Pub Date : 2025-03-12DOI: 10.1016/j.array.2025.100380
Yongjiao Sun, Xueyan Ma, Anrui Han
With the proliferation of mobile smart devices and wireless communication technologies, Mobile CrowdSensing (MCS) has emerged as a significant data collection method. MCS faces two key challenges: selecting high-quality data sellers with unknown reliability and determining fair compensation that addresses device wear and privacy risks. We introduce two novel contributions. First, the MAB-RS algorithm leverages multi-armed bandit reinforcement learning and a data freshness model to dynamically optimize seller recruitment, efficiently balancing exploration of unknown sellers and exploitation of high-quality ones. Second, the MAB-RSP employs a Stackelberg game framework, enabling platforms and sellers to collaboratively maximize profits through strategic pricing and participation incentives. Experiments demonstrate that the algorithm improves revenue while ensuring balanced benefits for all participants.
{"title":"MAB-RSP: Data pricing based on Stackelberg game in MCS","authors":"Yongjiao Sun, Xueyan Ma, Anrui Han","doi":"10.1016/j.array.2025.100380","DOIUrl":"10.1016/j.array.2025.100380","url":null,"abstract":"<div><div>With the proliferation of mobile smart devices and wireless communication technologies, Mobile CrowdSensing (MCS) has emerged as a significant data collection method. MCS faces two key challenges: selecting high-quality data sellers with unknown reliability and determining fair compensation that addresses device wear and privacy risks. We introduce two novel contributions. First, the MAB-RS algorithm leverages multi-armed bandit reinforcement learning and a data freshness model to dynamically optimize seller recruitment, efficiently balancing exploration of unknown sellers and exploitation of high-quality ones. Second, the MAB-RSP employs a Stackelberg game framework, enabling platforms and sellers to collaboratively maximize profits through strategic pricing and participation incentives. Experiments demonstrate that the algorithm improves revenue while ensuring balanced benefits for all participants.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100380"},"PeriodicalIF":2.3,"publicationDate":"2025-03-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143643388","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Cardiovascular disease, a leading cause of mortality, underscores the critical need for precise diagnostic methods. Cardiac Magnetic Resonance (CMR) imaging is pivotal for diagnosing heart conditions, yet accurately segmenting the myocardium ring (MYO) remains a significant challenge. This study enhances the U-Net model with fuzzy pooling and evaluates the effects of different loss functions: cross-entropy loss, which evaluates the disparity between predicted and actual probability distributions; focal loss, which tackles class imbalance by reducing the weight of easily classified examples; dice loss, which emphasizes the overlap between predicted and actual segments; Lovász-Softmax loss, which is optimized for Intersection over Union (IoU); and CrossLov, which merges cross-entropy and Lovász-softmax, using the ACDC 2017 dataset. Focal loss achieved the lowest train loss scores of 0.0011% at epoch 95 and 0.0012% at epoch 96. Cross-entropy showed high dice scores but did not excel in boundary delineation. Dice loss showed moderate performance. Lovász-softmax excelled in IoU with an average score of 90.68%, while CrossLov exhibited balanced performance, achieving robust general segmentation results with an IoU score of 93.691%. Additionally, CrossLov attained the lowest Hausdorff Distance (HD), with an overall score of 2.816 mm and 1.309 mm for the MYO, indicating superior boundary precision. These findings highlight the role of loss function selection used together with fuzzy pooling for enhancing the robustness and precision of MYO segmentation, thereby contributing to improved diagnostic accuracy in cardiovascular care.
