complexity of cancer molecular classification into account, they ignore the internal correlation and synergism of genes. To solve this problem, we propose a multi-layer graph convolutional network model for breast cancer subtype classification combined with hierarchical attention network. This model constructs the graph embedding datasets of patients' genes, and develops a new end-to-end multi-classification model, which can effectively recognize molecular subtypes of breast cancer. A large number of test data prove the good performance of this new model in the classification of breast cancer subtypes. Compared to the original graph convolutional neural networks and two mainstream graph neural network classification algorithms, the new model has remarkable advantages. The accuracy, weight-F1-score, weight-recall, and weight-precision of our model in seven-category classification has reached 0.851 7, 0.823 5, 0.851 7 and 0.793 6 respectively. In the four-category classification, the results are 0.928 5, 0.894 9, 0.928 5 and 0.865 0 respectively. In addition, compared with the latest breast cancer subtype classification algorithms, the method proposed in this paper also achieved the highest classification accuracy. In summary, the model proposed in this paper may serve as an auxiliary diagnostic technology, providing a reliable option for precise classification of breast cancer subtypes in the future and laying the theoretical foundation for computer-aided tumor classification.

Development of a model to predict the risk of early graft failure after adult-to-adult living donor liver transplantation: An ELTR study.

Graft survival is a critical end point in adult-to-adult living donor liver transplantation (ALDLT), where graft procurement endangers the lives of healthy individuals. Therefore, ALDLT must be responsibly performed in the perspective of a positive harm-to-benefit ratio. This study aimed to develop a risk prediction model for early (3 months) graft failure (EGF) following ALDLT. Donor and recipient factors associated with EGF in ALDLT were studied using data from the European Liver Transplant Registry. An artificial neural network classification algorithm was trained on a set of 2073 ALDLTs, validated using cross-validation, tested on an independent random-split sample (n=518), and externally validated on United Network for Organ Sharing Standard Transplant Analysis and Research data. Model performance was assessed using the AUC, calibration plots, and decision curve analysis. Graft type, graft weight, level of hospitalization, and the severity of liver disease were associated with EGF. The model ( http://ldlt.shinyapps.io/eltr_app ) presented AUC values at cross-validation, in the independent test set, and at external validation of 0.69, 0.70, and 0.68, respectively. Model calibration was fair. The decision curve analysis indicated a positive net benefit of the model, with an estimated net reduction of 5-15 EGF per 100 ALDLTs. Estimated risks>40% and<5% had a specificity of 0.96 and sensitivity of 0.99 in predicting and excluding EGF, respectively. The model also stratified long-term graft survival ( p <0.001), which ranged from 87% in the low-risk group to 60% in the high-risk group. In conclusion, based on a panel of donor and recipient variables, an artificial neural network can contribute to decision-making in ALDLT by predicting EGF risk.

Tracking of moving human in different overlapping cameras using Kalman filter optimized

Tracking objects is a crucial problem in image processing and machine vision, involving the representation of position changes of an object and following it in a sequence of video images. Though it has a history in military applications, tracking has become increasingly important since the 1980s due to its wide-ranging applications in different areas. This study focuses on tracking moving objects with human identity and identifying individuals through their appearance, using an Artificial Neural Network (ANN) classification algorithm. The Kalman filter is an important tool in this process, as it can predict the movement trajectory and estimate the position of moving objects. The tracking error is reduced by weighting the filter using a fuzzy logic algorithm for each moving human. After tracking people, they are identified using the features extracted from the histogram of images by ANN. However, there are various challenges in implementing this method, which can be addressed by using Genetic Algorithm (GA) for feature selection. The simulations in this study aim to evaluate the convergence rate and estimation error of the filter. The results show that the proposed method achieves better results than other similar methods in tracking position in three different datasets. Moreover, the proposed method performs 8% better on average than other similar algorithms in night vision, cloud vision, and daylight vision situations.

Parametric Modeling and Deep Learning for Enhancing Pain Assessment in Postanesthesia.

