Model Training Research Articles

The application of artificial intelligence and machine learning in civil law protection of privacy rights

Machine learning has emerged as a core technology in domains such as big data, the Internet of Things (IoT), and cloud computing. The training of machine learning models typically requires extensive datasets, often gathered through crowdsourcing methods. These datasets frequently contain significant amounts of private information, including personally identifiable information (e.g., phone numbers, identification numbers) and sensitive data (e.g., financial, medical, and health records). The efficient and cost-effective protection of such data represents a pressing challenge. This article introduces machine learning and explores the concepts and threats associated with privacy within this context. It focuses on mainstream techniques for privacy protection in machine learning, outlining their underlying mechanisms and distinctive features. The discussion is organised around key frameworks such as differential privacy, homomorphic encryption, and secure multi-party computation, presenting a comprehensive review of recent advancements in the field. A comparative analysis of the strengths and limitations of these mechanisms is provided. Finally, the article examines the developmental trajectory of privacy protection in machine learning and proposes potential future research directions in this critical area.

Utility of Multiparametric Breast MRI Radiomics to Predict Cyclin D1 and TGF-β1 Expression.

To develop a machine learning model that integrates clinical features and multisequence MRI radiomics for noninvasively predicting the expression status of prognostic-related factors cyclin D1 and TGF-β1 in breast cancer, providing additional information for the clinical development of personalized treatment plans. A total of 123 breast cancer patients confirmed by surgical pathology were retrospectively enrolled in our Hospital from January 2016 to July 2022. The patients were randomly divided into a training group (87 cases) and a validation group (36 cases). Preoperative routine and dynamic contrast-enhanced magnetic resonance imaging scans of the breast were performed for treatment subjects. The region of interest was manually outlined, and texture features were extracted using AK software. Subsequently, the LASSO algorithm was employed for dimensionality reduction and feature selection to establish the MRI radiomics labels. The diagnostic efficacy and clinical value were assessed through receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA). In the cyclin D1 cohort, the area under the receiver operating characteristic (ROC) curve in the clinical prediction model training and validation groups was 0.738 and 0.656, respectively. The multisequence MRI radiomics prediction model achieved an AUC of 0.874 and 0.753 in these respective groups, while the combined prediction model yielded an AUC of 0.892 and 0.785. In the TGF-β1 cohort, the ROC AUC for the clinical prediction model was found to be 0.693 and 0.645 in the training and validation groups, respectively. For the multiseries MRI radiomics prediction model, it achieved an AUC of 0.875 and 0.760 in these respective groups; whereas for the combined prediction model, it reached an AUC of 0.904 and 0.833. Decision curve analysis (DCA) demonstrated that both cohorts indicated a higher clinical application value for the combined prediction model compared with both individual models-clinical prediction model alone or radiomics model. The integration of clinical features and multisequence MRI radiomics in a combined modeling approach holds significant predictive value for the expression status of cyclin D1 and TGF-β1. The model provides a noninvasive, dynamic evaluation method that provides effective guidance for clinical treatment.

Physical activity and the outcome of cognitive trajectory: a machine learning approach

BackgroundPhysical activity (PA) may have an impact on cognitive function. Machine learning (ML) techniques are increasingly used in dementia research, e.g., for diagnosis and risk stratification. Less is known about the value of ML for predicting cognitive decline in people with dementia (PwD). The aim of this study was to use an ML approach to identify variables associated with a multimodal PA intervention that may impact cognitive changes in PwD, i.e., by distinguishing between cognitive decliners and non-decliners.MethodsThis is a secondary, exploratory analysis using data from a Randomized Controlled Trial that included a 16-week multimodal PA intervention for the intervention group (IG) and treatment as usual for the control group (CG) in nursing homes. Predictors included in the ML models were related to the intervention (e.g., adherence), physical performance (e.g., mobility, balance), and pertinent health-related variables (e.g., health status, dementia form and severity). Primary outcomes were global and domain-specific cognitive performance (i.e., attention/ executive function, language, visuospatial skills, memory) assessed by standardized tests. A Support Vector Machine model was used to perform the classification of each primary outcome into the two classes of decline and non-decline. GridSearchCV with fivefold cross-validation was used for model training, and area under the ROC curve (AUC) and accuracy were calculated to assess model performance.ResultsThe study sample consisted of 319 PwD (IG, N = 161; CG, N = 158). The proportion of PwD experiencing cognitive decline, in the different domains measured, ranged from 27–48% in CG, and from 23–49% in IG, with no statistically significant differences and no time*group effects. ML models showed accuracy and AUC values ranging from 40.6–75.6. The strongest predictors of cognitive decline or non-decline were performance of activities of daily living in IG and CG, and adherence and mobility in IG.ConclusionsML models showed moderate performance, suggesting that the selected variables only had limited value for classification, with adherence and performance of activities of daily living appearing to be predictors of cognitive decline. While the study provides preliminary evidence of the potential use of ML approaches, larger studies are needed to confirm our observations and to include other variables in the prediction of cognitive decline, such as emotional health or biomarker abnormalities.

