ZFZhiwen Fang

Zhiwen Fang

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

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


Zhiwen Fang has a total of 51 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Dynamic positron emission tomography image reconstruction using spatiotemporal kernel method with deep image prior

    by Zhijun Zhao, Haoyu Zou, Da Liang, Han Liu, Zhiwen Fang, Qiyu Peng

    Background: The inherently ill-posed nature of tomographic reconstruction, combined with the low photon counts in short acquisition frames, poses significant challenges for achieving high-quality dynamic positron emission tomography (PET) images. Recent studies have shown that deep learning methods, particularly the deep image prior (DIP) framework, can effectively enhance PET image reconstruction without requiring external training data. This study aimed to develop a novel DIP-based reconstruction algorithm integrated with a spatiotemporal kernel and to validate its effectiveness in improving the quality of reconstructed dynamic PET images. Methods: We constructed a novel dynamic PET reconstruction algorithm that integrates a spatiotemporal kernel with a DIP framework. Specifically, the reconstruction objective function was formulated as a constrained optimization problem. This problem was solved using the alternating direction method of multipliers (ADMM), alternating between image reconstruction and network parameter updates. Additionally, the algorithm supports list-mode reconstruction, enabling full 3-dimensional (3D) imaging and scalability to large PET systems. Results: Simulation and preclinical studies demonstrated that the proposed method substantially improved image quality compared to conventional methods such as maximum-likelihood expectation maximization (MLEM) with Gaussian post-filtering, kernel expectation maximization (KEM), spatiotemporal kernel expectation maximization (STKEM), as well as deep learning-based methods including DIPRecon and NeuralKEM. It achieved higher signal-to-noise ratio (SNR) and structural similarity index (SSIM), demonstrated good stability during the iterative reconstruction process, and attained a comparable level of contrast recovery coefficient (CRC), while effectively mitigating noise amplification in later iterations. Importantly, the method also demonstrated robust performance under low-count conditions, preserving image quality when conventional methods degrade significantly. Conclusions: By leveraging intrinsic spatiotemporal information, the proposed method improves dynamic PET reconstruction accuracy without external priors. Its modular design enables seamless integration into existing workflows and adapts to diverse PET acquisition protocols. These results underscore its potential for clinical and preclinical dynamic PET imaging, particularly for low-dose or high-resolution scenarios.

  • Multilingual-Prompt-Guided Directional Feature Learning for Weakly Supervised Video Anomaly Detection

    by Chizhuo Xiao, Yang Xiao, Joey Tianyi Zhou, Zhiwen Fang

    Weakly supervised video anomaly detection has gained attention for its effective performance and cost-efficient annotation, using video-level labels to distinguish between normal and abnormal patterns. However, challenges arise from the diversity and incompleteness of anomalous events, complicating feature learning. Vision-language models offer promising approaches, but designing precise prompts remains difficult. This is because accommodating the diverse range of normal and anomalous scenarios in real-world settings is challenging, and the workload is significant. To tackle these issues, we propose integrating multilingualism and multiple prompts to improve feature learning. By utilizing prompts in various languages to define “anomaly” and “normalcy,” we tackle these concepts across different linguistic domains. In each domain, multiple prompts are employed for adaptive top-K prompt selection of snippets. To enhance visual feature learning, a multi-granularity attention module combining Transformer and Mamba is designed. Mamba’s long-range adaptation selection builds fine-grained temporal correlations among coarse-grained snippets, while Transformer enhances fine-grained information guided by coarse-grained information. Alongside a multilingual prompt guidance loss, we introduce a gradual directional loss to jointly optimize visual feature distribution and the top-K prompt selection. Our method demonstrates effectiveness on four video datasets and provides generalizability analyses on two medical datasets, including EMG and ECG temporal data.


