Machine Learning-Based Classification of White Matter Functional Changes in Stroke Patients Using Resting-State fMRI.

Area of Science:

• Neuroimaging
• Neuroscience
• Medical Imaging

Background:

• Neuroimaging studies commonly focus on gray matter in stroke patients, leaving white matter functional changes under-researched.
• Understanding white matter's role in stroke recovery, including repair and compensation mechanisms, is crucial but currently unclear.

Purpose of the Study:

• To investigate and demonstrate changes in white matter functional connectivity in stroke patients.
• To reveal the reorganization characteristics of white matter functional networks post-stroke.
• To provide potential biomarkers for stroke rehabilitation and new clinical insights.

Main Methods:

• Resting-state functional magnetic resonance imaging (rs-fMRI) was used on 36 stroke patients and 36 healthy controls.
• Regional Homogeneity (ReHo) and Degree Centrality (DC) were employed as feature vectors to assess white matter function.
• Support Vector Machine (SVM) classification with leave-one-out cross-validation (LOOCV) was used to differentiate between patient groups.

Main Results:

• Stroke patients showed significantly reduced white matter DC in the corpus callosum genu (GCC), corpus callosum body (BCC), and left anterior corona radiata (ACRL).
• Increased DC was observed in the left superior longitudinal fasciculus (SLF_L) in stroke patients.
• Reduced ReHo values were found in the GCC and BCC regions of stroke patients compared to controls.
• SVM classification achieved an AUC of 0.89 for DC and 0.98 for ReHo, demonstrating high discriminative power.

Conclusions:

• Stroke significantly alters functional connectivity within specific white matter regions.
• Weakened connectivity in GCC, BCC, and ACRL, alongside enhanced compensatory connectivity in SLF_L, indicates white matter network reorganization post-stroke.
• These findings offer potential biomarkers for rehabilitation and novel clinical insights for stroke patient treatment.