Collection Of Datasets Research Articles

Anastomotic leak after manual circular stapled left-sided bowel surgery: analysis of technology-, disease-, and patient-related factors.

Anastomotic leak rates after colorectal surgery remain high. In most left-sided colon and rectal resection surgeries, a circular stapler is utilized to create the primary bowel anastomosis. However, it remains unclear whether a relationship between circular stapler technology and anastomotic leak in left-sided colorectal surgery exists. A post-hoc analysis was conducted using a prospectively collected data set of patients from the 2017 European Society of Coloproctology snapshot audit who underwent elective left-sided resection (left hemicolectomy, sigmoid colectomy, or rectal resection) with a manual circular stapled anastomosis. Rates of anastomotic leak and unplanned intensive care unit stay in association with manual circular stapling were assessed. Patient-, disease-, geographical-, and surgeon-related factors as well as stapler brand were explored using multivariable regression models to identify predictors of adverse outcomes. Across 3305 procedures, 8.0% of patients had an anastomotic leak and 2.1% had an unplanned intensive care unit stay. Independent predictors of anastomotic leak were male sex, minimal-access surgery converted to open surgery, and anastomosis height C11 (lower third rectum) (all P < 0.050). Independent predictors of unplanned intensive care unit stay were minimal-access surgery converted to open surgery and American Society of Anesthesiologists grade IV (all P < 0.050). Stapler device brand was not a predictor of anastomotic leak or unplanned intensive care unit stay in multivariable regression analysis. There were no differences in rates of anastomotic leak and unplanned intensive care unit stay according to stapler head diameter, geographical region, or surgeon experience. In patients undergoing left-sided bowel anastomosis, choice of manual circular stapler, in terms of manufacturer or head diameter, is not associated with rates of anastomotic leak and unplanned intensive care unit stay.

Metafeature Selection via Multivariate Sparse-Group Lasso Learning for Automatic Hyperparameter Configuration Recommendation.

The performance of classification algorithms is mainly governed by the hyperparameter settings deployed in applications, and the search for desirable hyperparameter configurations usually is quite challenging due to the complexity of datasets. Metafeatures are a group of measures that characterize the underlying dataset from various aspects, and the corresponding recommendation algorithm fully relies on the appropriate selection of metafeatures. Metalearning (MtL), aiming to improve the learning algorithm itself, requires development in integrating features, models, and algorithm learning to accomplish its goal. In this article, we develop a multivariate sparse-group Lasso (SGLasso) model embedded with MtL capacity in recommending suitable configurations via learning. The main idea is to select the principal metafeatures by removing those redundant or irregular ones, promoting both efficiency and performance in the hyperparameter configuration recommendation. To be specific, we first extract the metafeatures and classification performance of a set of configurations from the collection of historical datasets, and then, a metaregression task is established through SGLasso to capture the main characteristics of the underlying relationship between metafeatures and historical performance. For a new dataset, the classification performance of configurations can be estimated through the selected metafeatures so that the configuration with the highest predictive performance in terms of the new dataset can be generated. Furthermore, a general MtL architecture combined with our model is developed. Extensive experiments are conducted on 136 UCI datasets, demonstrating the effectiveness of the proposed approach. The empirical results on the well-known SVM show that our model can effectively recommend suitable configurations and outperform the existing MtL-based methods and the well-known search-based algorithms, such as random search, Bayesian optimization, and Hyperband.

Trapezoidal neutrosophic teaching learning-based optimization in enhancing accuracy of diabetes prognosis

