EPISeg: Automated segmentation of the spinal cord on echo planar images using open-access multi-center data

bioRxiv preprint, 2025 · DOI: https://doi.org/10.1101/2025.01.07.631402 · Published: January 27, 2025

Simple Explanation

Functional magnetic resonance imaging (fMRI) of the spinal cord is important for understanding how we sense things, move, and control body functions. To analyze spinal cord fMRI data, the spinal cord needs to be identified in the images. However, spinal cord fMRI images are often of low quality due to distortions and other artifacts. This makes it hard to automatically identify the spinal cord, requiring manual effort. This study introduces a new method called EPISeg, which uses deep learning to automatically identify the spinal cord in these challenging images. The researchers also created a large, openly available dataset to help train and test the method.

Study Duration

Not specified

Participants

406 participants

Evidence Level

Not specified

Key Findings

1
EPISeg significantly improves spinal cord segmentation quality compared to existing methods.
2
The model is robust to different MRI scanning techniques and common image problems found in fMRI data.
3
EPISeg has been integrated into the Spinal Cord Toolbox, making it easily accessible to researchers.

Research Summary

This study introduces EPISeg, a deep learning-based method for automated spinal cord segmentation on gradient-echo EPI images, commonly used in fMRI. EPISeg addresses challenges related to heterogeneous resolutions, scanner settings, and motion artifacts. A large multi-site EPI spinal cord dataset was created and made publicly available on OpenNeuro to facilitate the development of a robust segmentation model. EPISeg demonstrates superior accuracy and robustness compared to state-of-the-art models like PropSeg, DeepSeg, and Contrast-agnostic segmentation and is readily available as a part of SCT v6.3 or higher.

Practical Implications

Improved fMRI Data Preprocessing

EPISeg provides a reliable and automated method for spinal cord segmentation, which is crucial for fMRI data preprocessing, especially for spatial normalization and group-level results.

Enhanced Research Reproducibility

By making the dataset and EPISeg publicly available, this study promotes transparency and reproducibility in spinal cord fMRI research.

Clinical Applications

Accurate spinal cord segmentation can aid in the diagnosis, monitoring, and treatment planning of various neurological conditions, such as multiple sclerosis, neuropathic pain, and spinal cord injury.

Study Limitations

1
The study may not cover all possible data domains due to variations in acquisition parameters and spinal cord conditions across different populations.
2
The method has solely been tested on data acquired at 3T, and segmentation performance could vary across other field strengths.
3
Effective motion correction is critical to the success of spinal cord fMRI segmentation, and inadequate motion correction could impact the accuracy of segmentation.

Your Feedback

Was this summary helpful?

Back to Neuroimaging