MSc Thesis Defence by Emily Barot


Optimisation of Voxelwise and Surface-Based Preprocessing Pipelines for PET Data


Exploratory analysis has become an integral part of Positron Emission Tomography (PET) research, allowing PET radiotracer binding to be explored in a voxelwise or surfacebased manner, as opposed to a more traditional regionwise analysis. All PET investigations are susceptible to noise, however this is far greater at the voxel-level, making noise management a particularly vital part of the processing workflows used in exploratory PET. Little is known about how the complex data processing workflows used in PET effect study outcomes. This thesis compares 24 different preprocessing pipelines across three different coordinate spaces; subject space (regionwise analysis), FSaverage space (surface-based analysis) and MNI305 (voxelwise analysis). The key finding of this study is that different preprocessing choices effect the bias and variance of PET data, particularly techniques of motion correction and partial volume correction. Additionally, this thesis shows that noise and variation is greater in voxel space than surface space. Suggesting that where possible, exploratory PET should be performed using surface-based approaches. If voxelwise analysis is used then preprocessing pipelines need to be carefully considered in order to correctly manage noise without increasing the bias and variance within the data.


  • Asst. Prof. Melanie Ganz-Benjaminsen, PhD
  • Post Doc, Martin Nørgaard, PhD

Morten Pol Engell-Nørregård,PhD

Melanie Ganz-Benjaminsen,