Zhicui Lu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zhicui Lu has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Jiaqi Zhao
Centre for Nursing Studies, Faculty of Health Sciences, Universiti Teknologi MARA (UiTM), Selangor, Malaysia
4 shared publications - Ahmad Zia Ul-Saufie
Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Malaysia
2 shared publications - Aisyah Mat Jasin
School of Physics and Electronic Engineering, Sichuan University of Science and Engineering, Zigong 643000, China
2 shared publications - Farizuwana Akma Zulkifle
Universiti Teknologi MARA
2 shared publications - Hongyu Yu
School of Nursing, Jinzhou Medical University, Jinzhou, Liaoning, China
2 shared publications - Sharifah Shafinaz Sh Abdullah
Faculty of Health Sciences, Universiti Teknologi MARA (UiTM), Selangor, Malaysia
2 shared publications
Latest Publications
- Open Access
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
A Dual-Stream Spatiotemporal Attention Framework for Pain Recognition on Long-tail Clinical Video Data
This study proposes a dual-stream spatiotemporal attention framework to address the challenges of pose interference and category imbalance in in-the-wild pain assessment, termed Residual Convolutional Block Attention Bidirectional Long Short-Term Memory Network (ResCBL-Net), for pain recognition on imbalanced video data. At the artificial intelligence algorithm level, the framework utilizes a Residual Network (ResNet) embedded with a Convolutional Block Attention Module (CBAM) to extract key spatial features, thereby enhancing focus on pain-related regions and suppressing irrelevant interference, while integrating a stacked Bidirectional Long Short-Term Memory (BiLSTM) network to capture temporal dynamic changes in facial expressions. To mitigate the long-tail distribution problem, an improved Dynamically Weighted Composite Focus (DWCF) loss function is introduced to optimize the discriminatory performance of minority classes from the algorithmic foundation. Concurrently, Gradient-weighted Class Activation Mapping (Grad-CAM) is integrated to provide intuitive and easily understandable visual explanations for the decision-making process of the framework. From an engineering application perspective, the study addresses the challenges of deployment and is very robust and transparent in a complex real world environment, creating a very good, non-invasive solution. Experimental results show that the method obtains a high accuracy of 99.67% on the balanced data set, and 85% on the imbalanced data set, which is enough to prove the effectiveness of the method in a real-world application for the long-tail dataset, thereby providing sufficient technical support for clinical intelligent monitoring.
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Dual-Stream Spatiotemporal Attention Framework for Pain Recognition on Long-tail Clinical Video Data
This study proposes a dual-stream spatiotemporal attention framework to address the challenges of pose interference and category imbalance in in-the-wild pain assessment, termed Residual Convolutional Block Attention Bidirectional Long Short-Term Memory Network (ResCBL-Net), for pain recognition on imbalanced video data. At the artificial intelligence algorithm level, the framework utilizes a Residual Network (ResNet) embedded with a Convolutional Block Attention Module (CBAM) to extract key spatial features, thereby enhancing focus on pain-related regions and suppressing irrelevant interference, while integrating a stacked Bidirectional Long Short-Term Memory (BiLSTM) network to capture temporal dynamic changes in facial expressions. To mitigate the long-tail distribution problem, an improved Dynamically Weighted Composite Focus (DWCF) loss function is introduced to optimize the discriminatory performance of minority classes from the algorithmic foundation. Concurrently, Gradient-weighted Class Activation Mapping (Grad-CAM) is integrated to provide intuitive and easily understandable visual explanations for the decision-making process of the framework. From an engineering application perspective, the study addresses the challenges of deployment and is very robust and transparent in a complex real world environment, creating a very good, non-invasive solution. Experimental results show that the method obtains a high accuracy of 99.67% on the balanced data set, and 85% on the imbalanced data set, which is enough to prove the effectiveness of the method in a real-world application for the long-tail dataset, thereby providing sufficient technical support for clinical intelligent monitoring.
