Efficient Model Research Articles

3W‐MultiHier: A Three Way Multi‐Hierarchical Model Enabled Deep Learning for Brain Tumor Classification in MRI Scans

AbstractAccurate brain tumor detection and classification are vital for effective diagnosis and treatment planning in medical imaging. Despite advancements in deep learning, challenges such as multimodal complexity, small lesion segmentation, limited training data, and variability in tumor characteristics hinder precise tumor analysis in MRI scans. To address these issues, we propose the Three Way Multi‐Hierarchical Model (3W‐MultiHier) for tumor classification in MRI. 3W‐MultiHier employs a hybrid Capsule‐Transformer UNet (Capsule‐TransUNet) architecture, integrating capsule and transformer networks within the U‐Net framework. This enables the model to capture spatial hierarchies, long‐range dependencies, and global context, ensuring accurate tumor boundary segmentation. The model also incorporates Residual Network Version 2 ‐ Squeeze‐and‐Excitation Network (ResNetV2‐SENet), which excels at extracting complex features through deep hierarchical structures and feature recalibration. Additionally, the Vision Transformer ‐ Transfer Learning (ViT‐TL) pipeline enhances classification accuracy by leveraging fine‐grained hierarchical representations. Extensive evaluations on BraTS (2019, 2020, 2021) datasets demonstrate the superior performance of 3W‐MultiHier, achieving 99.8% accuracy with rapid training and low loss. These results highlight the model's efficiency in handling diverse datasets and its potential to improve clinical diagnostics by enabling precise, reliable brain tumor classification.

Integrating AI and Deep Learning for Efficient Drug Discovery and Target Identification

Artificial intelligence (AI) shows great potential in medical diagnosis and drug analysis. Through deep learning technology, AI can automatically extract features from large amounts of complex medical data, significantly improving diagnostic accuracy and drug development efficiency. Especially in drug target discovery and antiviral peptide classification, AI technology can accelerate data processing and prediction, helping researchers identify potential therapeutic molecules more quickly and optimize the drug development process. This study proposes and validates a Deep learn-based model, deep-Avpiden, to improve the classification and discovery efficiency of antiviral peptides (AVPs). By using sequential convolutional networks (TCNs), the model outperforms traditional recurrent neural networks (RNNs) and long short-term memory networks (LSTMs) in capturing long-term dependencies and parallel computing capabilities. In terms of dataset, we used AVPs and non-AVPS samples from multiple databases, totaling 5,414 cleaned and de-weighted peptide sequences for model training after data preprocessing and embedding. The Deep-AVPiden model has been shown to outperform existing advanced classifiers in experiments, and its effectiveness has been verified by accuracy, precision, recall rate, and area under the ROC curve (OC-ROC). In addition, to accommodate computing resource constraints, we propose an optimized version of Deep-AvPIDen (DS), which utilizes Deep separation convolution technology to significantly reduce computing resource consumption. Through the online application platform, researchers can efficiently classify antiviral proteins and discover new AVPs. Future research could further optimize the model's computational efficiency, handle larger data sets, and expand its potential for biomedical problems such as drug combination prediction and new drug discovery.

Corrigendum to “Analytical models for crack-hitting probability and crack-healing efficiency by linear capsules in a material with parallelogram crack network” [Mater. Today Commun. 41 (2024) 111048

Corrigendum to “Analytical models for crack-hitting probability and crack-healing efficiency by linear capsules in a material with parallelogram crack network” [Mater. Today Commun. 41 (2024) 111048

3D and 2D-QSAR Studies on Natural Flavonoids for Nitric Oxide Production Inhibitory Activity

