Parkinson's disease (PD) is a neurodegenerative disorder characterized by a range of motor and non-motor symptoms. Despite extensive research, the neural structure of time distortion, remains unclear. This study aimed to determine the neurobiological origins of time distortion by analyzing dynamic features in PD patients compared to control participants.
Approach.
We used PD and control electroencephalography (EEG) signals to investigate brain function during time distortion. The EEG signal was recorded during an interval-timing task. Following artifact reduction and EEG signal segmentation, dynamic features were extracted from each frequency band across all the channels. Channels showing significant discrepancies between the two groups were selected by statistical analysis. The features are sent to an Artificial Neural Network (ANN) classifier to evaluate their discriminative potential.
Main results.
The results indicated lower values of Lyapunov Exponent and Approximate Entropy along with higher value of Fractal dimension in PD which presented higher level of irregularity and randomness, particularly in the CPz, P5, P6, and C5 channels. The ANN classifier achieved 90% accuracy, 89% sensitivity, 87% F1 score, and 95% specificity in a 10-fold cross-validation. 
Significance.
Significantly different channels were concentrated in the central and parietal areas of the brain and were linked to decision-making, maintenance, and retrieval of stored information, and working memory. Moreover, based on dynamic EEG analysis, it seems that disrupted connections between the basal ganglia and posterior parietal cortex in PD appear to compromise frontoparietal network dysfunction, which is associated with impaired temporal processing in patients with PD.&#xD.

This study aims to develop artificial intelligence (AI) models for predicting the compressive strength and flow value of cementitious systems containing fly ash, influenced by various high-range water-reducing admixtures (HRWRAs) that differ in molecular weight and chain length. A database comprising 180 mixes was created, encompassing cement and fly ash dosages, HRWRA characteristics (including molecular weight, main and side chain lengths) curing period, and flow time. Two AI-based modelling approaches were employed: a classical artificial neural network (ANN) and a new hybrid model that integrates ANN with biogeography-based optimisation (ANN–BBO). The modeling results showed that the hybrid model achieved a compressive strength performance with an R2 of approximately 0.99 and an RMSE of around 1.37 MPa, while the single ANN model attained an R2 of about 0.91 and an RMSE of 4.40 MPa. For flow value prediction, the ANN–BBO model also demonstrated high accuracy (R2 ≈ 0.98; RMSE ≈ 0.32 cm). Furthermore, the ANN–BBO model reduced the prediction error by approximately 60% across the evaluation criteria compared to the single ANN model, highlighting its enhanced performance. The importance of the input variables indicated that curing time and cement content have the greatest impact on compressive strength, while flow time and the molecular weight of the HRWRA significantly influence the flow value. Since AI models rely solely on virtual trials, they significantly reduce laboratory time and material usage while aiding in the design of mixes with lower water-to-binder ratios and higher fly ash content, which ultimately helps to reduce the CO2 footprint. The proposed models provide a practical route to low-clinker, FA-rich mix designs that satisfy strength/workability targets with less cement, supporting embodied-carbon reductions and straightforward integration into ready-mix/precast quality-control workflows.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-32972-1.

Electronic nose systems are widely used in environmental monitoring and other related fields. In recent years, systems based on gas sensor arrays have attracted considerable attention. However, relying solely on improvements in gas-sensitive materials has struggled to break through the bottleneck in recognition accuracy. To address this challenge, this study designs and validates an integrated machine learning framework for enhanced gas identification in electronic nose systems. Specifically, (1) a Butterworth low-pass filter is combined with principal component analysis (PCA) to suppress sensor noise; (2) the synthetic minority over-sampling technique (SMOTE) is utilized for training set data augmentation to further enhance the classification accuracy of the support vector machine (SVM); and (3) the relationship between single-component and mixed-gas responses is analyzed to construct an artificial neural network (ANN) regression model. Experimental results demonstrate that the SMOTE-augmented, PCA-optimized SVM model achieves a recognition accuracy of 0.93 ± 0.08 for most target gases, representing improvements of 19% and 7% over decision tree and ANN classifiers, respectively, and that the ANN regression model attains a correlation coefficient of 99.55% between predicted and measured values in mixed-gas experiments. Overall, the construction and optimization of this system demonstrate significant practical value for intelligent gas identification and the development of advanced e-nose devices.

