Related Experiment Video
Updated: Jan 12, 2026
09:41
A Pipeline for 3D Multimodality Image Integration and Computer-assisted Planning in Epilepsy Surgery
Published on: May 20, 2016
12.7K
Multi-modal medical image synthesis via dual-branch wavelet encoding and deformable feature interaction
Xuefeng Jia1, Biyuan Li2, Jinying Ma1
1School of Electronic Engineering, Tianjin University of Technology and Education, Tianjin, 300222, China.
Artificial Intelligence in Medicine
|October 30, 2025
Summary
This study introduces DWFI-GAN, a new method for generating missing medical images. It effectively synthesizes missing modalities by enhancing feature fusion and detail extraction, improving multi-modal imaging applications.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Multi-modal medical imaging is crucial for diagnosis and treatment planning, offering complementary anatomical and pathological insights.
- Missing or unavailable imaging modalities due to scan limitations or protocol differences hinder the application of multi-modal data.
- Existing medical image synthesis methods struggle to effectively integrate local details and global context across modalities.
Purpose of the Study:
- To develop a novel generative adversarial network for synthesizing missing medical imaging modalities.
- To address the challenge of synergistically extracting local and global features and fusing complementary information across modalities.
Main Methods:
- Proposed a dual-branch wavelet encoding and deformable feature interaction generative adversarial network (DWFI-GAN).
- Introduced a wavelet multi-scale downsampling (Wavelet-MS-Down) module for efficient multi-scale feature extraction.
- Designed a deformable cross-attention feature fusion (DCFF) module for interactive multi-scale feature fusion.
- Incorporated an episodic bottleneck structure with frequency-space enhanced (FSE) modules for enriched feature representation.
Main Results:
- DWFI-GAN demonstrated superior performance compared to state-of-the-art methods on two medical imaging datasets.
- Qualitative and quantitative experiments confirmed the effectiveness of the proposed method.
- Ablation studies validated the contribution of each module to feature fusion and detail enhancement.
Conclusions:
- DWFI-GAN effectively synthesizes missing medical imaging modalities by improving feature fusion and detail extraction.
- The proposed method enhances the utility of multi-modal medical imaging despite data limitations.
- The developed DWFI-GAN offers a promising solution for reconstructing complete multi-modal datasets in clinical practice.
