Training Model Research Articles

Iteration and Reconstruction: Research on the Cultivation of Network and New Media Talents in the Context of New Liberal Arts

As a new major in journalism and communication, network and new media is born to meet the needs of the development of the digital information age and the requirements of the integration trend of mobile Internet media. Professional construction and personnel training should be more closely related to the needs of The Times and social progress. At present, there are some problems in the cultivation of network and new media professionals in three aspects: concept, model and mechanism. The construction of the major needs to grasp the historical opportunity of the construction of the new liberal arts and upgrade the professional talent training system from the perspective of keeping pace with The Times. Specifically, the construction of network and new media major should adhere to the new era education concept of “intelligence and wisdom”, pay more attention to the “multi-dimensional” teaching and training model, and try to strengthen school-enterprise cooperation from the mechanism level, and build a new media “campus factory”.

79 THE IMPACT OF ENGLISH LANGUAGE THERAPY ON LANGUAGE FUNCTION IN PATIENTS WITH SCHIZOPHRENIA

Abstract Background Schizophrenia is a serious mental disorder, with clinical manifestations including abnormalities in emotions, thinking, perception, and behavior. Language dysfunction is one of the common symptoms in patients with schizophrenia, manifested as reduced speech, unclear expression, and difficulty understanding. Long term use of antipsychotic drugs can cause damage to patients’ cognitive function, and some studies suggest that cognitive rehabilitation training can help improve their social functioning. In recent years, language therapy and healing models have gradually received attention. English language training can not only improve patients’ language organization ability, but also help them better express emotions and enhance psychological resilience. Therefore, this study aims to explore the impact of the English language combined therapy model on the language function of patients with schizophrenia, in order to provide reference value for clinical intervention. Methods Sixty patients with schizophrenia admitted to a rehabilitation hospital of a certain mental health center were selected as the research subjects. The admitted subjects met the diagnostic criteria for schizophrenia and were able to independently complete the experimental content during the training process. According to the numerical random table method, they were divided into an intervention group and a control group, with 30 cases in each group. Both groups of patients received routine care, with the intervention group receiving English language training and a healing mode. English language training refers to patients receiving English language listening, speaking, reading, and writing training three times a week for two hours each time, with mostly positive and interesting topics selected. The healing mode involves communication between rehabilitation therapists and patients, creating a relaxed and comfortable chat environment, guiding patients to communicate independently and open up their hearts. During the experiment, the General Self Efficacy Scale and Language Function Assessment were used as evaluation tools to assess the intervention effects of the two groups of patients before and after training. Results In Table 1, the effective rate of the intervention group after intervention was significantly higher than that of the control group, and the data difference was significant. Discussion Research has shown that the English language combined therapy model has a significant improvement effect on language function in patients with chronic schizophrenia, with an intervention effectiveness rate (96.67%) much higher than the control group receiving conventional treatment alone (50%). Through English language training, patients can gradually improve their language abilities, enhance their confidence and social skills, and the application of healing models has created a relaxed and inspiring environment for patients to market, creating favorable conditions for the recovery of language function. By integrating English language training and healing models, patients can gradually recover their language abilities, improve their social skills, and provide new design ideas for clinical treatment. Funding No. PHKT2312.

Digital Physiotherapeutic Scoliosis-Specific Exercises for Adolescent Idiopathic Scoliosis