Melting-like transitions in clusters are normally identified by a peak in the heat capacity curve C(T) at T = Tc. Computing C(T) requires costly simulations with millions of steps. We discuss four easily calculated functions of temperature that help detect and characterize melting-like transitions. The first is WU, the width of the potential energy distribution, which shows an abrupt increase near Tc. The other three are statistics of the ordered set of N(N - 1)/2 interatomic distances rij: (i) a measure of dissimilarity to the lowest energy configuration, or global minimum; (ii) the number of rij's found in a small interval centered around (r1 + r2)/2, where r1, r2 are the positions of the first two peaks in the pair distribution function; and (iii) a measure of non-uniformity in the distribution of the rij's. Numerical tests with empirical potentials that model three types of bonding (van der Waals, covalent, and metallic) show that these four functions produce estimates for the middle of the melting region in general agreement with Tc. An artificial neural network classifier is used to calculate the solid fraction FS(T) and find the solid-liquid coexistence region between freezing and melting temperatures, [Tf, Tm]. Inflection points in the third function and FS(T) are very sensitive indicators of phase transitions. Estimates of Tc obtained from them converge one to three orders of magnitude faster, in simulation time, than those obtained with C(T).

Knee osteoarthritis (OA) is a degenerative condition that significantly limits mobility and quality of life, particularly among older adults. Although radiographic assessment is the standard method for identifying structural changes, its accuracy can vary because interpretations depend heavily on the clinician's experience. Recent advances in artificial intelligence, intense learning, have enabled more objective and reliable analysis of medical images. In this work, we propose a hybrid diagnostic framework that combines a Convolutional Neural Network (CNN) to extract detailed radiographic features with an Artificial Neural Network (ANN) classifier to determine OA severity. The approach was trained and evaluated using publicly available knee X-ray datasets and demonstrated superior performance compared with traditional machine-learning techniques and standalone CNN models. Our findings indicate that integrating ANN-based decision layers with deep learning feature extractors can improve diagnostic consistency and assist healthcare professionals in detecting OA at earlier stages

Early and reliable diagnostic tools are critical for slowing the progression of Alzheimer's disease (AD), a neurodegenerative disorder characterized by memory loss and cognitive decline. This study introduces, ViTTL, lightweight deep learning framework for assessing the severity of AD using MRI data. ViTTL integrates Vision Transformers (ViT) with pre-trained convolutional neural networks utilized in transfer learning mode, to extract informative features from 2D MRI slices. Among the evaluated combinations, the ViT-DenseNet201 model integrated with an artificial neural network (ANN) classifier achieved the highest classification accuracy (99.89%) on the OASIS dataset. To ensure interpretability, we incorporated LIME and GRAD-CAM method, which consistently focus on cortical and hippocampal regions known to be associated with Alzheimer's pathology. The average Dice similarity coefficient across runs was 0.85 with a standard deviation of 0.03, indicating high consistency in the model's focus regions against ground truth annotations by expert radiologists. ViTTL also achieved a substantial reduction in model size from 83.0MB to 6.47MB enabling deployment in resource-limited environments without compromising performance. Validation on an independent dataset (Kaggle) and comparative performance analysis against state-of-the-art methods further support the robustness and generalizability. These findings demonstrate that ViTTL is a promising tool for accurate, interpretable, and resource-efficient AD diagnosis, with strong potential for clinical translation and patient outcome improvement. The related codes are available at https://github.com/RuhikaSharma/enhanced-alzheimer-risk-assessment .

In In today’s fast-paced e-commerce landscape, pricing strategies play a pivotal role in driving customer acquisition, maximizing revenue, and adapting to volatile market conditions. This paper introduces an AI-powered dynamic pricing framework that combines supervised machine learning with real-time market intelligence to recommend optimal product prices. The system integrates Random Forest and Artificial Neural Network (ANN) classifiers to predict consumer purchase behavior, achieving an average accuracy of 94.6% and an F1-score of 0.92. While Random Forest delivered a precision of 93.1%, ANN exhibited superior recall, underscoring their complementary classification capabilities. For precise price estimation, a multivariate linear regression model was implemented, attaining a high coefficient of determination (R² = 0.917) and a low Mean Absolute Error (MAE) of 2.87. The framework maintains prediction latency under 300 ms, ensuring suitability for real-time applications. Market data ingestion from platforms such as BigBasket was automated using Python-based web scraping, while an intuitive Gradio-powered GUI enables seamless user interaction. Designed to be scalable and modular, the system adapts effectively to evolving market dynamics and product-specific attributes. This research demonstrates how the synergy of machine learning, real-time analytics, and automation can drive intelligent pricing strategies that enhance competitiveness, customer targeting, and operational efficiency in modern e-commerce ecosystems.