Diabetes is one of chronic diseases in which blood glucose (sugar) level soar up high where human body are incapable to absorb it properly. It is important to have an appropriate diagnosis for proper management and treatment. The aim of this manuscript is to provide a more accurate diabetes prediction model through the new adaptive Trapezoidal Neutrosophic Teaching Learning-Based Optimization (TLBO) method. In order to address the inherent uncertainties and imprecisions in medical data, the suggested model makes use of the resilience of Trapezoidal Neutrosophic sets. The Trapezoidal Neutrosophic set theory provides a suitable basis for developing rule/knowledge-based systems in the medical field. The present investigation makes use of the dataset acquired from the Pima Indians Diabetes Database (PIDD) website, which has an extensive global collection of diabetes datasets. The performance of our model is evaluated against several existing methodologies, including Intuitionistic Neuro-Fuzzy System (INFS) Structure, Fuzzy Logic based Diabetes Diagnosis System (FLDDS), Fuzzy Verdict Mechanism (FVM) for Diabetes Decision, (Fuzzy Expert System) FES, and Hierarchical Neuro-Fuzzy Binary space partitioning System (HNFB-1). Quantitative analysis validates that proposed methodology achieves an exceptional predictive accuracy of 99.89 %, which is substantially higher than the comparative methodologies, namely INFS Structure (88.76 %), FLDDS (87.2 %), FVM for Diabetes Decision (85.03 %), FES (81.7 %), and HNFB-1 (78.26 %). These enhancements demonstrate show how well the suggested model works to lower diagnostic errors and increase dependability.

Quarantine Order Violators System using Face Recognition (FACID)

Face recognition technology is commonly utilized for user authentication and verification by analysing a digital image of a person's face and matching it against a database of faces for identification purposes. The COVID-19 pandemic has led to mandatory quarantines and the use of quarantine bracelets for some individuals, which can be time-consuming and require a lot of effort. Face recognition technology can help raise awareness about the current situation and alleviate some of the burden associated with quarantine measures. The aim of this research is to concentrate on the face recognition of individuals with a history of quarantine, as a measure to prevent the spread of COVID-19 in educational institutions like universities, colleges, and schools. This research study concentrates on individuals with a history of quarantine orders. The system will employ the Histograms of Oriented Gradients (HOG) algorithm for face detection. Additionally, the system will utilize the Face Landmark Algorithm to compare the 128-d vector with images stored locally. The system will make use of the Helen face collection dataset for its data requirements. The aim of this study is to explore and identify techniques for detecting individuals who violate quarantine measures and issuing notifications through the proposed system. By implementing this system, it could contribute to creating a more secure environment within the educational institution and potentially reduce the spread of the virus.

Forecasting international tourist arrivals in South Korea: a deep learning approach

Purpose This study aims to develop a robust long short-term memory (LSTM)-based forecasting model for daily international tourist arrivals at Incheon International Airport (ICN), incorporating multiple predictors including exchange rates, West Texas Intermediate (WTI) oil prices, Korea composite stock price index data and new COVID-19 cases. By leveraging deep learning techniques and diverse data sets, the research seeks to enhance the accuracy and reliability of tourism demand predictions, contributing significantly to both theoretical implications and practical applications in the field of hospitality and tourism. Design/methodology/approach This study introduces an innovative approach to forecasting international tourist arrivals by leveraging LSTM networks. This advanced methodology addresses complex managerial issues in tourism management by providing more accurate forecasts. The methodology comprises four key steps: collecting data sets; preprocessing the data; training the LSTM network; and forecasting future international tourist arrivals. The rest of this study is structured as follows: the subsequent sections detail the proposed LSTM model, present the empirical results and discuss the findings, conclusions and the theoretical and practical implications of the study in the field of hospitality and tourism. Findings This research pioneers the simultaneous use of big data encompassing five factors – international tourist arrivals, exchange rates, WTI oil prices, KOSPI data and new COVID-19 cases – for daily forecasting. The study reveals that integrating exchange rates, oil prices, stock market data and COVID-19 cases significantly enhances LSTM network forecasting precision. It addresses the narrow scope of existing research on predicting international tourist arrivals at ICN with these factors. Moreover, the study demonstrates LSTM networks’ capability to effectively handle multivariable time series prediction problems, providing a robust basis for their application in hospitality and tourism management. Originality/value This research pioneers the integration of international tourist arrivals, exchange rates, WTI oil prices, KOSPI data and new COVID-19 cases for forecasting daily international tourist arrivals. It bridges the gap in existing literature by proposing a comprehensive approach that considers multiple predictors simultaneously. Furthermore, it demonstrates the effectiveness of LSTM networks in handling multivariable time series forecasting problems, offering practical insights for enhancing tourism demand predictions. By addressing these critical factors and leveraging advanced deep learning techniques, this study contributes significantly to the advancement of forecasting methodologies in the tourism industry, aiding decision-makers in effective planning and resource allocation.

Application of artificial intelligence sensor and visual image technology in the analysis of hydrophilic space landscape characteristics

With the continuous development of intelligent sensor technology and artificial intelligence, the application of artificial intelligence feature recognition at the level of intelligent sensor information acquisition to urban landscape analysis has become a current research trend and hot spot, but also brings us new research directions and challenges. In this paper, the multi-sensor collaborative filtering algorithm is optimized, and a method based on evidence theory is proposed to deal with the fuzzy and unclear information generated in the process of multi-sensor collaboration. Then, the analysis and recognition of multi-sensor fusion information using AI technology are analyzed in detail. This paper discusses the problems in multi-sensor information fusion and related solutions. Combined with the key nodes of 3D object AI feature recognition, multi-sensor collaborative Dempster Shafer evidence theory and 3D convolutional neural network waterfront space landscape feature recognition sub-model are constructed, and the waterfront space landscape recognition analysis model is tested and analyzed. The results show that the multi-sensor information collection Kalman filtering fusion algorithm effectively realizes the recognition of landscape feature information and the filtering of irrelevant interference information combined with an artificial intelligence algorithm to build a landscape feature recognition model. Among all kinds of landscape recognition, the recognition effect of beach inflow is the best, and the recognition accuracy, recall, and F1 value are all above 95. However, the identification effect of hydrophilic plank roads and underwater breakwater is relatively poor. The recognition accuracy of the hydrophilic platform is the lowest at 83.16 % among the eight landscape types, and the recall rate of underwater breakwater is the lowest at 79.96 % among the eight landscape types. In general, the multi-sensor Kalman filtering algorithm information fusion model designed in this paper has higher recognition accuracy and better classification performance for the collection and classification of the entire landscape dataset. The work of this paper provides a new idea and direction for the design and application of waterfront landscape identification and analysis systems based on intelligent sensors and artificial intelligence technology and lays a certain foundation for the intelligent design of urban waterfront landscapes.

Defect Rate Prediction in Manufacturing Process Using K-Nearest Neighbor Algorithm

The efficiency and effectiveness in the manufacturing industry are significantly impacted by artificial intelligence technology. An important application involves the improvement of product quality, which is measurable through the defects occurring during the production process. This research is aimed at predicting defects in the manufacturing process using the K-Nearest Neighbor (KNN) algorithm with various distance measurement methods, namely Euclidean, Minkowski, and Manhattan distances. The research methodology is composed of four stages: dataset collection, data preprocessing, modeling, and evaluation. The focus of this research is on the optimal K value and the conditions that yield the highest accuracy, considering various scenarios of training and test data splitting ratios and different random state values. The test results indicate that the Minkowski distance method, with a data division ratio of 80% for training data, 20% for test data, and a random state value of 32, provides the best performance, with an optimal K value of 10 and an accuracy of 86.41%.

Checking the inventory: Illustrating different methods for individual participant data meta-analytic structural equation modeling.

Researchers may have at their disposal the raw data of the studies they wish to meta-analyze. The goal of this study is to identify, illustrate, and compare a range of possible analysis options for researchers to whom raw data are available, wanting to fit a structural equation model (SEM) to these data. This study illustrates techniques that directly analyze the raw data, such as multilevel and multigroup SEM, and techniques based on summary statistics, such as correlation-based meta-analytical structural equation modeling (MASEM), discussing differences in procedures, capabilities, and outcomes. This is done by analyzing a previously published collection of datasets using open source software. A path model reflecting the theory of planned behavior is fitted to these datasets using different techniques involving SEM. Apart from differences in handling of missing data, the ability to include study-level moderators, and conceptualization of heterogeneity, results show differences in parameter estimates and standard errors across methods. Further research is needed to properly formulate guidelines for applied researchers looking to conduct individual participant data MASEM.

Comparing glaucoma risk in children receiving low-dose and high-dose glucocorticoid treatment after cataract surgery.

