University of Wollongong
Browse

Development of Novel Dose Quantification Methods for Heavy Ion Radiotherapy Using In-Beam Positron Emission Tomography Imaging

Download (6.71 MB)
thesis
posted on 2025-01-22, 05:03 authored by Harley J. Rutherford

Ion therapy is a radiotherapy technique that employs accelerated ions, such as helium or carbon, to deliver a therapeutic dose to a target volume. This method delivers a highly conformal dose distribution, resulting in steep dose gradients surrounding the treated area. Consequently, minor errors in treatment delivery can lead to substantial discrepancies in dose delivered both to the target volume and healthy tissue.

During helium and carbon ion therapy, some ions in the beam undergo nuclear inelastic collisions with atoms along the beam path, generating various fragment particles, including positron-emitting radioisotopes. Images of the distribution of positron-emitting fragments can be acquired using positron-emission tomography, which can be compared to a predicted distribution to verify the treatment location. However, a more clinically relevant and valuable measurement of treatment quality is the quantitative estimation of the deposited dose. Calculating the deposited dose using distributions of positron-emitting fragments obtained during and immediately after helium or carbon ion therapy is challenging due to the non-linear relationship between the two quantities.

This Thesis evaluates and compares two methods for dose quantification in heavy ion therapy using the distribution of positron-emitting fragments. First, an iterative dose estimation procedure is developed for carbon ion therapy and evaluated using both Monte Carlo simulations and experimental irradiations in a homogeneous medium. This method is then extended to experimental helium ion irradiations to assess the model’s performance for the case where positron-emitting fragments are created only through target fragmentation. A complementary study examines the impact of heterogeneities in the treated volume. Subsequently, a deep learning approach, based on an InceptionTime-InceptionResNet hybrid model is then designed and implemented for simulated and experimental carbon ion treatments. The two dose quantification approaches are compared and contrasted based on accuracy and speed.

History

Year

2024

Thesis type

  • Doctoral thesis

Faculty/School

School of Physics

Language

English

Disclaimer

Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.

Usage metrics

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC