|
Activity Number:
|
352
- Recent Development in Imaging Data Analysis
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Imaging
|
|
Abstract #307227
|
Presentation
|
|
Title:
|
A New Adaptive Signal Detection Method for Neuroimage Analysis
|
|
Author(s):
|
M Inlow* and S Cong and Shen Li
|
|
Companies:
|
Indiana State University and ECE Department, Purdue University and University of Pennsylvania
|
|
Keywords:
|
neuroimage analysis;
statistical parametric map;
cross-validation;
permutation;
random field theory;
ADNI
|
|
Abstract:
|
Here we present a new method for testing the global null hypothesis of no relationship between any voxels and the covariate of interest. The test statistic is the studentized average of all statistics in the statistical parametric map exceeding an adaptively chosen threshold. The threshold is determined using cross-validation so as to maximize the test statistic value. Permutation is then used to estimate the standard deviation for studentizing the average. If the permutation distribution is normal the p-value is computed using the t distribution with degrees of freedom based on the number of permutations. We present simulation study results demonstrating that the new method is substantially more powerful than peak-level and cluster-extent random field theory methods, especially for weak signals. We conclude by presenting results of using our method to detect differences in hippocampal morphometry between healthy controls and early mild cognitive impairment subjects in an ADNI study.
|
Authors who are presenting talks have a * after their name.