Treatment with glucocorticoids following paediatric cataract surgery is crucial to prevent inflammation, but may lead to secondary glaucoma, and hypothalamic-pituitary-adrenal axis suppression. We wish to compare glaucoma outcomes following high-dose and low-dose glucocorticoid treatment after paediatric cataract surgery. This cohort study included Danish children undergoing cataract surgery before 10 years of age, receiving either a low-dose or high-dose postoperative glucocorticoid treatment. Case identification and collection of a standardized dataset were retrospective, from 1 January 2010 to 31 December 2016, and prospective thereafter, until 31 December 2021. High-dose treatment included 0.5-1.0 mg subconjunctival depot dexamethasone or methylprednisolone, followed by 6-8 drops of dexamethasone for 1 week, tapered by one drop weekly. Low-dose treatment included 6 drops for 3 days, followed by 3 drops for 18 days. Sustained (>3 months) ocular hypertension or glaucoma was compared between the two groups. Overall, 267 children (388 eyes) were included in the study. Ninety-five children (133 eyes) had received high-dose treatment and had a median follow-up time of 89 months (IQR: 57.2-107.4), while 173 children (255 eyes) had received the low-dose treatment and had a median follow-up time of 40.5 months (IQR: 22.9-60.4). Survival curves showed a lower risk of glaucoma in the low-dose group for children with axial lengths ≥18 mm. Low-dose glucocorticoid treatment was associated with a lower risk of glaucoma in children with axial lengths ≥18 mm. The same effect was not observed in children with shorter eyes. High-dose glucocorticoid should be limited in children undergoing cataract surgery.

Performance Evaluation of Structural Health Monitoring System Applied to Full-Size Composite Wing Spar via Probability of Detection Techniques.

Probability of detection (POD) is an acknowledged mean of evaluation for many investigations aiming at detecting some specific property of a subject of interest. For instance, it has had many applications for Non-Destructive Evaluation (NDE), aimed at identifying defects within structural architectures, and can easily be used for structural health monitoring (SHM) systems, meant as a compact and more integrated evolution of the former technology. In this paper, a probability of detection analysis is performed to estimate the reliability of an SHM system, applied to a wing box composite spar for bonding line quality assessment. Such a system is based on distributed fiber optics deployed on the reference component at specific locations for detecting strains; the attained data are then processed by a proprietary algorithm whose capability was already tested and reported in previous works, even at full-scale level. A finite element (FE) model, previously validated by experimental results, is used to simulate the presence of damage areas, whose effect is to modify strain transfer between adjacent parts. Numerical data are used to verify the capability of the SHM system in revealing the presence of the modeled physical discontinuities with respect to a specific set of loads, running along the beam up to cover its complete extension. The POD is then estimated through the analysis of the collected data sets, wide enough to assess the global SHM system performance. The results of this study eventually aim at improving the current strategies adopted for SHM for bonding analysis by identifying the intimate behavior of the system assessed at the date. The activities herein reported have been carried out within the RESUME project.

(Invited) Quantum Well Nanotube Emission for Cancer Research and Diagnostics

Quantum well defects (QWDs) on single-walled carbon nanotubes (SWCNTs) enable expanded analyte sensitization and functionalization for biosensing applications. We are using these capabilities to develop new research and diagnostic methods for cancers, infectious disease, and other indications. For instance, we are addressing issues with serum biomarker measurements, which are widely used for diagnostics, but these markers often fail to result in a clear diagnosis of complex disease indications. We developed a method using QWD-modified SWCNTs (QWNTs) to identify a “spectral fingerprint” through the collection of large data sets of molecular binding interactions to an array of QWNTs. We found that a library of QWNTs exhibit differentiated spectral variations in response to an ensemble of molecular binding events in patient serum. Via machine learning algorithms, we built prediction models of nanosensor responses that can reliably identify ovarian cancer, and we have expanded this approach to other indications without known biomarkers, providing a general method to identify clinical indications.

Examining the link between coach-athlete relationship and athlete burnout among college soccer players: the mediating role of training satisfaction.

