Authors:
Jaroslav Rokicki
1
;
Hiyoshi Kazuko
2
;
Francois-Benoit Vialatte
3
;
Andrius Ušinskas
4
and
Andrzej Cichocki
5
Affiliations:
1
Vilnius Gediminas Technical University, Brain Science Institute and RIKEN, Lithuania
;
2
Vilnius Gediminas Technical University and Kyoto University Graduate School of Medicine, Lithuania
;
3
ESPCI ParisTech, France
;
4
Vilnius Gediminas Technical University, Lithuania
;
5
Brain Science Institute and RIKEN, Japan
Keyword(s):
Alzheimer’s Disease, Brain Atrophy, Segmentation of Brain Subnetworks, Hippocampus, Amygdala, Entorhinal Cortex, Multi-volume, Classification, LDA, Early Detection.
Abstract:
Alzheimer’s disease is neurodegenerative disorder believed to affect 24.3 million people worldwide. Proposed
MRI based disease progression markers have shown ability to perform the classification between the
Alzheimer’s Disease (AD), Mild Cognitive Impariment (MCI) and Normal Cognitive (NC) subjects. We exploited
two approaches, first one is to use single sub-network volume as a feature, second to use a network of
most discriminative sub-networks. Multi-feature approach showed improvement by 4.5% in AD/NC classification
case, and 1.5 % in MCI/NC case. Study was summarized for 48 AD, 119 MCI and 66 NC subjects.