Related Experiment Video
Updated: May 29, 2025
Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities
Published on: October 27, 2023
Multimodal Cross Global Learnable Attention Network for MR images denoising with arbitrary modal missing.
Mingfu Jiang1, Shuai Wang2, Ka-Hou Chan3
1Faculty of Applied Sciences, Macao Polytechnic University, R. de Luís Gonzaga Gomes, Macao, 999078, Macao Special Administrative Region of China; College of Information Engineering, Xinyang Agriculture and Forestry University, No. 1 North Ring Road, Pingqiao District, Xinyang, 464000, Henan, China.
This study introduces a novel network for denoising Magnetic Resonance Imaging (MRI) scans, effectively handling missing image sequences. The method enhances diagnostic accuracy by preserving crucial image details while reducing noise.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Image Processing
Background:
- Magnetic Resonance Imaging (MRI) produces multimodal images crucial for disease diagnosis.
- Image noise degrades MRI quality, impacting diagnostic accuracy.
- Existing denoising methods often fail to leverage multimodal relationships or balance feature extraction.
Purpose of the Study:
- To develop a novel denoising network for MRI that addresses multimodal relationships and arbitrary modal missing.
- To improve the balance between denoising strength and preservation of image texture details.
Main Methods:
- A controllable Multimodal Cross-Global Learnable Attention Network (MMCGLANet) was proposed.
- Utilized an Encoder for shallow feature extraction and Convolutional Long Short-Term Memory (ConvLSTM) for intra-modal feature extraction.
- Employed Cross Global Learnable Attention Network (CGLANet) for inter-modal and intra-modal feature fusion, with sequence codes for handling missing modalities.
Main Results:
- The proposed MMCGLANet effectively denoises MR images across different modalities.
- The network demonstrates robust performance even with missing image sequences.
- Achieved superior denoising results compared to existing methods on public and real datasets.
Conclusions:
- MMCGLANet offers a significant advancement in MRI denoising, particularly for multimodal datasets with missing sequences.
- The method enhances the reliability of MRI for clinical diagnosis by improving image quality.
- The attention-based fusion and handling of missing modalities represent a key innovation.
