Premium
P3‐251: Novel Toolbox for Performing Voxel‐Wise Generalized Linear Regression with Mulitple Volumetric Covariates in Longitudinal Data
Author(s) -
Mathotaarachchi Sulantha S.,
Wang Seqian,
Shin Monica,
Pascoal Tharick A.,
Benedet Andrea Lessa,
Kang Min Su,
Beaudry Thomas,
Fonov Vladimir S.,
Gauthier Serge,
Labbe Aurelie,
Rosa-Neto Pedro
Publication year - 2016
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2016.06.1914
Subject(s) - covariate , computer science , generalized linear model , binomial regression , linear regression , regression analysis , voxel , artificial intelligence , data mining , statistics , machine learning , mathematics
additional scans not used in training. Results: Both Early Frame Amyloid (EFA) classifiers produced a pattern that captured a late timeframe-like amyloid accumulation pattern among its features (Figure 1). Leave-One-Out testing of the Am+/classifier produced perfect separation between groups (Figure 2). Among additional independent test scans, EFA classification agreed with late timeframe PET measurement even when CSF was negative or SUVR was threshold. Test results of the amyloid progression classifier produced a cascade that increased with increasing late timeframe SUVR (Figure 3). Conclusions:These preliminary results show the ability to obtain, using discrete dynamic frames and multivariate discriminant analysis, a measure of amyloid burden within a short timeframe post-injection that can complement the functional information obtained through the earliest frames of the same scan. This offers the potential to characterize, in an efficient manner, the amyloid status and neurodegenerative progression that are key determinants of preclinical and prodromal staging.