The problem of reliable and widely accepted measures of pain is still open. It follows the objective of this work as pain estimation through post-surgical trauma modeling and classification, to increase the needed reliability compared to measurements only. This article proposes (i) a recursive identification method to obtain the frequency response and parameterization using fractional-order impedance models (FOIM), and (ii) deep learning with convolutional neural networks (CNN) classification algorithms using time-frequency data and spectrograms. The skin impedance measurements were conducted on 12 patients throughout the postanesthesia care in a proof-of-concept clinical trial. Recursive least-squares system identification was performed using a genetic algorithm for initializing the parametric model. The online parameter estimates were compared to the self-reported level by the Numeric Rating Scale (NRS) for analysis and validation of the results. Alternatively, the inputs to CNNs were the spectrograms extracted from the time-frequency dataset, being pre-labeled in four intensities classes of pain during offline and online training with the NRS. The tendency of nociception could be predicted by monitoring the changes in the FOIM parameters' values or by retraining online the network. Moreover, the tissue heterogeneity, assumed during nociception, could follow the NRS trends. The online predictions of retrained CNN have more specific trends to NRS than pain predicted by the offline population-trained CNN. We propose tailored online identification and deep learning for artefact corrupted environment. The results indicate estimations with the potential to avoid over-dosing due to the objectivity of the information. Models and artificial intelligence (AI) allow objective and personalized nociception-antinociception prediction in the patient safety era for the design and evaluation of closed-loop analgesia controllers.

An ImprovedD2GAN‐based oversampling algorithm for imbalanced data classification

AbstractTo address the problems of pattern collapse, uncontrollable data generation and high overlap rate when generative adversarial network (GAN) oversamples imbalanced data, we propose an imbalanced data oversampling algorithm based on improved dual discriminator generative adversarial nets (D2GAN). First, we integrate the positive class attribute information into the generator and the discriminator to ensure that the generator only generates the samples for positive class samples, which overcomes the problem of uncontrollable data generation by the generator. Second, we introduce a classifier into D2GAN for discriminating the generated samples and the original data, which avoids the overlap among the generated samples and the negative class samples, and ensures the diversity of the generated samples, the problem of pattern collapse is solved. Finally, the performance of the proposed algorithm is evaluated on 9 datasets by using SVM and neural network classification algorithm for oversampling experiments, the results show that the proposed algorithm effectively improve the classification performance of imbalanced data.

Application of Convolutional Neural Network Classification Algorithm in Online-offline Blended Teaching of Chinese Painting

Application of Convolutional Neural Network Classification Algorithm in Online-offline Blended Teaching of Chinese Painting

Qualitative and Quantitative Recognition of Volatile Organic Compounds in Their Liquid Phase Based on Terahertz Microfluidic EIT Meta-Sensors

Volatile organic compounds (VOCs) are directly associated with human health concerns and environment safety. Therefore, it is urgent to achieve an accurate detection of VOCs both qualitatively and quantitatively. In this work, the qualitative detection of ethyl benzene (EB), isopropyl alcohol (IPA), ethyl acetate (EA) – three pure VOCs in liquid phase were discussed using terahertz (THz) microfluidic Electromagnetic Induced Transparency (EIT) meta-sensor. The THz response illustrated that with an increase in VOCs’ volumes (1~6μL), resonant frequencies of dual transmission dips (0.855, 1.724 THz) and the EIT peak (1.213THz) exhibited red shift. The limit of detections for pure IPA, EA, EB can achieve 5.45μg, 13.46μg, 4.35μg, respectively. The multivariate fusion (MF) model based on the EIT responses to VOCs was utilized to improve the accuracy of trace detection and classification of VOCs. Furthermore, the above method combined with principal component analysis with Gaussian mixture model (PCA-GMM) and the neural network classification algorithm support vector machine (SVM) was applied to the recognition of VOCs. In addition, the THz method is not feasible to detect trace amounts of VOCs (typically 0.3mg/L) in waste water because water is highly absorbable in THz band and VOCs will evaporate if water is removed. Here, IPA, EA, EB in soil were detected and classified by PCA-GMM combined with MF. Our results provide a new terahertz meta-sensor platform to trace detection of VOCs in liquid phase and in soil and may be used to identify hazardous wastes in illegal dumping.