Athlete burnout is an adverse factor that usually decreases athletes' sports performance and reduces their passion for athletic career development. The present study examined the association of coach-athlete relationship, training satisfaction, and athlete burnout, and then explored the training satisfaction as a mediating role of the effect of coach-athlete relationship on athlete burnout among college soccer players. Soccer players from seven Chinese higher education institutions were recruited for the study. The participants were selected using stratified random sampling and included 218 participants. The Coach-Athlete Relationship Scale, Training Satisfaction Scale, and Athlete Burnout Questionnaire were used to collect data from the participants. Pearson's product-moment correlation analysis, linear regression analysis, and bootstrap method were conducted to analyze the collected data set. The results showed that (1) there was a significant correlation between coach-athlete relationship, training satisfaction, and athlete burnout. (2) coach-athlete relationship and training satisfaction significantly negatively predicted athlete burnout (β = -0.39, p < 0.001; β = - 0.29, p < 0.001). (3) training satisfaction had a significantly mediating effect on the relationship between coach-athlete relationship and athlete burnout (β = -0.15, p < 0.001, ES = 0.28). Coach-athlete relationship can not only directly negative impact athlete burnout but can also negatively influence athlete burnout through the mediating effect of training satisfaction. Therefore, we suggest the following to alleviate athlete burnout: on one hand, coaches can provide more communication opportunities for athletes to improve the coach-athlete relationship. On the other hand, teams can enhance training satisfaction by building team culture.

Prediction of product properties and identification of key influencing parameters in microwave pyrolysis of microalgae using machine learning

Exploring the effects of operating parameters and microalgae composition on product distribution and properties using machine learning is an effective and economical approach. However, there is currently a lack of research utilizing machine learning to comprehensively describe the microwave pyrolysis of microalgae. In this study, machine learning was employed to accurately model microalgae microwave pyrolysis based on microalgae composition, operating parameters, and the yields and properties of three-phase products. Using the collected data set, machine learning models, including Support Vector Regressor, Random Forest Regressor, Gradient Boosting Regressor, and eXtreme Gradient Boosting, were employed to model microalgae microwave pyrolysis. Results showed that the Gradient Boosting Regressor model outperformed the other three models, with average coefficients of determination (R2) of 0.998 and 0.762 in the training and testing phases, respectively. SHapley Additive exPlanations analysis revealed that reaction temperature, reaction time, and microwave power significantly influenced the predictions of the yields and properties of three-phase products. The findings of this study can effectively save costs and time in the future study on microwave pyrolysis of microalgae, and serve as an important tool for guiding subsequent experiments and process optimization.

Multimodal Retinal Imaging Classification for Parkinson's Disease Using a Convolutional Neural Network.

Changes in retinal structure and microvasculature are connected to parallel changes in the brain. Two recent studies described machine learning algorithms trained on retinal images and quantitative data that identified Alzheimer's dementia and mild cognitive impairment with high accuracy. Prior studies also demonstrated retinal differences in individuals with PD. Herein, we developed a convolutional neural network (CNN) to classify multimodal retinal imaging from either a Parkinson's disease (PD) or control group. We trained a CNN to receive retinal image inputs of optical coherence tomography (OCT) ganglion cell-inner plexiform layer (GC-IPL) thickness color maps, OCT angiography 6 × 6-mm en face macular images of the superficial capillary plexus, and ultra-widefield (UWF) fundus color and autofluorescence photographs to classify the retinal imaging as PD or control. The model consists of a shared pretrained VGG19 feature extractor and image-specific feature transformations which converge to a single output. Model results were assessed using receiver operating characteristic (ROC) curves and bootstrapped 95% confidence intervals for area under the ROC curve (AUC) values. In total, 371 eyes of 249 control subjects and 75 eyes of 52 PD subjects were used for training, validation, and testing. Our best CNN variant achieved an AUC of 0.918. UWF color photographs were the most effective imaging input, and GC-IPL thickness maps were the least contributory. Using retinal images, our pilot CNN was able to identify individuals with PD and serves as a proof of concept to spur the collection of larger imaging datasets needed for clinical-grade algorithms. Developing machine learning models for automated detection of Parkinson's disease from retinal imaging could lead to earlier and more widespread diagnoses.