Background: Nitric oxide (NO), an important second messenger molecule, regulates numerous physiological responses, while excessive NO generates negative effects on the circulatory, nervous and immune systems. Recently, some natural flavonoids were reported to possess the capability of inhibiting LPS-induced NO production. To fully understand the nature of their own NO inhibitory activity, it is necessary to address the structural requirements of flavonoids as NO inhibitors. Objective: The objective of this work was to develop efficient QSAR models for predicting the NOinhibitory activity of new flavonoids and improving insights into the critical properties of the chemical structures that were required for the ideal NO production inhibitory activities. Methods: To provide insights into the structural basis of flavonoids as NO inhibitors, 3D quantitative structure-activity relationship (3D-QSAR) and 2D-QSAR models were developed on a dataset of 55 flavonoids using comparative molecular field analysis (CoMFA), comparative molecular similarity indices analysis (CoMSIA) and hologram quantitative structure-activity relationship (HQSAR) approaches. Results: The statistically significant models for CoMFA, CoMSIA and HQSAR resulted in crossvalidated coefficient (q2) values of 0.523, 0.572 and 0.639, non-cross-validated coefficient (r2) values of 0.793, 0.828 and 0.852, respectively. The robustness of these models was further affirmed using a test set of 18 compounds, which resulted in predictive correlation coefficients (r2 pred) of 0.968, 0.954 and 0.906. Furthermore, the models-derived contour maps were appraised for activity trends for the molecules analyzed. Conclusion: The 3D and 2D-QSAR models constructed in this paper were efficient in estimating the NO inhibitory activities of flavonoids and facilitating the design of flavonoid-derived NO production inhibitors.

Integrating YOLOv7 with FixMatch for Enhancing Vehicle Detection Performance in Mixed Traffic Environments

A major challenge in the development of object detection technology is the significant reliance on large labeled datasets, which requires substantial time and memory for manual annotation—especially in complex, mixed traffic environments with varied vehicle types, congestion levels, and unpredictable motion patterns. This study addresses this issue by integrating the semi-supervised learning technique, FixMatch, into the YOLOv7 object detection model, utilizing 4000 transportation-related datasets. The FixMatch technique enables the model to detect unlabeled objects effectively through strong and weak augmentation methods. In this study, the detected objects in the mixed traffic environment include public transportation, pedicabs, cars, motorcycles, and trucks. This study achieved an impressive 97.5% detection accuracy by leveraging unlabeled data, demonstrating the model's efficiency and effectiveness in identifying vehicles under diverse traffic conditions. Consequently, integrating the FixMatch method into YOLOv7 provides a practical and efficient solution for object detection in situations where collecting labeled data is challenging, such as in dynamic and highly variable traffic environments.

Residual Pix2Pix networks: streamlining PET/CT imaging process by eliminating CT energy conversion.

Objective
Attenuation correction of PET data is commonly conducted through the utilization of a secondary imaging technique to produce attenuation maps. The customary approach to attenuation correction, which entails the employment of CT images, necessitates energy conversion. However, the present study introduces a novel deep learning-based method that obviates the requirement for CT images and energy conversion.
Methods
This study employs a residual Pix2Pix network to generate attenuation-corrected PET images using the 4033 2D PET images of 37 healthy adult brains for train and test. The model, implemented in TensorFlow and Keras, was evaluated by comparing image similarity, intensity correlation, and distribution against CT-AC images using metrics such as PSNR and SSIM for image similarity, while a 2D histogram plotted pixel intensities. Differences in standardized uptake values (SUV) demonstrated the model's efficiency compared to the CTAC method.
Results
The residual Pix2Pix demonstrated strong agreement with the CT-based attenuation correction, the proposed network yielding MAE, MSE, PSNR, and MS-SSIM values of 3×10-3, 2×10-4, 38.859, and 0.99, respectively. The residual Pix2Pix model's results showed a negligible mean SUV difference of 8×10-4(P-value = 0.10), indicating its accuracy in PET image correction. The residual Pix2Pix model exhibits high precision with a strong correlation coefficient of R2 = 0.99 to CT-based methods. The findings indicate that this approach surpasses the conventional method in terms of precision and efficacy.
Conclusions
The proposed residual Pix2Pix framework enables accurate and feasible attenuation correction of brain F-FDG PET without CT. However, clinical trials are required to evaluate its clinical performance. The PET images reconstructed by the framework have low errors compared to the accepted test reliability of PET/CT, indicating high quantitative similarity.&#xD.