Low-Power FPGA Realization of Lightweight Active Noise Cancellation with CNN Noise Classification

Active noise cancellation (ANC) is the most important function in an audio device because it removes unwanted ambient noise. As many audio devices are increasingly equipped with digital signal processing (DSP) circuits, the need for low-power and high-performance processors has arisen because of hardware resource restrictions. Low-power design is essential because wireless audio devices have limited batteries. Noise cancellers process the noise in real time, but they have a short secondary path delay in conventional least mean square (LMS) algorithms, which makes implementing high-quality ANC difficult. To solve these problems, we propose a fixed-filter noise cancelling system with a convolutional neural network (CNN) classification algorithm to accommodate short secondary path delay and reduce the noise ratio. The signal-to-noise ratio (SNR) improved by 2.3 dB with CNN noise cancellation compared to the adaptive LMS algorithm. A frequency-domain noise classification and coefficient selection algorithm is introduced to cancel the noise for time-varying systems. Additionally, our proposed ANC architecture includes an even–odd buffer that efficiently computes the fast Fourier transform (FFT) and overlap-save (OLS) convolution. The simulation results demonstrate that the proposed even–odd buffer reduces processing time by 20.3% and dynamic power consumption by 53% compared to the single buffer.

An Improved COVID-19 Lung X-Ray Image Classification Algorithm Based on ConvNeXt Network

Aiming at the new coronavirus that appeared in 2019, which has caused a large number of infected patients worldwide due to its high contagiousness, in order to detect the source of infection in time and cut off the chain of transmission, we developed a new Chest X-ray (CXR) image classification algorithm with high accuracy, simple operation and fast processing for COVID-19. The algorithm is based on ConvNeXt pure convolutional neural network, we adjusted the network structure and loss function, added some new Data Augmentation methods and introduced attention mechanism. Compared with other classical convolutional neural network classification algorithms such as AlexNet, ResNet-34, ResNet-50, ResNet-101, ConvNeXt-tiny, ConvNeXt-small and ConvNeXt-base, the improved algorithm has better performance on COVID dataset.

A Survey of Fuzzy Pattern Tree Classification Algorithms

Data classification algorithm is the core content of big data mining. Its main function is to extract valuable knowledge and information,analyze the characteristics of all kinds of information, and provide data basis for further research. With the wide application of data mining technology, data classification algorithms continue to emerge and gradually improve. The classic classification algorithms include decision tree classification algorithm, naive Bayes algorithm, support vector machine classification algorithm, artificial neural network classification algorithm, fuzzy pattern tree (FPT) and so on. This paper summarizes several common algorithms in data classification algorithms, and analyzes their characteristics to understand their algorithm principles and application scenarios.

ConvNeXt-Haze: A Fog Image Classification Algorithm for Small and Imbalanced Sample Dataset Based on Convolutional Neural Network

Aiming at the different abilities of the defogging algorithms in different fog concentrations, this paper proposes a fog image classification algorithm for a small and imbalanced sample dataset based on a convolution neural network, which can classify the fog images in advance, so as to improve the effect and adaptive ability of image defogging algorithm in fog and haze weather. In order to solve the problems of environmental interference, camera depth of field interference and uneven feature distribution in fog images, the CutBlur-Gauss data augmentation method and focal loss and label smoothing strategies are used to improve the accuracy of classification. It is compared with the machine learning algorithm SVM and classical convolution neural network classification algorithms alexnet, resnet34, resnet50 and resnet101. This algorithm achieves 94.5% classification accuracy on the dataset in this paper, which exceeds other excellent comparison algorithms at present, and achieves the best accuracy. It is proved that the improved algorithm has better classification accuracy.

Photonics-aided 335 GHz PS-64QAM wireless transmission over 200 m employing complex-valued NN classification and random sampling techniques.