The effectiveness of a training support program of individually tailored physiotherapeutic scoliosis-specific exercises (PSSEs) delivered via a digital care system is unclear. To determine the effectiveness of a digital care program in which patients receive PSSE supervision and guidance via the Healbone Intelligent Rehabilitation System and educational videos compared with a conventional intervention involving 3 treatment sessions conducted by qualified physiotherapists at outpatient clinics or specialized orthopedic institutions and unsupervised home-based PSSE training. This randomized clinical trial recruited patients aged 9 to 17 years from Peking Union Medical College Hospital in Beijing, China, who had adolescent idiopathic scoliosis with a primary curve Cobb angle of 10° or greater and skeletal immaturity (as determined via the Risser grading system). The trial was conducted from June 1, 2023, to August 10, 2024. Patients were randomly assigned to a digital care (DC) group, in which each patient received fully remote and home-based PSSE training supported by a digital care system, or a usual care group, in which each patient completed 3 treatment sessions conducted by qualified physiotherapists at outpatient clinics or specialized orthopedic institutions and home-based PSSE training without supervision. The primary outcome was a change in the Cobb angle of the major curve between baseline and month 6 of the intervention. Primary and secondary outcomes were evaluated using an independent samples t test. The proportion of patients who exhibited disease progression or improvement in the Cobb angle between the 2 groups was compared using a χ2 test. Of 591 patients assessed for eligibility, 436 did not meet the eligibility criteria, 12 declined to participate, and 15 withdrew before randomization. The remaining 128 patients were randomized to 2 parallel groups, with 64 patients in each group (mean [SD] age, 11.1 (2.2) years; 97 female [75.8%]), all of whom completed the baseline assessment and 6-month follow-up. The mean difference in the improvement of the Cobb angle between the DC group and the usual care group after a 6-month intervention was -4.23° (95% CI, -6.08° to -2.39°) in the intention-to-treat analysis and -4.01° (95% CI, -5.68° to -2.35°) in the per-protocol analysis. The posttreatment Cobb angle was adjusted by using analysis of covariance, and the mean of the difference in the adjusted posttreatment Cobb angle between 2 groups was -4.24° (95% CI, -6.09° to -2.38°). In this randomized clinical trial, the effectiveness of individually tailored PSSE training support delivered via a digital care system was superior to that of a conventional PSSE training model in improving the Cobb angle. This mode of exercise may be an effective and convenient alternative for individuals with adolescent idiopathic scoliosis. Chinese Clinical Trial Registry Identifier: ChiCTR2300076563.

Training on a 3D-Printed Simulation Model Improves Accuracy in External Ventricular Drain Placement.

External ventricular drain (EVD) placement is a critical, lifesaving procedure in cranial neurosurgery, often performed manually using anatomical landmarks that vary between individuals. This study evaluates the efficacy of a 3-dimensional (3D)-printed EVD training model designed to improve the accuracy of this procedure. Computed tomography scans from 3 patients were used to create 3D-printed head models with narrow, wide, and normal ventricles. Twenty-five neurosurgeons participated in a three-round training protocol: pre-training, training with neuronavigation and a standardized protocol, and post-training. The accuracy of EVD placement was measured using an optical navigation system, and participants' confidence levels were assessed through questionnaires. Training significantly enhanced EVD placement accuracy. Pre-training, only 55.3% of placements were intraventricular (Kakarla grade 1), which increased to 84.0% post-training (P < .001). The distance to the ideal entry point improved from 5.8 mm (SD, ±3.7 mm) to 4.1 mm (SD, ±1.5 mm), and the distance to the target point improved from 12.6 mm (SD, ±5.8 mm) to 8.3 mm (SD, ±4.0 mm) (P < .001 for both). The time to identify entry points and puncture the ventricles also improved significantly. Left-sided EVDs were more frequently misplaced. In addition, right-handed participants (n = 24) performed better when placing left-sided EVDs with their right hand. Participants with more than 6 years of experience were more likely to misplace the EVD and overestimate their placement accuracy compared with less experienced participants. Post-training, both experienced and less experienced neurosurgeons achieved similar success rates. Confidence in EVD placement and puncture direction significantly increased post-training. A standardized training protocol using a 3D-printed model significantly improves the accuracy and confidence of neurosurgeons in EVD placement. Regular training is recommended to maintain high clinical performance, emphasizing the need for standardized procedures and the use of neuronavigation for complex cases.

Comparative study of peripheral intravenous catheter insertions with capillary, notched, and a grooved needle flashback design.

Evidence of the costly effects of first-attempt peripheral intravenous catheter (PIVC) insertion failures continues to mount. This study was conducted to determine if a unique catheter design can improve operative error, promote PIVC first-stick success and reduce the costs of first-attempt failures. In Phase One of this comparative simulation use in vitro study, 16 nurses from acute care hospitals inserted four PIVC types into a training model, each type characterized by distinct timing of flashback occurrence and needle design. Each nurse performed three attempts per catheter type (12 total per inserter). Insertions were video-recorded and analyzed for the effect of the needle on a vessel during insertion, double punctures and catheter placement failure or success. In Phase Two, 100 nurses and 25 purchasing agents nationwide identified items used and associated costs for PIVC insertions. In Phase One, nurses using a grooved needle flashback with a thin-tipped needle were 20% more successful with catheter placement than those using a notched needle flashback and 22% more successful than those placing a capillary flashback design. Catheter placement success with a grooved needle flashback was 15% higher than with a capillary flashback and 13% higher than with a notched needle flashback. Double punctures were highest among nurses using a capillary flashback catheter, and catheter placement failure was highest among nurses using notched needle flashback. In Phase Two, nurses reported an average of 51% first-attempt success. The authors calculated the estimated annual cost of first-stick failure to be $US 35,919.15 per nurse, including labor and materials used in the second and third attempts. In this study, the authors found the insertion technique with grooved flashback needle design reduced operative error, double punctures and improved first-attempt success. This flashback needle design could significantly improve patient outcomes, speed time to treatment and reduce hospital costs.

