Thesis Format

Monograph

Optimizing Preprocessing of fMRI Data to Maximize Correspondence of Functional Anatomy Across Individuals

Nargess Ghazaleh, The University of Western OntarioFollow

Degree

Master of Science

Program

Neuroscience

Supervisor

Johnsrude, Ingrid

Abstract

In movie-activation fMRI, intersubject correlation (ISC) indicates a functional correspondence across viewers. Brains di↵er in shape; spatial normalization and smoothing enhance inter-subject functional overlap. We compare three normalization methods and six smoothing levels to discover which method yields the best functional overlap, indexed by ISC. This is key to optimizing data analysis in clinical studies using movie-activation fMRI in future. In a 3T scanner, 44 healthy subjects watched an 8-min movie. Both normalization and smoothing a↵ected the strength and extent of the ISC. ISC values were more robust for ANTs and DARTELthanforSPM12andwere(asymptotically)thestrongestat12mmsmoothing. When image data are preprocessed with high-dimensional volumetric spatial registration methods, such as ANTs, and 12mm smoothing, the sensitivity of the movie-fMRI paradigm will be optimal for detecting abnormalities in presurgical evaluation of neurological patients, providing lessvariableandmorereliableISCmeasures.

Summary for Lay Audience

Almosttwoinevery1000oftheworldpopulationhasaneurologicdisorderthatrequiresbrain surgery. E↵ective surgical treatment requires that the lesion be localized precisely, and resection does not result in unexpected cognitive decline. It is usually accomplished through preoperativeassessmentofpatientsusingimplantedelectrodesorfunctionalMRI(fMRI). Current methods are not very sensitive to abnormalities in cognitive function. We propose to use fMRI to rapidly and noninvasively identify and map brain function. When neurologicallyhealthyindividualswatchanengagingﬁlmwhileundergoingfMRI(movie-drivenfMRI), reliable and distinctive time-locked ﬂuctuations in fMRI signals are observed throughout the brain. These ﬂuctuations are di↵erent in di↵erent brain regions, reﬂecting the perceptual and cognitivedemandsoftheﬁlm,butareverysimilaracrosspeoplewithoutneurologicaldisorders (controls). The degree to which a patient’s data matches predictions derived from control data can be tested statistically, region by region, to reveal functionally important cortex and local abnormalities of brain function. Because di↵erent brains di↵er in shape and size they should beregisteredtoastandardtemplateforcomparison. Thewaythepatient’sdataisregisteredto thetemplatea↵ectsthesensitivityofthemethodtoabnormalities. We asked people without any neurological disorder to watch a short, suspenseful movie while they underwent a functional brain scan. We - for the ﬁrst time - evaluated three di↵erent ways to register the movie-driven fMRI data. We found that the registration method that usesmorepointsindatatomatchitwiththetemplatewasmoresensitivetoﬂuctuationsofthe fMRI signal and as a result more sensitive to abnormalities. We also compiled this data set as anormativepatternofactivityinaneurologicallyhealthypopulation. Movie-driven fMRI and this method of registration may be a valuable adjunct assessment method in the presurgical evaluation of neurologic patients. Future studies can compare the fMRI data from any individual patient with a neurologic disorder, such as focal epilepsy, with thisnormativepatternofbrainsignalﬂuctuation.

Recommended Citation

Ghazaleh, Nargess, "Optimizing Preprocessing of fMRI Data to Maximize Correspondence of Functional Anatomy Across Individuals" (2020). Electronic Thesis and Dissertation Repository. 7286.
https://ir.lib.uwo.ca/etd/7286