Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Dakun Lai has a total of 167 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    DCPA-SNN, Direct-Coding-Physics-Aware Spiking Neural Network: A Framework for Wearable ECG Denoising Under Dynamic-Noise Conditions

    by Yukun Ren, Hongyou Zuo, Yuhang Cai, Shenghua Wang, Guihao Ran, Dakun Lai

    Smart wearable electrocardiogram (ECG) monitoring enables continuous cardiac assessment beyond clinical settings, but ECG signals are often degraded by baseline wander, muscle artifacts, and electrode motion artifacts. At present, pure end-to-end spiking neural networks (SNNs), with the advantage of low computational complexity, have rarely been explored for ECG noise suppression, particularly under dynamic conditions. To address this gap, this study proposes Direct-Coding-Physics-Aware Spiking Neural Network (DCPA-SNN) for wearable ECG denoising. The proposed method integrates a direct-coding SNN, channel attention, residual noise learning, and a physics-aware multi-domain loss function to preserve diagnostically important waveform structures. Clean ECG signals from the MIT-BIH Arrhythmia Database and real-noise segments from the MIT-BIH Noise Stress Test Database were used to construct single-noise and mixed-noise evaluation scenarios with input SNRs ranging from −6 dB to 4 dB to reflect the noise characteristics of wearable devices. Experimental results demonstrate that DCPA-SNN achieves robust denoising performance under different noise conditions. In the mixed-noise scenario, which serves as the primary evaluation setting of this study, the average denoised SNR reached 5.80 dB, with an average SNR improvement of 6.80 dB, while the R-peak detection rate increased from 90.71% to 95.72%. These results demonstrate that the proposed model, DCPA-SNN, provides a promising approach for wearable ECG denoising with potential for low-power deployment.

  • Open Access

    Physiological Monitoring Applications of Wearable Multimodal Fusion Systems Based on ECG and PPG: A Comprehensive Review

    by Chamod Rathnayake, Wenjing Chen, Sahan Jayawickrama, Dakun Lai

    Wearable technology has become popular today not only for clinical uses but also among the general public for everyday activities such as fitness tracking and activity recognition. It’s no secret that wearable devices need to be more comfortable and that there is increasing attention to their reliability and accuracy. Therefore, researchers are making various attempts to improve these specifications. Recent research has employed multimodal fusion, rather than relying on results from a single source, such as electrocardiography (ECG) or photoplethysmography (PPG), because it enables a more comprehensive understanding of physical conditions. There are very few extensive reviews of ECG and PPG wearable multimodal fusion application systems, and they are limited to a single practical use. However, this review fills that gap and provides a comprehensive overview of recent wearable physiological monitoring technologies based on ECG and PPG, along with their modern multimodal fusion applications in the medical field. These include cuffless blood pressure estimation, stress monitoring, heart rate and heart rate variability monitoring. This presents the theoretical background, including the characteristics of ECG and PPG signals, recently developed wearable monitoring techniques for ECG and PPG, and the advantages and disadvantages of multimodal fusion. It also provides a comprehensive and comparative analysis of recent studies employing modern multimodal fusion approaches in the aforementioned medical field, as well as a discussion of the limitations and challenges of ECG-PPG wearable multimodal fusion systems reported in the literature. Therefore, this review will enable researchers to gain a complete and comprehensive understanding of the development of wearable multimodal fusion applications based on ECG and PPG.


  • Open Access

    UDC-SNN: An Uncertainty-Aware Dynamic Cascading Framework with Spiking Neural Network for Balancing Performance and Energy in Multimodal Emotion Recognition

    by Guihao Ran, Shengzhe Li, Zhiwen Jiang, Han Zhang, Xinyuan Long, Dakun Lai

    The aim of this study is to propose an uncertainty-aware dynamic cascading framework based on spiking neural network (UDC-SNN) for multimodal emotion recognition, particularly to address the inherent trade-off between recognition performance and energy efficiency. An asymmetric dynamic routing mechanism was proposed to enable demand-driven activation of the high-power electroencephalogram (EEG) branch, coupled with preliminary inference on a low-power electrocardiogram (ECG) branch and uncertainty quantification via Shannon entropy. Meanwhile, a parameter-free log-linear aggregation strategy was developed to transform modality-specific entropy into dynamic Bayesian weights through an exponential decay function, effectively mitigating the negative transfer effects induced by unimodal noise. The UDC-SNN was evaluated on the multimodal affective dataset DREAMER, comprising 23 subjects (170,660 segments). The averaged recognition accuracy and energy consumption across the three dimensions of valence, arousal, and dominance were 90.75% and 4.62 μJ, respectively. The obtained results suggest that the proposed framework could potentially achieve a favorable balance between high emotion recognition and low energy consumption, thereby establishing its applicability for real-time monitoring in resource-constrained scenarios.