Combining reference-star and angular differential imaging for high-contrast imaging of extended sources

Context. High-contrast imaging (HCI) is a technique designed to observe faint signals near bright sources, such as exoplanets and circumstellar disks. The primary challenge in revealing the faint circumstellar signal near a star is the presence of quasi-static speckles, which can produce patterns on the science images that are as bright, or even brighter, than the signal of interest. Strategies such as angular differential imaging (ADI) or reference-star differential imaging (RDI) aim to provide a means of removing the quasi-static speckles in post-processing. Aims. In this paper, we present and discuss the adaptation of state-of-the-art algorithms, initially designed for ADI, to jointly leverage angular and reference-star differential imaging (ARDI) for the direct HCI of circumstellar disks. Methods. Using a collection of HCI datasets, we assessed the performance of ARDI in comparison to ADI and RDI based on iterative principal component analysis (IPCA). These diverse datasets were acquired under various observing conditions and include the injection of synthetic disk models at various contrast levels. We also considered reference stars with different levels of correlation with the science targets. Results. Our results demonstrate that ARDI with IPCA improves the quality of recovered disk images and the sensitivity to planets embedded in disks, compared to ADI or RDI individually. This enhancement is particularly pronounced when dealing with extended sources exhibiting highly ambiguous structures that cannot be accurately retrieved using ADI alone, and when the quality of the reference frames is suboptimal, leading to an underperformance of RDI. We finally applied our method to a sample of real observations of protoplanetary disks taken in star-hopping mode, and propose to revisit the protoplanetary claims associated with these disks. Among eight proposed protoplanets claimed through velocity kinks or direct imaging, none of them were re-detected in our new processed images.

Klasifikasi Kualitas Buah Apel Dengan Algoritma K-Nearest Neighbor (K-NN) Menggunakan Bahasa Pemrograman Python

Fruit is an important intake for the human body, apples are included in the fruit favored by the people of Indonesia. For this reason, it is necessary to provide apples of good quality, so that they can benefit the body. By using the k-NN method that is considered able to train data quickly and effectively for training data and testing data in large quantities. This study began from the collection of datasets obtained from https://www.kaggle.com/, then perform a preprocessing process followed by separating the training data and testing data with a composition of 25% testing data and 75% training data. Then the k-NN method is applied to this study to be classified based on several existing criteria, so as to obtain the results of performance evaluation K-NN with the value of accuracy that has been calculated with python programming. In implementing datamining using Python programming language by utilizing the library that has been provided as a process to facilitate the implementation of machine learning. From The Matrix confution test, there are 441 data predicted with true data, and 440 data predicted incorrectly. As for the 54 and 65 data predicted to be less precise than 1000 testing data. So that the accuracy value obtained by the k-NN method is equal to 0.88 or 88%. It is seen that the k-NN method can work well, quickly and efficiently in training large amounts of data.

Meaning, concept and application of bioinformatics

Bioinformatics, originating from the term coined in 1979 by Paulien Hogeweg and Ben Hesper, is a burgeoning interdisciplinary field at the nexus of biology, computer science, and information technology. It encompasses the collection, classification, storage, and analysis of vast biological datasets using computational methods, aiming to unravel the complexities of biological systems. This abstract provides a comprehensive overview of bioinformatics, emphasizing its definition, scope, branches, aims, research areas, databases, and applications, particularly highlighting its role in drug discovery through structural bioinformatics. Bioinformatics integrates diverse branches such as computational biology, genomics, proteomics, structural biology, systems biology, pharmacogenomics, and bioprogramming. Each branch applies mathematical modeling, molecular interactions analysis, or genomic data interpretation to advance understanding in biology and medicine. Key aims include applying advanced computational technologies to biological problems, presenting complex data clearly, and providing robust statistical tools for genomic analysis. The field also fosters collaborations across academic, commercial, and government sectors to leverage resources and expertise. Research areas span computational evolutionary analysis, genome annotation, gene expression analysis, cancer mutation studies, and comparative genomics, illuminating disease mechanisms and evolutionary relationships. Bioinformatics databases play a crucial role by storing and organizing biological data. Primary repositories like GenBank and the Protein Data Bank (PDB), secondary databases such as SWISS-Prot, and specialized databases support research and clinical applications by providing access to genomic, proteomic, and structural information. In drug discovery, structural bioinformatics techniques like homology modeling, molecular docking, and simulations predict protein structures, analyze interactions, and aid in designing therapies. These computational methods accelerate drug development, optimize drug-target interactions, and reduce costs associated with bringing new drugs to market. Overall, bioinformatics drives innovation in biological sciences and healthcare by leveraging computational methods to interpret biological data. As technology advances, bioinformatics will continue to play a pivotal role in addressing global health challenges, advancing personalized medicine, and facilitating groundbreaking discoveries in biology and medicine. The evolving field of bioinformatics promises continued growth and impact, shaping the future of biological research and its applications in improving human health and understanding the natural world.

