Premium
P2‐058: Rubidium is altered from early stages of AD: A postmortem study using neutron activation analysis
Author(s) -
Leite Renata E.P.,
Mitiko Saiki,
Grinberg Lea T.,
Ferretti Renata E.L.,
Farfel José M.,
Santos Erika B.,
Alho Ana T.D.L.,
Andrade Mara P.,
Polichiso Livia,
Santos Edilaine T.,
Lima Maria C.,
CaetanoJunior Antonio,
Oliveira Katia C.,
Pasqualucci Carlos A.G.,
Nitrini Ricardo,
JacobFilho Wilson
Publication year - 2009
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.2009.04.368
Subject(s) - vascular dementia , dementia , hippocampus , clinical dementia rating , postmortem studies , neutron activation analysis , medicine , pathogenesis , alzheimer's disease , pathology , disease , chemistry , radiochemistry
Olshen, and Stone, 1983) provides a series of selected predictors and cut-off values to classify subjects into two or more groups. Classification trees provide an easily interpretable classification rule and are adept at handling large dimensional problems. Classification trees are also invariant to monotone transformations of the covariates. Trees were fit assuming ‘‘unit cost,’’ or equal cost for each type of misclassification error, and subsequently assuming misclassification costs that favor, in turn, sensitivity and specificity. We used the ‘‘1 SE’’ rule (Breiman et al., 1983) to prune the decision trees to find the simplest trees which best identify the converters. Results: Table 1 summarizes the various classification trees’ characteristics for predicting the 128 MCI to AD conversions among the 383 MCI subjects using 1) all available baseline data, 2) baseline biomarker data, and 3) all available baseline and contemporaneous follow-up data. Figure 1 gives the details of the unit cost tree for MCI to AD conversions considering 519 baseline biomarker and imaging variables. Figure 2 give the details of a high specificity tree considering 2,770 baseline and follow-up variables. Conclusions: CART is an effective tool for mining good predictors of conversion out of the high dimensional ADNI data set. In designing the HBA study, CART was useful in confirming the importance of functional measures in signally probably conversions in the MCI population. CART is also a useful tool for finding the biomarker and neuroimaging covariates which best predict disease progression.