FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation

To solve the problems of existing hybrid networks based on convolutional neural networks (CNN) and Transformers, we propose a new encoder–decoder network FI‐Net based on CNN‐Transformer for medical image segmentation. In the encoder part, a dual‐stream encoder is used to capture local details and lo...

Full description

Saved in:
Bibliographic Details
Main Authors: Yuhan Ding, Jinhui Liu, Yunbo He, Jinliang Huang, Haisu Liang, Zhenglin Yi, Yongjie Wang
Format: Article
Language:English
Published: Wiley 2024-12-01
Series:Advanced Intelligent Systems
Subjects:
Online Access:https://doi.org/10.1002/aisy.202400201
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850243543575035904
author Yuhan Ding
Jinhui Liu
Yunbo He
Jinliang Huang
Haisu Liang
Zhenglin Yi
Yongjie Wang
author_facet Yuhan Ding
Jinhui Liu
Yunbo He
Jinliang Huang
Haisu Liang
Zhenglin Yi
Yongjie Wang
author_sort Yuhan Ding
collection DOAJ
description To solve the problems of existing hybrid networks based on convolutional neural networks (CNN) and Transformers, we propose a new encoder–decoder network FI‐Net based on CNN‐Transformer for medical image segmentation. In the encoder part, a dual‐stream encoder is used to capture local details and long‐range dependencies. Moreover, the attentional feature fusion module is used to perform interactive feature fusion of dual‐branch features, maximizing the retention of local details and global semantic information in medical images. At the same time, the multi‐scale feature aggregation module is used to aggregate local information and capture multi‐scale context to mine more semantic details. The multi‐level feature bridging module is used in skip connections to bridge multi‐level features and mask information to assist multi‐scale feature interaction. Experimental results on seven public medical image datasets fully demonstrate the effectiveness and advancement of our method. In future work, we plan to extend FI‐Net to support 3D medical image segmentation tasks and combine self‐supervised learning and knowledge distillation to alleviate the overfitting problem of limited data training.
format Article
id doaj-art-b9ae2866e15243e0bcc8cdc9eb018a27
institution OA Journals
issn 2640-4567
language English
publishDate 2024-12-01
publisher Wiley
record_format Article
series Advanced Intelligent Systems
spelling doaj-art-b9ae2866e15243e0bcc8cdc9eb018a272025-08-20T01:59:57ZengWileyAdvanced Intelligent Systems2640-45672024-12-01612n/an/a10.1002/aisy.202400201FI‐Net: Rethinking Feature Interactions for Medical Image SegmentationYuhan Ding0Jinhui Liu1Yunbo He2Jinliang Huang3Haisu Liang4Zhenglin Yi5Yongjie Wang6School of Computer Science and Engineering Central South University Changsha 410000 ChinaDepartments of Urology Xiangya Hospital Central South University Changsha 410008 ChinaDepartments of Urology Xiangya Hospital Central South University Changsha 410008 ChinaDepartments of Urology Xiangya Hospital Central South University Changsha 410008 ChinaDepartments of Urology Xiangya Hospital Central South University Changsha 410008 ChinaDepartments of Urology Xiangya Hospital Central South University Changsha 410008 ChinaDepartment of Burns and Plastic Surgery Xiangya Hospital Central South University Changsha 410008 ChinaTo solve the problems of existing hybrid networks based on convolutional neural networks (CNN) and Transformers, we propose a new encoder–decoder network FI‐Net based on CNN‐Transformer for medical image segmentation. In the encoder part, a dual‐stream encoder is used to capture local details and long‐range dependencies. Moreover, the attentional feature fusion module is used to perform interactive feature fusion of dual‐branch features, maximizing the retention of local details and global semantic information in medical images. At the same time, the multi‐scale feature aggregation module is used to aggregate local information and capture multi‐scale context to mine more semantic details. The multi‐level feature bridging module is used in skip connections to bridge multi‐level features and mask information to assist multi‐scale feature interaction. Experimental results on seven public medical image datasets fully demonstrate the effectiveness and advancement of our method. In future work, we plan to extend FI‐Net to support 3D medical image segmentation tasks and combine self‐supervised learning and knowledge distillation to alleviate the overfitting problem of limited data training.https://doi.org/10.1002/aisy.202400201convolutional neural networkfeature interactionmedical image segmentationmulti‐scale featuretransformer
spellingShingle Yuhan Ding
Jinhui Liu
Yunbo He
Jinliang Huang
Haisu Liang
Zhenglin Yi
Yongjie Wang
FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation
Advanced Intelligent Systems
convolutional neural network
feature interaction
medical image segmentation
multi‐scale feature
transformer
title FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation
title_full FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation
title_fullStr FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation
title_full_unstemmed FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation
title_short FI‐Net: Rethinking Feature Interactions for Medical Image Segmentation
title_sort fi net rethinking feature interactions for medical image segmentation
topic convolutional neural network
feature interaction
medical image segmentation
multi‐scale feature
transformer
url https://doi.org/10.1002/aisy.202400201
work_keys_str_mv AT yuhanding finetrethinkingfeatureinteractionsformedicalimagesegmentation
AT jinhuiliu finetrethinkingfeatureinteractionsformedicalimagesegmentation
AT yunbohe finetrethinkingfeatureinteractionsformedicalimagesegmentation
AT jinlianghuang finetrethinkingfeatureinteractionsformedicalimagesegmentation
AT haisuliang finetrethinkingfeatureinteractionsformedicalimagesegmentation
AT zhenglinyi finetrethinkingfeatureinteractionsformedicalimagesegmentation
AT yongjiewang finetrethinkingfeatureinteractionsformedicalimagesegmentation