Machine learning in public health informatics: Evidence that complex sampling structures may not be needed for prediction models with imbalanced outcomes.

The objective of this study is to investigate the predictive ability of machine learning models for imbalanced outcomes from national survey data without the use of sampling weights. We evaluated the predictive performance of machine learning models on imbalanced outcomes from the US National Health and Nutrition Examination Survey (USNHANES) without using sampling weights. Four machine learning models (support vector machine, random forest, least absolute shrinkage and selection operator regression, and deep neural network) were compared with a logistic model that incorporates the survey's complex sampling design. Three resampling methods (oversampling, undersampling, and combined) were used to address class imbalance during the model training process. For all models, the balanced accuracy was similar (ranging from 0.72 to 0.76) and the specificity was smaller than sensitivity except for random forest. The support vector machine and neural networks performed best with sensitivity (ranging from 0.79 to 0.83), while the random forest had the largest specificity (ranging from 0.86 to 0.96), with one exception. PR-AUC scores and Brier scores were low ranging from 0.2529 to 0.3313 (lower scores worse) and 0.1005-0.3245 (lower scores better), respectively CONCLUSIONS: The machine learning models had overall similar predictive capacity to the recommended methods which integrate the complex sampling design for the prediction of osteoarthritis occurrence with USNHANES.

Clinical Application Of Deep Learning-assisted Needles Reconstruction In Prostate Ultrasound Brachytherapy.

High dose rate (HDR) prostate brachytherapy (BT) procedure requires image-guided needle insertion. Given that general anesthesia is often employed during the procedure, minimizing overall planning time is crucial. In this study, we explore the clinical feasibility and time-saving potential of artificial intelligence (AI)-driven auto-reconstruction of transperineal needles in the context of US-guided prostate BT planning. This study included a total of 102 US-planned BT images from a single institution and split into three groups: 50 for model training and validation, 11 to evaluate reconstruction accuracy (test set), and 41 to evaluate the AI tool in a clinical implementation (clinical set). Reconstruction accuracy for the test set was evaluated by comparing the performance of AI-derived and manually reconstructed needles from 5 medical physicists on the 3D-US scans after treatment. The needle total reconstruction time for the clinical set was defined as the timestamp difference from scan acquisition to the start of dose calculations and was compared to values recorded before the clinical implementation of the AI-assisted tool. A mean error of (0.44±0.32)mm was found between the AI-reconstructed and the human consensus needle positions in the test set, with 95.0% of AI needle points falling below 1mm from their human-made counterparts. Post-hoc analysis showed only one of the human observers' reconstructions were significantly different from the others including the AI's. In the clinical set, the AI algorithm achieved a true positive reconstruction rate of 93.4% with only 4.5% of these needles requiring manual corrections from the planner before dosimetry. Total time required to perform AI-assisted catheter reconstruction on clinical cases was on average 15.2min lower (p < 0.01) compared to procedure without AI assistance. This study demonstrates the feasibility of an AI-assisted needle reconstructing tool for 3D-US based HDR prostate BT. This is a step toward treatment planning automation and increased efficiency in HDR prostate BT.

