SZShusen Zhou

Shusen Zhou

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Shusen Zhou has a total of 27 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • MRAST-Net: A Lightweight Multi-Receptive-Field Residual Temporal Network for 12-Lead ECG Classification

    by Mingliang Wang, Mujun Zang, Tong Liu, Shusen Zhou, Chanjuan Liu, Qingjun Wang

    Electrocardiogram (ECG) signals play an important role in cardiovascular disease screening, clinical monitoring, and remote healthcare. However, existing high-performing ECG classification methods usually rely on complex deep learning architectures, which often involve a large number of parameters, high computational cost, and limited deployability in resource-constrained scenarios. To balance classification performance and lightweight deployment, this paper proposes MRAST-Net, a lightweight multi-receptive-field residual temporal network for 12-lead ECG classification. The proposed model first employs a lead mixing layer to enable cross-lead information interaction. It then uses a lightweight multi-scale residual depthwise convolution module to extract local waveform morphology features under different receptive fields, while an efficient channel attention mechanism is introduced to enhance informative channel responses. On this basis, a temporal convolutional network is further incorporated to model long-range dynamic dependencies, and attentive statistics pooling is used to aggregate global statistical features for final multi-label classification. Experiments on two public 12-lead ECG datasets, PTB-XL and CPSC-2018, show that the proposed model achieves AUC scores of 93.12 $\pm$ 0.16\%, 92.29 $\pm$ 0.80\%, and 92.62 $\pm$ 0.51\%, and ACC scores of 89.17 $\pm$ 0.09\%, 96.40 $\pm$ 0.02\%, and 97.72 $\pm$ 0.07\% on the 5-superclass, 23-subclass, and 44-diagnostic-class tasks of PTB-XL, respectively. On CPSC-2018, MRAST-Net achieves an AUC of 96.57 $\pm$ 0.43\% and an ACC of 96.59 $\pm$ 0.26\% under 5$\times$3 nested cross-validation. These results demonstrate that MRAST-Net achieves competitive classification performance while maintaining a low parameter scale, indicating its potential for deployment in resource-constrained scenarios such as portable monitoring devices, real-time assisted diagnosis, and primary healthcare.

  • KDM-Net: a large-kernel and bi-mamba integrated network for 12-lead atrial arrhythmia classification

    by Yinghao Sun, Mujun Zang, Tong Liu, Shusen Zhou, Chanjuan Liu, Qingjun Wang

    Objective. Automatic recognition of atrial arrhythmias is important for early clinical screening and computer-aided diagnosis. Distinguishing atrial fibrillation from atrial flutter remains challenging because of their similar electrocardiographic manifestations and high risk of misclassification. This study aimed to develop an effective model for atrial arrhythmia classification with improved multi-lead feature utilization, long-range temporal modeling, and cross-patient generalization. Approach. We propose KDM-Net, which takes 12-lead electrocardiogram signals as input. The network employs a large-kernel K module to enhance the perception of waveform morphology and rhythm patterns, a D module to recalibrate and fuse multi-lead features, and a bidirectional Mamba-based M module to model long-range temporal dependencies. KDM-Net was evaluated on the combined CUSPH and PTB-XL dataset under an inter-patient protocol. Main results. In five-fold cross-validation, KDM-Net achieved an average accuracy and F1-score of 96.04% and 89.21%, respectively. On the independent patient-level test set, the accuracy and F1-score were 96.41% and 88.31%, respectively. On the CinC 2020 external test set, the model achieved an accuracy and F1-score of 91.14% and 80.41%, respectively. KDM-Net outperformed several representative models in comparative experiments. Significance. KDM-Net effectively integrates multi-lead spatial information and long-range temporal dependencies, demonstrating strong recognition performance and promising cross-patient generalization. It provides an effective solution for the intelligent diagnosis of atrial arrhythmias.

  • FBDA ‐Net: Medical Image Segmentation via Foreground–Background Decoupled Attention Aggregation

