| dc.description.abstract | Non-melanoma skin cancer (NMSC) can be treated clinically using Photodynamic Therapy (PDT), a nurse-led treatment method which is an attractive alternative to invasive removal methods, which are currently common practice. PDT is significantly cheaper, easier for the patient and notably does not result in cosmetic damage. The treatment works by first applying a topical exogenous photosensitiser (PS) which is then selectively uptaken by malignant cells. The PS in the presence of molecular oxygen produces reactive oxygen species (ROS), which cause local cytotoxic damage, when exposed to light of an appropriate wavelength. Targeted removal of the lesion is achieved as the damaging ROS are localised to malignant cells, leaving healthy tissue relatively unharmed.
However, the efficacy of PDT is limited to thin (< 2mm) lesions. Penetration of activating light is limited by the optical properties of biological tissue, which are strongly scattering at the visible wavelengths typically used during PDT. Availability of molecular oxygen also may become a limiting factor during treatment, and the concentration of PpIX (the PS molecule) is not assumed to be uniformly distributed throughout the skin. The PpIX molecule itself forms part of a cell’s natural synthesis cycle, and so is naturally converted into Heam (the end product) when Iron (the other component) is available. Chelating agents (such as CP94) may also be topically applied to remove free iron from the local volume, suppressing the final step of Haem synthesis and allowing PpIX levels to accumulate.
Practical investigations into PDT are common, and the fundamental aspects of the reaction mechanism are well understood. Whilst clinical studies will always remain the final and most important stage of investigation, they are limited in terms of how many can be conducted, and what may ethically attempted in the first place. Measurement and analysis in a practical setting is also limited by the nature of reality making measurements in practice is usually an invasive procedure in and of itself. The end result is that detailed analysis of a system (beyond a limited number of scalar measurements) is not practically possible. Alternatively, numerical methods allow us to form the well understood details of the system the reaction mechanisms and the way in which species travel throughout the system and structure them into a model which we can continuously measure (with omnipotence) as it evolves through simulated time. After calibration, simulations (the product of the numerical method and the model) allow us to run experiments beyond those that can be conducted practically. The detailed output of these simulations may then be used to investigate the inner dynamics of the system, which in turn lead to better informed treatment plans in future practice.
This thesis concerns the development of a novel library of code (completely documented) which provides the framework for several simulations and tools, which themselves are capable of simulating the entire PDT system in a fully three-dimensional skin model. The PDT mechanism is a complex processes which requires modelling of chemical interactions and diffusion, as well as light scattering in turbid media. The three of these mechanisms require different simulation methodologies, and must all interact fluidly with one another and the underlying input model. These simulations should also scalable, capable of running a complete model within an hour on a clinic laptop, up to HPC workstations with high-core CPUs and substantially more memory which can produce highly-resolved data for detailed analysis. Altogether these tools can be used to create a complete virtual laboratory, which simulates the PDT treatment process in its entirety, supporting the investigation into how PDT may be improved in general practice. Parameterisation of the model is straightforward and simple to update as empirical values are refined. Furthermore, if a patients particular presentation can be measured to a high enough level of fidelity then a model can be parameterised specifically for their unique case opening the door to informed personalised medicine.
Source code and parameterisations are presented alongside the written document of the thesis, and can be downloaded here. Details of what’s happening will be discussed in context during the later chapters, but for the impatient running the Bash script in the /input folder will first download the dependencies (NetCDF, HDF5 and the Rust compiler), and then proceed to produce all data afresh under the /output directory.
iii
In short: first we will explore the pathology and histology of NMSC, and the benefits PDT offers as a method of treatment and how it operates on a biological, chemical and physical level. Secondly, the theory behind each of the numerical methods utilised is described and explored such that a future researcher could derive even better software without viewing any of the presented code. Thirdly, the implementation of these numerical methods within the general arctk library is then discussed, before being composed into the numerical software tools which are actually used to realise a set of novel results. The fourth chapter explicitly walks through how each of the tools may be parameterised and combined into a complete pipeline which models the standard PDT protocol for a selection of tumours of different penetrating depths. Chapter five subsequently applies these tools in a photo-thermal (PTT) setting, investigating how surface tumours may be treated using gold nanorods (GNRs) to locally modify the absorption optical properties of a lesion such that exposure to thermal photons induces hyperthermal cell death. Chapter six investigates how the standard PDT method may be improved to allow treatment of thicker lesions, how the light source may be modified spectrally to increase efficacy at depth, how fibre optics may be introduced to increase light fluence at depth, and how chelating agents may be used to increase the threshold concentration of PpIX within the skin, resulting in increased ROS generation at depth. Finally chapter seven will discuss some of the other ways that the arctk library is currently being used in research with Jeynes’ investigations into PTT, Moran using Raman spectroscopy to develop a non-invasive method of breast cancer detection, and Morell using the MCRT tool to determine the impact of artificial light on ecology. We will then conclude with some remarks on how machine learning, in particular Gaussian processes, may be used to better chart the parameter space which simulations allow us to explore in the first place.
To the best of our knowledge, no prior work has captured the skin model geometry and the evolution of constituent chemical species to the same level of detail as presented within this work. The numerical methods detailed allow for a robust and extensive investigation of the problem, and the software implementation enables results to be computed within relatively short timescales, something which has also not been previously achieved. The results of these simulations provide significant contributions to the state of medical physics as they allow us to quantify the impacts of an altered treatment method, in an accurate
iv
and repeatable manner. | en_GB |