Intelligent Monitoring of Tunnel Fire Smoke Based on Improved YOLOX and Edge Computing

To overcome the defects of traditional fire detection methods that have a high false alarm rate and long delay, a smart tunnel fire monitoring method based on a YOLOX deep convolutional neural network and edge computing is proposed. This method first improves the detection accuracy by analyzing the relationship between frequency domain and convolutional neural networks and the use of wavelet transform. Then, based on the smoke features observed in the experiments, a fuzzy loss method is proposed to accelerate the model convergence speed. To address the issue of a weak computing power of edge devices, the training model is optimized by using knowledge distillation and model quantization, thereby improving the running speed on edge devices. At the same time, a series of related lightweight methods are adopted to optimize the model, reduce the computational cost, and improve the detection speed. Finally, the accuracy of flame and smoke detection on a self-built dataset reaches 85%, which is about 1.8% higher than the baseline method YOLOX and achieves a balance between the speed and accuracy of the model.

Experiences in a Clinical Innovation Pharmacy Fellowship: A Novel Model of Ambulatory Care Training and Practice Advancement.

The University of Utah Clinical Innovation Fellowship models novel partnerships between third-party payers, clinical practices, and academia. While healthcare costs continue to increase unabated and physician burnout leads to provider shortages, this fellowship focuses on both crises by training pharmacists to establish new practices in ambulatory clinic spaces using funding provided by third-party payers. Not only does this fellowship represent a future in which pharmacists are able to address third-party payers' need to reduce healthcare costs and clinics' need to address provider shortages, it also successfully trained fellows to pursue jobs in ambulatory care and academia. Payers, clinics, providers and patients all expressed a high degree of satisfaction with the work of the fellows. In multiple clinics where fellows established new pharmacy services, those services led directly to new job approvals funded by the clinics themselves. The purpose of this paper is to serve as a model by which fellowship programs elsewhere can be designed, as well as to show that partnerships between ambulatory clinics, payers, and pharmacists are both sustainable and beneficial to all parties including, most importantly, the patients who receive better care for their complex chronic disease states. While this paper is descriptive in nature, work is ongoing to objectively measure the impact of the fellows on patients, providers, and third-party payers. A sampling of outcomes is presented, describing the impact of the pharmacist fellows' efforts to improve medication management in primary care. Even with limited objective measures of success, we are able conclude that over the past 3 years, the fellowship has accomplished its aim of preparing fellows for future roles in ambulatory care, practice design, and academia while also demonstrating that a funding model aligning payers, clinics, and academia is sustainable.

Multi-Scale Debris Flow Warning Technology Combining GNSS and InSAR Technology

The dynamic loads of fluid impact and static loads, such as the gravity of a rock mass during the formation of debris flows, exhibit a coupled effect of mutual influence. Under this coupling effect, surface monitoring points in disaster areas experience displacement. However, existing methods do not consider the dynamic–static coupling effects of debris flows on the surface. Instead, they rely on GNSS or InSAR technology for dynamic or static single-scale monitoring, leading to high Mean Absolute Percentage Error (MAPE) values and low warning accuracy. To address these limitations and improve debris flow warning accuracy, a multi-scale warning method was proposed based on Global Navigation Satellite System (GNSS) and Synthetic Aperture Radar Interferometry (InSAR) technology. GNSS technology was utilized to correct coordinate errors at monitoring points, thereby enhancing the accuracy of monitoring data. Surface deformation images were generated using InSAR and Small Baseline Subset (SBAS) technology, with time series calculations applied to obtain multi-scale deformation data of the surface in debris flow disaster areas. A debris flow disaster morphology classification model was developed using a support vector mechanism. The actual types of debris flow disasters were employed as training labels. Digital Elevation Model (DEM) files were utilized to extract datasets, including plane curvature, profile curvature, slope, and elevation of the monitoring area, which were then input into the training model for classification training. The model outputted the classification results of the hidden danger areas of debris flow disasters. Finally, the dynamic and static coupling variables of surface deformation were decomposed into valley-type internal factors (rock mass static load) and slope-type triggering factors (fluid impact dynamic load) using the moving average method. Time series prediction models for the variable of the dynamic–static coupling effects on surface deformation were constructed using polynomial regression and particle swarm optimization (PSO)–support vector regression (SVR) algorithms, achieving multi-scale early warning of debris flows. The experimental results showed that the error between the predicted surface deformation results using this method and the actual values is less than 5 mm. The predicted MAPE value reached 6.622%, the RMSE value reached 8.462 mm, the overall warning accuracy reached 85.9%, and the warning time was under 30 ms, indicating that the proposed method delivered high warning accuracy and real-time warning.