  • CrossGLG: LLM Guides One-Shot Skeleton-Based 3D Action Recognition in a Cross-Level Manner

    by Tingbing Yan, Wenzheng Zeng, Yang Xiao, Xingyu Tong, Bo Tan, Zhiwen Fang, Zhiguo Cao, Joey Tianyi Zhou


  • Rhythmer: Ranking-Based Skill Assessment With Rhythm-Aware Transformer

    by Zhuang Luo, Yang Xiao, Feng Yang, Joey Tianyi Zhou, Zhiwen Fang

    Ranking-based skill assessment is an essential component of video understanding. In this task lacking precise procedure annotations, existing methods place greater emphasis on evaluating the procedure quality via manually normalizing the execution duration. However, the inherent duration-related procedural patterns will undergo alteration. Experimentally, we discover that distinct duration biases are prevalent in duration-sensitive skills, such as those in medical and everyday life. Hence, duration information is crucial for ranking-based skill assessment when dealing with varying durations. Additionally, similar execution processes tend to have closer execution durations. Thus, another critical factor lies in extracting duration-related procedural information alongside similar durations. It is defined as mining rhythm patterns, which are inspired by music rhythms including various duration and duration-related procedures. In our work, a rhythm-aware transformer is proposed to mine the rhythm patterns adaptively. Given pairwise inputs, a coattention module is designed to mutually highlight duration-related procedure information when comparing pairwise input videos with similar durations, and adaptively attenuate the efficacy when confronted with pairwise inputs featuring significantly different durations. A rhythm-encoding module further embeds duration information into the concatenation of raw features and co-attention features. Following these features, the transformer decoder is designed to learn duration-related queries supervised by a novel duration grouping loss among various duration groups. The experimental results demonstrate that the rhythm-aware transformer is effective for ranking-based skill assessment.


  • Open Access

    CrossGLG: LLM Guides One-shot Skeleton-based 3D Action Recognition in a Cross-level Manner

    by Tingbing Yan, Wenzheng Zeng, Yang Xiao, Xingyu Tong, Bo Tan, Zhiwen Fang, Zhiguo Cao, Joey Tianyi Zhou

    Most existing one-shot skeleton-based action recognition focuses on raw low-level information (e.g., joint location), and may suffer from local information loss and low generalization ability. To alleviate these, we propose to leverage text description generated from large language models (LLM) that contain high-level human knowledge, to guide feature learning, in a global-local-global way. Particularly, during training, we design $2$ prompts to gain global and local text descriptions of each action from an LLM. We first utilize the global text description to guide the skeleton encoder focus on informative joints (i.e.,global-to-local). Then we build non-local interaction between local text and joint features, to form the final global representation (i.e., local-to-global). To mitigate the asymmetry issue between the training and inference phases, we further design a dual-branch architecture that allows the model to perform novel class inference without any text input, also making the additional inference cost neglectable compared with the base skeleton encoder. Extensive experiments on three different benchmarks show that CrossGLG consistently outperforms the existing SOTA methods with large margins, and the inference cost (model size) is only $2.8$\% than the previous SOTA. CrossGLG can also serve as a plug-and-play module that can substantially enhance the performance of different SOTA skeleton encoders with a neglectable cost during inference. The source code will be released soon.

  • C$^{2}$Net: content-dependent and -independent cross-attention network for anomaly detection in videos

    by Jiafei Liang, Yang Xiao, Joey Tianyi Zhou, Feng Yang, Ting Li, Zhiwen Fang


  • TaiChiNet: Negative-Positive Cross-Attention Network for Breast Lesion Segmentation in Ultrasound Images

    by Jinting Wang, Jiafei Liang, Yang Xiao, Joey Tianyi Zhou, Zhiwen Fang, Feng Yang

