Similar articles

Abstract: Background and Objective: medical image segmentation is critical for clinical diagnosis, but the scarcity of annotated data limits robust model training, making few-shot learning indispensable. Existing methods, such as those featured in the paper "Anatomical prior guided spatial contrastive learning for few-shot medical image segmentation", often suffer from two issues: performance degradation due to significant inter-class variations in pathological structures, and an overreliance on attention mechanisms with high computational complexity (O(n²)), which hinders the efficient modeling of long-range dependencies. In contrast, the state space model (SSM) offers linear complexity (O(n)) and superior efficiency, making it a key solution. To address these challenges, we propose PPFFR (Parallel prototype filter and Feature Refinement) for few-shot medical image segmentation. PPFFR includes two core modules: (1) a Parallel prototype filter (PPF) to reduce inter-class variations by enhancing interactions between support and query prototypes; (2) a Feature Refinement (FR) module that leverages SSM (via state transition h_t = Ah_t-1 + Bx_t) for efficient long-range dependency modeling, integrated with multi-head attention (MHA) to preserve spatial details. Experimental results on the Abd-MRI dataset demonstrate that FR with MHA outperforms FR alone in segmenting Left Kidney, Right Kidney, Liver, Spleen, and Mean accuracy, confirming MHA's role in improving precision. As the critical component, SSM ensures that PPFFR balances performance with efficiency. In conclusion, PPFFR effectively enhances inter-class consistency and computational efficiency for few-shot medical image segmentation. Methods: the proposed framework comprises three key modules. First, we propose the Prototype Refinement (PG) module to construct refined class subgraphs from encoder-extracted features of both support and query images. This can generate support prototypes with minimized inter-class variation. We then propose the PPF module to suppress background interference and enhance the correlation between support and query prototypes. Finally, we implement the FR module to further enhance segmentation accuracy and accelerate model convergence with SSM's robust long-range dependency modeling capability. Results: in extensive experiments conducted on three public datasets under the 1-way 1-shot setting, PPFFR achieves Dice scores of 87.62%, 86.68%, and 79.71%, consistently surpassing state-of-the-art few-shot medical image segmentation methods. Conclusions: ablation studies validate the effectiveness of the PG, PPF, and FR modules. The results indicate that explicit inter-class variation reduction and SSM-based feature refinement can enhance accuracy without heavy computational overhead. This work provides new insights for few-shot learning in medical imaging and inspires lightweight architecture designs for clinical deployment.