Rapidly growing and urbanizing countries like India need forecasting of pollutants in the air to manage air quality. Traditional Artificial Neural Networks (ANN) have limited representation capacity for prediction, especially for complex, non-linear interactions among parameters in environmental data. To forecast hourly concentrations of PM2.5, this study presents an advanced deep learning model that combines Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and a numerical-based ANN.The study measures impact factors such as PM10, NO_2,SO_2, and O_3 in a few chosen cities of India, influencing air pollution and determines the transition from traditional ANN to deep learning models. The result shows that the hybrid CNN-LSTM models reduce Root Mean Squared Error by 22-35% across pollutants, giving better results than classical ANN models.

ABSTRACT Recent advancements in deep learning and the utilization of pre‐trained convolutional neural network (CNN) architectures have led to enhancements in classification tasks. However, these architectures often entail millions of training parameters, posing challenges for real‐world deployment. In this work, we propose an iterative Gaussian feature extractor with a custom 3‐layer CNN network (IGF‐CNN) coupled with a feedforward artificial neural network (ANN) classifier. The input images undergo pre‐processing before being fed to the proposed IGF‐CNN and then ANN classifies the input into Covid‐19, non‐Covid‐19 and pneumonia classes. The suggested model demands considerably fewer parameters and reduces training time substantially and achieves accuracies of 99.80%, 98.78%, 99.0%, respectively, across three different benchmark datasets. We have also performed cross‐dataset validation and obtained consistently good results, further demonstrating the robustness of the proposed approach. The proposed architecture is accurate and efficient and can be integrated with real‐time systems.

ABSTRACT A fine‐tuned pattern net artificial neural network (PNANN) is explored for design of porous ceramic matrix (PCM) based burner by using classification model. The PNANN is attempted to fine‐tune by comparing two different performance functions: mean squared error and sum absolute error. Three different numbers of hidden neurons are also tested while using scaled conjugate gradient as training algorithm. The data for the classification model is obtained by simultaneously solving the governing equations of two phases for the porous ceramic matrix (PCM) based burner. Different values of two critical parameters are used in the numerical model. The critical parameters considered for influencing the performance of the PCM based burner are: convective coupling and extinction coefficient. Radiative heat flux is also incorporated in the numerical model by solving radiative transfer equation by discrete transfer method. Based on the difference between the temperature profiles of the two phases (solid and gas), binary array are constructed and used in the classification model. Ten different classes of data signify ten different regime of operation, each having a unique range of values for the critical parameters. All PNANNs with performance function mse are able to give correct identification above 41.3% and upto 81.1%.

Nowadays, online connectivity is increasing with the rapid growth of the world wide web. Consequently, content shared across numerous platforms varies in appropriateness. it is necessary to ensure the suitability of the content since children are among the consumers of online content. A lot of children watch videos on YouTube these days, and such platforms can contain useful content. However, such videos can also have a negative impact on children. The suitability of these videos can be determined through sentiment analysis to refine the content for children on YouTube, by classifying the posted comments as either positive or negative. Therefore, this study utilizes natural language processing methods, machine learning classifiers (MLCs) and deep learning models (DLMs) to detect and classify negative user comments using the proposed dataset. Different MLCs such as random forest (RF), logistic regression (LR), multinomial Naïve Bayes (MNB), decision tree (DT), K-nearest neighbour (KNN), AdaBoost, and support vector machine (SVM) have been used. Additionally, DLMs were also used such as artificial neural network (ANN), convolutional neural network (CNN) and long short-term memory (LSTM). Overall, the experimental results showed that the LR, RF, AdaBoost, ANN and LSTM classifiers outperformed all the other classifiers in terms of accuracy.

Cold atmospheric pressure plasma (CAP) technology has vast potential in several technological domains, including biomedical engineering. CAP, also known as non-thermal plasma, is characterized by high-energy electrons while the bulk gas remains near room temperature, allowing for effective plasma treatment without thermal damage—critical for biomedical applications. This paper presents a coupled machine learning and statistical technique-based process modeling and optimization approach for a novel floating-helix electrode-based cold plasma device, operating strictly within the cold plasma regime. An artificial neural network (ANN) model was developed to describe the relationship between the process parameters—supply voltage (SV) and frequency (SF)—and performance parameters—power consumption (P), and jet lengths with and without an end ring (JwER and JwoER). The generality and robustness of the ANN model were confirmed through experimental validation and extrapolative predictions. For multi-response optimization, the composite desirability method was employed. Finally, machine learning models for logistic regression—namely, ANN classifier, K-Nearest Neighbor, and Support Vector Machine—were developed to classify the discharge type within the cold plasma operating range, ensuring its suitability for biomedical applications. The proposed system may hold potential for biomedical use, contingent upon further validation through biological testing.

