Audrunas Gruslys1,
Julio Acosta-Cabronero2, Peter J. Nestor3, Guy B.
Williams4, Richard E. Ansorge1
1Department
of Physics, University of Cambridge, Cambridge, England, United Kingdom; 2Department
of Clinical Neurosciences, University of Cambridge, Cambridge, England,
United Kingdom; 3German Center for Neurodegenerative Diseases
(DZNE), Magdeburg, Germany; 4Wolfson Brain Imaging Centre,
University of Cambridge, Cambridge, England, United Kingdom
Our goal was to develop a fully-automated algorithm capable of detecting and quantifying the progression of brain atrophy in single dementia patients using sequential clinical scans. We used our non-linear image registration program Ezys to measure longitudinal local volume changes between each two scans of the same subject and used regression to estimate yearly volume atrophy at each voxel. We tested our method on 19 subjects: controls (6), Alzheimer's disease (6) and semantic dementia (7). The results were consistent with prior knowledge about each disease progression. Each condition could be visually identified by looking to detected atrophy maps only.