edoc

Ultrasound-based Motion Modelling for the Lungs in Scanned Proton Therapy

Giger, Alina. Ultrasound-based Motion Modelling for the Lungs in Scanned Proton Therapy. 2021, Doctoral Thesis, University of Basel, Faculty of Medicine.

[img] PDF
Restricted to Repository staff only until 26 January 2022.

47Mb

Official URL: https://edoc.unibas.ch/84616/

Downloads: Statistics Overview

Abstract

Respiratory motion poses a major challenge in image acquisition and image-guided interventions of thoracic and abdominal organs, such as the liver or lungs. In the field of radiotherapy, accurate knowledge of the organ motion is essential for precise radiation of the target volume while sparing surrounding healthy tissue and organs at risk. In this thesis, we present different tools and methods towards ultrasound-guided lung tumour tracking in scanned proton therapy with the main focus on respiratory motion modelling.
We start off with introducing an ultrasound-based 4D magnetic resonance imaging (4D MRI) method for which simultaneously acquired ultrasound and partial MRI data is used to retrospectively reconstruct a time-resolved volumetric MR image. In the following, different motion modelling approaches are presented where 2D abdominal ultrasound images serve as a surrogate signal to predict complete lung motion information. First, we propose a novel approach based on a conditional generative adversarial network (cGAN) in conjunction with a state-of-the-art navigator-based 4D MRI. Second, we investigate the performance of a polynomial regression model when subject to ultrasound probe repositioning as required for fractionated treatments. Third, we propose a motion model based on Gaussian process regression (GPR) and analyse the impact of prediction errors on proton dose distributions with and without tumour tracking. All of these studies are based on simultaneously acquired ultrasound and 4D MRI data sets of two to eight healthy volunteers. For the dosimetric analysis, the motion patterns extracted from 4D MRI of healthy volunteers were combined with computed tomography (CT) scans of two lung cancer patients.
In general, the mean or median prediction error was found to be below 3mm for intra-fractional motion modelling. Moreover, motion predictions based on GPR were shown to translate into clinically acceptable dose distributions, emphasising the great potential of ultrasound-guidance for motion mitigation in scanned proton therapy. From a treatment point of view, however, the dosimetric benefits of tumour tracking were found to be limited. Tumour tracking alone may not always be sufficient to restore clinically acceptable dose distributions and should be combined with other motion mitigation techniques such as rescanning.
Advisors:Cattin, Philippe Claude
Committee Members:Lomax, Antony J. and Göksel, Orçun and Jud, Christoph
Faculties and Departments:03 Faculty of Medicine > Departement Biomedical Engineering > Imaging and Computational Modelling > Center for medical Image Analysis & Navigation (Cattin)
UniBasel Contributors:Cattin, Philippe Claude and Jud, Christoph
Item Type:Thesis
Thesis Subtype:Doctoral Thesis
Thesis no:14424
Thesis status:Complete
Number of Pages:172
Language:English
Identification Number:
  • urn: urn:nbn:ch:bel-bau-diss144243
edoc DOI:
Last Modified:10 Nov 2021 05:30
Deposited On:09 Nov 2021 15:03

Repository Staff Only: item control page