Classification of multiple sclerosis women with voiding dysfunction using machine learning: Is functional...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Multiple Sclerosis (MS) often leads to lower urinary tract symptoms, including voiding dysfunction (VD).
• Understanding the neural basis of VD in MS is crucial for targeted treatment.
• Brain connectivity, both functional (FC) and structural (SC), offers insights into neurological disorders.

Purpose of the Study:

• To investigate the utility of machine learning (ML) in classifying VD in female MS patients.
• To determine the relative importance of functional and structural brain connectivity for VD classification.
• To identify distinct brain connectivity patterns associated with VD in MS.

Main Methods:

• Recruited 27 ambulatory MS patients with lower urinary tract dysfunction, divided into voiders (V) and VD groups.
• Utilized concurrent functional MRI and urodynamics testing for data acquisition.
• Applied machine learning algorithms, including Partial Least Squares (PLS) and Random Forest (RF), to analyze FC and SC data.

Main Results:

• Random Forest (RF) achieved the highest Area Under the Curve (AUC) of 0.96 using combined FC and SC data.
• RF using SC alone yielded an AUC of 0.93, while PLS using FC alone achieved an AUC of 0.86.
• Top 10 predictors for classification were associated with FC, suggesting compensatory mechanisms in grey matter despite white matter changes.

Conclusions:

• Distinct brain connectivity patterns differentiate MS patients with and without VD during a voiding task.
• Functional connectivity (grey matter) is more critical than structural connectivity (white matter) for classifying VD in MS.
• Identifying these connectivity patterns can aid in phenotyping MS patients for centrally focused therapeutic interventions.