MOBILENET PERFORMANCE IMPROVEMENTS FOR DEEPFAKE IMAGE IDENTIFICATION USING ACTIVATION FUNCTION AND REGULARIZATION

Deepfake images are often used to spread false information, manipulate public opinion, and harm individuals by creating fake content, making developing deepfake detection technology essential to mitigate these potential dangers. This study utilized the MobileNet architecture by applying regularization and activation function methods to improve detection accuracy. ReLU (Rectified Linear Unit) enhances the model's efficiency and ability to capture non-linear features, while Dropout and L2 regularization help reduce overfitting by penalizing large weights, thereby improving generalization. Based on experimental results, the MobileNet model optimized with ReLU and Dropout achieved an accuracy of 99.17% in the training phase, 85.34% in validation, and 70.60% in testing, whereas the MobileNet model optimized with ReLU and L2 showed lower accuracy in the training and validation phases compared to Dropout but achieved higher accuracy in testing at 72.18%. This study recommends MobileNet with ReLU and L2 due to its better generalization ability when testing data (resulting from reduced overfitting).

Performance Assessment of Some Count Data Models to Immunization Coverage Data

This research evaluates the performance of various count data models, including Poisson Regression (PR), Zero-Inflated Poisson Regression (ZIP), Zero-Truncated Poisson Regression (ZTP), Truncated Negative Binomial Poisson Regression (TNBP), and Negative Binomial Poisson Regression (NBP), using immunization coverage data from the National Primary Health Care Development Agency (NPHCDA). The study focuses on children under 12 months, assessing model fit using Likelihood Ratio (LR), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC) criteria. Analysis conducted with STATA indicates that the Truncated Negative Binomial Poisson Regression (TNBP) outperformed other models in fit and efficiency. Both the ZTeeP and TNBP models demonstrated the best fit, with lower AIC (1959.107) and BIC (2037.649) values and higher Pseudo R-squared values (0.0677 for ZTP and 0.0590 for TNBP), compared to standard models. Age was identified as a significant predictor, negatively associated with immunization status, implying that older infants in the under-12-month category are less likely to receive all vaccinations. The ZTP model showed significant positive effects for antigens such as HepB0, OPV0, BCG, and Measles, with age having a significant negative association. The findings highlight the importance of selecting appropriate statistical models for accurate public health data analysis, enhancing decision-making in immunization programs.

A Distributed Mobile Edge Computing Based Dynamic Resource Allocation in 5G Network Using Green Anaconda Optimization Based Deep Learning Network

ABSTRACTMobile edge computing (MEC) facilitates storage, cloud computing, and analysis capabilities near to the users in 5G communication systems. MEC and deep learning (DL) are combined in 5G networks to enable automated network management that provides resource allocation (RA), energy efficiency (EE), and adaptive security, thereby reducing computational costs and enhancing user services. A hybrid quantum‐classical convolutional neural network (HQCCNN) with simplicial attention network (SAN) is presented in the study that allocates appropriate resources for various users in the network. First, the green anaconda optimization (GAO) algorithm is used to optimize the objective function for effective RA. Consequently, the neural network receives the optimized objective functions to allocate resources. In the study, the suggested HQCCNN‐GAO model assesses the degree of need for every user and, based on those needs, allots resources to every user in the 5G network while preserving higher throughput and EE. Throughput, latency, mean square errors, processing time, bit error rates, and EE are used to measure the proposed model's efficiency. A few of the RA models that are now in use are contrasted with the outcomes of the suggested method. From the obtained outcomes, it is noticed that the suggested model provides a low latency of 0.08 s and a high throughput of 790 kbps for a range of network users.

DV-DETR: Improved UAV Aerial Small Target Detection Algorithm Based on RT-DETR.

For drone-based detection tasks, accurately identifying small-scale targets like people, bicycles, and pedestrians remains a key challenge. In this paper, we propose DV-DETR, an improved detection model based on the Real-Time Detection Transformer (RT-DETR), specifically optimized for small target detection in high-density scenes. To achieve this, we introduce three main enhancements: (1) ResNet18 as the backbone network to improve feature extraction and reduce model complexity; (2) the integration of recalibration attention units and deformable attention mechanisms in the neck network to enhance multi-scale feature fusion and improve localization accuracy; and (3) the use of the Focaler-IoU loss function to better handle the imbalanced distribution of target scales and focus on challenging samples. Experimental results on the VisDrone2019 dataset show that DV-DETR achieves an mAP@0.5 of 50.1%, a 1.7% improvement over the baseline model, while increasing detection speed from 75 FPS to 90 FPS, meeting real-time processing requirements. These improvements not only enhance the model's accuracy and efficiency but also provide practical significance in complex, high-density urban environments, supporting real-world applications in UAV-based surveillance and monitoring tasks.

