Weighting for neuroimaging Christina Chen

Format:

Book

Thesis/Dissertation

Author/Creator:

Chen, Christina, author.

Contributor:

University of Pennsylvania. Bioengineering., degree granting institution.

Language:

English

Subjects (All):

Medical imaging.

Biostatistics.

Neurosciences.

0574.

0308.

0317.

Local Subjects:

Medical imaging.

Biostatistics.

Neurosciences.

0574.

0308.

0317.

Genre:

Academic theses

Physical Description:

1 online resource (112 pages)

Contained In:

Dissertations Abstracts International 87-12B

Place of Publication:

Ann Arbor : ProQuest Dissertations and Theses, 2026

Language Note:

English

Summary:

Neuroimaging, especially MRI, constructs informative and convenient brain maps. These rich data sources share common high-dimensional data features that justify classical analysis methods but also introduce unique problems that warrant new approaches. My dissertation comprises three projects that invoke weighting to refine neuroimaging analyses probing biology or disease. My first project proposes a method, called LaxKAT, for detecting brain-behavior associations. LaxKAT confers power advantages and flexibility by accommodating a pre-specified subspace of weight vectors that encode probable signal location or structure. I applied LaxKAT to cortical thickness data in the Alzheimer's Disease Neuroimaging Initiative (ADNI) and showed that it exhibits better statistical properties in a variety of simulation settings and discovers more brain regions exhibiting sex-specific differences compared to other methods. My second project introduces a method to compute subject-specific precision weights for volumetric estimates from structural MRI. This method implements bootstrapping and exploits the unique properties of multi-atlas label fusion. I computed precision weights for hippocampal volume estimates from ADNI subjects and showed that incorporating these weights in regression analyses increases the power to detect mean hippocampal volume differences between different disease groups. My third project extends previous work to enable calibrating standard errors for functional connectivity (FC) estimates from scrubbed fMRI time series. I applied this method, called scrubbing-aware xDF, to resting state fMRI data from the Human Connectome Project (HCP) and showed that weighting by scrubbing-aware xDF standard errors ensures better type I error rate control in individual-level inference and additionally highlights interesting links between FC estimate variability and biological covariates. These projects demonstrate the versatility of weighting for analyzing different types of MRI data and more broadly, promise the potential of adapting classical statistical ideas to understand the brain from neuroimaging data

Notes:

Source: Dissertations Abstracts International, Volume: 87-12, Section: B.

Advisors: Shinohara, Russell T.; Yushkevich, Paul A. Committee members: Wolk, David A.; Shou, Haochang

Ph.D. University of Pennsylvania 2026

Vendor supplied data

Local Notes:

School code: 0175

ISBN:

9798247973263

Access Restriction:

Restricted for use by site license