- Open Access
- 28 June 2026
- Published by Belitung Raya Foundation in Belitung Nursing Journal
- Vol. 12 (3) , 288-297
Network structure of multidimensional frailty among long-term care residents in China: A cross-sectional network analysis
Background: Multidimensional frailty is highly prevalent among older adults residing in long-term care facilities (LTCFs). However, the structural interrelationships among physical, psychological, social, cognitive, and functional health domains remain insufficiently understood. Objective: To characterize the network structure of multidimensional frailty domains and related health indicators among institutionalized older adults. Methods: This cross-sectional study included 272 LTCF residents from Liaoning, China. Eleven theoretically selected health indicators were modeled using a Mixed Graphical Model with ℓ1-regularized neighborhood regression and model selection based on the Extended Bayesian Information Criterion (γ = 0.25). This approach estimates sparse conditional dependence structures among mixed data types. Strength centrality was used as the primary descriptive metric. Network accuracy and stability were evaluated using non-parametric and case-dropping bootstrap procedures. Results: The strongest conditional association emerged between age and cognitive impairment (edge weight = 0.68). Substantial associations were also observed between physical frailty and ADL dependency (edge weight = 0.43), and between psychological frailty and depressive symptoms (edge weight = 0.45). Physical frailty, ADL dependency, and depression exhibited relatively high strength centrality values; however, bootstrap difference tests indicated that these nodes were not statistically more central than several other highly connected indicators. The correlation stability coefficient for strength centrality was 0.75, indicating good stability of strength centrality estimates within this sample. Conclusion: The network revealed an interconnected pattern linking functional dependency, physical frailty, and psychosocial health among LTCF residents. Because the data were cross-sectional, the observed conditional associations should not be interpreted as causal or temporal pathways. These findings highlight the potential value of multidomain nursing assessment in long-term care settings and may inform hypothesis generation for future longitudinal or interventional nursing studies.
- 28 June 2026
- Published by Belitung Raya Foundation in Belitung Nursing Journal
- Vol. 12 (3) , 288-297
Open AccessNetwork structure of multidimensional frailty among long-term care residents in China: A cross-sectional network analysis
Background: Multidimensional frailty is highly prevalent among older adults residing in long-term care facilities (LTCFs). However, the structural interrelationships among physical, psychological, social, cognitive, and functional health domains remain insufficiently understood. Objective: To characterize the network structure of multidimensional frailty domains and related health indicators among institutionalized older adults. Methods: This cross-sectional study included 272 LTCF residents from Liaoning, China. Eleven theoretically selected health indicators were modeled using a Mixed Graphical Model with ℓ1-regularized neighborhood regression and model selection based on the Extended Bayesian Information Criterion (γ = 0.25). This approach estimates sparse conditional dependence structures among mixed data types. Strength centrality was used as the primary descriptive metric. Network accuracy and stability were evaluated using non-parametric and case-dropping bootstrap procedures. Results: The strongest conditional association emerged between age and cognitive impairment (edge weight = 0.68). Substantial associations were also observed between physical frailty and ADL dependency (edge weight = 0.43), and between psychological frailty and depressive symptoms (edge weight = 0.45). Physical frailty, ADL dependency, and depression exhibited relatively high strength centrality values; however, bootstrap difference tests indicated that these nodes were not statistically more central than several other highly connected indicators. The correlation stability coefficient for strength centrality was 0.75, indicating good stability of strength centrality estimates within this sample. Conclusion: The network revealed an interconnected pattern linking functional dependency, physical frailty, and psychosocial health among LTCF residents. Because the data were cross-sectional, the observed conditional associations should not be interpreted as causal or temporal pathways. These findings highlight the potential value of multidomain nursing assessment in long-term care settings and may inform hypothesis generation for future longitudinal or interventional nursing studies.
- Open Access
- 27 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 102-107
Explainable Machine Learning for Predicting Depressive Symptoms in Older Adults Living Alone: A CHARLS-Based Study
- 27 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 102-107
Open AccessExplainable Machine Learning for Predicting Depressive Symptoms in Older Adults Living Alone: A CHARLS-Based Study
- Open Access
- 1 January 2026
- Published by Wiley in Journal of Advanced Transportation
- Vol. 2026 (1)
Toward Class Imbalance in Facial Pain Recognition for Transportation Safety: An Enhanced ResNet–BiLSTM Framework
Pain is a complex experience involving both sensations and emotions, and its accurate assessment is essential in healthcare and intelligent transportation. Where diverse traffic environments can heighten driver discomfort, leading to fatigue, reduced focus, and increased accident risk. In recent years, extensive studies have demonstrated the potential of deep learning techniques for automatically predicting pain intensity by analyzing pain‐related facial expressions. However, two fundamental challenges remain in real‐world applications: the uncertainties arising from nonstandardized data collection and the class imbalance problem in pain assessment datasets. Addressing this issue is essential to improving recognition accuracy and robustness. This study proposes a novel pain level recognition framework based on a ResNet–BiLSTM architecture. First, the Squeeze‐and‐Excitation attention mechanism is integrated to enhance the model’s ability to capture basic pain‐related facial features, thereby increasing robustness against data uncertainties. Additionally, the I‐center loss function is introduced during model optimization, enhancing intraclass compactness and interclass separability to improve class‐imbalanced model learning. Experimental results on the BioVid Heat Pain Database demonstrate that this approach outperforms state‐of‐the‐art baselines, achieving 99.67% accuracy. Furthermore, ablation studies validate the effectiveness of the proposed components in our framework.