  • Multimodal ECG–PPG Wearable Technologies and Modern Fusion Methods for Atrial Fibrillation Detection: A Review

    by Chamod Rathnayake, Wenjing Chen, Guihao Ran, Han Zhang, B. L. Sanjaya Thilakarathne, Jamal- E- Fatima, Dakun Lai

    Atrial fibrillation (AF) is recognized as the most common and clinically significant cardiac arrhythmia worldwide. Early detection of atrial fibrillation is essential to prevent serious cardiovascular complications such as stroke, heart attack, heart failure, cardiomyopathy, and systemic thromboembolism. Noninvasive methods such as electrocardiography (ECG) and photoplethysmography (PPG) are used to detect AF, and wearable technologies have expanded the opportunities for continuous AF monitoring. However, each of these methods faces limitations in real-world environments, including ECG electrode noise and PPG motion artifacts, which reduce accuracy when used independently. In recent years, researchers have focused on multimodal ECG–PPG fusion approaches that combine electrical and mechanical cardiac signals, as these enable more advanced AF detection under various conditions, including noise, motion, and real-life scenarios. The novelty of this review is its focus on modern ECG-PPG fusion systems, which demonstrate that combining ECG and PPG is more effective for detecting AF than using either signal alone. This addresses an important research gap in AF diagnosis. This review fills this gap by covering the physiology of ECG and PPG, their features in AF, modern ECG–PPG wearable designs, sensor technologies, fusion pipelines, real-world challenges, and future directions. Therefore, this review is intended to be of significant value to researchers developing wearable multimodal fusion systems or other ECG-and PPG-based applications for atrial fibrillation detection.


  • [Automatic detection and visualization of myocardial infarction in electrocardiograms based on an interpretable deep learning model].

    by Yue Zhang, Yifei Zhang, Baojie Xie, Dakun Lai

    Automated detection of myocardial infarction (MI) is crucial for preventing sudden cardiac death and enabling early intervention in cardiovascular diseases. This paper proposes a deep learning framework based on a lightweight convolutional neural network (CNN) combined with one-dimensional gradient-weighted class activation mapping (1D Grad-CAM) for the automated detection of MI and the visualization of key waveform features in single-lead electrocardiograms (ECGs). The proposed method was evaluated using a total of 432 records from the Physikalisch-Technische Bundesanstalt Diagnostic ECG Database (PTBDB) and the Normal Sinus Rhythm Database (NSRDB), comprising 334 MI and 98 normal ECGs. Experimental results demonstrated that the model achieved an accuracy, sensitivity, and specificity of 95.75%, 96.03%, and 95.47%, respectively, in MI detection. Furthermore, the visualization results indicated that the model's decision-making process aligned closely with clinically critical features, including pathological Q waves, ST-segment elevation, and T-wave inversion. This study confirms that the proposed deep learning algorithm combined with explainable technology performs effectively in the intelligent diagnosis of MI and the visualization of critical ECG waveforms, demonstrating its potential as a useful tool for early MI risk assessment and computer-aided diagnosis.

  • Open Access

    Lightweight Fused Siamese CNN for Multi-Modal Cardiac Anomaly Detection from ECG and PCG Spectrograms

