Yanbin Gong

Yanbin Gong

Auto-generated profile, based on publicly available information

Yanbin Gong

CSE, The Hong Kong University of Science and Technology, Hong Kong SAR

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 August 2024 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
WXWentao Xie
Wentao Xie
CXChi Xu
Chi Xu
YGYanbin Gong
Yanbin Gong
YWYu Wang
Yu Wang
YLYuxin Liu
Yuxin Liu
SYShifang Yang
Shifang Yang
Practicing breathing exercises is crucial for patients with chronic obstructive pulmonary disease (COPD) to enhance lung function. Breathing mode (chest or belly breathing) and lung volume are two important metrics for supervising breathing exercises. Previous works propose that these metrics can be sensed separately in a contactless way, but they are impractical with unrealistic assumptions such as distinguishable chest and belly breathing patterns, the requirement of calibration, and the absence of body motions. In response, this research proposes DeepBreath, a novel depth camera-based breathing exercise assessment system, to overcome the limitations of the existing methods. DeepBreath, for the first time, considers breathing mode and lung volume as two correlated measurements and estimates them cooperatively with a multitask learning framework. This design boosts the performance of breathing mode classification. To achieve calibration-free lung volume measurement, DeepBreath uses a data-driven approach with a novel UNet-based deep-learning model to achieve one-model-fit-all lung volume estimation, and it is designed with a lightweight silhouette segmentation model with knowledge transferred from a state-of-the-art large segmentation model that enhances the estimation performance. In addition, DeepBreath is designed to be resilient to involuntary motion artifacts with a temporal-aware body motion compensation algorithm. We collaborate with a clinical center and conduct experiments with 22 healthy subjects and 14 COPD patients to evaluate DeepBreath. The experimental result shows that DeepBreath can achieve high breathing metrics estimation accuracy but with a much more realistic setup compared with previous works.

3 November 2025
Verified
Conference Paper
Open access
CXChi Xu
Chi Xu
WXWentao Xie
Wentao Xie
BYBaichen Yang
Baichen Yang
YZYizhen Zhang
Yizhen Zhang
YGYanbin Gong
Yanbin Gong
QZQian Zhang
Qian Zhang

9 June 2025 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
YGYanbin Gong
Yanbin Gong
WXWentao Xie
Wentao Xie
CXChi Xu
Chi Xu
QZQian Zhang
Qian Zhang
SYShifang Yang
Shifang Yang
Airway clearance is essential for managing Muco-Obstructive Lung Diseases (MOLDs). Percussion, a widely used airway clearance technique (ACT) in community and home care settings, is favored for its ease of implementation compared to other complex techniques. However, percussion is time-consuming and physically demanding for both caregivers and patients, as caregivers typically perform percussion on the entire back to avoid missing accumulated sputum when its exact location is unknown. Therefore, accurate sputum localization can significantly enhance the percussion experience. Current clinical methods for sputum localization typically rely on imaging techniques, which are costly, expose patients to radiation, and are usually performed only once during diagnosis, thereby limiting their application to inpatient settings. Alternatively, some medical professionals combine auscultation with other clinical assessments, but this approach requires substantial clinical experience and is impractical for community or home care settings where medical experts are unavailable. To address these limitations, we introduce SputumLocator, an innovative sputum localization system based on digital stethoscopes. SputumLocator leverages standard auscultation procedures to detect accumulated sputum in the four quadrants of the back, which is straightforward and highly practical. SputumLocator comprises two components: SputumEmbedder, which extracts key abnormal sounds and their spatial features using a Transformer-based feature extractor, and SputumClassifier, which maps these features to determine sputum presence in each region via a Convolutional Block Attention Module (CBAM). Given the limited availability of annotated sputum data, we developed a pretraining method based on Embedding on Masked Data (EOM) and enhanced model robustness through a Teacher-Student Architecture (TSA) that integrates noisy data. In collaboration with a medical institution, we evaluate SputumLocator on 43 patients with diverse physiological characteristics and under varying recording conditions. Experimental results demonstrate that SputumLocator achieves high accuracy with an overall sensitivity of 0.97, specificity of 0.82, and F1-Score of 0.83, maintaining robustness across different thoracic regions, genders, and disease types.