    by Honggang Xie, Tong Liu, Mujun Zang, Shusen Zhou, Chanjuan Liu, Qingjun Wang

    Medical image segmentation is a fundamental task in computer‐aided diagnosis, playing an important role in organ measurement, lesion delineation, and surgical planning. However, due to the limitations of imaging conditions and the complexity of tissue characteristics, foreground targets and the background often exhibit a “soft boundary,” and the two are easily confused in ambiguous transition regions, which limits further improvement of segmentation accuracy. To address this, we propose FBDA‐Net, a segmentation network based on foreground–background decoupled attention (FBDA). Its core module, the FBDA, decouples the input features into foreground and background at the feature level using prior masks, computes attention with a shared Query and three independent sets of Key/Value, and employs a joint Softmax so that the foreground, background, and global branches competitively allocate weights within the same probability space. Building upon this, we construct a top‐down progressive decoupling decoder that converts the prediction at each level into prior masks and passes them downward layer by layer, driving the decoupled priors to be progressively reconstructed and refined along the decoding process, and we adopt a combinatorial supervision loss to strengthen the cross‐level consistency of multi‐level predictions. Experiments on three benchmark datasets—Synapse multi‐organ CT, ACDC cardiac MRI, and ISIC 2016 dermoscopy—demonstrate competitive overall segmentation performance. Zero‐shot external evaluation on PH2 without fine‐tuning further supports the cross‐dataset generalization of FBDA‐Net. The source code is publicly available at https://github.com/TongLiuLDU/FBDANet.git .

  • Fourier-Enhanced Kolmogorov–Arnold Network With Attention for Drug–Target Interaction Prediction

    by Jinlong Li, Shusen Zhou, Chanjuan Liu, Qingjun Wang, Tong Liu, Mujun Zang

    Identifying drug–target interactions (DTI) is a fundamental yet costly step in drug discovery, motivating the development of accurate and efficient computational prediction methods. In this paper, we propose FKAN-a, a DTI prediction framework that integrates Fourier-enhanced Kolmogorov–Arnold networks (KAN) with attention mechanisms under a contrastive learning paradigm. Drug molecules and protein sequences are preprocessed and encoded using pretrained representations to capture interaction-relevant features. The resulting embeddings are further transformed through KAN with learnable Fourier bases to model complex nonlinear relationships. A cross-modality attention module is introduced to enhance the modeling of fine-grained drug–protein associations. Experiments conducted on three public benchmark datasets demonstrate that FKAN-a consistently outperforms representative state-of-the-art methods in terms of prediction performance and computational efficiency. These results indicate that the proposed framework provides an effective solution for DTI prediction and practical candidate prioritization in drug discovery.


  • Deep contrastive learning method for drug-target interactions prediction

    by Jinlong Li, Shusen Zhou, Tong Liu, Chanjuan Liu, Mujun Zang, Qingjun Wang

    1. Drug development is a lengthy process. In its initial stages, it is necessary to identify certain compounds from a vast array of chemicals that can potentially interact with the target as drug c...


  • GDT-Net: Multi-level feature extraction network for precise diagnosis of atrial and ventricular fibrillation

    by Zhihao Wang, Mujun Zang, Tong Liu, Shusen Zhou, Chanjuan Liu, Qingjun Wang

  • Multiple instance learning method based on convolutional neural network and self-attention for early cancer detection

    by Junjiang Liu, Shusen Zhou, Mujun Zang, Chanjuan Liu, Tong Liu, Qingjun Wang

    Early cancer detection using T-cell receptor sequencing (TCR-seq) and multiple instances learning methods has shown significant effectiveness. We introduce a multiple instance learning method based on convolutional neural networks and self-attention (MICA). First, MICA preprocesses TCR-seq using word vectors and then extracts features using convolutional neural networks. Second, MICA uses an enhanced self-attention mechanism to extract relational features of instances. Finally, MICA can extract the crucial TCR-seq. After cross-validation, MICA achieves an area under the curve (AUC) of 0.911 and 0.946 on the lung and thyroid cancer datasets, which are 7.1% and 2.1% higher than other methods, respectively.


  • Parallel Convolutional Contrastive Learning Method for Enzyme Function Prediction

    by Xindi Yu, Shusen Zhou, Mujun Zang, Qingjun Wang, Chanjuan Liu, Tong Liu

    The function labeling of enzymes has a wide range of application value in the medical field, industrial biology and other fields. Scientists define enzyme categories by enzyme commission (EC) numbers. At present, although there are some tools for enzyme function prediction, their effects have not reached the application level. To improve the precision of enzyme function prediction, we propose a parallel convolutional contrastive learning (PCCL) method to predict enzyme functions. First, we use the advanced protein language model ESM-2 to preprocess the protein sequences. Second, PCCL combines convolutional neural networks (CNNs) and contrastive learning to improve the prediction precision of multifunctional enzymes. Contrastive learning can make the model better deal with the problem of class imbalance. Finally, the deep learning framework is mainly composed of three parallel CNNs for fully extracting sample features. we compare PCCL with state-of-art enzyme function prediction methods based on three evaluation metrics. The performance of our model improves on both two test sets. Especially on the smaller test set, PCCL improves the AUC by 2.57%. The source code can be downloaded from https://github.com/biomg/PCCL