Structural brain changes versus self-report: machine-learning classification of chronic fatigue syndrome patients

Area of Science:

• Neuroimaging
• Medical Informatics
• Psychiatry

Background:

• Chronic Fatigue Syndrome (CFS) is characterized by profound fatigue, pain, and often presents with detectable structural and functional brain abnormalities.
• Magnetic Resonance Imaging (MRI) has shown potential in identifying these neurological differences between CFS patients and healthy individuals.

Purpose of the Study:

• To evaluate the efficacy of structural MRI (sMRI) abnormalities in classifying Chronic Fatigue Syndrome (CFS) patients versus healthy controls (HC).
• To compare the performance of sMRI-based machine learning (ML) classification with ML classification derived from self-report (SR) measures.

Main Methods:

• Utilized T1-weighted sMRI data from 18 CFS patients and 15 HC, segmented using FreeSurfer to analyze 61 brain regions.
• Employed a linear support vector machine (SVM) with bootstrap optimism correction for classification tasks.
• Compared ML models based on sMRI regional estimates against models using visual analogue scale ratings of fatigue, pain, anxiety, depression, anger, and sleep quality.

Main Results:

• The sMRI-based ML model achieved a classification accuracy of 79.58%.
• The ML model based on self-report (SR) ratings demonstrated superior performance with 95.95% accuracy.
• Brain regions associated with cognition, emotion, and memory significantly contributed to the classification accuracy in both models.

Conclusions:

• Structural MRI abnormalities are valuable for differentiating CFS patients from healthy controls.
• Self-report measures are currently more effective than sMRI for ML-based classification of CFS.
• This study represents the first ML-based group classification of CFS, highlighting the potential of integrating neuroimaging and clinical data.