Adaptive Robot Motion Planning for Smart Manufacturing Based on Digital Twin and Bayesian Optimization-Enhanced Reinforcement Learning

Abstract Advanced motion planning is crucial for safe and efficient robotic operations in various scenarios of smart manufacturing, such as assembling, packaging and palletizing. Compared to traditional motion planning methods, Reinforcement Learning (RL) shows better adaptability to complex and dynamic working environments. However, the training of RL models is often time-consuming and the determination of well-behaved reward function parameters is challenging. To tackle these issues, we propose an adaptive robot motion planning approach based on digital twin and reinforcement learning. The core idea is to adaptively select geometry-based or RL-based method for robot motion planning through a real-time distance detection mechanism, which can reduce the complexity of RL model training and accelerate the training process. In addition, we integrate Bayesian Optimization within RL training to refine the reward function parameters. We validate the approach with a Digital Twin enabled robot system through five kinds of tasks (Pick and Place, Drawer Open, Light Switch, Button Press, Cube Push) in dynamic environments. Experiment results show that our approach outperforms traditional RL-based method with improved training speed and guaranteed task performance. This work contributes to the practical deployment of adaptive robot motion planning in smart manufacturing.

An Enhanced Approach Using AGS Network for Skin Cancer Classification

Skin cancer accounts for over 40% of all cancer diagnoses worldwide. However, accurately diagnosing skin cancer remains challenging for dermatologists, as multiple types of skin cancer often appear visually similar. The diagnostic accuracy of dermatologists ranges between 62% and 80%. Although AI models have shown promise in assisting with skin cancer classification in various studies, obtaining the large-scale medical image datasets required for AI model training is not straightforward. To address this limitation, this study proposes the AGS network, designed to overcome the challenges of small datasets and enhance the performance of skin cancer classifiers. The AGS network integrates three key modules: Augmentation (A), GAN (G), and Segmentation (S). It was evaluated using eight deep learning classifiers—GoogLeNet, DenseNet201, ResNet50, MobileNet V3, EfficientNet B0, ViT, EfficientNet V2, and Swin Transformers—on the HAM10000 dataset. Five model configurations were also tested to assess the contribution of each module. The results showed that all eight classifiers demonstrated consistent performance improvements with the AGS network. In particular, EfficientNet V2 + AGS achieved the most significant performance gains over the baseline model, with an increase of +0.1808 in Accuracy and +0.1674 in F1-Score. Among all configurations, ResNet50+AGS achieved the best overall performance, with an Accuracy of 95.87% and an F1-Score of 95.73%. While most previous studies focused on single augmentation methods, this study demonstrates the effectiveness of combining multiple augmentation techniques within an integrated framework. The AGS network demonstrates how integrating diverse methods can improve the performance of skin cancer classification models.

A Robust Anomaly Detection System for Nuclear Power Plants Under Varying Environmental Conditions and Malfunction Levels

Nuclear power plants (NPPs) require rigorous monitoring systems for their safety and efficiency, thus extensive data are acquired continuously from instrumentation and control systems. Anomaly detection is one of the most widely used machine learning approaches for monitoring data, especially when available data for model training are limited or imbalanced, as are data from NPPs. This research presents an anomaly detection system for centralized online monitoring in NPPs that is composed of two modules: data reconstruction and anomaly determination. Considering the large feature dimensions of the data, and leveraging their sequential characteristics in the time domain, four different autoencoder models for data reconstruction, long short-term memory, convolutional neural network, fully connected neural network, and principal component analysis are employed and compared. Two anomaly determination methods are presented and analyzed from the perspective of the characteristics of residuals from the data reconstruction models. The developed system is validated with simulation data containing 239 process variables (sensors) from different subsystems in a NPP. This paper highlights the effectiveness of simulations not only in overcoming the limited amount of data acquired from real plants in malfunctioning status, but also for evaluating the performance of the given models in a more quantitative way by comparing the performance at different malfunction levels. Weighted areas under the receiver operating characteristic curves are suggested as metrics for the validation of the given models and methods, and performance metrics, which can reflect engineers’ preferences, are demonstrated as well.

