Â鶹´«Ã½Ó³»­

Â鶹´«Ã½Ó³»­

Explore the latest content from across our publications
Join The Â鶹´«Ã½Ó³»­

Log In

Forgot Password?
Create New Account

Loading... please wait

Abstract Details

A Comparison of Automated and Manually-Segmented White Matter Lesions in Primary Progressive Aphasia
Aging, Dementia, and Behavioral Neurology
P10 - Poster Session 10 (5:30 PM-6:30 PM)
10-005
This study aimed to 1) evaluate the performance of an automated, multi-atlas detection localization (MADL) pipeline in segmenting white matter (WM) lesions in individuals with primary progressive aphasia (PPA) and 2) determine which scan factors are related to overlap of automated and manually-delineated lesions. 
Primary progressive aphasia (PPA) is a disorder where progressive left perisylvian cortical atrophy leads to language decline over time (Gorno-Tempini et al., 2011). Leukoaraiosis, or WM degeneration, occurs with aging in ostensibly healthy individuals but may be more severe in neurodegenerative disease, including PPA. Automated WM lesion segmentation pipelines offer better precision, accuracy and reliability than often-used visual rating scales (e.g., Fazekas et al., 1987), yet their performance has not been evaluated for use in PPA.
Forty participants with PPA (24 women; mean age=68.8±8.97 years) completed Fluid Attenuated Inversion Recovery (FLAIR) scans (3T/1.5T: n=24/19; voxel size range: 0.517-4.036mm3). WM lesions were automatically segmented according to the MADL pipeline described in Wu et al. (2019). Two authors (S.P., n=40 and E.M., n=10) manually edited segmentations in ROIEditor. Overlap between lesion volumes was determined using DSI (i.e. 2|x|∩|y|/|x|+|y|). We conducted Pearson correlations to determine relationships between DSI scores and scan/lesion parameters.
Mean DSI between automated and (S.P.) manually-edited volumes was 0.313 (range: 0.006-0.851). Mean DSI between the two sets of manually-edited lesions was 0.784 (range: 0.597-0.924). Higher DSIs were related to greater WM lesion volume (r=0.591, p<0.001) and higher subjective ratings of scan clarity (t=3.095, p=0.004) but not magnet strength (p=0.074) or contrast range (p=0.494).  
The automated pipeline did not adequately segment WMLs compared to manual segmentations, due to lesion (i.e., WM lesion volume) and scan quality (i.e., scan clarity) variables. Future work entails determining if combining semi-automated, manually-edited WM lesion volumes and metrics of cortical atrophy will assist in classifying patients with PPA versus controls.
Authors/Disclosures

PRESENTER 		No disclosure on file
No disclosure on file
Andreia V. Faria, MD, PhD (Johns Hopkins University) No disclosure on file
Argye E. Hillis, MD, MA (Johns Hopkins Hospital) Dr. Hillis has received personal compensation for serving as an employee of Johns Hopkins University. Dr. Hillis 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 American Heart Association. The institution of Dr. Hillis has received research support from NIH. Dr. Hillis has received publishing royalties from a publication relating to health care. Dr. Hillis has received personal compensation in the range of $500-$4,999 for serving as a NIDCD Council Member with NIH.