Prediction of the Topography of the Corticospinal Tract on T1-Weighted MR Images Using Deep-Learning-Based...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Tractography is crucial for brain tumor surgery planning and studying neurological conditions.
• Accurate white matter tract localization is essential for preserving brain function.
• Current manual segmentation methods can be time-consuming and subjective.

Purpose of the Study:

• To evaluate a deep learning model for predicting white matter tract topography.
• To compare deep learning segmentation performance against manual segmentation.
• To assess the utility of T1-weighted MRI for deep learning-based tractography.

Main Methods:

• Utilized T1-weighted MRI from 190 healthy subjects across 6 datasets.
• Reconstructed corticospinal tracts using deterministic diffusion tensor imaging.
• Trained a deep learning segmentation model (nnU-Net) on 90 subjects and validated on 100 subjects.

Main Results:

• Developed a deep learning model capable of predicting corticospinal pathway topography on T1-weighted images.
• Achieved an average Dice score of 0.5479 on the validation dataset.
• Demonstrated the model's performance across diverse datasets.

Conclusions:

• Deep learning-based segmentation offers a viable approach for predicting white matter pathway locations.
• This technique holds potential for future applications in clinical neuroimaging.
• Further research can refine deep learning models for enhanced tractography accuracy.