Malassezia pachydermatis is a zoophilic yeast found on the skin and in the outer ear canal of many mammals. It normally maintains a commensal lifestyle but can cause dermatitis and otitis in predisposed hosts, particularly in atopic dogs. M. pachydermatis is genetically diverse, with strains clustering into at least three phylogroups based on molecular typing, a pattern we now confirm through whole-genome sequencing (WGS). Accurate species and strain-level identification is essential for understanding its epidemiology, pathogenic potential, and response to treatment. In this study, we established Fourier Transform Infrared (FTIR) spectroscopy as a rapid, cost-effective method for distinguishing M. pachydermatis from other Malassezia species, including M. globosa, M. furfur, M. restricta, and M. sympodialis. Within M. pachydermatis, FTIR spectroscopy resolved even closely related strains with high accuracy producing clusters congruent with WGS-based phylogeny. The incorporation of an Artificial Neural Network classifier further enhanced the discriminatory power, enabling robust and automated strain assignment. These findings demonstrate the potential of FTIR spectroscopy as a practical tool for large-scale epidemiological surveillance of M. pachydermatis and for clinical and veterinary applications where strain-level identification could inform treatment and management of Malassezia-associated diseases.

Parameter estimation for solar photovoltaic panels is a popular research topic in green energy. Model parameters can be used for fault diagnosis in solar panels. Artificial neural network (ANN) approaches have been developed to estimate the model parameters of solar panels. In this study, an ANN and Adaptive Particle Swarm Optimization (APSO) approach for model parameter estimation of solar panel is proposed. Load perturbation is injected at the output of the solar PV panel, and the load voltage and current time series are measured. The current and voltage vectors are used as inputs for an ANN, which is used as a classifier for the ranges of the model parameters. The population of the APSO is initialized according to the results of the ANN classifier, and the APSO algorithm is then used to estimate the model parameters of the PV panel. Simulations and experimental studies show that the proposed method has better performance than conventional PSO, and it requires a smaller number of generations to achieve an average parameter estimation error of less than 5%.

Accurate prediction of a robot’s dynamic parameters, including mass and moment of inertia, is essential for adequate motion planning and control in autonomous systems. Traditional methods often depend on manual computation or physics-based modelling, which can be time-consuming and approximate for intricate, real-world environments. Recent advances in machine learning, primarily through artificial neural networks (ANNs), offer profitable alternatives. However, the potential of quantum-inspired models in this context remains largely uncharted. The current research assesses the predictive performance of a classical artificial neural network (CANN) and a quantum-inspired artificial neural network (QANN) in estimating a car-like mobile robot’s mass and moment of inertia. The predictive accurateness of the models was considered by minimizing a cost function, which was characterized as the RMSE between the predicted and actual values. The outcomes indicate that while both models demonstrated commendable performance, QANN consistently surpassed CANN. On average, QANN achieved a 9.7% reduction in training RMSE, decreasing from 0.0031 to 0.0028, and an 84.4% reduction in validation RMSE, dropping from 0.125 to 0.0195 compared to CANN. These enhancements highlight QANN’s singular predictive accuracy and greater capacity for generalization to unseen data. In contrast, CANN displayed overfitting tendencies, especially during the training phase. These findings emphasize the significance of quantum-inspired neural networks in enhancing prediction precision for involved regression tasks. The QANN framework has the potential for wider applications in robotics, including autonomous vehicles, uncrewed aerial vehicles, and intelligent automation systems, where accurate dynamic modelling is necessary.

This article is concerned with the proportional-integral-observer (PIO)-based fusion estimation problem for a class of artificial neural networks (ANNs) equipped with multiple sensors, which are constrained by bandwidth and subjected to unknown-but-bounded noises (UBBNs). For the purpose of efficient information communication, an approach known as the one-bit encoding mechanism (OBEM) is proposed that enables the encoding of scalar data using merely a single bit. Then, a local PIO-based set-membership estimator is devised for each sensor node, with the aim of achieving the desired estimation task while considering the possible data distortion due to OBEM and the existence of UBBNs. Subsequently, sufficient conditions are established to ensure the existence and effectiveness of the PIO-based set-membership estimator. Moreover, to enhance the global estimation performance, an ellipsoid-based fusion rule is introduced for all local PIO-based set-membership estimators. The performance of fusion estimation is then analyzed using set theory and the optimization method, leading to the determination of relevant parameters. Finally, the effectiveness and advantages of the proposed estimation algorithm are demonstrated through a simulation example.