A robust multimodal brain MRI-based diagnostic model for migraine: validation across different migraine phases and longitudinal follow-up data

Inter-individual variability in symptoms and the dynamic nature of brain pathophysiology present significant challenges in constructing a robust diagnostic model for migraine. In this study, we aimed to integrate different types of magnetic resonance imaging (MRI), providing structural and functional information, and develop a robust machine learning model that classifies migraine patients from healthy controls by testing multiple combinations of hyperparameters to ensure stability across different migraine phases and longitudinally repeated data. Specifically, we constructed a diagnostic model to classify patients with episodic migraine from healthy controls, and validated its performance across ictal and interictal phases, as well as in a longitudinal setting. We obtained T1-weighted and resting-state functional MRI data from 50 patients with episodic migraine and 50 age- and sex-matched healthy controls, with follow-up data collected after one year. Morphological features, including cortical thickness, curvature, and sulcal depth, and functional connectivity features, such as low-dimensional representation of functional connectivity (gradient), degree centrality, and betweenness centrality, were utilized. We employed a regularization-based feature selection method combined with a random forest classifier to construct a diagnostic model. By testing the models with varying feature combinations, penalty terms, and spatial granularities within a strict cross-validation framework, we found that the combination of curvature, sulcal depth, cortical thickness, and functional gradient achieved a robust classification performance. The model performance was assessed using the test dataset and achieved 87% accuracy and 0.94 area under the curve (AUC) at distinguishing migraine patients from healthy controls, with 85%, 0.97 and 84%, 0.93 during the interictal and ictal/peri-ictal phases, respectively. When validated using follow-up data, which was not included during model training, the model achieved 91%, 94%, 89% accuracies and 0.96, 0.94, 0.98 AUC for the total, interictal, and ictal/peri-ictal phases, respectively, confirming its robustness. Feature importance and clinical association analyses exhibited that the somatomotor, limbic, and default mode regions could be reliable markers of migraine. Our findings, which demonstrate a robust diagnostic performance using multimodal MRI features and a machine-learning framework, may offer a valuable approach for clinical diagnosis across diverse cohorts and help alleviate the decision-making burden for clinicians.

Machine Learning-Assisted Prediction of Photothermal Metal-Phenolic Networks.

Photothermal therapy (PTT) demonstrates significant potential in cancer treatment, wound healing, and antibacterial therapy, with its efficacy largely depending on the performance of photothermal agents (PTAs). Metal-phenolic network (MPN) materials are ideal PTA candidates due to their low cost, good biocompatibility and excellent ligand-to-metal charge transfer properties. However, not all MPNs exhibit significant photothermal properties, and the vast chemical space of MPNs (over 700,000 potential combinations) complicates the screening of high-photothermal materials. To address this challenge, this study introduces machine learning (ML) methods for predicting the photothermal performance of MPNs. A database of photothermal properties of 80 modular MPNs was constructed, and the ML process was optimized through feature engineering and model training. The selected extreme gradient boosting model (XGBoost) successfully identified 1,654 high photothermal MPNs from a virtual database of 44,438. Subsequent experimental validation revealed a remarkable success rate of 70% in predicting high photothermal MPNs. Additionally, several previously unreported high photothermal MPNs were discovered, demonstrating advantages in photothermal antibacterial applications. This study offers an innovative ML-driven approach for the efficient screening of MPN materials, providing a solid foundation for PTA design in PTT and other biomedical applications.

Tissue concepts: Supervised foundation models in computational pathology.

Due to the increasing workload of pathologists, the need for automation to support diagnostic tasks and quantitative biomarker evaluation is becoming more and more apparent. Foundation models have the potential to improve generalizability within and across centers and serve as starting points for data efficient development of specialized yet robust AI models. However, the training of foundation models themselves is usually very expensive in terms of data, computation, and time. This paper proposes a supervised training method that drastically reduces these expenses. The proposed method is based on multi-task learning to train a joint encoder, by combining 16 different classification, segmentation, and detection tasks on a total of 912,000 patches. Since the encoder is capable of capturing the properties of the samples, we term it the Tissue Concepts encoder. To evaluate the performance and generalizability of the Tissue Concepts encoder across centers, classification of whole slide images from four of the most prevalent solid cancers - breast, colon, lung, and prostate - was used. The experiments show that the Tissue Concepts model achieve comparable performance to models trained with self-supervision, while requiring only 6% of the amount of training patches. Furthermore, the Tissue Concepts encoder outperforms an ImageNet pre-trained encoder on both in-domain and out-of-domain data. The pre-trained models and will be made available under https://github.com/FraunhoferMEVIS/MedicalMultitaskModeling.