26 June 2023 in Digital Health
Verified
Journal Article
Open access
JQJia Qiao
Jia Qiao
YJYuan-Tong Jiang
Yuan-Tong Jiang
YDYong Dai
Yong Dai
YGYan-Bin Gong
Yan-Bin Gong
MDMeng Dai
Meng Dai
ZDZu-Lin Dou
Zu-Lin Dou
Objective: This study aims to establish a real-time dynamic monitoring system for silent aspiration (SA) to provide evidence for the early diagnosis of and precise intervention for SA after stroke. Methods: Multisource signals, including sound, nasal airflow, electromyographic, pressure and acceleration signals, will be obtained by multisource sensors during swallowing events. The extracted signals will be labeled according to videofluoroscopic swallowing studies (VFSSs) and input into a special dataset. Then, a real-time dynamic monitoring model for SA will be built and trained based on semisupervised deep learning. Model optimization will be performed based on the mapping relationship between multisource signals and insula-centered cerebral cortex–brainstem functional connectivity through resting-state functional magnetic resonance imaging. Finally, a real-time dynamic monitoring system for SA will be established, of which the sensitivity and specificity will be improved by clinical application. Results: Multisource signals will be stably extracted by multisource sensors. Data from a total of 3200 swallows will be obtained from patients with SA, including 1200 labeled swallows from the nonaspiration category from VFSSs and 2000 unlabeled swallows. A significant difference in the multisource signals is expected to be found between the SA and nonaspiration groups. The features of labeled and pseudolabeled multisource signals will be extracted through semisupervised deep learning to establish a dynamic monitoring model for SA. Moreover, strong correlations are expected to be found between the Granger causality analysis (GCA) value (from the left middle frontal gyrus to the right anterior insula) and the laryngeal rise time (LRT). Finally, a dynamic monitoring system will be established based on the former model, by which SA can be identified precisely. Conclusion: The study will establish a real-time dynamic monitoring system for SA with high sensitivity, specificity, accuracy and F1 score.

16 March 2026 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
QHQingyong Hu
Qingyong Hu
YZYuxuan Zhou
Yuxuan Zhou
JWJinjian Wang
Jinjian Wang
YGYanbin Gong
Yanbin Gong
YZYizhen Zhang
Yizhen Zhang
GLGuihua Li
Guihua Li
As the global population ages, the prevalence of cognitive impairment continues to rise, highlighting the urgent need for accessible and low-burden cognitive assessment. Current assessments based on clinical scales are often hindered by subjectivity, significant user burden, and practice effects, limiting their applicability. We observe that passive visual stimuli can engage multiple cognitive domains while minimizing the need for active participation. Considering the lack of related datasets, we establish EasyCog, the first large-scale multimodal dataset for low-burden cognitive assessments. EasyCog collects synchronized forehead/ear EEG and contactless eye tracking data while participants passively view a short, cognitively structured video followed by an eyes-closed rest. The dataset includes 101 participants spanning healthy controls and patients with PD, AD, and VaD, with clinician-administered MoCA/MMSE scores collected in daily settings. We provide detailed collection procedures, quality validation, implementation, and benchmark baselines. Results indicate assessment feasibility while highlighting generalization challenges. By integrating passive visual stimuli with affordable sensing, EasyCog provides a foundation for future research in accessible and scalable cognitive monitoring in both clinical and community settings.

30 September 2026 in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Verified
Journal Article
Open access
CXChi Xu
Chi Xu
WXWentao Xie
Wentao Xie
HTHuaning Tan
Huaning Tan
ZHZirui Huang
Zirui Huang
YGYanbin Gong
Yanbin Gong
QZQian Zhang
Qian Zhang
Cardiorespiratory fitness (CRF) is crucial for maintaining overall health, with VO 2 max (maximal oxygen consumption) and VO 2 AT (oxygen consumption at anaerobic threshold) serving as key physiological indicators. The current gold standard for assessing these metrics is Cardiopulmonary Exercise Testing (CPET), which requires bulky equipment, trained personnel, and significant time, putting it out of reach for most people. In this study, we introduce CaReFit, an innovative and low-cost system that utilizes a single-lead electrocardiogram (ECG) and muscle oxygenation (SmO 2 ) sensors to estimate VO 2 max and VO 2 AT. Our method leverages the coupled behavior of ECG and SmO 2 signals during exercise to predict these critical performance metrics. To mitigate exercise-induced noise in ECG signals, we implement a diffusion-based ECG denoising model trained with a tailored ECG-guided loss function on combined public datasets. Then, our multimodal neural network is specifically designed to leverage physiological insights for accurate CRF predictions. We conducted a deployment of CaReFit with 50 patients in a clinical setting, and evaluated CaReFit under a strict subject-disjoint protocol. Under this leakage-free evaluation, CaReFit achieves a mean absolute percentage error (MAPE) of 9.30% (95% CI: 7.06-11.80) for VO 2 max and 9.49% (95% CI: 7.02-12.12) for VO 2 AT. We further conduct a head-to-head comparison with a commercial portable device, and our system demonstrates similar VO 2 max performance trends under the same session protocol while additionally supporting VO 2 AT estimation. Furthermore, we make our dataset and evaluation code publicly available to facilitate further research and development in this domain.