O-DAT: A novel framework integrating optimized dual attention for medical image segmentation

This paper presents O-DAT (Optimized DA-TransUNet), a deep learning framework for medical image segmentation. Accurate segmentation is essential for clinical diagnosis but challenging due to complex anatomical structures. O-DAT enhances U-Net with optimized dual attention (ODA) modules for better sp...

Full description

Saved in:
Bibliographic Details
Main Authors: Chen Guo, Hongyuan Ren, Haiying Qi, Xue Zhang, Xiaolin Gu, Jingjing Liu, Yuefan Liu
Format: Article
Language:English
Published: Elsevier 2025-09-01
Series:Alexandria Engineering Journal
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1110016825008452
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper presents O-DAT (Optimized DA-TransUNet), a deep learning framework for medical image segmentation. Accurate segmentation is essential for clinical diagnosis but challenging due to complex anatomical structures. O-DAT enhances U-Net with optimized dual attention (ODA) modules for better spatial and channel feature extraction, and uses patch expansion layers in the decoder for efficient upsampling. This approach combines the strengths of U-Net and Transformers, improving both accuracy and efficiency. Experiments show O-DAT outperforms existing methods, achieving a DSC of 80.30% and an HD of 25.88 mm on the Synapse dataset. Ablation studies confirm the effectiveness of ODA blocks and patch expansion layers. O-DAT sets a new benchmark for medical image segmentation, with potential to enhance clinical diagnosis and guide future research.
ISSN:1110-0168