{"title":"Investigating diverse loss functions for myocardium ring segmentation in Cardiac Magnetic Resonance images using fuzzy pooling","authors":"Riandini , Eko Mulyanto Yuniarno , I. Ketut Eddy Purnama , Masayoshi Aritsugi , Mauridhi Hery Purnomo","doi":"10.1016/j.array.2025.100382","DOIUrl":"10.1016/j.array.2025.100382","url":null,"abstract":"<div><div>Cardiovascular disease, a leading cause of mortality, underscores the critical need for precise diagnostic methods. Cardiac Magnetic Resonance (CMR) imaging is pivotal for diagnosing heart conditions, yet accurately segmenting the myocardium ring (MYO) remains a significant challenge. This study enhances the U-Net model with fuzzy pooling and evaluates the effects of different loss functions: cross-entropy loss, which evaluates the disparity between predicted and actual probability distributions; focal loss, which tackles class imbalance by reducing the weight of easily classified examples; dice loss, which emphasizes the overlap between predicted and actual segments; Lovász-Softmax loss, which is optimized for Intersection over Union (IoU); and CrossLov, which merges cross-entropy and Lovász-softmax, using the ACDC 2017 dataset. Focal loss achieved the lowest train loss scores of 0.0011% at epoch 95 and 0.0012% at epoch 96. Cross-entropy showed high dice scores but did not excel in boundary delineation. Dice loss showed moderate performance. Lovász-softmax excelled in IoU with an average score of 90.68%, while CrossLov exhibited balanced performance, achieving robust general segmentation results with an IoU score of 93.691%. Additionally, CrossLov attained the lowest Hausdorff Distance (HD), with an overall score of 2.816 mm and 1.309 mm for the MYO, indicating superior boundary precision. These findings highlight the role of loss function selection used together with fuzzy pooling for enhancing the robustness and precision of MYO segmentation, thereby contributing to improved diagnostic accuracy in cardiovascular care.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100382"},"PeriodicalIF":2.3,"publicationDate":"2025-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143628378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-03-05DOI: 10.1016/j.array.2025.100381
Abdulatif Alabdulatif
In the rapidly evolving landscape of cybersecurity, privacy-preserving anomaly detection has become crucial, particularly with the rise of sophisticated privacy attacks in distributed learning systems. Traditional centralized anomaly detection systems face challenges related to data privacy and scalability, making federated learning a promising alternative. However, federated learning models remain vulnerable to several privacy attacks, such as inference attacks, model inversion, and gradient leakage. To address these threats, this paper presents GuardianAI, a novel federated anomaly detection framework that incorporates advanced differential privacy techniques, including Gaussian noise addition and secure aggregation protocols, specifically designed to mitigate these attacks. GuardianAI aims to enhance privacy while maintaining high detection accuracy across distributed nodes. The framework effectively prevents attackers from extracting sensitive data from model updates by introducing noise to the gradients and securely aggregating updates across nodes. Experimental results show that GuardianAI achieves a testing accuracy of 99.8 %, outperforming other models like Logistic Regression, SVM, and Random Forest, while robustly defending against common privacy threats. These results demonstrate the practical potential of GuardianAI for secure deployment in various network environments, ensuring privacy without compromising performance.
{"title":"GuardianAI: Privacy-preserving federated anomaly detection with differential privacy","authors":"Abdulatif Alabdulatif","doi":"10.1016/j.array.2025.100381","DOIUrl":"10.1016/j.array.2025.100381","url":null,"abstract":"<div><div>In the rapidly evolving landscape of cybersecurity, privacy-preserving anomaly detection has become crucial, particularly with the rise of sophisticated privacy attacks in distributed learning systems. Traditional centralized anomaly detection systems face challenges related to data privacy and scalability, making federated learning a promising alternative. However, federated learning models remain vulnerable to several privacy attacks, such as inference attacks, model inversion, and gradient leakage. To address these threats, this paper presents GuardianAI, a novel federated anomaly detection framework that incorporates advanced differential privacy techniques, including Gaussian noise addition and secure aggregation protocols, specifically designed to mitigate these attacks. GuardianAI aims to enhance privacy while maintaining high detection accuracy across distributed nodes. The framework effectively prevents attackers from extracting sensitive data from model updates by introducing noise to the gradients and securely aggregating updates across nodes. Experimental results show that GuardianAI achieves a testing accuracy of 99.8 %, outperforming other models like Logistic Regression, SVM, and Random Forest, while robustly defending against common privacy threats. These results demonstrate the practical potential of GuardianAI for secure deployment in various network environments, ensuring privacy without compromising performance.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100381"},"PeriodicalIF":2.3,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143620801","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-19DOI: 10.1016/j.array.2025.100378
Dianwei Chi , Tiantian Huang , Zehao Jia , Sining Zhang
Due to the high semantic flexibility of Chinese text, the difficulty of word separation, and the problem of multiple meanings of one word, a sentiment analysis model based on the combination of BERT dynamic semantic coding with temporal convolutional neural network (TCN), bi-directional long- and short-term memory network (BiLSTM), and Self-Attention mechanism (Self-Attention) is proposed. The model uses BERT pre-training to generate word vectors as model input, uses the causal convolution and dilation convolution structures of TCN to obtain higher-level sequential features, then passes to the BiLSTM layer to fully extract contextual sentiment features, and finally uses the Self-Attention mechanism to distinguish the importance of sentiment features in sentences, thus improving the accuracy of sentiment classification. The proposed model demonstrates superior performance across multiple datasets, achieving accuracy rates of 89.4 % and 91.2 % on the hotel review datasets C1 and C2, with corresponding F1 scores of 0.898 and 0.904. These results, which surpass those of the comparative models, validate the model's effectiveness across different datasets and highlight its robustness and generalizability in sentiment analysis. It also shows that BERT-based coding can improve the model's performance more than Word2Vec.
