Logistic Regression Algorithm Differentiates Gulf War Illness (GWI) Functional Magnetic Resonance Imaging (fMRI) Data...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Gulf War Illness (GWI) presents with debilitating cognitive, pain, fatigue, and somatic symptoms without a clear pathology.
• Objective biomarkers, such as functional Magnetic Resonance Imaging (fMRI) activity patterns, are crucial for validating GWI diagnostic criteria.
• Exertional exhaustion is a hallmark GWI symptom, necessitating studies that model its impact on cognitive function.

Purpose of the Study:

• To identify objective neuroimaging biomarkers for Gulf War Illness (GWI) using fMRI.
• To investigate the impact of exertional exhaustion on cognitive brain activity patterns in GWI patients.
• To differentiate GWI and its subtypes (START, STOPP, POTS) from healthy controls using machine learning.

Main Methods:

• Functional Magnetic Resonance Imaging (fMRI) was used to assess brain activity during a 2-Back working memory task in GWI patients (n=80) and sedentary controls (n=31).
• Subjects underwent submaximal exercise stress tests on two consecutive days, with fMRI scans performed before and after exercise.
• Machine learning algorithms, including recursive feature selection and logistic regression, were employed to analyze fMRI data and classify subject groups.

Main Results:

• Machine learning models identified distinct patterns of brain activation differentiating GWI patients from controls, with improved classification accuracy after exercise (70% pre-exercise, 85% post-exercise).
• Ten brain regions, including the right anterior insula, orbital frontal cortex, and thalamus, consistently distinguished GWI from controls across both days.
• The models successfully classified GWI subgroups (START, STOPP, POTS) with 67-69% accuracy, demonstrating the potential for subtyping.

Conclusions:

• fMRI combined with machine learning can differentiate GWI patients from controls and identify distinct brain activity patterns associated with exertional exhaustion.
• This approach offers promising objective biomarkers for GWI diagnosis and subtyping, potentially improving clinical validity.
• The study highlights the utility of neuroimaging in understanding the complex pathophysiology of Gulf War Illness and its subtypes.