Research on Fine-Tuning Optimization Strategies for Large Language Models in Tabular Data Processing.

Recent advancements in natural language processing (NLP) have been significantly driven by the development of large language models (LLMs). Despite their impressive performance across various language tasks, these models still encounter challenges when processing tabular data. This study investigates the optimization of fine-tuning strategies for LLMs specifically in the context of tabular data processing. The focus is on the effects of decimal truncation, multi-dataset mixing, and the ordering of JSON key-value pairs on model performance. Experimental results indicate that decimal truncation reduces data noise, thereby enhancing the model's learning efficiency. Additionally, multi-dataset mixing improves the model's generalization and stability, while the random shuffling of key-value pair orders increases the model's adaptability to changes in data structure. These findings underscore the significant impact of these strategies on model performance and robustness. The research provides novel insights into improving the practical effectiveness of LLMs and offers effective data processing methods for researchers in related fields. By thoroughly analyzing these strategies, this study aims to establish theoretical foundations and practical guidance for the future optimization of LLMs across a broader range of application scenarios.

Day-Ahead Optimization of Proton Exchange Membrane Electrolyzer Operations Considering System Efficiency and Green Hydrogen Production Constraints Imposed by the European Regulatory Framework

Clean hydrogen (H2) use (i.e., produced using either renewable or low-carbon energy sources) can help decarbonize energy-intensive industries, the transport sector, and the power sector. The European regulatory framework establishes that the production of green H2 must be supported either by the electricity grid through a power purchase agreement (PPA) or by intermittent renewable energy source (RES) plants owned by the hydrogen producer. Although the issue of the optimization of hydrogen production costs has already been approached, constraints related to the current regulatory framework and the modeling of nonlinear electrolyzer efficiency still represent open problems. In this paper, a mixed-integer linear programming (MILP) problem, assuming as the objective function the overall cost minimization of the allowed energy mix for green H2 production, is formulated. Two approaches are compared: in the first one, electrolyzers can only operate at 100% load, whereas the second one allows for more flexible electrolyzer scheduling, by enabling partial-load working operations. The simulation results of several scenarios considering different H2 production targets, forecasted RES production, and cost for PPAs demonstrate the effectiveness of the proposed methodology.

A queueing system for analysing traffic throughput in mixed traffic with semi- and fully-autonomous vehicles

As autonomous vehicles (AVs) integrate into transportation systems, a transitional period will emerge where semi-autonomous vehicles (semi-AVs), which still require an attentive driver for steering, coexist with fully-autonomous vehicles (fully-AVs). Focusing on this phase, this study proposes a novel framework that incorporates Markov chains into an M / G n / n max / n max state-dependent queueing system to capture the distinct platooning dynamics and their impacts on the long-term travel efficiency of mixed traffic. The developed Markov chain-based traffic queueing system enables the derivation of traffic throughput in mixed semi- and fully-AV traffic, accounting for their diverse platooning and car-following behaviours. Numerical experiments are used to assess the influence of fully- and semi-AVs on throughput and driving experience in the mixed traffic. This study advances the modelling and analysis of traffic flow efficiency during the transitional phase of mixed traffic, offering insights for managing and controlling semi-AV traffic in the near future.

How efficient are schools in South-East Asia? An analysis through OECD PISA 2018 data

ABSTRACT This study assesses school efficiency in six South-East Asian countries using OECD PISA 2018 data. The average PISA scores were 425.4 in mathematics, 429.5 in science, and 406.2 in reading, with Singapore scoring the highest and the Philippines lowest. Using stochastic frontier analysis, the study model's efficiency across these domains, finding average efficiency scores of 38.47% in math, 48.83% in science, and 49.82% in reading. The study also examines how ICT affects scores and efficiency, suggesting that internet-connected computers can improve PISA scores, while general computer availability may boost efficiency. This analysis offers a benchmark for schools in the region, providing insights into factors that influence both PISA performance and efficiency.