{"title":"Research on sentiment analysis of hotel review text based on BERT-TCN-BiLSTM-attention model","authors":"Dianwei Chi , Tiantian Huang , Zehao Jia , Sining Zhang","doi":"10.1016/j.array.2025.100378","DOIUrl":"10.1016/j.array.2025.100378","url":null,"abstract":"<div><div>Due to the high semantic flexibility of Chinese text, the difficulty of word separation, and the problem of multiple meanings of one word, a sentiment analysis model based on the combination of BERT dynamic semantic coding with temporal convolutional neural network (TCN), bi-directional long- and short-term memory network (BiLSTM), and Self-Attention mechanism (Self-Attention) is proposed. The model uses BERT pre-training to generate word vectors as model input, uses the causal convolution and dilation convolution structures of TCN to obtain higher-level sequential features, then passes to the BiLSTM layer to fully extract contextual sentiment features, and finally uses the Self-Attention mechanism to distinguish the importance of sentiment features in sentences, thus improving the accuracy of sentiment classification. The proposed model demonstrates superior performance across multiple datasets, achieving accuracy rates of 89.4 % and 91.2 % on the hotel review datasets C1 and C2, with corresponding F1 scores of 0.898 and 0.904. These results, which surpass those of the comparative models, validate the model's effectiveness across different datasets and highlight its robustness and generalizability in sentiment analysis. It also shows that BERT-based coding can improve the model's performance more than Word2Vec.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100378"},"PeriodicalIF":2.3,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143464923","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-17DOI: 10.1016/j.array.2025.100377
Qinxia Wang , Yue Qiu , Weiqiang Qu , Dianhui Wang
With the fast development of advanced science and technology, the urban rail transit continues to develop rapidly, with the industry pays more attention to the operation safety and maintenance of trains. In this paper, an improved 2D convolutional stochastic configuration network (2DConSCN) based method is proposed to deal with traffic video for foreign object recognition. Comparing with the existing stochastic configuration networks,the proposed method retains the stochastic configured mechanism for the convolutional kernel weights. Moreover, a feature selection method is presented to improve the image representation ability. The proposed improved 2DConSCN method greatly reduces the number of parameters, and the trained model can quickly obtain results on test data. Experiments are performed on a rail transit dataset, the comparison results show that the proposed method gets better performance in the recognition task, showing its great potential to meet the requirement of railway monitoring.
