Wearable technologies have the potential to become a valuable influence on human daily life where they may enable observing the world in new ways,including,for example,using augmented reality(AR)applications.Wearable ...Wearable technologies have the potential to become a valuable influence on human daily life where they may enable observing the world in new ways,including,for example,using augmented reality(AR)applications.Wearable technology uses electronic devices that may be carried as accessories,clothes,or even embedded in the user's body.Although the potential benefits of smart wearables are numerous,their extensive and continual usage creates several privacy concerns and tricky information security challenges.In this paper,we present a comprehensive survey of recent privacy-preserving big data analytics applications based on wearable sensors.We highlight the fundamental features of security and privacy for wearable device applications.Then,we examine the utilization of deep learning algorithms with cryptography and determine their usability for wearable sensors.We also present a case study on privacy-preserving machine learning techniques.Herein,we theoretically and empirically evaluate the privacy-preserving deep learning framework's performance.We explain the implementation details of a case study of a secure prediction service using the convolutional neural network(CNN)model and the Cheon-Kim-Kim-Song(CHKS)homomorphic encryption algorithm.Finally,we explore the obstacles and gaps in the deployment of practical real-world applications.Following a comprehensive overview,we identify the most important obstacles that must be overcome and discuss some interesting future research directions.展开更多
Despite the widespread use of Decision trees (DT) across various applications, their performance tends to suffer when dealing with imbalanced datasets, where the distribution of certain classes significantly outweighs...Despite the widespread use of Decision trees (DT) across various applications, their performance tends to suffer when dealing with imbalanced datasets, where the distribution of certain classes significantly outweighs others. Cost-sensitive learning is a strategy to solve this problem, and several cost-sensitive DT algorithms have been proposed to date. However, existing algorithms, which are heuristic, tried to greedily select either a better splitting point or feature node, leading to local optima for tree nodes and ignoring the cost of the whole tree. In addition, determination of the costs is difficult and often requires domain expertise. This study proposes a DT for imbalanced data, called Swarm-based Cost-sensitive DT (SCDT), using the cost-sensitive learning strategy and an enhanced swarm-based algorithm. The DT is encoded using a hybrid individual representation. A hybrid artificial bee colony approach is designed to optimize rules, considering specified costs in an F-Measure-based fitness function. Experimental results using datasets compared with state-of-the-art DT algorithms show that the SCDT method achieved the highest performance on most datasets. Moreover, SCDT also excels in other critical performance metrics, such as recall, precision, F1-score, and AUC, with notable results with average values of 83%, 87.3%, 85%, and 80.7%, respectively.展开更多
This study aimed to develop a multimodal imaging histological model based on computed tomography(CT)images and carcinoembryonic antigen(CEA)values to predict the efficacy of preoperative neoadjuvant therapy in rectal ...This study aimed to develop a multimodal imaging histological model based on computed tomography(CT)images and carcinoembryonic antigen(CEA)values to predict the efficacy of preoperative neoadjuvant therapy in rectal cancer patients.Data were obtained from the Database of Colorectal Cancer of West China Hospital of Sichuan University.A total of 155 patients were enrolled and categorized into good and poor response groups based on pathological evaluation using the tumor regression grade system.Radiomics features were extracted from CT images using PyRadiomics software,and CEA data were collected and processed.Three types of models—a clinical model,a pure radiomics model,and an integrated model—were constructed using logistic regression,support vector machine,random forest(RF),and XGBoost algorithms.The results showed that the integrated model,particularly the RF and XGBoost models,demonstrated the best predictive performance.The RF model achieved an area under the curve(AUC)value of 0.96 in the test set,with accuracy,sensitivity,and specificity of 0.88,0.50,and 1.00,respectively.The XGBoost model had the highest AUC value of 0.97 in the test set,with accuracy,sensitivity,and specificity of 0.91,0.70,and 0.97,respectively.This model can be integrated into existing clinical practice to provide clinicians with additional insights for guiding treatment decisions.Future studies should recruit a larger and more diverse patient population to validate and refine the model,and prospective validation is needed to assess its real-world applicability.展开更多