    by Jiahui Cheng, Yukun Ren, Hongyou Zuo, Han Zhang, Dakun Lai

    Electrocardiogram (ECG) and phonocardiogram (PCG) are critical for the early prevention and diagnosis of cardiovascular diseases (CVDs). With the rapid advancement of deep learning techniques, automated CVD detection from these signals has attracted considerable research attention. However, existing methods are predominantly based on unimodal data, failing to leverage the complementary electrophysiological and mechanical information. Furthermore, current multi-modal fusion strategies, often employing separate encoders, suffer from parameter redundancy, high computational complexity, and ineffective cross-modal feature alignment in the latent space. It is therefore crucial to develop an efficient multi-modal learning method for accurate and deployable CVD screening. In this study, we propose a novel Lightweight Fused Siamese Convolutional Neural Network (LFS-CNN) for cardiac anomaly detection from synchronized ECG and PCG signals. Our model employs a weight-shared siamese backbone to process time-frequency representations of both modalities, facilitating the learning of modality-invariant features while dramatically reducing parameters. An R-peak anchored data augmentation strategy is also designed to ensure physiological consistency. Experimental results on the PhysioNet/CinC Challenge 2016 dataset demonstrate that the performance of our proposed LFS-CNN is superior to those of unimodal methods and conventional fusion alternatives. Our fused model achieves an F1-score of 97.02%, an accuracy of 95.83%, and a sensitivity of 97.13%, while significantly enhancing the specificity of the PCG model from 62.50% to 92.50%, validating its effectiveness and strong potential for deployment in wearable cardiac monitoring systems.

  • Open Access

    Balancing Energy Consumption and Detection Accuracy in Cardiovascular Disease Diagnosis: A Spiking Neural Network-Based Approach with ECG and PCG Signals

    by Guihao Ran, Yijing Wang, Han Zhang, Jiahui Cheng, Dakun Lai

    Electrocardiogram (ECG) and phonocardiogram (PCG) signals are widely used in the early prevention and diagnosis of cardiovascular diseases (CVDs) due to their ability to accurately reflect cardiac conditions from different physiological perspectives and their ease of acquisition. Currently, some studies have explored the joint use of ECG and PCG signals for disease screening, but few studies have considered the trade-off between classification performance and energy consumption in model design. In this study, we propose a multimodal CVDs detection framework based on Spiking Neural Networks (SNNs), which integrates ECG and PCG signals. A differential fusion strategy at the signal level is employed to generate a fused EPCG signal, from which time–frequency features are extracted using the Adaptive Superlets Transform (ASLT). Two separate Spiking Convolutional Neural Network (SCNN) models are then trained on the ECG and EPCG signals, respectively. A confidence-based dynamic decision-level (CDD) fusion strategy is subsequently employed to perform the final classification. The proposed method is validated on the PhysioNet/CinC Challenge 2016 dataset, achieving an accuracy of 89.74%, an AUC of 89.08%, and an energy consumption of 209.6 μJ. This method not only achieves better balancing performance compared to unimodal signals but also realizes an effective balance between model energy consumption and classification effect, which provides an effective idea for the development of low-power, multimodal medical diagnostic systems.


  • Open Access

    Unsupervised detection of high-frequency oscillations in intracranial electroencephalogram: promoting a valuable automated diagnostic tool for epilepsy

    by Wenjing Chen, Tongzhou Kang, Belal Bin Heyat, Jamal E. Fatima, Yuanning Xu, Dakun Lai

    Objective: This study aims to develop an unsupervised automated method for detecting high-frequency oscillations (HFOs) in intracranial electroencephalogram (iEEG) signals, addressing the limitations of manual detection processes.Method: The proposed method utilizes an unsupervised convolutional variational autoencoder (CVAE) model in conjunction with the short-term energy method (STE) to analyze two-dimensional time-frequency representations of iEEG signals. Candidate HFOs are identified using STE and transformed into time-frequency maps using the continuous wavelet transform (CWT). The CVAE model is trained for dimensionality reduction and feature reconstruction, followed by clustering of the reconstructed maps using the K-means algorithm for automated HFOs detection.Results: Evaluation of the proposed unsupervised method on clinical iEEG data demonstrates its superior performance compared to traditional supervised models. The automated approach achieves an accuracy of 93.02%, sensitivity of 94.48%, and specificity of 92.06%, highlighting its efficacy in detecting HFOs with high accuracy.Conclusion: The unsupervised automated method developed in this study offers a reliable and efficient solution for detecting HFOs in iEEG signals, overcoming the limitations of manual detection processes of traditional supervised models. By providing clinicians with a clinically useful diagnostic tool, this approach holds promise for enhancing surgical resection planning in epilepsy patients and improving patient outcomes.