{"title":"Objects recognition from traffic video data using improved 2D convolutional stochastic configuration networks","authors":"Qinxia Wang , Yue Qiu , Weiqiang Qu , Dianhui Wang","doi":"10.1016/j.array.2025.100377","DOIUrl":"10.1016/j.array.2025.100377","url":null,"abstract":"<div><div>With the fast development of advanced science and technology, the urban rail transit continues to develop rapidly, with the industry pays more attention to the operation safety and maintenance of trains. In this paper, an improved 2D convolutional stochastic configuration network (2DConSCN) based method is proposed to deal with traffic video for foreign object recognition. Comparing with the existing stochastic configuration networks,the proposed method retains the stochastic configured mechanism for the convolutional kernel weights. Moreover, a feature selection method is presented to improve the image representation ability. The proposed improved 2DConSCN method greatly reduces the number of parameters, and the trained model can quickly obtain results on test data. Experiments are performed on a rail transit dataset, the comparison results show that the proposed method gets better performance in the recognition task, showing its great potential to meet the requirement of railway monitoring.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100377"},"PeriodicalIF":2.3,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143437113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-07DOI: 10.1016/j.array.2025.100375
Gazi Hasan Al Masud , Rejaul Islam Shanto , Ishmam Sakin , Muhammad Rafsan Kabir
Depression is an increasingly prevalent issue, particularly among young people, significantly impacting their well-being and causing persistent distress. Early detection is crucial to address this growing concern. This study utilizes various machine learning, deep learning, and language models to detect depression among Bangladeshi university students. To address data imbalance in the employed dataset, resampling techniques such as SMOTE and Cluster Centroids are applied. Additionally, exhaustive hyperparameter optimization is performed to enhance classification performance. Our results indicate that machine learning algorithms, particularly Random Forest, effectively predict depression with an accuracy of 91.1% and an F1-score of 91.6%. Language models like RoBERTa also achieve strong results, with a recall score of 98.6%. Moreover, explainable AI (XAI) methods, including SHAP and LIME, are employed to interpret model predictions, underscoring the importance of transparency in machine learning. This work contributes to the early identification of depression by integrating machine learning, deep learning, natural language processing, and XAI techniques. While this study focuses on Bangladeshi or similar demographic groups, the proposed approaches are adaptable and can be applied to other populations for generalization.
{"title":"Effective depression detection and interpretation: Integrating machine learning, deep learning, language models, and explainable AI","authors":"Gazi Hasan Al Masud , Rejaul Islam Shanto , Ishmam Sakin , Muhammad Rafsan Kabir","doi":"10.1016/j.array.2025.100375","DOIUrl":"10.1016/j.array.2025.100375","url":null,"abstract":"<div><div>Depression is an increasingly prevalent issue, particularly among young people, significantly impacting their well-being and causing persistent distress. Early detection is crucial to address this growing concern. This study utilizes various machine learning, deep learning, and language models to detect depression among Bangladeshi university students. To address data imbalance in the employed dataset, resampling techniques such as SMOTE and Cluster Centroids are applied. Additionally, exhaustive hyperparameter optimization is performed to enhance classification performance. Our results indicate that machine learning algorithms, particularly Random Forest, effectively predict depression with an accuracy of 91.1% and an F1-score of 91.6%. Language models like RoBERTa also achieve strong results, with a recall score of 98.6%. Moreover, explainable AI (XAI) methods, including SHAP and LIME, are employed to interpret model predictions, underscoring the importance of transparency in machine learning. This work contributes to the early identification of depression by integrating machine learning, deep learning, natural language processing, and XAI techniques. While this study focuses on Bangladeshi or similar demographic groups, the proposed approaches are adaptable and can be applied to other populations for generalization.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100375"},"PeriodicalIF":2.3,"publicationDate":"2025-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-06DOI: 10.1016/j.array.2025.100374
Ying Liu , Xiaohua Huang , Liwei Xiong , Ruyu Chang , Wenjing Wang , Long Chen
Stock price prediction presents significant challenges owing to the highly volatile and nonlinear nature of financial markets, which are influenced by various factors including macroeconomic conditions, policy changes, and market sentiment. Traditional prediction models such as ARIMA and classic linear regression models are often inadequate for capturing the complex dynamics of stock prices. The advent of deep learning has led to substantial improvements in prediction accuracy, with various recurrent neural networks widely employed for representation learning from stock sequences. However, recurrent networks such as LSTM and GRU may exhibit susceptibility to overfitting the training data, leading to suboptimal performance in real-world predictions due to the inherent noise and volatility of stock market data. Recent research has demonstrated that temporal convolutional networks (TCN) exhibit impressive capabilities in stock price prediction. A TCN can achieve extensive sequence memory by utilizing dilated convolutions, enabling it to capture long-term dependencies in time-series data, as well as causal convolution, ensuring that the model does not utilize future information when predicting future values, which is particularly crucial for time-series prediction. Nevertheless, stock market data typically contain substantial noise to which TCNs may be overly sensitive, thereby affecting the accuracy of the predictions. To address this issue, we propose a novel stock price prediction method based on the Generative Adversarial Networks (GANs) framework, utilizing an Attentive Temporal Convolutional Network (ATCN) as the generator, termed Attentive Temporal Convolution-based Generative Adversarial Network (ATCGAN). This approach employs a GAN framework to generate stock price data using an attentive temporal convolutional network as a generator, whereas a CNN-based discriminator evaluates the authenticity of the data. Adversarial training facilitates the model’s learning of the complex distribution of stock price data. Within the GAN framework, the TCN effectively captures long-term dependencies, combined with an attention mechanism for generating representative feature combinations, thereby enhancing prediction accuracy. Compared to the traditional ARIMA forecasting method, ACTGAN achieved a 78.29% reduction in Mean Absolute Error (MAE). Furthermore, when compared to the deep learning method GRU, ACTGAN reduced the Mean Absolute Error (MAE) by 51.01%. The experimental results demonstrate that the proposed GAN-based approach significantly outperforms the traditional methods and deep learning techniques.
