Multiresolution Image Analysis
: Innovative applications for Positron Emission Tomography in clinical practice

Student thesis: Doctoral ThesisDoctor of Philosophy


Positron Emission Tomography is an excellent tool to image physiological processes in vivo and it is of great potential when it comes to disease staging for targeted therapies. However, the potential of PET imaging is somewhat limited by its low spatial resolution with resulting significant partial volume effect (PVE) that deteriorates the accuracy of the quantification of the physiological process under scrutiny. In this context, the use of multimodality imaging is very convenient to resolve this limitation. Using novel techniques based on a multiresolution approach, it is possible to recover PET resolution by a synergistic coupling of the PET images with the anatomical counterpart, either CT or MRI. The multiresolution analysis is performed through a wavelet decomposition of both functional and anatomical images which has been used already in the past with similar purposes.

The aim of this thesis is to present novel multiresolution partial volume correction (PVC) techniques that target two different clinical applications. The first part of the project aims to correct for PVE in order to improve the clinical assessment of [18F]Fluoride PET/CT imaging in presence of bone metastasis from prostate and breast cancer. In the second part of the project we develop a different PVC multiresolution approach aiming to improve the quantification of [11C]PIB PET/MR brain myelin imaging in Multiple Sclerosis (MS) patients. The algorithms validation was performed using either phantom data or clinical images of human controls.

The main result of this work is that application of the PVC methodology resulted in a very significant gain in image resolution without any detectable increase of image noise. Lesions sharpness and detectability improved as well with a resulting increase in quantification accuracy. The algorithms developed and presented in this thesis proved to be straightforward tools to improve PET quantification in routine clinical practice.
Date of Award2016
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorFederico Turkheimer (Supervisor), Charalampos Tsoumpas (Supervisor) & Gary Cook (Supervisor)

Cite this