Transcription factor prediction using protein 3D secondary structures.

Transcription factors (TFs) are DNA-binding proteins that regulate gene expression. Traditional methods predict a protein as a TF if the protein contains any DNA-binding domains (DBDs) of known TFs. However, this approach fails to identify a novel TF that does not contain any known DBDs. Recently proposed TF prediction methods do not rely on DBDs. Such methods use features of protein sequences to train a machine learning model, and then use the trained model to predict whether a protein is a TF or not. Because the 3-dimensional (3D) structure of a protein captures more information than its sequence, using 3D protein structures will likely allow for more accurate prediction of novel TFs. We propose a deep learning-based TF prediction method (StrucTFactor), which is the first method to utilize 3D secondary structural information of proteins. We compare StrucTFactor with recent state-of-the-art TF prediction methods based on ∼525 000 proteins across 12 datasets, capturing different aspects of data bias (including sequence redundancy) possibly influencing a method's performance. We find that StrucTFactor significantly (p-value<0.001) outperforms the existing TF prediction methods, improving the performance over its closest competitor by up to 17% based on Matthews correlation coefficient. Data and source code are available at https://github.com/lieboldj/StrucTFactor and on our website at https://apps.cosy.bio/StrucTFactor/. Included.

Shaping current European mitochondrial haplogroup frequency in response to infection: the case of SARS-CoV-2 severity

The frequency of mitochondrial DNA haplogroups (mtDNA-HG) in humans is known to be shaped by migration and repopulation. Mounting evidence indicates that mtDNA-HG are not phenotypically neutral, and selection may contribute to its distribution. Haplogroup H, the most abundant in Europe, improved survival in sepsis. Here we developed a random forest trained model for mitochondrial haplogroup calling using data procured from GWAS arrays. Our results reveal that in the context of the SARS-CoV-2 pandemic, HV branch were found to represent protective factors against the development of critical SARS-CoV-2 in an analysis of 14,349 patients. These results highlight the role of mtDNA in the response to infectious diseases and support the proposal that its expansion and population proportion has been influenced by selection through successive pandemics.

Multimodal deep learning for predicting in-hospital mortality in heart failure patients using longitudinal chest X-rays and electronic health records.

Amid an aging global population, heart failure has become a leading cause of hospitalization among older people. Its high prevalence and mortality rates underscore the importance of accurate mortality prediction for swift disease progression assessment and better patient outcomes. The evolution of artificial intelligence (AI) presents new avenues for predicting heart failure mortality. Yet current research has predominantly leveraged structured data and unstructured clinical notes from electronic health records (EHR), underutilizing the prognostic value of chest X-rays (CXRs).This study aims to harness deep learning methodologies to explore the feasibility of enhancing the precision of predicting in-hospital all-cause mortality in heart failure patients using CXRs data.We propose a novel multimodal deep learning network based on the spatially and temporally decoupled Transformer (MN-STDT) for in-hospital mortality prediction in heart failure by integrating longitudinal CXRs and structured EHR data. The MN-STDT captures spatial and temporal information from CXRs through a Hybrid Spatial Encoder and a Distance-Aware Temporal Encoder, ultimately fusing features from both modalities for predictive modeling. Initial pre-training of the spatial encoder was conducted on CheXpert, followed by full model training and evaluation on the MIMIC-IV and MIMIC-CXR datasets for mortality prediction tasks. In a comprehensive view, the MN-STDT demonstrated the best performance, with an AUC-ROC of 0.8620, surpassing all baseline models. Comparative analysis revealed that the AUC-ROC of the multimodal model (0.8620) was significantly higher than that of models using only structured data (0.8166) or chest X-ray data alone (0.7479).This study demonstrates the value of CXRs in the prognosis of heart failure, showing that the combination of longitudinal CXRs with structured EHR data can significantly improve the accuracy of mortality prediction in heart failure. Feature importance analysis based on SHAP provides interpretable decision support, paving the way for potential clinical applications.

