Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation
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...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-03-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/25/7/2176 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849730263989354496 |
|---|---|
| author | Lina Ni Yang Liu Zekun Zhang Yongtao Li Jinquan Zhang |
| author_facet | Lina Ni Yang Liu Zekun Zhang Yongtao Li Jinquan Zhang |
| author_sort | Lina Ni |
| collection | DOAJ |
| description | 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. |
| format | Article |
| id | doaj-art-7d322c6cac7d41c092eb6936dc5f45f3 |
| institution | DOAJ |
| issn | 1424-8220 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj-art-7d322c6cac7d41c092eb6936dc5f45f32025-08-20T03:08:56ZengMDPI AGSensors1424-82202025-03-01257217610.3390/s25072176Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image SegmentationLina Ni0Yang Liu1Zekun Zhang2Yongtao Li3Jinquan Zhang4College of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaCollege of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaSchool of Computer Science, University of Glasgow, Glasgow G12 8QQ, UKCollege of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaCollege of Computer Science and Engineering, Shandong University of Science and Technology, Qingdao 266590, ChinaFew-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.https://www.mdpi.com/1424-8220/25/7/2176few-shot learningmedical image segmentationprototype learning |
| spellingShingle | Lina Ni Yang Liu Zekun Zhang Yongtao Li Jinquan Zhang Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation Sensors few-shot learning medical image segmentation prototype learning |
| title | Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation |
| title_full | Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation |
| title_fullStr | Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation |
| title_full_unstemmed | Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation |
| title_short | Dual-Filter Cross Attention and Onion Pooling Network for Enhanced Few-Shot Medical Image Segmentation |
| title_sort | dual filter cross attention and onion pooling network for enhanced few shot medical image segmentation |
| topic | few-shot learning medical image segmentation prototype learning |
| url | https://www.mdpi.com/1424-8220/25/7/2176 |
| work_keys_str_mv | AT linani dualfiltercrossattentionandonionpoolingnetworkforenhancedfewshotmedicalimagesegmentation AT yangliu dualfiltercrossattentionandonionpoolingnetworkforenhancedfewshotmedicalimagesegmentation AT zekunzhang dualfiltercrossattentionandonionpoolingnetworkforenhancedfewshotmedicalimagesegmentation AT yongtaoli dualfiltercrossattentionandonionpoolingnetworkforenhancedfewshotmedicalimagesegmentation AT jinquanzhang dualfiltercrossattentionandonionpoolingnetworkforenhancedfewshotmedicalimagesegmentation |