MS Lesion Visualization Assisted Segmentation Daniel Biediger COSC 6397 – Scientific Visualization
Multiple Sclerosis • Autoimmune disease of the central nervous system • Affects 2.5 million people globally, ~100 per 100,000 • Damage to the insulating myelin sheaths around brain cells • Appears as a range of cognitive and physical disability • Progresses at different rates with remission and relapse Myelin: ~40% water, ~40-50% lipids
Multiple Sclerosis • Identified in-vivo with MRI images • Normally 3 tissues: GM, WM, CSF (SG, SB, LCR) • Lesion appear hyper- or hypo-intense (bright or dark) • Reflects a difference in relaxation times for lipids/water • Detected as an outlier to the normal WM • FLAIR MRI (Fluid Attenuated Inversion Recovery) best
MS Detection Challenges • WM Lesion detection is a challenge • Overlap in intensity between tissue classes • 3D brain structure (sulcus, ventricles, multiple tissues) • Partial Volume Effects (PVE) • Artifacts from fluid/patient motion, tumors, etc • Random noise, bias field, sun spots, etc • Requires an expert to identify lesions • Tissue identification (T1-weighted MRI) • Lesion detection (FLAIR MRI) • Time consuming, difficult (multi-modality), and subjective
MS Lesion Appearance in Flair This? Is this? or This? or This?
Result Comparison FLAIR MRI Base Method Expert Me Under Segmentation Correct Segmentation Over Segmentation Results DICE = 0.65