AI and radiomics for improving the medical workflow for Multiple Sclerosis

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Medicine
• Radiomics

Background:

• Multiple Sclerosis (MS) is a chronic neurological disorder impacting patient quality of life.
• Magnetic Resonance Imaging (MRI) is crucial for MS diagnosis and monitoring.
• Variability in MRI data acquisition and interpretation hinders reproducibility.

Purpose of the Study:

• To design a standardized pipeline integrating AI and radiomics for MS evaluation.
• To improve the precision, reproducibility, and clinical utility of MS assessment using MRI.
• To overcome limitations of current MRI interpretation in MS.

Main Methods:

• Developed a pipeline with preprocessing for data harmonization and synthesis of missing MRI modalities.
• Implemented automatic lesion segmentation and registration with anatomical/connectomic atlases.
• Extracted novel radiomic features quantifying lesion characteristics and involvement of neural structures.

Main Results:

• The pipeline stabilized MRI resolution and contrast, reducing inter-examination variability.
• Eliminated rater dependency in lesion segmentation, enabling consistent radiomic feature extraction.
• New radiomics captured lesion position, orientation, and neural involvement, proving more reproducible than lesion volume.

Conclusions:

• Introduced a structured, reproducible AI and radiomics pipeline for clinical MS workflow.
• The pipeline supports predictive modeling by stabilizing imaging data and enabling advanced radiomic analyses.
• External validation is necessary before clinical application in MS diagnosis or therapy planning.