Radiological segmentation of knee meniscus ultrasound images based on boundary constraints and multi-scale fusion network

Knee osteoarthritis (KOA) is one of the most popular joint diseases endangering human health because of its high incidence, disability, and younger onset. However, there is no suitable method for early diagnosis, evaluation, and treatment of KOA. In recent years, some clinical studies have found that ultrasound can detect early changes in KOA in advance, and the automatic segmentation of ultrasound images can achieve rapid and effective quantitative research on KOA. However, in ultrasound images, the soft tissue boundaries of the lesion are blurred, making lesion segmentation difficult. Although the U-Net family is one of the best networks for image segmentation, they still have defects such as blurred segmentation boundaries, distorted morphology, and insufficient accuracy when segmenting ultrasound images. To address this issue, we added attention, atrous spatial pyramid pooling (ASSP), and edge loss function terms into the Unet3+ network, which improved the contour clarity and accuracy of output images (the improved Unet3+ Dice acc = 78.74%. Then, we extract the key features of the improved Unet3+ for outputting meniscus images: calculating meniscus area and distance, where the average accuracy of the area is: area_ avg_ acc = 91.12%, with an average distance accuracy of distance_ avg_ acc = 91.14%. This thesis creates a new dataset collection from West China Hospital, Sichuan University, and automated measurement of knee meniscus protrusion area has been achieved for the first time. This article is the first to apply deep learning to ultrasound image segmentation of the knee meniscus, helping doctors conduct qualitative and quantitative analysis of the diagnosis and treatment of early KOA. The results indicate that the improved Unet3+ can assist doctors in automatically diagnosing and evaluating KOA based on meniscus ultrasound images, which is beneficial for guiding early clinical intervention.

Addressing data imbalance in neural network spam detection with insights from SMS spam collection

In cybersecurity, the persistent challenge of spam detection remains paramount. Traditional methods reliant on human scrutiny or rule-based algorithms are proving inadequate against the constantly evolving tactics employed by spammers. Machine learning emerges as a promising solution, leveraging vast datasets to swiftly and objectively discern patterns and traits within spam messages. By uncovering subtle correlations among message elements, machine learning enhances the precision and efficacy of spam detection systems, offering a dependable and economical approach to combat spam. This study aims to investigate the impact of different strategies for addressing data imbalance on neural network-based spam detection performance. Using the SMS Spam Collection Dataset, four methods for mitigating data imbalance are evaluated against an untreated scenario. Notably, despite inherent data imbalance, the unprocessed scenario exhibits the highest overall performance. Stratified sampling emerges as the most effective technique for accurately identifying spam, while SMOTE excels in preserving legitimate messages (ham) while filtering out spam. These results contribute significantly to peoples understanding of the intricate dynamics in controlling data imbalance in spam detection and offer insightful information for future studies and real-world applications.

Calibrating Low-Cost Smart Insole Sensors with Recurrent Neural Networks for Accurate Prediction of Center of Pressure.

This paper proposes a scheme for predicting ground reaction force (GRF) and center of pressure (CoP) using low-cost FSR sensors. GRF and CoP data are commonly collected from smart insoles to analyze the wearer's gait and diagnose balance issues. This approach can be utilized to improve a user's rehabilitation process and enable customized treatment plans for patients with specific diseases, making it a useful technology in many fields. However, the conventional measuring equipment for directly monitoring GRF and CoP values, such as F-Scan, is expensive, posing a challenge to commercialization in the industry. To solve this problem, this paper proposes a technology to predict relevant indicators using only low-cost Force Sensing Resistor (FSR) sensors instead of expensive equipment. In this study, data were collected from subjects simultaneously wearing a low-cost FSR Sensor and an F-Scan device, and the relationship between the collected data sets was analyzed using supervised learning techniques. Using the proposed technique, an artificial neural network was constructed that can derive a predicted value close to the actual F-Scan values using only the data from the FSR Sensor. In this process, GRF and CoP were calculated using six virtual forces instead of the pressure value of the entire sole. It was verified through various simulations that it is possible to achieve an improved prediction accuracy of more than 30% when using the proposed technique compared to conventional prediction techniques.