Related Experiment Video
Updated: Jul 2, 2025
07:50
A Rapid Method for Multispectral Fluorescence Imaging of Frozen Tissue Sections
Published on: March 30, 2020
8.3K
Exploring subtypes of multiple sclerosis through unsupervised machine learning of automated fiber quantification
Xueheng Liang1,2, Zichun Yan1, Yongmei Li3
1Department of Radiology, The First Affiliated Hospital of Chongqing Medical University, No 1 Youyi Road, Yuzhong District, Chongqing, 40016, China.
Japanese Journal of Radiology
|February 26, 2024
Summary
Unsupervised machine learning identified three distinct multiple sclerosis (MS) subtypes based on white matter (WM) fiber tracts. These subtypes correlate with significant differences in cognitive function and disability progression, aiding personalized treatment strategies.
Area of Science:
- Neuroimaging
- Machine Learning
- Neurology
Background:
- Multiple sclerosis (MS) is a chronic neurological disease characterized by demyelination and axonal damage in the central nervous system.
- White matter (WM) integrity is crucial for neurological function, and its damage is a hallmark of MS.
- Understanding MS heterogeneity is vital for developing effective, personalized treatment strategies.
Purpose of the Study:
- To subtype multiple sclerosis (MS) patients using unsupervised machine learning applied to white matter (WM) fiber tracts.
- To investigate the relationship between identified MS subtypes and cognitive function and disability outcomes.
- To explore the potential of WM fiber tract abnormalities as biomarkers for predicting MS disease progression.
Main Methods:
- Automated Fiber Quantification (AFQ) was used to extract 18 WM fiber tracts from 103 MS patients.
- Unsupervised machine learning (cluster analysis) was employed to identify distinct patient subtypes.
- Clinical data, diffusion tensor imaging (DTI) metrics (FA, MD, AD, RD), and survival analysis were used to compare subtypes and predict outcomes.
Main Results:
- Three distinct MS subtypes were identified based on WM fiber tract analysis.
- Significant differences in fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) were observed among subtypes.
- Subtype 3 exhibited the most rapid disability progression and cognitive decline, followed by Subtype 1, with Subtype 2 showing a slower progression rate.
Conclusions:
- Unsupervised machine learning on WM fiber tracts effectively subtypes MS patients, revealing distinct clinical and disability trajectories.
- WM abnormalities identified through this approach can serve as predictive biomarkers for MS outcomes.
- This subtyping strategy holds promise for personalized medicine, enabling tailored treatment and prognostic predictions in MS care.