Automatic segmentation and visualization of cortical and marrow bone in mandibular condyle on CBCT: a preliminary exploration of clinical application.

To develop a deep learning-based automatic segmentation method for cortex and marrow in mandibular condyle on cone-beam computed tomography (CBCT) images and explore its clinical application. 825 condyles of 490 CBCT images from 3 centers of Stomatology hospitalaffliated to Zhejiang University School of Medicinewere collected. A deep learning model was developed for simultaneous segmentation of cortex and marrow in mandibular condyle. It included a region of interest extraction network and a segmentation network based on 3D U-net, with modifications made to improve the segmentation boundaries. To evaluate its clinical potential, the model's segmentation efficiency and accuracy were compared with those of both junior and senior oral and maxillofacial radiologists. Additionally, the model's ability to assist junior radiologists in diagnosis through visualization and quantitative analysis of the generated 3D model was also assessed. The Dice similarity coefficient of the deep learning model was 0.901 (cortex), 0.969 (marrow), and 0.982 (entire condyle). Hausdorff distance was 0.755mm (cortex), 0.826mm (marrow), and 0.760mm (entire condyle). The model outperformed radiologists across all segmentation metrics, completing the task in merely 15.06s. With the assistance of visualization and quantitative analysis generated from the model's segmentation, the diagnostic accuracy of junior radiologists significantly improved. The proposed deep learning-based model achieved accurate and efficient segmentation for mandibular condylar cortex and marrow. It possessed capability to generate precise 3D models, facilitating visual quantitative measurement and aiding in the diagnosis of condylar bony changes. This model holds potential for clinical applications in orthognathic surgery, orthodontic treatment, and other TMJ-related interventions.

The Role of Machine Learning in Improving Robotic Perception and Decision Making

Machine learning, specifically through Convolutional Neural Networks (CNNs) and Reinforcement Learning (RL), significantly enhances robotic perception and decision-making capabilities. This research explores the integration of CNNs to improve object recognition accuracy and employs sensor fusion for interpreting complex environments by synthesizing multiple sensory inputs. Furthermore, RL is utilized to refine robots real-time decision-making processes, which reduces task completion times and increases decision accuracy. Despite the potential, these advanced methods require extensive datasets and considerable computational resources for effective real-time applications. The study aims to optimize these machine learning models for better efficiency and address the ethical considerations involved in autonomous systems. Results indicate that machine learning can substantially advance robotic functionality across various sectors, including autonomous vehicles and industrial automation, supporting sustainable industrial growth. This aligns with the United Nations Sustainable Development Goals, particularly SDG 9 (Industry, Innovation, and Infrastructure) and SDG 8 (Decent Work and Economic Growth), by promoting technological innovation and enhancing industrial safety. The conclusion suggests that future research should focus on improving the scalability and ethical application of these technologies in robotics, ensuring broad, sustainable impact.

Machine learning-based EEG analysis for enhanced schizophrenia classification and diagnosis

Abstract. Schizophrenia can have a significant impact on patients' lives, studies, and work. Patients may experience delusions, hallucinations, disorganized thinking, and abnormal behavior, among other symptoms. Therefore, research related to schizophrenia is of great importance. This paper proposes the design of a classifier based on machine learning for diagnosing schizophrenia. The classifier extracts N100 features, P300 features, and power features across different frequency bands from both time-domain and frequency-domain characteristics from Electroencephalography (EEG) signals. Given the small dataset, a Support Vector Machine (SVM) machine learning algorithm was chosen to process and classify the selected features. To address the limitations of SVM in handling high-dimensional data and nonlinear problems, this study introduces a comprehensive improvement method based on Bayesian optimization, Recursive Feature Elimination (RFE), and data augmentation. Bayesian optimization was used to find the best combination of hyperparameters for the model, thereby improving its performance; RFE was employed to assess feature importance and remove the least important features, enhancing the model's training efficiency and generalization capability. Additionally, data augmentation was included to increase the sample size and introduce diversity, thereby improving the model's robustness. The study found that these methods effectively improved classification accuracy and generalization ability.

Non-small cell lung cancer detection through knowledge distillation approach with teaching assistant.