    Breast lesion segmentation in ultrasound images is essential for computer-aided breast-cancer diagnosis. To improve the segmentation performance, most approaches design sophisticated deep-learning models by mining the patterns of foreground lesions and normal backgrounds simultaneously or by unilaterally enhancing foreground lesions via various focal losses. However, the potential of normal backgrounds is underutilized, which could reduce false positives by compacting the feature representation of all normal backgrounds. From a novel viewpoint of bilateral enhancement, we propose a negative-positive cross-attention network to concentrate on normal backgrounds and foreground lesions, respectively. Derived from the complementing opposites of bipolarity in TaiChi, the network is denoted as TaiChiNet, which consists of the negative normal-background and positive foreground-lesion paths. To transmit the information across the two paths, a cross-attention module, a complementary MLP-head, and a complementary loss are built for deep-layer features, shallow-layer features, and mutual-learning supervision, separately. To the best of our knowledge, this is the first work to formulate breast lesion segmentation as a mutual supervision task from the foreground-lesion and normal-background views. Experimental results have demonstrated the effectiveness of TaiChiNet on two breast lesion segmentation datasets with a lightweight architecture. Furthermore, extensive experiments on the thyroid nodule segmentation and retinal optic cup/disc segmentation datasets indicate the application potential of TaiChiNet.


  • Multi-Template Meta-Information Regularized Network for Alzheimer’s Disease Diagnosis Using Structural MRI

    by Kangfu Han, Gang Li, Zhiwen Fang, Feng Yang

    Structural magnetic resonance imaging (sMRI) has been widely applied in computer-aided Alzheimer’s disease (AD) diagnosis, owing to its capabilities in providing detailed brain morphometric patterns and anatomical features in vivo. Although previous works have validated the effectiveness of incorporating metadata (e.g., age, gender, and educational years) for sMRI-based AD diagnosis, existing methods solely paid attention to metadata-associated correlation to AD (e.g., gender bias in AD prevalence) or confounding effects (e.g., the issue of normal aging and metadata-related heterogeneity). Hence, it is difficult to fully excavate the influence of metadata on AD diagnosis. To address these issues, we constructed a novel Multi-template Meta-information Regularized Network (MMRN) for AD diagnosis. Specifically, considering diagnostic variation resulting from different spatial transformations onto different brain templates, we first regarded different transformations as data augmentation for self-supervised learning after template selection. Since the confounding effects may arise from excessive attention to meta-information owing to its correlation with AD, we then designed the modules of weakly supervised meta-information learning and mutual information minimization to learn and disentangle meta-information from learned class-related representations, which accounts for meta-information regularization for disease diagnosis. We have evaluated our proposed MMRN on two public multi-center cohorts, including the Alzheimer’s Disease Neuroimaging Initiative (ADNI) with 1,950 subjects and the National Alzheimer’s Coordinating Center (NACC) with 1,163 subjects. The experimental results have shown that our proposed method outperformed the state-of-the-art approaches in both tasks of AD diagnosis, mild cognitive impairment (MCI) conversion prediction, and normal control (NC) vs . MCI vs . AD classification.


  • Open Access

    Real-time Multi-person Eyeblink Detection in the Wild for Untrimmed Video

    by Wenzheng Zeng, Yang Xiao, Sicheng Wei, Jinfang Gan, Xintao Zhang, Zhiguo Cao, Zhiwen Fang, Joey Tianyi Zhou

    Real-time eyeblink detection in the wild can widely serve for fatigue detection, face anti-spoofing, emotion analysis, etc. The existing research efforts generally focus on single-person cases towards trimmed video. However, multi-person scenario within untrimmed videos is also important for practical applications, which has not been well concerned yet. To address this, we shed light on this research field for the first time with essential contributions on dataset, theory, and practices. In particular, a large-scale dataset termed MPEblink that involves 686 untrimmed videos with 8748 eyeblink events is proposed under multi-person conditions. The samples are captured from unconstrained films to reveal "in the wild" characteristics. Meanwhile, a real-time multi-person eyeblink detection method is also proposed. Being different from the existing counterparts, our proposition runs in a one-stage spatio-temporal way with end-to-end learning capacity. Specifically, it simultaneously addresses the sub-tasks of face detection, face tracking, and human instance-level eyeblink detection. This paradigm holds 2 main advantages: (1) eyeblink features can be facilitated via the face's global context (e.g., head pose and illumination condition) with joint optimization and interaction, and (2) addressing these sub-tasks in parallel instead of sequential manner can save time remarkably to meet the real-time running requirement. Experiments on MPEblink verify the essential challenges of real-time multi-person eyeblink detection in the wild for untrimmed video. Our method also outperforms existing approaches by large margins and with a high inference speed.