Radiomics-Based Differentiation of Primary Central Nervous System Lymphoma and Solitary Brain Metastasis Using Contrast-Enhanced T1-Weighted Imaging: A Retrospective Machine Learning Study.

The broken rotor bar is the fault that most often occurs in induction motors. This paper proposes a method to identify the broken rotor bar fault using a combination of Empirical Mode Decomposition and Artificial Neural Networks (ANNs). The motor current signal is processed using EMD analysis resulting in the Intrinsic Mode Function (IMF) signal. The zero crossing point of the IMF signal is recorded to obtain the Time Successive between Zero Crossing (TSZC). The Probability Density Function (PDF) of the TZSC is used as the ANN classifier input. The PDF properties of peak, width, and standard deviation are selected as the input variables. Two ANNs were designed as fault detection and severity identification systems. The experimental testing also considers the load level variation. The experiment of the broken rotor bar fault diagnostics shows that the ANN-based fault detection system is able to detect faults with accuracy up to 94.2%. Moreover, the ANN-based severity identification successfully identified 76.09% of the cases. In addition, the experiment on load variations reveals that the fault diagnostic is more effective at higher loads.

Smart farming systems, while enhancing agricultural productivity, are increasingly vulnerable to cyber threats due to their reliance on interconnected devices and networks. However, existing Intrusion Detection Systems (IDS) often suffer from high computational costs and suboptimal detection accuracy due to irrelevant features. To address this challenge, this study proposes a novel hybrid filter-based feature selection method designed to optimize feature selection for artificial neural network (ANN)-based IDS in smart farming environments. Unlike conventional methods that rely solely on chi-square, mutual information, or mean absolute deviation, the proposed method combines these techniques to leverage their complementary strengths. Furthermore, a comprehensive smart farming system was established to collect extensive data, creating a dedicated dataset named Smart-Farm-IDS for binary classification, distinguishing between normal operations and anomalies. Both shallow and deep ANN models were employed to detect these anomalies, with their performances compared in detail. Experimental results demonstrate that the proposed hybrid feature selection method enhances detection accuracy while reducing computational overhead compared to existing methods. This study offers a robust approach for improving the security and resilience of smart farming systems, providing a foundation for more secure agricultural operations.

Abstract This study proposed to compare Sentinel‐2 and SPOT 7 multispectral instrument (MSI) images to create a forest canopy density model (FCDm) in the Dalki Dadin area in the South Zagros forests of Fars province, Iran. First, a forest and non‐forest area map was prepared, and then an FCDm was prepared in four categories: 5%–25%, 25%–50%, 50%–75%, and >75%. In this research, the classification of satellite images was done using a parallelepiped classifier, traditional Mahalanobis distance classifier (MDC), maximum‐likelihood classification (MLC), and artificial neural network (ANN) algorithms using appropriate band sets in ENVI 5.3 software. To classify correctly, the ground reality map was implemented based on the interpretation of ortho digital photos of the 80s with a scale of 1:25,000‐scale topographic map. The overall accuracy and kappa coefficient for Sentinel‐2 images (with band combination [BC] of principal component analysis (PCA)‐1‐8 using the MLC algorithm) and the SPOT 7 image (with the BC of PCA‐1‐3 and the use of ANN classification algorithm) were obtained equal to 96.3%, 0.91%, and 87.57%, 0.70, respectively. Therefore, the Sentinel‐2 image has had better results compared to the SPOT 7 image to prepare the forest and non‐forest classification map. Furthermore, the overall accuracy and kappa coefficient for the Sentinel‐2 image (with BC of PCA‐3‐8 using the MLC algorithm) and the SPOT 7 image (with the BC of 2‐3‐4 and the use of ANN classification algorithm) were obtained equal to 88.36%, 0.72%, and 78.74%, 0.64, respectively. Therefore, the Sentinel‐2 image has had better results compared to the SPOT 7 image to provide the forest classification map. Also, after the integration of SPOT7 and SPOT7‐Pan image, the map obtained by PCA method using an ANN classifier with BC of PCA‐2‐4 with a kappa coefficient of 0.75 and accuracy of 89.26% had the highest accuracy. Also, the maps obtained from forest classification into four density classes obtained by PCA method using ANN with BC of PCA‐2‐4 and with a kappa coefficient of 0.37 and accuracy of 59.60% had the highest accuracy. The overall results showed that, according to the extracted information, the Sentinel‐2 image has more appropriate accuracy for producing FCDm in four density classes.