Xiaoxiao Cui

Zero-shot video action recognition aims to recognize unseen action categories by transferring knowledge from seen categories through vision-language alignment. However, existing CLIP-based methods inherit image-centric representations, limiting their ability to capture hierarchical temporal dependencies and coherent action semantics across video sequences. To address this limitation, we propose HTM-CLIP, a unified vision-language framework that enhances frozen CLIP through hierarchical temporal modeling and conditional semantic enhancement. Specifically, hierarchical temporal modeling captures short-, mid-, and long-range dependencies in a progressive manner, enabling efficient video representation learning without global attention. Meanwhile, conditional semantic enhancement facilitates early visual-semantic interaction and category-aware feature refinement via gated prompt modulation. Together, these components learn temporally coherent and semantically discriminative video representations, improving vision-language alignment and zero-shot generalization. Extensive experiments demonstrate that HTM-CLIP consistently outperforms state-of-the-art methods while maintaining competitive computational efficiency.

Multi-view echo quality assessment in domain adaptation (MEA-DA) has not been explored due to the challenges posed by the significant differences between cardiac anatomical structures and cross-domain distributions. To cope with the robustness, specificity, and diversity limitations encountered when extending the existing DA algorithms to MEA-DA, we introduce a Hybrid Attention-Aware Domain-Invariant Transformer (HEED-former), which offers three main contributions. (i) Locally-invariant Consistency Alignment provides enhanced robustness by constraining the cross-domain multiscale semantic structural similarity registration to learn locally-invariant knowledge. This approach enhances the correlation and consistency of locally-invariant semantic knowledge, thereby enabling the generation of stable locally-invariant features that accommodate the substantial transferable differences caused by subtle structural variances and thus improving robustness.(ii) Hybrid Attention Excitation yields improved specificity by complementarily stimulating significant multidimensional information. It suppresses regions with low contributions and aggregates significant multidimensional feature representations, enhancing specificity. (iii) Globally-invariant aggregation representation improves diversity by modeling global mutually invariant relationships to learn global topological semantic consistency. It aggregates globally-invariant knowledge with local structural similarity to capture the cross-domain feature correlations, thus enhancing diversity. Our method achieves highest performance on two challenging tasks. It demonstrates nearly 100% classification accuracy on both the 10-views and 7-views echo datasets. For quality assessment, it achieves an optimal mean absolute error of 0.1354 on the homogeneous experiments and a robust PLCC of 0.9140 on the heterogeneous experiments with significant quality variations. We believe that this novel framework will provide a powerful MEA-DA solution.

Precise medical image segmentation of regions of interest (ROIs) is crucial for accurate disease diagnosis and progression assessment. However, acquiring high-quality annotated data at the pixel level poses a significant challenge due to the resource-intensive nature of this process. This scarcity of high-quality annotated data results in few-shot scenarios, which are highly prevalent in clinical applications. To address this obstacle, this paper introduces Agent-Guided SAM (AGSAM), an innovative approach that transforms the Segment Anything Model (SAM) into a fully automated segmentation method by automating prompt generation. Capitalizing on the pre-trained feature extraction and decoding capabilities of SAM-Med2D, AGSAM circumvents the need for manual prompt engineering, ensuring adaptability across diverse segmentation methods. Furthermore, the proposed feature augmentation convolution module (FACM) enhances model accuracy by promoting stable feature representations. Experimental evaluations demonstrate AGSAM’s consistent superiority over other methods across various metrics. These findings highlight AGSAM’s efficacy in tackling the challenges associated with limited annotated data while achieving high-quality medical image segmentation.

Clinical workflow of cardiac assessment on 2D echocardiography requires both accurate segmentation and quantification of the Left Ventricle (LV) from paired apical 4-chamber and 2-chamber. Moreover, uncertainty estimation is significant in clinically understanding the performance of a model. However, current research on 2D echocardiography ignores this vital task while joint segmentation with quantification, hence motivating the need for a unified optimization method. In this paper, we propose a multitask model with Task Relation Spatial co-Attention (referred as TRSA-Net) for joint segmentation, quantification, and uncertainty estimation on paired 2D echo. TRSA-Net achieves multitask joint learning by novelly exploring the spatial correlation between tasks. The task relation spatial co-attention learns the spatial mapping among task-specific features by non-local and co-excitation, which forcibly joints embedded spatial information in the segmentation and quantification. The Boundary-aware Structure Consistency (BSC) and Joint Indices Constraint (JIC) are integrated into the multitask learning optimization objective to guide the learning of segmentation and quantification paths. The BSC creatively promotes structural similarity of predictions, and JIC explores the internal relationship between three quantitative indices. We validate the efficacy of our TRSA-Net on the public CAMUS dataset. Extensive comparison and ablation experiments show that our approach can achieve competitive segmentation performance and highly accurate results on quantification.

Segmentation of the left ventricular (LV) myocardium in 2-D echocardiography is essential for clinical decision making, especially in geometry measurement and index computation. However, segmenting the myocardium is a time-consuming process and challenging due to the fuzzy boundary caused by the low image quality. The ground-truth label is employed as pixel-level class associations or shape regulation in segmentation, which works limit for effective feature enhancement for 2-D echocardiography. We propose a training strategy named multiconstrained aggregate learning (referred to as MCAL), which leverages anatomical knowledge learned through ground-truth labels to infer segmented parts and discriminate boundary pixels. The new framework encourages the model to focus on the features in accordance with the learned anatomical representations, and the training objectives incorporate a boundary distance transform weight (BDTW) to enforce a higher weight value on the boundary region, which helps to improve the segmentation accuracy. The proposed method is built as an end-to-end framework with a top-down, bottom-up architecture with skip convolution fusion blocks and carried out on two datasets (our dataset and the public CAMUS dataset). The comparison study shows that the proposed network outperforms the other segmentation baseline models, indicating that our method is beneficial for boundary pixels discrimination in segmentation.