- 1 January 2026
- Published by Wiley in Journal of Advanced Transportation
- Vol. 2026 (1)
Open AccessToward Class Imbalance in Facial Pain Recognition for Transportation Safety: An Enhanced ResNet–BiLSTM Framework
Pain is a complex experience involving both sensations and emotions, and its accurate assessment is essential in healthcare and intelligent transportation. Where diverse traffic environments can heighten driver discomfort, leading to fatigue, reduced focus, and increased accident risk. In recent years, extensive studies have demonstrated the potential of deep learning techniques for automatically predicting pain intensity by analyzing pain‐related facial expressions. However, two fundamental challenges remain in real‐world applications: the uncertainties arising from nonstandardized data collection and the class imbalance problem in pain assessment datasets. Addressing this issue is essential to improving recognition accuracy and robustness. This study proposes a novel pain level recognition framework based on a ResNet–BiLSTM architecture. First, the Squeeze‐and‐Excitation attention mechanism is integrated to enhance the model’s ability to capture basic pain‐related facial features, thereby increasing robustness against data uncertainties. Additionally, the I‐center loss function is introduced during model optimization, enhancing intraclass compactness and interclass separability to improve class‐imbalanced model learning. Experimental results on the BioVid Heat Pain Database demonstrate that this approach outperforms state‐of‐the‐art baselines, achieving 99.67% accuracy. Furthermore, ablation studies validate the effectiveness of the proposed components in our framework.
- Open Access
- 20 December 2025
- Published by FOREX Publication in International Journal of Electrical and Electronics Research
- Vol. 13 (4) , 820-829
Handling Class Imbalance in Video-Based Pain Intensity Estimation Using ResNetBiLSTM: A Study on Loss Functions and Optimizers
Accurate pain intensity recognition is vital for improving clinical care, especially in scenarios where patients cannot self-report. However, although existing datasets are often imbalanced, the main challenge is that current optimization and loss functions lack sensitivity to minority classes, adapt poorly to intra-class variability, and are less robust under imbalance, leading to biased recognition performance. To address these challenges, this study proposes a hybrid deep learning model based on ResNet-50 and BiLSTM to capture both spatial and temporal features from facial expression videos while incorporating strategies to mitigate the imbalance issue. To address the challenge of category imbalance inherent in real-world datasets, the study constructs an imbalanced version of the BioVid Part A database, reflecting realistic pain distribution with limited high-intensity samples. The study systematically compares three loss functions—Cross Entropy Loss, Focal Loss, and Cross Entropy Focal Loss (CEFL)—and evaluates their effectiveness in enhancing minority class recognition. A comprehensive benchmarking of four state-of-the-art optimization algorithms, AdamW, LAMB, AdaBelief, and NovoGrad, is conducted across a range of learning rates to systematically evaluate their convergence dynamics and generalization performance within the proposed framework. Experimental results show that the Focal Loss combined with LAMB or NovoGrad achieves superior performance, with the best accuracy reaching up to 84.89%, significantly outperforming traditional configurations. This research highlights the importance of tailored training strategies for imbalanced facial pain recognition and provides a robust baseline for future work. Future directions include expanding to real-world, in-the-wild pain assessment scenarios and integrating multimodal signals to enhance robustness and accuracy.
- 20 December 2025
- Published by FOREX Publication in International Journal of Electrical and Electronics Research
- Vol. 13 (4) , 820-829
Open AccessHandling Class Imbalance in Video-Based Pain Intensity Estimation Using ResNetBiLSTM: A Study on Loss Functions and Optimizers
Accurate pain intensity recognition is vital for improving clinical care, especially in scenarios where patients cannot self-report. However, although existing datasets are often imbalanced, the main challenge is that current optimization and loss functions lack sensitivity to minority classes, adapt poorly to intra-class variability, and are less robust under imbalance, leading to biased recognition performance. To address these challenges, this study proposes a hybrid deep learning model based on ResNet-50 and BiLSTM to capture both spatial and temporal features from facial expression videos while incorporating strategies to mitigate the imbalance issue. To address the challenge of category imbalance inherent in real-world datasets, the study constructs an imbalanced version of the BioVid Part A database, reflecting realistic pain distribution with limited high-intensity samples. The study systematically compares three loss functions—Cross Entropy Loss, Focal Loss, and Cross Entropy Focal Loss (CEFL)—and evaluates their effectiveness in enhancing minority class recognition. A comprehensive benchmarking of four state-of-the-art optimization algorithms, AdamW, LAMB, AdaBelief, and NovoGrad, is conducted across a range of learning rates to systematically evaluate their convergence dynamics and generalization performance within the proposed framework. Experimental results show that the Focal Loss combined with LAMB or NovoGrad achieves superior performance, with the best accuracy reaching up to 84.89%, significantly outperforming traditional configurations. This research highlights the importance of tailored training strategies for imbalanced facial pain recognition and provides a robust baseline for future work. Future directions include expanding to real-world, in-the-wild pain assessment scenarios and integrating multimodal signals to enhance robustness and accuracy.