Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation.

Journal: Sensors (Basel, Switzerland)
Published Date:

Abstract

Few-shot learning has demonstrated remarkable performance in medical image segmentation. However, existing few-shot medical image segmentation (FSMIS) models often struggle to fully utilize query image information, leading to prototype bias and limited generalization ability. To address these issues, we propose the dual-filter cross attention and onion pooling network (DCOP-Net) for FSMIS. DCOP-Net consists of a prototype learning stage and a segmentation stage. During the prototype learning stage, we introduce a dual-filter cross attention (DFCA) module to avoid entanglement between query background features and support foreground features, effectively integrating query foreground features into support prototypes. Additionally, we design an onion pooling (OP) module that combines eroding mask operations with masked average pooling to generate multiple prototypes, preserving contextual information and mitigating prototype bias. In the segmentation stage, we present a parallel threshold perception (PTP) module to generate robust thresholds for foreground and background differentiation and a query self-reference regularization (QSR) strategy to enhance model accuracy and consistency. Extensive experiments on three publicly available medical image datasets demonstrate that DCOP-Net outperforms state-of-the-art methods, exhibiting superior segmentation and generalization capabilities.

Authors

  • Lina Ni
    College of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China.
  • Yang Liu
    Department of Computer Science, Hong Kong Baptist University, Hong Kong, China.
  • Zekun Zhang
    From the Department of Radiology, Hôpital Fondation A. de Rothschild, 25 rue Manin, 75019 Paris, France (L.D.); Faculty of Medicine, Université de Paris, Paris, France (L.D., A. Feydy); Gleamer, Paris, France (A.D., C.A., N.C., Z.Z., N.N., E.L., A.P., N.E.R.); Department of Biostatistics, CHU Rouen, Rouen, France (A.G.); Department of Radiology, Hôpital Hôtel-Dieu, Assistance Publique-Hôpitaux de Paris, Paris, France (J.L.); Department of Radiology, Hôpital Ambroise-Paré, Assistance Publique-Hôpitaux de Paris, Boulogne-Billancourt, France (A. Felter); Department of Radiology, Hôpital Raymond-Poincaré, Assistance Publique-Hôpitaux de Paris, Garches, France (A. Felter); and Department of Radiology B, Hôpital Cochin, Assistance Publique-Hôpitaux de Paris, Paris, France (L.L., N.E.R., A. Feydy).
  • Yongtao Li
    College of Chemistry and Chemical Engineering, Southwest Petroleum University, Chengdu 610050, PR China.
  • Jinquan Zhang
    College of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, China.