VSPNet: a vehicle speed prediction model incorporating transformer and BiLSTM

Abstract In recent years, with the increasing adoption of hybrid vehicles, energy management strategies &amp;#xD;have become a prominent research focus. Accurate Vehicle Speed Prediction (VSP) is a critical &amp;#xD;prerequisite for achieving optimal results in predictive energy management strategies. However, &amp;#xD;existing speed prediction algorithms fail to fully leverage vehicle data to enhance prediction &amp;#xD;accuracy. Therefore, a novel Vehicle Speed Prediction Net (VSPNet) is proposed in this study. &amp;#xD;Firstly, we constructed a combined cycle condition for model training through comprehensive &amp;#xD;analysis and analysed the vehicle feature parameters through the Random Forest (RF) algorithm &amp;#xD;and Pearson correlation analysis to select the best input feature parameters. Then a VSPNet &amp;#xD;speed prediction model is proposed based on the Transformer model. In the encoder part, firstly, &amp;#xD;by assigning weights to the input feature parameters and incorporating the temporal attention &amp;#xD;mechanism, the model is made to make better use of the input features from two dimensions, &amp;#xD;and at the same time the Transformer model's encoder based on positional coding combined &amp;#xD;with Bi-directional Long Short-Term Memory (BiLSTM) belonging to Recurrent Neural &amp;#xD;Networks(RNN), which is used as a decoder to better catch and handle long-term dependencies &amp;#xD;in sequence data. Finally, a comparative experiment between VSPNet and the classical speed &amp;#xD;prediction models was carried out. The proposed VSPNet model reduces the RMSE by 37%, &amp;#xD;22%, and 20% and MAE by 39%, 25, and 24% compared to the LSTM model for the prediction &amp;#xD;time horizons of 3s, 5s, and 8s. The RMSE is reduced by 47%, 28%, and 7%, and the MAE is &amp;#xD;reduced by 47%, 30, and 9% compared to the Transformer model for the prediction time &amp;#xD;horizons of 3s, 5s, and 8s. The experimental results demonstrate the superiority of this speed &amp;#xD;prediction model.

A Study on Canopy Volume Measurement Model for Fruit Tree Application Based on LiDAR Point Cloud

The accurate measurement of orchard canopy volume serves as a crucial foundation for wind regulation and dosage adjustments in precision orchard management. However, existing methods for measuring canopy volume fail to satisfy the high precision and real-time requirements necessary for accurate variable-rate applications in fruit orchards. To address these challenges, this study develops a canopy volume measurement model for orchard spraying using LiDAR point cloud data. In the domain of point cloud feature extraction, an improved Alpha Shape algorithm is proposed for extracting point cloud contours. This method improves the validity judgment for contour line segments, effectively reducing contour length errors on each 3D point cloud projection plane. Additionally, improvements to the mesh integral volume method incorporate the effects of canopy gaps in height difference calculations, significantly enhancing the accuracy of canopy volume estimation. For feature selection, a random forest-based recursive feature elimination method with cross-validation was employed to filter 10 features. Ultimately, five key features were retained for model training: the number of point clouds, the 2D point cloud contour along the X- and Z-projection directions, the 2D width in the Y-projection direction, and the 2D length in the Z-projection direction. During model construction, the study optimized the hyperparameters of partial least squares regression (PLSR), backpropagation (BP) neural networks, and gradient boosting decision trees (GBDT) to build canopy volume measurement models tailored to the dataset. Experimental results indicate that the PLSR model outperformed other approaches, achieving optimal performance with three principal components. The resulting canopy volume measurement model achieved an R2 of 0.9742, an RMSE of 0.1879, and an MAE of 0.1161. These results demonstrate that the PLSR model exhibits strong generalization ability, minimal prediction bias, and low average prediction error, offering a valuable reference for precision control of canopy spraying in orchards.

