Unsupervised Pattern Analysis to Differentiate Multiple Sclerosis Phenotypes Using Principal Component Analysis on...

Area of Science:

• Neuroimaging
• Neurology
• Biomedical Engineering

Background:

• Multiple sclerosis (MS) presents with distinct phenotypes: relapsing-remitting MS (RRMS) and progressive MS (PMS).
• Accurate differentiation of MS phenotypes is crucial for tailored treatment strategies but challenging with conventional MRI.
• Existing MRI techniques often struggle to reliably distinguish between RRMS and PMS, impacting clinical management.

Purpose of the Study:

• To investigate the utility of scaled subprofile modelling using principal component analysis (SSM/PCA) for differentiating MS phenotypes.
• To evaluate the effectiveness of various MRI sequences, including myelin-sensitive quantitative methods, in distinguishing RRMS from PMS.
• To determine which MRI sequences provide optimal discriminatory power between MS phenotypes using SSM/PCA.

Main Methods:

• MRI scans were acquired from patients diagnosed with RRMS (n=30) and PMS (n=20).
• Standard MRI sequences (T1w, T2w, T2w-FLAIR) and myelin-sensitive sequences (MTR, qMT, ihMTR, qihMT) were utilized.
• Scaled subprofile modelling using principal component analysis (SSM/PCA) was applied to analyze the MRI data for phenotype classification.

Main Results:

• SSM/PCA analysis of quantitative inhomogeneous MT (qihMT) images demonstrated the highest specificity (87%) and positive predictive value (PPV) (83%) for differentiating PMS from RRMS.
• T1-weighted (T1w) imaging analysis yielded the highest sensitivity (93%) and negative predictive value (NPV) (92%) for identifying RRMS.
• Concordant classification between T1w and qihMT analyses in a subset of patients (57%) significantly improved predictive accuracy (100% sensitivity, 88% specificity, 90% PPV, 100% NPV).

Conclusions:

• SSM/PCA effectively reveals distinct MRI patterns associated with different MS phenotypes.
• Quantitative inhomogeneous MT (qihMT) sequences are optimal for identifying progressive MS (PMS), while T1-weighted (T1w) sequences excel at identifying relapsing-remitting MS (RRMS).
• Combined analysis of qihMT and T1w data through SSM/PCA enhances the accuracy of MS phenotype prediction.