{"title":"Stock price prediction with attentive temporal convolution-based generative adversarial network","authors":"Ying Liu , Xiaohua Huang , Liwei Xiong , Ruyu Chang , Wenjing Wang , Long Chen","doi":"10.1016/j.array.2025.100374","DOIUrl":"10.1016/j.array.2025.100374","url":null,"abstract":"<div><div>Stock price prediction presents significant challenges owing to the highly volatile and nonlinear nature of financial markets, which are influenced by various factors including macroeconomic conditions, policy changes, and market sentiment. Traditional prediction models such as ARIMA and classic linear regression models are often inadequate for capturing the complex dynamics of stock prices. The advent of deep learning has led to substantial improvements in prediction accuracy, with various recurrent neural networks widely employed for representation learning from stock sequences. However, recurrent networks such as LSTM and GRU may exhibit susceptibility to overfitting the training data, leading to suboptimal performance in real-world predictions due to the inherent noise and volatility of stock market data. Recent research has demonstrated that temporal convolutional networks (TCN) exhibit impressive capabilities in stock price prediction. A TCN can achieve extensive sequence memory by utilizing dilated convolutions, enabling it to capture long-term dependencies in time-series data, as well as causal convolution, ensuring that the model does not utilize future information when predicting future values, which is particularly crucial for time-series prediction. Nevertheless, stock market data typically contain substantial noise to which TCNs may be overly sensitive, thereby affecting the accuracy of the predictions. To address this issue, we propose a novel stock price prediction method based on the Generative Adversarial Networks (GANs) framework, utilizing an Attentive Temporal Convolutional Network (ATCN) as the generator, termed Attentive Temporal Convolution-based Generative Adversarial Network (ATCGAN). This approach employs a GAN framework to generate stock price data using an attentive temporal convolutional network as a generator, whereas a CNN-based discriminator evaluates the authenticity of the data. Adversarial training facilitates the model’s learning of the complex distribution of stock price data. Within the GAN framework, the TCN effectively captures long-term dependencies, combined with an attention mechanism for generating representative feature combinations, thereby enhancing prediction accuracy. Compared to the traditional ARIMA forecasting method, ACTGAN achieved a 78.29% reduction in Mean Absolute Error (MAE). Furthermore, when compared to the deep learning method GRU, ACTGAN reduced the Mean Absolute Error (MAE) by 51.01%. The experimental results demonstrate that the proposed GAN-based approach significantly outperforms the traditional methods and deep learning techniques.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100374"},"PeriodicalIF":2.3,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143351005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-30DOI: 10.1016/j.array.2025.100376
Tazkia Mim Angona, M. Rubaiyat Hossain Mondal
Brain Tumors are a life-threatening cancer type. Due to the varied types and aggressive nature of these tumors, medical diagnostics faces significant challenges. Effective diagnosis and treatment planning depends on identifying the brain tumor areas from MRI images accurately. Traditional methods tend to use manual segmentation, which is costly, time consuming and prone to errors. Automated segmentation using deep learning approaches has shown potential in detecting tumor region. However, the complexity of the tumor areas which contain various shapes, sizes, fuzzy boundaries, makes this process difficult. Therefore, a robust automated segmentation method in brain tumor segmentation is required. In our paper, we present a hybrid model, 3-Dimension (3D) ResAttU-Net-Swin, which combines residual U-Net, attention mechanism and swin transformer. Residual blocks are introduced in the U-Net structure as encoder and decoder to avoid vanishing gradient problems and improve feature recovery. Attention-based skip connections are used to enhance the feature information transition between the encoder and decoder. The swin transformer obtains broad-scale features from the image data. The proposed hybrid model was evaluated on both the BraTS 2020 and BraTS 2019 datasets. It achieved an average Dice Similarity Coefficients (DSC) of 88.27 % and average Intersection over Union (IoU) of 79.93 % on BraTS 2020. On BraTS 2019, the model achieved an average DSC of 89.20 % and average IoU of 81.40 %. The model obtains higher DSC than the existing methods. The experiment result shows that the proposed methodology, 3D ResAttU-Net-Swin can be a potential for brain tumor segmentation in clinical settings.