NEURAL NETWORKS FOR DETECTING FAKE NEWS AND MISINFORMATION: AN AI-POWERED FRAMEWORK FOR SECURING DIGITAL MEDIA AND SOCIAL PLATFORMS

The growing concern of fake news and information in contemporary society threatens the integrity of democracy and global security. Social media and on-line news websites are now considered to be some of the primary channels of fake news dissemination since they are supported by engagement-based content promotion algorithms and bot accounts from adversaries. Organic fact-checking cannot cope with the current flood of fake news and thus there is a need for machine learning (ML)-based solutions. As much as this research focus on general neural networks, this work mainly concentrates on deep learning models in dealing with fake news and misinformation detection; CNNs, RNNs, LSTM, BERT, GPT-3, and RoBERTa. The performance of these models is assessed by employing the benchmark datasets including Fake Newsnet, LIAR, PHEME, and PolitiFact, and the evaluation is made based on accuracy and computational time along with studying model’s compatibility with various types of fake news. Empirical evidence shows that the Transformer-based models improve on the traditional machine learning resulting in more than 95 % precision with enhanced contextual meaning. However, computational cost is still a drawback, and in order to overcome this, better and more efficient hybrid models are needed. Likewise, the current study also addresses some of the critical linguistic and metadata elements such as sentiments, source reliability, and social interactions that define this phenomenon. In terms of error analysis, this research finds that political misinformation represents the most significant area of difficulty for AI models while underlining the importance of domain-specific training and non-stopping model updates. The proposed AI-based framework uses NLP and social network analysis to improve the process of real-time misinformation detection, which can solve the problem of security in digital media and platforms. This study advances knowledge on fake news detection using artificial intelligence and paves way for new approaches on the further development of artificial intelligence fact-checking, ethical issues concerning artificial intelligence, and integration of explainable artificial intelligence in the fight against fake news.

Exploring Essential Acoustic Features for Early Parkinson’s Disease Classification: A Machine Learning Study

Parkinson’s disease (PD) is a neurological condition that affects approximately 10 million individuals globally and is ranked as the second most prevalent neurodegenerative condition after Alzheimer’s disease. Vocal disorders can be identified in approximately 90% of PD patients in the early stages of the disease. In this study, 19 machine learning (ML) algorithms were applied to a database of voice recordings of healthy individuals and individuals with PD obtained from a public repository. Different feature selection (FS) and hyperparameter optimization techniques were applied to all models for training, testing, and validation data. Among the ML algorithms, support vector machine with radial kernel, Naïve Bayes (NB), and Gaussian process classifier (GPC) yielded promising results when considering all features. Linear discriminant analysis, K neighbors classifier (KNN), extra trees classifier, GPC, and NB demonstrated excellent performance on the testing data after employing FS techniques. Decision tree classifier, KNN, and GPC emerged as the top performers when applied to the validation dataset. Our findings, derived from an extensive and chronological review of studies utilizing the same dataset, which surpass previous benchmarks, provide a comprehensive understanding of ML’s application in voice analysis to support accurate clinical decision-making.

