Related Experiment Video
Updated: Jun 7, 2025
10:05
DTI of the Visual Pathway - White Matter Tracts and Cerebral Lesions
Published on: August 26, 2014
14.0K
Tractography-Based Automated Identification of Retinogeniculate Visual Pathway With Novel Microstructure-Informed
Sipei Li1,2, Wei Zhang1, Shun Yao3,4
1School of Information and Communication Engineering, University of Electronic Science and Technology of China, Chengdu, China.
Human Brain Mapping
|November 20, 2024
Summary
DeepRGVP, a novel deep learning framework, accurately identifies the retinogeniculate visual pathway (RGVP) from diffusion MRI tractography. This automated method overcomes limitations of manual selection, offering faster and more reliable visualization for brain disease research.
Area of Science:
- Neuroscience
- Medical Imaging
- Computer Science
Background:
- The retinogeniculate visual pathway (RGVP) transmits visual information from the retina to the lateral geniculate nucleus.
- Accurate RGVP mapping is crucial for understanding visual system anatomy and treating related neurological disorders.
- Current manual streamline selection in diffusion MRI (dMRI) tractography is labor-intensive and prone to variability.
Purpose of the Study:
- To introduce DeepRGVP, a deep learning framework for automated and precise identification of the RGVP from dMRI tractography data.
- To develop a novel microstructure-informed supervised contrastive learning approach for enhanced streamline classification.
- To address data imbalance issues in tractography datasets using streamline-level data augmentation.
Main Methods:
- Developed DeepRGVP, a deep learning framework utilizing microstructure-informed supervised contrastive learning.
- Implemented a novel streamline-level data augmentation technique to handle imbalanced RGVP data.
- Compared DeepRGVP against state-of-the-art tractography parcellation methods and evaluated on patient data with pituitary tumors.
Main Results:
- DeepRGVP demonstrated superior accuracy and F1 scores in identifying the RGVP compared to existing deep learning methods.
- The framework successfully identified RGVPs in patient data, even in the presence of lesions affecting the pathway.
- Achieved significantly higher accuracy and F1 scores compared to state-of-the-art methods.
Conclusions:
- DeepRGVP offers a fast, accurate, and automated solution for RGVP identification from dMRI tractography.
- The proposed deep learning approach has high potential for clinical applications in visual system research and disease management.
- Automated RGVP identification using deep learning can reduce expert labor costs and inter-observer variability.
