IVIM-DWI-Based Radiomics for Lesion Phenotyping and Clinical Status Prediction in Relapsing-Remitting Multiple...

Area of Science:

• Radiology and Medical Imaging
• Neuroscience
• Artificial Intelligence in Medicine

Background:

• Multiple sclerosis (MS) is a central nervous system autoimmune disorder causing myelin degradation and neurological symptoms.
• Current assessment of MS disability relies on clinical scales like the Expanded Disability Status Scale (EDSS).
• Radiomics and advanced imaging analysis offer potential for objective disease assessment.

Purpose of the Study:

• To evaluate the predictive value of Intravoxel Incoherent Motion (IVIM) diffusion parameters (D, D*, f) using radiomics.
• To correlate IVIM parameters with disability severity (EDSS) and mobility in relapsing-remitting MS patients.
• To explore the application of machine learning (ML) for predicting MS progression and functional impairment.

Main Methods:

• Retrospective analysis of MRI data from 197 relapsing-remitting MS patients.
• Quantitative IVIM parameter calculation using a 1.5 Tesla MRI scanner.
• Application of ML algorithms (XGB, Random Forest, ANN, RNN) to correlate IVIM radiomics with clinical data (EDSS, mobility, disease duration, DMT status).

Main Results:

• IVIM radiomics demonstrated high accuracy in lesion phenotyping (e.g., Random Forest: 96% accuracy for enhancement detection).
• IVIM-D and D* radiomics strongly correlated with EDSS, with CNN achieving 90% accuracy (AUC=0.95).
• Mobility impairment prediction was highly accurate (RNN: 96% accuracy, AUC=0.99) using IVIM-f features.

Conclusions:

• ML analysis of IVIM metrics provides independent predictors of functional impairment and disability in MS.
• This novel radiomic approach shows promise for enhancing diagnostic accuracy in MS.
• The findings support the development of personalized treatment strategies for MS patients based on objective imaging biomarkers.