THz fiber-wireless technique can overcome the bandwidth bottleneck of electrical devices and has been popularized in different application scenarios. Furthermore, the probabilistic shaping (PS) technique can optimize both the transmission capacity and the distance, and has been extensively used in the optical fiber communication field. However, the probability of the point in the PS m-ary quadrature-amplitude-modulation (m-QAM) constellation varies with the amplitude, which leads to the class imbalance and degrades the performances of all supervised neural network classification algorithms. In this paper, we propose a novel complex-valued neural network (CVNN) classifier coupled with balanced random oversampling (ROS), which can be trained to restore the phase information simultaneously and overcome the class imbalance caused by PS. Based on this scheme, the fusion of oversampled features in complex domain increases the amount of the effective information of few classes, and thus improves the recognition accuracy effectively. It also has less requirement on the sample size than NN-based classifiers and largely simplifies the neural network architecture. By using our proposed ROS-CVNN classification method, single-lane 10 Gbaud 335 GHz PS-64QAM fiber-wireless transmission over 200 m free-space distance is experimentally realized, and the experimental results show that the efficient data rate is 44 Gbit/s considering the soft-decision forward-error-correction (SD-FEC) with 25% overhead. The results show that ROS-CVNN classifier outperforms the other real-valued NN equalizers and traditional Volterra-series by average 0.5 to 1 dB in receiver sensitivity at the bit error rate (BER) of 6 × 10-2 magnitude. Therefore, we believe that the combination of ROS and NN supervised algorithms has an application prospect for the future 6 G mobile communication.

Heterogeneous Structure Omnidirectional Strain Sensor Arrays With CognitivelyLearned Neural Networks.

Mechanically stretchable strain sensors gain tremendous attention for bioinspired skin sensation systems and artificially intelligent tactile sensors. However, high-accuracy detection of both strain intensity and direction with simple device/array structures is still insufficient. To overcome this limitation, an omnidirectional strain perception platform utilizing a stretchable strain sensor array with triangular-sensor-assembly (three sensors tilted by 45°) coupled with machine learning (ML) -based neural network classification algorithm, is proposed. The strain sensor, which is constructed with strain-insensitive electrode regions and strain-sensitive channel region, can minimize the undesirable electrical intrusion from the electrodes by strain, leading to a heterogeneous surface structure for more reliable strain sensing characteristics. The strain sensor exhibits decent sensitivity with gauge factor (GF) of ≈8, a moderate sensing range (≈0-35%), and relatively good reliability (3000 stretching cycles). More importantly, by employing a multiclass-multioutput behavior-learned cognition algorithm, the stretchable sensor array with triangular-sensor-assembly exhibits highly accurate recognition of both direction and intensity of an arbitrary strain by interpretating the correlated signals from the three-unit sensors. The omnidirectional strain perception platform with itsneural network algorithm exhibits overall strain intensity and direction accuracy around 98% ± 2% over a strain range of ≈0-30% in various surface stimuli environments.

Deep learning based Brain Tumour Classification based on Recursive Sigmoid Neural Network based on Multi-Scale Neural Segmentation

Brain tumours are malignant tissues in which cells replicate rapidly and indefinitely, and tumours grow out of control. Deep learning has the potential to overcome challenges associated with brain tumour diagnosis and intervention. It is well known that segmentation methods can be used to remove abnormal tumour areas in the brain. It is one of the advanced technology classification and detection tools. Can effectively achieve early diagnosis of the disease or brain tumours through reliable and advanced neural network classification algorithms. Previous algorithm has some drawbacks, an automatic and reliable method for segmentation is needed. However, the large spatial and structural heterogeneity between brain tumors makes automated segmentation a challenging problem. Image tumors have irregular shapes and are spatially located in any part of the brain, making their segmentation is inaccurate for clinical purposes a challenging task. In this work, propose a method Recursive SigmoidNeural Network based on Multi-scale Neural Segmentation (RSN2-MSNS) for image proper segmentation. Initially collets the image dataset from standard repository for brain tumour classification. Next, pre-processing method that targets only a small part of an image rather than the entire image. This approach reduces computational time and overcomes the over complication. Second stage, segmenting the images based on the Enhanced Deep Clustering U-net (EDCU-net) for estimating the boundary points in the brain tumour images. This method can successfully colour histogram values are evaluating segment complex images that contain both textured and non-textured regions. Third stage, Feature extraction for extracts the features from segmenting images using Convolution Deep Feature Spectral Similarity (CDFS2) scaled the values from images extracting the relevant weights based on its threshold limits. Then selecting the features from extracting stage, this selection is based on the relational weights. And finally classified the features based on the Recursive Sigmoid Neural Network based on Multi-scale Neural Segmentation (RSN2-MSNS) for evaluating the proposed brain tumour classification model consists of 1500 trainable images and the proposed method achieves 97.0% accuracy. The sensitivity, specificity, detection accuracy and F1 measures were 96.4%, 952%, and 95.9%, respectively.