Use of FET PET Uptake Dynamics for Spatial Characterisation of Glioblastoma, By: Dane Lynch

Dane Lynch | Final Masters Presentation | Medical Physics | The University of Western Australia

Use of FET PET Uptake Dynamics for Spatial Characterisation of Glioblastoma

Supervisors:
Martin Ebert (Medical Physics Specialist, Radiation oncology department, SCGH, WA)
Pejman Rowshanfarzad (UWA, Medical Physics Group)

=========================
For more information visit our
website: https://www.uwamedicalphysics.org
Weblog: http://www.uwamedicalphysics.com
Facebook:   / groups  
Instagram:   / uwa_medical_physics  
Twitter: @MedicalUwa
=========================

Abstract
Introduction/Aim: Glioblastoma (GBM) is a highly aggressive brain cancer, that becomes detri-
mental on the lives of those who are diagnosed due to its high mortality and devastating rapid degra-
dation in brain function. Utilising quantitative imaging methods on GBM patient positron emission
tomography (PET) scans can provide deeper understanding on the extend of the disease progres-
sion while highlighting precisely how a patient is responding to treatment on a physiological level.
Development of eectively targeted amino acid radiotracers in brain PET, such as 18F-
uoroethylty-
rosine (FET), has led to the rise of more comprehensive kinetic modelling of tracers in dynamic PET
imaging. A method for calculation of quantitative PET parameters time-to-peak (TTP), in
ux rate
constant (Ki) and volume of distribution (V d) was developed in MATLAB, with the repeatability of
each parameter tested through Bland-Altman analysis.
Methods: A graphical technique for analysis of tracer uptake dynamics was developed and tested on
24 dynamic 18F-FET PET patient image data sets, with 9 of those patients having test/retest scans
5-9 days apart. TTP of the tissue time activity curves was calculated in MATLAB by finding the time
from injection that the first maximum signal occurs. The input function was generated by segmenting
the superior sagittal sinus vein of each brain scan using the earliest PET frames of high plasma-blood
activity in 3D Slicer. A kinetic compartment solution was implemented by extracting the input func-
tion and voxel concentration activities to form Patlak plots. Linear regression was used to solve for the
gradient (in
ux rate Ki) and intercept (volume of distribution V d) of these plots. The repeatability
of these calculated PET parameters was evaluated by calculating levels of agreement (LoA) between
patient test/retest scans through Bland-Altman analysis on a whole volume and voxel-level basis.
Results: Regions of high in
ux constant within the Ki0 maps and high volume of distribution within
the V d0 maps corresponded with tumourous regions of the scans, with the mean tumour Ki0 and
V d0 across patients being 0.061 min􀀀1 (95% CI 0.052-0.071) and 0.88 (95% CI 0.80-0.95) respectively.
Tumour regions had earlier time-to-peak than the background regions of the brain on average, with
the mean whole tumour TTP being 24 minutes (95% CI 22 mins - 26 mins). Tumour LoA ratios were
[0.32, 2.82] for Kimax, [0.82, 1.50] for V dmax and [0.48, 1.47] for TTP. Tumour-to-background ratio
LoA for Ki and V d were [0.62, 1.63] and [0.86, 1.45] respectively. Tumour voxel-wise dierence LoA
was [-0.050, 0.035] for Ki and [-0.21, 0.32] for V d.
Conclusion: The presented approach for calculating parametric maps of dynamic 18F-FET PET
scans and tumour TTP allowed for enhanced spatial characterisation of glioblastoma by highlighting
areas of increased uptake
ux and early TTP, with dynamic image noise and lack of early image frames
being a major source of uncertainty. Volume of distribution V d was found to be vastly more stable
than in
ux rate constant Ki on both a whole volume and voxel-wise basis.