Collaboration of IoT devices in smart home scenarios: algorithm research based on graph neural networks and federated learning

Traditional IoT device collaboration is usually static and cannot adjust the collaboration mode between devices according to various changes, which limits work efficiency. To this end, an IoT device collaboration optimization algorithm based on graph neural network and federated learning is studied. This method abstracts various IoT device nodes and their communication relationships into graph structured data for storage, and then uses federated learning to train the graph convolutional network with graph structured data. The obtained model can be used to optimize the collaboration mode of IoT devices. During the training process, the total average MSE (mean square error) between the output and the label of the graph convolutional network model based on federated learning is 0.968; the total standard deviation of MSE is 0.0353; the total time from training to model convergence is 435.82 s, of which data transmission time accounts for 27.1% and model training time accounts for 72.9%. In a 2-h practical experiment, the graph convolutional network model based on federated learning was used to optimize the collaboration mode of smart homes, achieving a target environment residence time of 87 min and a total power consumption reduction of 0.69 kW·h. The results show that this method can effectively optimize the collaboration efficiency of IoT devices, reduce training time and network overhead, but it fails to improve the prediction accuracy of the model and may also lead to a decrease in stability.

Sistem Chabot Layanan Informasi Mahasiswa Menggunakan Algoritma Long Short-Term Memory

In the era of globalization and rapid information flow, the demand for efficient and accurate information, especially within academic institutions, is rising. Students often face challenges in accessing educational resources and real-time information, particularly outside official working hours. Existing online information services have limitations in providing continuous access. This research focuses on developing and evaluating a student information service chatbot system at Universitas Jenderal Achmad Yani Yogyakarta (UNJAYA) using the Long Short-Term Memory (LSTM) algorithm. The primary objective is to create a system that delivers real-time, accurate, and efficient information services to students. The Machine Learning Development Cycle (MLDC) is employed in the model development process, including stages such as data collection, processing, model training, evaluation, and implementation. The system's performance is tested using a questionnaire distributed to students, with responses measured on a Likert scale. The results demonstrate a chatbot with a 97.76% accuracy rate, 98.34% precision, and 97.76% recall. The overall system evaluation yielded an average score of 3.87, categorized as good. This research concludes that the LSTM-based chatbot successfully enhances information services at the Faculty of Engineering and Information Technology, providing an innovative solution to meet student needs in real-time

Predictive Modeling of Chronic Kidney Disease Progression with Ensemble Learning Techniques

The present study aims to tackle the significant issue of prompt identification of chronic kidney disease (CKD), a highly prevalent and potentially fatal medical illness. Given the crucial function of the kidneys in maintaining homeostasis, we put forth a novel ensemble learning model to forecast the onset of chronic kidney disease (CKD). Utilizing an extensive dataset, the study employs ten carefully designed stages, covering data analysis, missing data management, normalization, and training of machine learning models. The model that we have proposed exhibits superior performance compared to the existing approaches, attaining a noteworthy accuracy rate of 98.74%. Additionally, it demonstrates a sensitivity rate of 100%, a specificity rate of 96.54%, and an F1 score of 99.02%. The visual representation of the confusion matrix effectively showcases the strong performance of the model. The results of this study indicate that our ensemble technique holds promise as a valuable tool for the prompt detection of chronic kidney disease (CKD). It has the potential to improve diagnostic accuracy in clinical settings and alleviate the financial burden associated with advanced CKD treatments.

Ultrasound radiomics model based on grayscale transrectal ultrasound-guided biopsy for diagnosing prostate cancer and predicting distant metastasis.

A prostate ultrasound (US) imaging omics model was established to assess its effectiveness in diagnosing prostate cancer (PCa), predicting Gleason score (GS), and determining the likelihood of distant metastasis. US images of patients with prostate pathology confirmed by biopsy or surgery at our hospital were retrospectively analyzed. Regions of interest (ROI) segmentation, feature extraction, feature screening, and the construction and training of the radiomics model were performed. Area under the curve (AUC) for the magnetic resonance imaging Prostate Imaging Reporting and Data System (MRI PI-RADS) classification, radiomics alone, and radiomics combined with prostate-specific antigen (PSA) in diagnosing PCa were 70.74%, 71.13%, and 90.47%, respectively. AUCs for the MRI PI-RADS classification, radiomics alone, and radiomics combined with PSA in predicting the GS of PCa were 75.6%, 74.7%, and 88.9%, respectively. Furthermore, AUCs for MRI PI-RADS classification and radiomics alone in predicting distant metastasis of PCa were 66.7% and 90.8%, respectively. The combination of ultrasonic imaging omics and serum PSA significantly improves the efficiency of PCa diagnosis, GS prediction, and distant metastasis prediction. This method is an important tool for PCa screening and follow-up.