Beyond Random Sampling: Distribution-Aware Alignment for Semi-Supervised Medical Image Segmentation

GroundMed-SAM: Prompt-based Zero-shot Medical Image Segmentation

Medical image segmentation is a key component of computer-aided diagnosis and treatment planning. Despite substantial progress in deep learning–based models, most existing approaches depend heavily on large annotated datasets and often fail to generalize across heterogeneous clinical environments, limiting their deployment in real-world settings characterized by domain shifts and scarce expert annotations. This paper presents a zero-shot learning framework named GroundMed-SAM for medical image segmentation. The framework integrates GroundingDINO for prompt-based region localization and MedSAM for mask generation. To address the weak alignment between visual features and medical semantics in GroundingDINO, which is pretrained on general domain image-text pairs, we introduce learnable medical text embeddings that explicitly parameterize domain-specific terminology in a continuous semantic space. These embeddings are optimized during training to better align medical concepts with visual representations, thereby strengthening text-image correspondence and improving detection-guided segmentation. The proposed framework preserves true zero-shot capability, enabling segmentation of previously unseen anatomical structures without task-specific labels. Extensive experiments on multiple public datasets across diverse modalities and clinical contexts demonstrate that our method achieves competitive segmentation performance in-domain while exhibiting superior robustness under cross-domain evaluation. Although supervised baselines outperform the proposed framework by only 3–5% on in-domain datasets, they experience substantial performance degradation when evaluated on unseen domains. Additionally, the framework achieves an AUC of 98.9 in endoscopic polyp detection, highlighting the effectiveness of the proposed medical-aware textual embeddings in guiding region localization. These results demonstrate the effectiveness of the proposed framework in improving cross-domain generalization for medical image segmentation with limited annotations.

Unifying Active Learning and Semi-Supervised Learning for Medical Image Segmentation

RegAL is proposed, a unified active semi-supervised framework governed by a shared topology-aware Pareto optimization that couples sample acquisition with unlabeled data utilization and consistently outperforms state-of-the-art AL, SSL, and active semi-supervised baselines across Dice and boundary-distance metrics under extreme annotation scarcity.

Distribution-Constrained Semi-Supervised Learning for Medical Organ Segmentation

VCDP: Variation-Conditioned Distributional Proxy Learning for Semi-Supervised Medical Image Segmentation

Semi-supervised 3D medical image segmentation reduces the need for dense voxel-level annotations by exploiting unlabeled volumes. Although existing methods such as consistency regularization, pseudo-labeling, and co-training improve prediction-level robustness, they often provide insufficient feature-space organization for anatomically complex structures, especially small organs and ambiguous boundary regions with large intra-class variations. To address this issue, we propose Variation-Conditioned Distributional Proxy Learning (VCDP), a plug-and-play training-only regularization module for semi-supervised 3D medical image segmentation. VCDP represents each class with a learnable Gaussian distribution for shared class semantics and multiple variation prototypes for fine-grained intra-class patterns. A unified variation-conditioned compatibility score is further formulated to fuse distributional similarity and soft variation aggregation, guiding voxel embeddings to align with both global organ identity and local anatomical variations. VCDP is attached to decoder features during training and removed during inference, introducing no additional inference cost. Experiments on multi-organ segmentation benchmarks show that VCDP improves most evaluated baselines, particularly for small, ambiguous, and highly variable organs. Our anonymous code is released at https://anonymous.4open.science/r/VCDP_code-41ED.

Robust unsupervised domain adaptation for medical image segmentation via frequency-conditioned graph diffusion

Cross-domain variability in medical imaging, arising from differences in scanners, acquisition protocols, and patient populations, remains a major challenge for reliable semantic segmentation. Existing unsupervised domain adaptation (UDA) methods predominantly rely on image-level transformations or feature alignment, which often fail to preserve anatomical consistency under large domain shifts. In this work, we propose a novel structured latent UDA framework that performs domain alignment in a topology-aware representation space rather than directly modifying image appearance. Specifically, we introduce a Frequency-Conditioned Graph Diffusion paradigm, where convolutional features are transformed into anatomical graphs to explicitly capture structural relationships. A latent diffusion process then progressively refines these graph embeddings, guided by frequency-aware contextual cues, enabling robust cross-domain alignment. To further enhance generalization, we integrate structural consistency regularization with adversarial latent alignment, eliminating the need for labeled target data. A dedicated decoder reconstructs dense segmentation maps, while stochastic diffusion sampling provides uncertainty estimates for improved potential clinical reliability. Extensive experiments on multiple public medical imaging benchmarks demonstrate that our method consistently outperforms state-of-the-art UDA approaches, achieving superior segmentation accuracy and robustness under significant domain shifts. These results highlight the effectiveness of structured latent modeling and diffusion-based learning for robust domain-adaptive segmentation.

Semi-supervised Medical Image Segmentation via Perturbation-Aware Mutual Learning and Edge-Aware Uncertainty Loss for Accurate Anatomical Delineation.

This paper proposes a novel framework that effectively leverages unlabeled data to improve segmentation performance in cardiac structures and applies a novel consistency constraint by a dual fine-grained boundary loss that provide global characteristics-based guidance from the transition of the boundary region and an edge-aware uncertainty loss.