Zero-trust based dynamic access control for cloud computing

Most corporations and organizations rely heavily on access control to protect data accessibility and enable resource sharing across networks and departments. However, with the development of cloud computing, traditional boundary protection struggles to mitigate the increasing attacks and threats. In addition, most existing dynamic access control methods match static rules with dynamic metrics, which cause system damage through their delayed responses to threats and attacks. The zero-trust architecture (ZTA) provides continuous authentication and dynamic authorization for all users to accommodate the security demands of cloud computing. Drawing inspiration from the ZTA, we first present a TBAC (Trust-based Access Control) model and design a trust assessment methodology to update user trustworthiness. Then, we introduce dynamic rules in the TBAC model to implement a dynamic access control system DR-TBAC (TBAC with Dynamic Rule). We apply the DQN (Deep Q-Network) algorithm to dynamically update the trust thresholds based on static rules comparing dynamic trust with predefined trust thresholds to achieve adaptive access control policies. In this paper, we rebuild the cloud security access environment from the perspective of dynamic trust and rule optimization and strengthen the constraints on user behaviors throughout the access control lifecycle of cloud computing. Finally, a thorough analysis and assessment regarding offline training models and the online deployment of the DR-TBAC system into the cloud platform highlight its security and accuracy relative to baseline models.

How to improve drivers’ hazard perception ability? an interactive training method based on VR technology and the Unity 3D Platform

ABSTRACT Effective hazard perception training is crucial for improving road safety, yet the factors influencing its efficacy remain underexplored. This study assessed the efficacy of an interactive hazard perception training program utilizing Virtual Reality (VR) technology and the Unity 3D Platform. A driving simulator experiment was conducted with 34 licensed drivers, who were exposed to 20 high-risk urban driving scenarios. Data were collected over three test sessions: baseline, immediately post-training, and 7 days post-training. The study focused on two types of conflict scenarios—horizontal and vertical. Our analysis showed that drivers’ HPTs followed a Weibull distribution, which allowed us to develop an accelerated failure time (AFT) model. The results indicated that scene conflict type and driver age positively affected HPT, while test session and initial driving speed showed an inverse relationship. Importantly, the training significantly improved drivers’ ability to identify hazards, with notable improvements in both horizontal and vertical conflict scenarios observed immediately after training and 7 days later. This study demonstrates that VR-based hazard perception training effectively enhances drivers’ hazard detection skills. These findings contribute to the development of more standardized and effective models for hazard perception training, offering potential for wider application in driver education and road safety programs.

Federated learning-driven IoT system for automated freshness monitoring in resource-constrained vending carts

Street vendors in developing regions often lack access to portable and affordable cold storage, leading to accelerated food spoilage, financial losses, and health risks. Traditional refrigeration solutions are bulky and costly, while manual freshness assessment is error-prone. This study proposes a smart vending cart integrating IoT sensors and federated learning (FL) to address these challenges, offering real-time environmental monitoring, freshness classification, and privacy-preserving data handling. The smart vending cart incorporates IoT sensors to monitor temperature, humidity, and gas emissions. A Peltier cooling module and a humidifier maintain optimal conditions. Machine learning models classify food freshness, while federated learning ensures vendor privacy by training models locally on each cart. The study explores nine federated learning approaches to train machine learning models across multiple carts without sharing raw data, thus preserving vendor privacy. The Stacking Ensemble approach outperformed all other methods, achieving the highest accuracy, F1-Score, and Cohen’s Kappa (0.99964), with the lowest log loss (0.0022). MetaLearning and Weighted Aggregation also demonstrated high performance but with marginally higher log loss values. Personalized models performed well in heterogeneous data environments but were less effective than ensemble methods. The developed smart vending cart system effectively reduces food spoilage and enhances vendor profitability through automated freshness classification and real-time environmental control. The integration of federated learning ensures privacy, while ensemble techniques improve robustness in resource-constrained settings, offering a scalable solution for street vendors.

EFEITOS DO TREINAMENTO COMBINADO NA COMPOSIÇÃO CORPORAL, FORÇA MUSCULAR E POTÊNCIA AERÓBIA DE MULHERES: UMA REVISÃO SISTEMÁTICA E META-ANÁLISE

