Detection Limits of Automated MRI Morphometry in Rodent Brains (University of Cambridge, 2010)

Overview

This project investigated the detection limits of automated MRI morphometry for identifying structural brain changes in rodent models of neurological disease. The work was conducted during a competitive summer research appointment at the Wolfson Brain Imaging Centre, University of Cambridge, under the supervision of Adrian Carpenter and Steve Sawiak.

I presented this work as a poster at the Amgen Scholars European Symposium 2010, held at the University of Cambridge, as part of the global Amgen Scholars Programme, which had an acceptance rate of approximately 6% that year.

Methodological Focus

The project focused on evaluating how effectively voxel-based morphometry (VBM) can detect subtle, spatially localized brain changes in high-resolution rodent MRI, a common challenge in preclinical phenotyping studies.

Using simulated 3D phantom images with controlled morphometric deformations, we systematically studied how key experimental and statistical factors influence VBM sensitivity and specificity, including:

• Effect size and spatial extent of anatomical changes
• Image smoothing parameters
• Sample size
• Noise characteristics and anatomical variability

This simulation-driven approach allowed us to isolate methodological limits independently of biological confounds.

Experimental Framework

The poster detailed a complete pipeline for:

• Generating synthetic MRI datasets with known ground-truth deformations
• Preprocessing and segmentation of rodent brain MRI
• Performing voxel-wise statistical inference under varying experimental conditions

By varying one factor at a time, we quantified how design choices impact statistical power and false-positive behavior in small-animal morphometry studies.

Contributions and Impact

This work provided methodological guidance for designing and interpreting rodent MRI phenotyping experiments, clarifying when automated morphometric pipelines are likely—or unlikely—to detect subtle neuroanatomical changes.

The results helped inform:

• Preclinical imaging study design
• Interpretation of negative findings in rodent MRI experiments
• Limitations of VBM-based approaches for small effect sizes

More broadly, the project contributed to understanding how algorithmic sensitivity, experimental design, and biological variability interact in quantitative neuroimaging.