Open access
Autor(in)
Datum
2020Typ
- Doctoral Thesis
ETH Bibliographie
yes
Altmetrics
Abstract
In the radiation therapy of cancer, proton therapy offers the opportunity to conform the delivered dose precisely to the tumour, sparing healthy tissues more than conventional photon therapy. This is due to protons delivering most of their dose at a well-defined depth, in the so-called Bragg peak. In order to spread this narrow peak over the whole tumour, in the Pencil Beam Scanning (PBS) technique, the narrow pencil beam is deflected laterally using dipole magnets and the depth of its peak in the patient is varied by changing the proton energy. However, interference between this highly dynamic delivery and respiratory motion of the tumour, for instance if the tumour is in the thorax or abdomen, can lead to substantial distortions of the delivered dose, which need to be mitigated in order to guarantee an effective treatment. One way to do so is to adapt the pencil beam position and energy during the delivery in order to follow the tumour motion, a technique known as ‘tumour tracking’. For this, knowledge of the three-dimensional tumour position in real-time is required, which is the main challenge holding back the clinical implementation of tracking for proton therapy. As such, this thesis investigates the feasibility and benefits of a respiratory motion model, based on real-time abdominal ultrasound, for tracking lung tumours using PBS proton therapy.
This thesis is divided into four parts. Part I introduces the motion problem for PBS proton therapy, and describes the main tools and methodologies used in the work, which are then validated in Part II. Part III then presents three individual studies in which different motion mitigation techniques are described and validated, leading towards the ultimate goal of ultrasound-guided lung tumour tracking. Part IV finally summarises the main findings of the work, as well as it provides an outlook to future work.
After the introductory chapters of Part I, Part II starts with a chapter (Chapter 3) which validates the so-called 4DCT(MRI) approach, which combines CT and 4DMRI data sets to create temporally varying, simulated CT volumes. In this work it is found that under ideal conditions, 4DCT(MRI) can reproduce the original 4DCT with high accuracy. However, careful consideration needs to be given to the quality of the 4DMRI data and the deformable registration approach used to extract motions. Such data sets are important to provide the geometrical representations of the anatomical motion of patients for 4D dose calculations (4DDC), which simulate the delivered dose to the patient. As such, the 4DDC used in this work is experimentally validated in Chapter 4. Under various motion and delivery conditions, the dose is measured using a scintillating CCD dosimeter and a water phantom, and compared to the corresponding 4D dose distributions calculated using the 4DDC. The results show very high agreement between measurements and calculations when both the motion and delivery dynamics are well known. When this not the case, residual differences can however be reduced using motion mitigation techniques. Finally in Part II, Chapter 5 presents a detailed commissioning of the code used to simulate tumour tracking. In this, computational phantoms of increasing complexity have been used to test every component of the beam adaptation simulation code and confirmed that the developed tracking code properly adapts the beam positions and energies under the well known motion and geometrical conditions provided by the developed numerical phantoms.
Part III opens with Chapter 6, which investigates how information about variable breathing patterns affect proton dose distributions and how they can be included in the treatment planning process. As such, a novel, probabilistic target definition approach is introduced and validated, which is shown to provide reliable target coverage whilst reducing dose to the healthy lung. A respiratory motion model for the lung, based on abdominal ultrasound imaging, is then introduced in Chapter 7. This study investigates the geometrical accuracy of such a model and analyses the impact of geometrical errors on proton dose distributions. The results show a good geometrical agreement between model estimation and ground truth motion, which also results in clinically acceptable dose accuracy. Finally, Chapter 8 makes use of this respiratory motion model as an input into proton beam tracking simulations, which are in turn compared to tracking simulations using ‘ground truth’ motions. From this work, ultrasound guided motion modelling has been found to lead to results which are very similar to ‘ground truth’ tracking. However, regardless of the motion information used, tracking alone is not always capable of restoring clinically acceptable dose distributions, especially when the tumour compresses or stretches during respiration. Mehr anzeigen
Persistenter Link
https://doi.org/10.3929/ethz-b-000420475Publikationsstatus
publishedExterne Links
Printexemplar via ETH-Bibliothek suchen
Beteiligte
Referent: Lomax, Antony
Referent: Zhang, Ye
Referent: Dissertori, Günther
Referent: Knopf, Antje
Verlag
ETH ZurichThema
Proton Therapy; Pencil Beam Scanning; Tumour Tracking; Ultrasound-guidance; 4DMRI; Motion Modelling; Lung tumoursOrganisationseinheit
03593 - Dissertori, Günther / Dissertori, Günther
ETH Bibliographie
yes
Altmetrics