Statistical Analysis Of Image Features

National Institute On Alcohol Abuse And Alcoholism

Investigators

Linked publications & trials

Abstract

Currently three major results have been developed with regards to Fourier statistical analysis developed for fMRI.[unreadable] [unreadable] First, statistical analysis in the temporal domain based on traditional time series analysis in the Fourier domain have been developed and given similar results in terms of localization of the signal in fMRI blood flow studies to other less rigorous and generalizable techniques. Its advantage over these other less rigorous techniques is that noise in the fMRI data is modeled more generally and closer to that actually seen in the data. Any input function is allowed regardless of the timing (unlike that for de-convolution methods in the time domain). Thus allowing statistically correct analysis of data acquired at very fast image collection rates (for example TR = 400 msec) as previously demonstrated as well as that for data collected at the more standard rate, TR = 2 seconds. Also, non-parametric estimates of the transfer function are part of this statistical procedure. Finally this statistical theory is both rigorous and transparent as to its mathematical assumptions and implementation, and appropriate for single subject analysis.[unreadable] [unreadable] Secondly, this technique is being used to analyze experimental designs that have slides of visual stimuli designed to elicit different emotions or alcohol craving in normal and alcoholic subjects under drug and placebo conditions. Furthermore this technique has been extended to incorporate a full complex general linear model. In particular statistical inference test have been applied to both multiple input and multiple output (multiple fMRI runs on the same subject under different conditions) fMRI data with striking results for single subjects.[unreadable] Thirdly, this statistical methodology has been extended to measure partial coherences for various structures in the brain after correcting for the linear effect of the input stimuli. This methodology gives insight into which brain structures are processing information among themselves during a specific task.[unreadable] [unreadable] Finally, novel research into using the single subject results in analysis of groups of subjects is being investigated. This methodology based on non-parametric, randomization statistical techniques will be developed next.