Meeting Banner
Abstract #3752

Automated Multi-Atlas Segmentation of Anatomical Brain MR Images from Elderly Subjects.

MAGNA25Aikaterini Kotrotsou1, Niranjini Rajendran2, David A. Bennett2, Konstantinos Arfanakis1, 2

1Department of Biomedical Engineering, Illinois Institute of Technology, Chicago, IL, United States; 2Rush Alzheimer's Disease Center, Rush University Medical Center, Chicago, IL, United States

An increasing number of studies use brain MRI to assess volumetric changes due to neurodegenerative diseases in the elderly. Crucial to these studies is segmentation of anatomical brain MRI data. Multi-atlas segmentation is one of the approaches used for automated labeling; however the performance of this method in subjects with age-related atrophy has not been thoroughly investigated. In this work, the performance of multi-atlas segmentation in data from elderly subjects (>80 years of age) was compared to that of FreeSurfer. It was demonstrated that multi-atlas segmentation provides results that are similar to those of FreeSurfer, in a fully automated manner.

Abstract PDF E-Poster Video

Keywords

according affine agreement almost anatomical appearance applied applying approaches assess assessment atlas atlases atrophy automated biomedical biometrics brain chosen classified coefficient collected conducted consisted corrections crucial dataset datasets degrees dementia described detected disease diseases distribution divided educ elderly engineering equation extraction fast final flirt freedom frequency fully grants history human illustrate illustrates improved included increasing institute investigated investigating label labeled labeling labels linear literature location manner manual maps materials matrices meas medical occurrence often outcome oxford perfect performance plot previously probability propagating published randomly registered registration regression related relationship represents requires resolution restricted rigid rule rush scanner segmentation segmented selected space steps studies subject subjects suggest suggested suggests supported surfer table technology terms thereby thoroughly tissue toolbox transformation type unique validation ventricles volume volumes volumetric vote years