4D offline PET-based treatment verification in ion beam therapy

Due to the accessible sharp dose gradients, external beam
radiotherapy with protons and heavier ions enables a highly
conformal adaptation of the delivered dose to arbitrarily shaped
tumour volumes. However, this high conformity is accompanied by an
increased sensitivity to potential uncertainties, e.g., due to
changes in the patient anatomy. Additional challenges are imposed
by respiratory motion which does not only lead to rapid changes of
the patient anatomy, but, in the cased of actively scanned ions
beams, also to the formation of dose inhomogeneities. Therefore, it
is highly desirable to verify the actual application of the
treatment and to detect possible deviations with respect to the
planned irradiation. At present, the only clinically implemented
approach for a close-in-time verification of single treatment
fractions is based on detecting the distribution of β+-emitter
formed in nuclear fragmentation reactions during the irradiation by
means of positron emission tomography (PET). For this purpose, a
commercial PET/CT (computed tomography) scanner has been installed
directly next to the treatment rooms at the Heidelberg Ion-Beam
Therapy Center (HIT). Up to present, the application of this
treatment verification technique is, however, still limited to
static target volumes. This thesis aimed at investigating the
feasibility and performance of PET-based treatment verification
under consideration of organ motion. In experimental irradiation
studies with moving phantoms, not only the practicability of
PET-based treatment monitoring for moving targets, using a
commercial PET/CT device, could be shown for the first time, but
also the potential of this technique to detect motion-related
deviations from the planned treatment with sub-millimetre accuracy.
The first application to four exemplary hepato-cellular carcinoma
patient cases under substantially more challenging clinical
conditions indicated potential for improvement by taking organ
motion into consideration, particularly for patients exhibiting
motion amplitudes of above 1cm and a sufficiently large number of
detected true coincidences during their post-irradiation PET scan.
Despite the application of an optimised PET image reconstruction
scheme, as retrieved from a dedicated phantom imaging study in the
scope of this work, the small number of counts and the resulting
high level of image noise were identified as a major limiting
factor for the detection of motion-induced dose inhomogeneities
within the patient. Moreover, the biological washout modelling of
the irradiation-induced isotopes proved to be not sufficiently
accurate and thereby impede a quantitative analysis of measured and
simulated data under consideration of target motion. In future,
improvements are particularly foreseen through dedicated
noise-robust time-resolved (4D) image reconstruction algorithms, an
improved tracking of the organ motion, e.g., by ultrasound (US)
imaging, as implemented for the first time in 4D PET imaging in the
scope of this work, as well as by patient-specific washout models.