文摘Wearable technologies have the potential to become a valuable influence on human daily life where they may enable observing the world in new ways,including,for example,using augmented reality(AR)applications.Wearable technology uses electronic devices that may be carried as accessories,clothes,or even embedded in the user's body.Although the potential benefits of smart wearables are numerous,their extensive and continual usage creates several privacy concerns and tricky information security challenges.In this paper,we present a comprehensive survey of recent privacy-preserving big data analytics applications based on wearable sensors.We highlight the fundamental features of security and privacy for wearable device applications.Then,we examine the utilization of deep learning algorithms with cryptography and determine their usability for wearable sensors.We also present a case study on privacy-preserving machine learning techniques.Herein,we theoretically and empirically evaluate the privacy-preserving deep learning framework's performance.We explain the implementation details of a case study of a secure prediction service using the convolutional neural network(CNN)model and the Cheon-Kim-Kim-Song(CHKS)homomorphic encryption algorithm.Finally,we explore the obstacles and gaps in the deployment of practical real-world applications.Following a comprehensive overview,we identify the most important obstacles that must be overcome and discuss some interesting future research directions.
文摘Despite the widespread use of Decision trees (DT) across various applications, their performance tends to suffer when dealing with imbalanced datasets, where the distribution of certain classes significantly outweighs others. Cost-sensitive learning is a strategy to solve this problem, and several cost-sensitive DT algorithms have been proposed to date. However, existing algorithms, which are heuristic, tried to greedily select either a better splitting point or feature node, leading to local optima for tree nodes and ignoring the cost of the whole tree. In addition, determination of the costs is difficult and often requires domain expertise. This study proposes a DT for imbalanced data, called Swarm-based Cost-sensitive DT (SCDT), using the cost-sensitive learning strategy and an enhanced swarm-based algorithm. The DT is encoded using a hybrid individual representation. A hybrid artificial bee colony approach is designed to optimize rules, considering specified costs in an F-Measure-based fitness function. Experimental results using datasets compared with state-of-the-art DT algorithms show that the SCDT method achieved the highest performance on most datasets. Moreover, SCDT also excels in other critical performance metrics, such as recall, precision, F1-score, and AUC, with notable results with average values of 83%, 87.3%, 85%, and 80.7%, respectively.
基金supported by the 1-3-5 projects for artificial intelligence(Grant No.:ZYAI24067)West China Hospital,Sichuan University and the medical research project(Grant No.:S2024045),Sichuan Medical Association.
文摘This study aimed to develop a multimodal imaging histological model based on computed tomography(CT)images and carcinoembryonic antigen(CEA)values to predict the efficacy of preoperative neoadjuvant therapy in rectal cancer patients.Data were obtained from the Database of Colorectal Cancer of West China Hospital of Sichuan University.A total of 155 patients were enrolled and categorized into good and poor response groups based on pathological evaluation using the tumor regression grade system.Radiomics features were extracted from CT images using PyRadiomics software,and CEA data were collected and processed.Three types of models—a clinical model,a pure radiomics model,and an integrated model—were constructed using logistic regression,support vector machine,random forest(RF),and XGBoost algorithms.The results showed that the integrated model,particularly the RF and XGBoost models,demonstrated the best predictive performance.The RF model achieved an area under the curve(AUC)value of 0.96 in the test set,with accuracy,sensitivity,and specificity of 0.88,0.50,and 1.00,respectively.The XGBoost model had the highest AUC value of 0.97 in the test set,with accuracy,sensitivity,and specificity of 0.91,0.70,and 0.97,respectively.This model can be integrated into existing clinical practice to provide clinicians with additional insights for guiding treatment decisions.Future studies should recruit a larger and more diverse patient population to validate and refine the model,and prospective validation is needed to assess its real-world applicability.