{"title":"An attention based residual U-Net with swin transformer for brain MRI segmentation","authors":"Tazkia Mim Angona, M. Rubaiyat Hossain Mondal","doi":"10.1016/j.array.2025.100376","DOIUrl":"10.1016/j.array.2025.100376","url":null,"abstract":"<div><div>Brain Tumors are a life-threatening cancer type. Due to the varied types and aggressive nature of these tumors, medical diagnostics faces significant challenges. Effective diagnosis and treatment planning depends on identifying the brain tumor areas from MRI images accurately. Traditional methods tend to use manual segmentation, which is costly, time consuming and prone to errors. Automated segmentation using deep learning approaches has shown potential in detecting tumor region. However, the complexity of the tumor areas which contain various shapes, sizes, fuzzy boundaries, makes this process difficult. Therefore, a robust automated segmentation method in brain tumor segmentation is required. In our paper, we present a hybrid model, 3-Dimension (3D) ResAttU-Net-Swin, which combines residual U-Net, attention mechanism and swin transformer. Residual blocks are introduced in the U-Net structure as encoder and decoder to avoid vanishing gradient problems and improve feature recovery. Attention-based skip connections are used to enhance the feature information transition between the encoder and decoder. The swin transformer obtains broad-scale features from the image data. The proposed hybrid model was evaluated on both the BraTS 2020 and BraTS 2019 datasets. It achieved an average Dice Similarity Coefficients (DSC) of 88.27 % and average Intersection over Union (IoU) of 79.93 % on BraTS 2020. On BraTS 2019, the model achieved an average DSC of 89.20 % and average IoU of 81.40 %. The model obtains higher DSC than the existing methods. The experiment result shows that the proposed methodology, 3D ResAttU-Net-Swin can be a potential for brain tumor segmentation in clinical settings.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100376"},"PeriodicalIF":2.3,"publicationDate":"2025-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143141157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-12-16DOI: 10.1016/j.array.2024.100373
Chen Li , Ye Zhu , Yang Cao , Jinli Zhang , Annisa Annisa , Debo Cheng , Yasuhiko Morimoto
The computation of the skyline provides a mechanism for utilizing multiple location-based criteria to identify optimal data points. However, the efficiency of these computations diminishes and becomes more challenging as the input data expands. This study presents a novel algorithm aimed at mitigating this challenge by harnessing the capabilities of Apache Spark, a distributed processing platform, for conducting area skyline computations. The proposed algorithm enhances processing speed and scalability. In particular, our algorithm encompasses three key phases: the computation of distances between data points, the generation of distance tuples, and the execution of the skyline operators. Notably, the second phase employs a local partial skyline extraction technique to minimize the volume of data transmitted from each executor (a parallel processing procedure) to the driver (a central processing procedure). Afterwards, the driver processes the received data to determine the final skyline and creates filters to exclude irrelevant points. Extensive experimentation on eight datasets reveals that our algorithm significantly reduces both data size and computation time required for area skyline computation.
