Abstract Details

Title Automated and Manual Hippocampal Segmentation Techniques: A Comparison of Results and Reproducibility

Topic Aging and Dementia

Presentation(s) P06 - (-)

Poster/Presentation
Number 053

Objective

Background BACKGROUND: Hippocampal atrophy is a widely accepted imaging biomarker for Alzheimer's disease (AD). Many semi-automated segmentation techniques have now become available. Most require human expert input to learn how to accurately segment new data.

Design/Methods DESIGN/METHODS: We analyzed the T1-weighted brain MRI data from 155 MCI subjects (mean MMSE: 27.53±1.85) from the ADCS Donepezil/Vitamin E clinical study. In addition to manual segmentations by Rater 1 (AG), we generated two automated segmentations - using manually delineated ADCS training sets by Rater 1 and Rater 2 (JS). We used Cronbach's alpha and Pearson's correlation statistics to compare hippocampal volumes. We also applied the radial distance method to compare how well each segmentation method can detect baseline hippocampal differences between subjects who converted to AD (MCIc) and who remained stable (MCInc). Multiple comparisons correction was applied by using permutation analysis at a threshold p<0.01.

Results RESULTS: Across the three samples, Cronbach's alpha was 0.94 for the right and 0.95 for the left hippocampus. Reliability and Pearson correlation analyses showed significant agreement between manual versus automated segmentations from Rater 1 (right ?=0.87, r=0.78; left ?=0.89, r=0.8), manual versus automated segmentations from Rater 2 (right ?=0.85, r=0.76; left ?=0.88, r=0.79), and automated segmentations from Rater 1 versus Rater 2 (right ?=0.99, r=0.97; left ?=0.99, r=0.97). All three methods performed well and identified baseline differences in the CA1 and subiculum between MCIc versus MCInc as described previously (manual: right pcorrected= 0.0043, left pcorrected= 0.0005; Rater 1 automated: right pcorrected= 0.0217, left pcorrected= 0.0283; Rater 2 automated: right pcorrected=0.0016, left pcorrected=0.0223).

Conclusions CONCLUSIONS: Our automated hippocampal segmentation technique is highly reliable and efficient when applied to large data sets.

Authors/Disclosures
PRESENTER	No disclosure on file
	No disclosure on file
Nicole Chow	No disclosure on file
	No disclosure on file
Johanne Somme, PhD (Hospital De Cruces)	No disclosure on file
Kristy Hwang	No disclosure on file
Clifford R. Jack, Jr., MD (Mayo Clinic)	The institution of Dr. Jack has received research support from NIH. The institution of Dr. Jack has received research support from Alexander Family Alzheimer's Disease Research Professorship of the Mayo Clinic.
Paul M. Thompson, PhD (USC)	No disclosure on file
Liana Apostolova, MD, FAAN (Indiana University School of Medicine)	Dr. Apostolova has received personal compensation in the range of $500-$4,999 for serving as a Consultant for NIH. Dr. Apostolova has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Eli Lilly. Dr. Apostolova has received personal compensation in the range of $5,000-$9,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Roche. Dr. Apostolova has received personal compensation in the range of $500-$4,999 for serving on a Scientific Advisory or Data Safety Monitoring board for siemens. Dr. Apostolova has received personal compensation in the range of $5,000-$9,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Eisai. Dr. Apostolova has received personal compensation in the range of $5,000-$9,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Alnylam. Dr. Apostolova has received personal compensation in the range of $10,000-$49,999 for serving as an Editor, Associate Editor, or Editorial Advisory Board Member for Alzheimer Association. The institution of Dr. Apostolova has received research support from Roche Diagnostics. The institution of Dr. Apostolova has received research support from NIA. The institution of Dr. Apostolova has received research support from Alzheimer Association. The institution of Dr. Apostolova has received research support from AVID radiopharmaceuticals. The institution of Dr. Apostolova has received research support from Life Molecular Imaging. Dr. Apostolova has a non-compensated relationship as a advisor with FDA that is relevant to AAN interests or activities.

��ɫ��

��ɫ��