Identifying Response to Therapy in Longitudinal PET Imaging Studies

Student thesis: Doctoral ThesisDoctor of Philosophy


18F-FDG PET can predict response using both qualitative and quantitative measures. PET Therapy Response Assessor (PETTRA) software was developed to allow users to view and analyse pre- and post- therapy images and compute quantitative measures for predicting response to therapy. Additionally, registration methodology was developed to register pre- and post- therapy PET/CT images. The methodology registers pre- and post- therapy PET/CT scans by registering CT scans using customised rigid and non-rigid registration performed by the Image Registration Toolkit (IRTK). Registration success was assessed using qualitative visual analysis and quantitative landmark analysis on a cohort of 20 lymphoma patients. Landmark analysis results found average misalignment on IRTK of ~10mm for rigid registration and ~6.5mm for non-rigid registration, in comparison with ~40mm with no registration applied. The effect of both rigid and non-rigid registration on transformed images was assessed. While rigid registration transformation caused minimal changes on intensity and tumour volume (<2%), non-rigid transformations caused changes of 11% and 21% respectively. PETTRA software was used to analyse quantitative parameters in 14 patients with mesothelioma and 85 patients with diffuse large B-cell lymphoma (DLBCL). For the 14 patients with mesothelioma, a range of parameters were used to assess response including SUVmax, SUVpeak, tumour volume (TV), total lesion glycolysis (TLG) and intensity volume histogram (IVH) parameters. TV and TLG were obtained using 13 fixed and 9 adapative threshold segmentation methods. Pre-and post- therapy SUVmax, SUVpeak, TV and TLG all showed promise in predicting survival. The comparison between TV and TLG obtained using different segmentation methods was negligible. For the 85 patients with DLBCL, SUVmax, TV and TLG struggled to predict response in patients according to ROC curves.
Date of Award1 Jan 2014
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorPaul Marsden (Supervisor) & Derek Hill (Supervisor)

Cite this