Harmonized Z-scores obtained from the large scale normal MRI databased to evaluate brain atrophy for neurodegenerative disorder

Abstract

Alzheimers disease (AD), the most common type of dementia in elderly individuals, slowly and progressively diminishes cognitive function. Mild cognitive impairment is also a significant risk factor to the onset of AD. Magnetic resonance imaging (MRI) images have become widely used to detection and understand the natural progression not only AD but also neurodegenerative disorders. For this purpose, construct a reliable cognitive normal database is important. However, difference in magnetic field strength, sex and age between normal database and evaluation data-set can be affect the accuracy of detection and evaluation of AD and other neurodegenerative disorders. To solve this problem, we suggest a harmonized Z-score considering differences filed strength, sex and age derived from large cognitive normal subjects dataset ((1235 subjects)) including 1.5 T and 3T T1 brain MRI. And we evaluate our harmonized Z-score of discriminative power of AD, and classification accuracy between stable MCI and progressive MCI. The harmonized Z-score of hippocampus achieved high accuracy (AUC=0.96) for detection AD and moderate accuracy (AUC=0.70) for classification stable MCI and progressive MCI. Theses results shows that our method not only can detect AD with high accuracy and high generalization capability, but also can be valid to classify stable MCI and progressive MCI.