This systematic review and meta-analysis verified the effects of different Combined Training (CT) programs on body composition, aerobic power and muscle strength performance in healthy women. A bibliographic search was performed in the following electronic databases: Medline, PubMed, Lilacs, Plataforma Capes and EMBASE from the oldest registry until November 2023. 14 eligible articles were found that met the following criteria: original randomized articles and non-randomized comparatives; experimental CT studies with at least four weeks of intervention and with healthy adult women (&gt;18 years); studies with the Control Group (CG) that restricted the practice of physical exercises; that analyzed at least one of the outcomes related to muscle strength, body composition (fat mass and lean mass) and aerobic power (VO2). We compared CT versus GC applying a random effects meta-analysis model. We estimated the standardized mean difference (SMD) for each group (CT and GC) comparing pre-post results. According to the results of the meta-analysis, all the variables analyzed were favorable to the CT compared to the CG, but Body Mass (SMD=0.05;CI: -0,02-0,13) and Aerobic Power (SMD=1.34; CI: -0,04-2,72) were not significant. Fat Mass (SMD=0.22; CI: 0.09-0.35), muscle strength (SMD= 0.83; CI: 0.59-1.41), and muscle mass (SMD=0.34; CI: 0.22-0.47) showed significant changes. These results suggest that the training models promoted increases in aerobic power and muscular strength, and significantly improved body composition compared to control group, regardless of the protocol used.

A diagnostic model for sepsis using an integrated machine learning framework approach and its therapeutic drug discovery

BackgroundSepsis remains a life-threatening condition in intensive care units (ICU) with high morbidity and mortality rates. Some biomarkers commonly used in clinic do not have the characteristics of rapid and specific growth and rapid decline after effective treatment. Machine learning has shown great potential in early diagnosis, subtype analysis, accurate treatment and prognosis evaluation of sepsis.MethodsGene expression matrices from GSE13904 and GSE26440 were combined into a training model after quality control and standardization. Then, the intersection genes were obtained by crossing the screened differentially expressed genes (DEGs) and the module genes with the strongest correlation obtained by WGCNA analysis. 113 combined machine learning algorithms to build a diagnosis model. Then the CIBERSORT algorithm is used to analyze the relationship between the change of core gene expression and immune response in sepsis. Construct nomogram, DCA and CIC to further verify the reliability of the diagnosis model. The potential molecular compounds interacting with key genes were searched from the Traditional Chinese Medicine Active Compound Library (TCMACL).ResultsWe screened 405 DEGs, including 334 up-regulated and 71 down-regulated genes. The 308 potential genes were obtained by intersection of MEturquoise module genes in WGCNA analysis and DEGs for subsequent machine learning analysis. GO and KEGG enrichment analysis showed that sepsis was mainly related to immune response and bacterial infection. Then 113 combined machine learning algorithms are applied to construct a diagnosis model to screen 22 hub genes. Four four key genes (CD177, GNLY, ANKRD22, and IFIT1) are obtained through further analysis of PPI network constructed by 22 hub genes. Subsequently, the diagnostic model is proved to have good predictive value by nomogram, DCA and CIC. Finally, molecular compounds (Dieckol, Grosvenorine and Tellimagrandin II) were screened out as potential drugs.Conclusion113 combinated machine learning algorithms screened out four key genes that can distinguish sepsis patients. At the same time, potential therapeutic molecular compounds interacting with key genes genes were screened out by molecular docking.

Using Deep Learning to Simultaneously Reduce Noise and Motion Artifacts in Brain MR Imaging.

To reduce motion artifacts (MA) and noise in brain MRI using deep learning to promote clinical utility. T1-weighted (T1W), T2-weighted (T2W), and fluid attenuated inversion recovery (FLAIR) images of the brain (including sagittal, coronal, and axial sections) of 20 healthy volunteers were collected using a 3.0T MR system. Simulated images with horizontal and vertical phase directions exhibiting varying white noise and MA (n = 115200) were created for each sequence and trained in deep learning (36000 pairs), validation (200 pairs), and testing (200 pairs, 2000 pairs) datasets. Images with MA and noise and images without MA and noise were included. A training model was constructed to remove noise and MA. The model's ability to remove noise and MA was evaluated by the structural similarity index (SSIM) and peek signal to noise ratio (PSNR). The SSIM and PSNR between the ground-truth and output images were calculated (SSIMout, PSNRout), and the SSIM and PSNR between the ground-truth and input images were calculated (SSIMinp, PSNRinp). The ratio of SSIMinp to SSIMout was then evaluated as the improvement ratio of SSIM (IMPRs) and the ratio of PSNRinp to PSNRout as the improvement ratio of PSNR (IMPRp). In addition, 10 radio technologists performed visual evaluation. The SSIMout were >0.95 and 33 dB, respectively, for T1W, T2W, and FLAIR images with different contrasts. The mean value of SSIMinp was 0.72. Noise and MA removal effects were observed in images, with an average value of 72 dB. Visual evaluation revealed that the reduction effect in the output image was higher than that observed in the input image. The method proposed herein, which uses separate training models for the T1W, T2W, and FLAIR sequences, is a valuable approach for removing MA and noise, independent of the imaging direction and artifact orientation. An improvement in image quality was achieved by generating a large amount of training data through simulation.