{"title":"Mining area skyline objects from map-based big data using Apache Spark framework","authors":"Chen Li , Ye Zhu , Yang Cao , Jinli Zhang , Annisa Annisa , Debo Cheng , Yasuhiko Morimoto","doi":"10.1016/j.array.2024.100373","DOIUrl":"10.1016/j.array.2024.100373","url":null,"abstract":"<div><div>The computation of the skyline provides a mechanism for utilizing multiple location-based criteria to identify optimal data points. However, the efficiency of these computations diminishes and becomes more challenging as the input data expands. This study presents a novel algorithm aimed at mitigating this challenge by harnessing the capabilities of Apache Spark, a distributed processing platform, for conducting area skyline computations. The proposed algorithm enhances processing speed and scalability. In particular, our algorithm encompasses three key phases: the computation of distances between data points, the generation of distance tuples, and the execution of the skyline operators. Notably, the second phase employs a local partial skyline extraction technique to minimize the volume of data transmitted from each executor (a parallel processing procedure) to the driver (a central processing procedure). Afterwards, the driver processes the received data to determine the final skyline and creates filters to exclude irrelevant points. Extensive experimentation on eight datasets reveals that our algorithm significantly reduces both data size and computation time required for area skyline computation.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100373"},"PeriodicalIF":2.3,"publicationDate":"2024-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143141155","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-12-09DOI: 10.1016/j.array.2024.100372
Md Shawmoon Azad, Shakirul Islam Leeon, Riasat Khan, Nabeel Mohammed, Sifat Momen
Depressive illness, influenced by social, psychological, and biological factors, is a significant public health concern that necessitates accurate and prompt diagnosis for effective treatment. This study explores the multifaceted nature of depression by investigating its correlation with various social factors and employing machine learning, natural language processing, and explainable AI to analyze depression assessment scales. Data from a survey of 520 Bangladeshi university students, encompassing socio-personal and clinical questions, was utilized in this study. Eight machine learning algorithms with optimized hyperparameters were applied to evaluate eight depression assessment scales, identifying the most effective one. Additionally, ten machine learning models, including five BERT-based and two generative large language models, were tested using three prompting approaches and assessed across four categories of social factors: relationship dynamics, parental pressure, academic contentment, and exposure to violence. The results showed that support vector machines achieved a remarkable 99.14% accuracy with the PHQ9 scale. While considering the social factors, the stacking ensemble classifier demonstrated the best results. Among NLP approaches, BioBERT outperformed other BERT-based models with 90.34% accuracy when considering all social aspects. In prompting approaches, the Tree of Thought prompting on Claude Sonnet surpassed other prompting techniques with 75.00% accuracy. However, traditional machine learning models outshined NLP methods in tabular data analysis, with the stacking ensemble model achieving the highest accuracy of 97.88%. The interpretability of the top-performing classifier was ensured using LIME.
{"title":"SAD: Self-assessment of depression for Bangladeshi university students using machine learning and NLP","authors":"Md Shawmoon Azad, Shakirul Islam Leeon, Riasat Khan, Nabeel Mohammed, Sifat Momen","doi":"10.1016/j.array.2024.100372","DOIUrl":"10.1016/j.array.2024.100372","url":null,"abstract":"<div><div>Depressive illness, influenced by social, psychological, and biological factors, is a significant public health concern that necessitates accurate and prompt diagnosis for effective treatment. This study explores the multifaceted nature of depression by investigating its correlation with various social factors and employing machine learning, natural language processing, and explainable AI to analyze depression assessment scales. Data from a survey of 520 Bangladeshi university students, encompassing socio-personal and clinical questions, was utilized in this study. Eight machine learning algorithms with optimized hyperparameters were applied to evaluate eight depression assessment scales, identifying the most effective one. Additionally, ten machine learning models, including five BERT-based and two generative large language models, were tested using three prompting approaches and assessed across four categories of social factors: relationship dynamics, parental pressure, academic contentment, and exposure to violence. The results showed that support vector machines achieved a remarkable 99.14% accuracy with the PHQ9 scale. While considering the social factors, the stacking ensemble classifier demonstrated the best results. Among NLP approaches, BioBERT outperformed other BERT-based models with 90.34% accuracy when considering all social aspects. In prompting approaches, the Tree of Thought prompting on Claude Sonnet surpassed other prompting techniques with 75.00% accuracy. However, traditional machine learning models outshined NLP methods in tabular data analysis, with the stacking ensemble model achieving the highest accuracy of 97.88%. The interpretability of the top-performing classifier was ensured using LIME.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"25 ","pages":"Article 100372"},"PeriodicalIF":2.3,"publicationDate":"2024-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143141156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号