Non-small cell lung cancer (NSCLC) exhibits a comparatively slower rate of metastasis in contrast to small cell lung cancer, contributing to approximately 85% of the global patient population. In this work, leveraging CT scan images, we deploy a knowledge distillation technique within teaching assistant (TA) and student frameworks for NSCLC classification. We employed various deep learning models, CNN, VGG19, ResNet152v2, Swin, CCT, and ViT, and assigned roles as teacher, teaching assistant and student. Evaluation underscores exceptional model performance in performance metrics achieved via cost-sensitive learning and precise hyperparameter (alpha and temperature) fine-tuning, highlighting the model's efficiency in lung cancer tumor prediction and classification. The applied TA (ResNet152) and student (CNN) models achieved 90.99% and 94.53% test accuracies, respectively, with optimal hyperparameters (alpha = 0.7 and temperature = 7). The implementation of the TA framework improves the overall performance of the student model. After obtaining Shapley values, explainable AI is applied with a partition explainer to check each class's contribution, further enhancing the transparency of the implemented deep learning techniques. Finally, a web application designed to make it user-friendly and classify lung types in recently captured images. The execution of the three-stage knowledge distillation technique proved efficient with significantly reduced trainable parameters and training time applicable for memory-constrained edge devices.

Predictive modeling of CO2 capture efficiency using piperazine solutions: a comparative study of white-box algorithms

The urgency to mitigate carbon dioxide (CO2) emissions and combat climate change has spurred the development of effective CO2 capture technologies. One of the industry’s most well-known CO2 collection processes is CO2 absorption utilizing amine solvents. However, designing a successful amine scrubbing system in power plants requires precise prediction of CO2 absorption in aqueous amine solutions under various operating circumstances. Using aqueous piperazine (PZ) solutions for chemical absorption is promising due to the favorable reactivity of PZ with CO2. In this study, a comprehensive evaluation of PZ solution performance in CO2 capturing, employing the white-box algorithms, namely, Genetic Programming (GP), Gene Expression Programming (GEP), and Group Method of Data Handling (GMDH), was performed. Through extensive experimentation and data analysis, several correlations were developed with high and acceptable R2 values, such as 0.933 for GP, 0.949 for GEP, and 0.889 for GMDH, which shows high accuracy and reliability in predicting the CO2 capture efficiency of PZ solutions under varying operating conditions. The results of sensitivity analysis revealed that CO2 partial pressure increased CO2 absorption, while PZ concentration and temperature had negative and decreasing effects. These insights provide essential guidance for optimizing process conditions to enhance the CO2 capture efficiency. Finally, the leverage method was used to assess the reliability of both experimental and predicted data from white-box algorithms. This analysis identified potential outliers and validated the accuracy of the model’s predictions, enhancing the credibility of developed correlations and demonstrating the robustness of the approach.

Real Time Object Detection Algorithm in Foggy Weather Based on WVIT-YOLO Model

Introduction: To address the challenges of low visibility, object recognition difficulties, and low detection accuracy in foggy weather, this paper introduces the WViT-YOLO real-time fog detection model, built on the YOLOv5 framework. The NVIT-Net backbone network, incorporating NTB and NCB modules, enhances the model's ability to extract both global and local features from images. Method: An efficient convolutional C3_DSConv module is designed and integrated with channel attention mechanisms and ShuffleAttention at each upsampling stage, improving the model's computational speed and its ability to detect small and blurry objects. The Wise-IOU loss function is utilized during the prediction stage to enhance the model's convergence efficiency. Result: Experimental results on the publicly available RTTS dataset for vehicle detection in foggy conditions demonstrate that the WViT-YOLO model achieves a 3.2% increase in precision, a 9.5% rise in recall, and an 8.6% improvement in mAP50 compared to the baseline model. Furthermore, WViT-YOLO shows a 9.5% and 8.6% mAP50 improvement over YOLOv7 and YOLOv8, respectively. For detecting small and blurry objects in foggy conditions, the model demonstrates approximately a 5% improvement over the benchmark, significantly enhancing the detection network's generalization ability under foggy conditions. Conclusion: This advancement is crucial for improving vehicle safety in such weather. The code is available at https://github.com/QinghuaZhang1/mode.