Comparison and Enlightenment of the Implementation Strategies of Basic Education Informatization in Mongolia and China in the Digital Era

In order to cope with the opportunities and challenges brought by the digital era to basic education and promote the improvement of education equity and quality, this paper uses literature analysis and comparative research to systematically analyze and compare the implementation strategies of basic education informatization in Mongolia and China. The study found that Mongolia has deficiencies in policy support and infrastructure construction, but pays attention to localized resource development and teacher training. China, on the other hand, has performed well in improving its policy system, widely applying technology, and building a resource sharing platform, but it faces the challenges of unbalanced regional development and differences in teachers' informatization capabilities. Through comparison, this paper reveals the similarities and differences between the two countries in the process of informatization of basic education, and proposes that Mongolia can learn from China's experience in policy coordination and resource integration, while China can draw inspiration from Mongolia's localization practice and teacher training model. The results of this study provide a theoretical basis and practical reference for further optimizing the informatization strategy of basic education in the two countries, and are of great significance for promoting regional education modernization.

Comparison of DeepLearning Models for Voice DisorderClassification Using Kymographic Images.

Comparison of DeepLearning Models for Voice DisorderClassification Using Kymographic Images.

Contrast quality control for segmentation task based on deep learning models-Application to stroke lesion in CT imaging.

Although medical imaging plays a crucial role in stroke management, machine learning (ML) has been increasingly used in this field, particularly in lesion segmentation. Despite advances in acquisition technologies and segmentation architectures, one of the main challenges of subacute stroke lesion segmentation in computed tomography (CT) imaging is image contrast. To address this issue, we propose a method to assess the contrast quality of an image dataset with a ML trained model for segmentation. This method identifies the critical contrast level below which the medical-imaging model fails to learn meaningful content from images. Contrast measurement relies on the Fisher's ratio, estimating how well the stroke lesion is contrasted from the background. The critical contrast is found-thanks to the following three methods: Performance, graphical, and clustering analysis. Defining this threshold improves dataset design and accelerates training by excluding low-contrast images. Application of this method to brain lesion segmentation in CT imaging highlights a Fisher's ratio threshold value of 0.05, and training validation of a new model without these images confirms this with similar results with only 60% of the training data, resulting in an almost 30% reduction in initial training time. Moreover, the model trained without the low-contrast images performed equally well with all images when tested on another database. This study opens discussion with clinicians concerning the limitations, areas for improvement, and strategies for enhancing datasets and training models. While the methodology was only applied to stroke lesion segmentation in CT images, it has the potential to be adapted to other tasks.

Robust pose estimation for non-cooperative space objects based on multichannel matching method

Accurate space object pose estimation is crucial for various space tasks, including 3D reconstruction, satellite navigation, rendezvous and docking maneuvers, and collision avoidance. Many previous studies, however, often presuppose the availability of the space object’s computer-aided design model for keypoint matching and model training. This work proposes a generalized pose estimation pipeline that is independent of 3D models and applicable to both instance- and category-level scenarios. The proposed framework consists of three parts based on deep learning approaches to accurately estimate space objects pose. First, a keypoints extractor is proposed to extract sub-pixel-level keypoints from input images. Then a multichannel matching network with triple loss is designed to obtain the matching pairs of keypoints in the body reference system. Finally, a pose graph optimization algorithm with a dynamic keyframes pool is designed to estimate the target pose and reduce long-term drifting pose errors. A space object dataset including nine different types of non-cooperative targets with 11,565 samples is developed for model training and evaluation. Extensive experimental results indicate that the proposed method demonstrates robust performance across various challenging conditions, including different object types, diverse illumination scenarios, varying rotation rates, and different image resolutions. To verify the demonstrated approach, the model is compared with several state-of-the-art approaches and shows superior estimation results. The mAPE and mMS scores of the proposed approach reach 0.63° and 0.767, respectively.