Improving machine learning classification of Alzheimer's disease and normal pressure hydrocephalus using MRE-derived compressibility metric

Motivation: Alzheimer's disease (AD) and normal pressure hydrocephalus (NPH) are dementia-causing neurologic disorders. NPH is highly treatable whereas AD is not. Clinically, it is important to differentiate the two disorders. Goal(s): To improve the MR Elastography-based classification of AD and NPH cases. Approach: Apply regional analysis of MR Elastography-derived mechanical properties (shear stiffness and compressibility metric) of the brain as features in a machine learning classifier to distinguish between AD, NPH, and healthy participants. Results: The derived compressibility metric is a complementary mechanical property to the shear stiffness for improving the classification of AD and NPH cases. Impact: We used a lobar atlas based regional values of MR Elastography derived mechanical properties to improve machine learning classification of cognitively unimpaired controls, and normal pressure hydrocephalus and Alzheimer's dementia patients.