This ARDC series aims to drive recognition of research software and its authors. Each month, we talk to a leading research software engineer (RSE) and share their experience creating, sustaining and improving software for research.
Professor Susanna Guatelli is an international expert in radiation physics and the use of Monte Carlo simulations to solve medical physics problems, such as how to improve a specific radiotherapy treatment or how to improve technology associated with radiotherapy and imaging. She is also an expert in simulations used in radiation protection to optimise the design of radiation shielding in facilities on earth or in space.
She conducts her research at the Centre For Medical Radiation Physics and teaches Physics in the Faculty of Engineering and Information Sciences in the School of Physics at the University of Wollongong. In 2024 she became the Interim Head of the School of Physics. In 2022 she became a member of the ARC (Australian Research Council) College of Experts. Susanna contributes to the development of Geant4
, one of the most used Monte Carlo codes in the world for radiation physics. Geant4
is developed by a large international collaboration based at CERN, Geneva, Switzerland, and she has been part of this international effort since 2003.
Geant4
is a powerful software tool for studying how particles interact with matter. For example, it can show what happens when particles (including photons, electrons, protons, etc.) hit a human cell or a piece of metal. This is important in designing better cancer treatments with radiation, radiation shielding solutions and improving particle detectors. Think of it as a virtual lab where scientists can test ideas without needing to run expensive or impossible physical experiments.
What are your contributions to develop Geant4
?
Geant4
was born at the end of the 1990’s to support the high-energy physics experiments of international nuclear physics laboratories, including CERN in Geneva. In 2002 while studying a Bachelor of Physics at the University of Genova, I had the exceptional opportunity to work as a Technical Student at CERN, where I conducted research as part of the Geant4 CERN team.
The goal of my project was to help extend Geant4
from high-energy physics to biomedical radiation physics. This included the development of physics models and software functionality needed to study and improve technology used in radiotherapy. For radiobiology applications, I also contributed to the development of state-of-the-art physics models designed to accurately describe particle interactions within biological media at the subcellular level. Since then, in the past twenty years, I have continued to develop and improve physics models and functionality in Geant4
to satisfy the evolving requirements of the biomedical physics research community.
Since 2002, I have been taking on more leadership roles in the Geant4 Collaboration. For example, I have been a member of the Geant4 Steering Board, which coordinates the overall development of this software from 2018 to 2024. Since 2018 I have also been the Coordinator of the Geant4 Medical Simulation Benchmarking Group, which currently accounts for the contribution of more than 50 scientists based at 33 institutions in 12 different countries.
Geant4
is open source and is distributed for free. One use of Geant4 in medical physics is to verify commercial treatment planning systems (TPSs). TPSs are software tools used in the clinics to define and optimise the radiotherapy delivery to the target tumor, while sparing surrounding healthy tissue. Monte Carlo radiation physics simulations, like those based on Geant4, are recognised to be the state of the art in terms of calculation accuracy and they are often used to verify clinical TPSs. However, Monte Carlo simulations are difficult to use directly in a clinical context because of their lengthy execution times.
A few years ago, with collaborators from the Centre For Medical and Radiation Physics and the Center for Artificial Intelligence at the University of Wollongong and from the Technical University Dortmund, we explored the synergy of Geant4
and machine learning solutions to develop a radiation dose calculation engine for Microbeam Radiation Therapy (MRT), which adds accuracy, typical of Monte Carlo codes, and a quick response .
MRT is a novel radiotherapy treatment being investigated at the Australian Synchrotron (ANSTO) in Clayton, Victoria. It uses a highly collimated and very intense beam of electromagnetic radiation emitted by high energy electrons when deflected in their trajectory. It is a very promising treatment option for radioresistant cancers, with the potential to significantly benefit adult patients with glioblastoma multiforme and paediatric patients with diffuse intrinsic pontine glioma.
MRT is still at the research stage and there is no commercial TPS, which will be needed to move from a research stage to a clinical implementation. Therefore we explored the idea of the development of our novel radiation dose engine for MRT, capitalising on the strong research direction focused on MRT at the Centre For Medical and Radiation Physics, University of Wollongong.
How are the radiation treatment simulations being applied?
The core idea, new at that time, was to develop a radiation dose calculation engine based on a 3D U-Net or convolutional neural network, trained with 3D radiation dose maps pre-calculated with Geant4
. Geant4
provided the calculation accuracy, while the neural network model provided the quick response required for a clinical context.
We successfully developed a dose engine[1], and in 2024 we obtained an NHMRC Idea Grant (GNT2028435) to further develop our machine learning engine to verify the treatment planning systems (TPSs) used in head and neck cancer treatments in hospitals in Australia.
What computational techniques are key to driving advancement in biomedical physics?
The Monte Carlo method has become a core instrument in radiation physics since it was conceptualised in the 1940s by Ulam, Metropolis and Von Neumann at the Los Alamos National Laboratory, New Mexico, US. The earliest Monte Carlo simulations of radiation transport for dosimetry are dated to the 1960s, and were in the field of nuclear medicine (Ellett et al., 1964 and 1965). Since then, they have been increasingly used in biomedical physics, medical imaging (e.g. X-ray imaging, computed tomography, positron emission tomography), radiotherapy, radiation protection and radiobiology.
In the past decade, artificial intelligence has become extremely popular in biomedical physics, from medical imaging to radiotherapy. When we started our project on the development of a deep-learning radiation dose engine for MRT about 6 years ago, only a few scientific articles were published in our field of research. We pioneered the application of neural networks internationally and now the scientific literature is rich in this research space.
What would you recommend to other people developing research software to include in successful grant applications?
It is essential to demonstrate how the proposed research software contributes to solving a significant problem for society, beyond the specific research outcome. It is also essential to substantiate a proposed research project with robust and convincing preliminary results.
Developing open source and free software is important for increasing the impact of research within the scientific community and this is often appreciated in grant applications. The challenges associated with maintaining these tools over an extended period of time can also be alleviated by a large community of researchers who support the developed tools.
Another advantage of software development that is often appreciated in grants applications is the opportunities in terms of public outreach to inspire the younger generation to pursue careers in STEM. Software is often an excellent tool for demos at school and public outreach events. It can be used to stimulate curiosity and engage students in scientific and technological disciplines.
Finally, it is important to show the long term vision of the project, in terms of both research development and mentorship/coaching of more junior researchers, to increase the local research momentum to tackle challenges of national interest.
It is essential to demonstrate how the proposed research software contributes to solving a significant problem for society, beyond the specific research outcome … Another advantage of software development that is often appreciated in grants applications is the opportunities in terms of public outreach to inspire the younger generation to pursue careers in STEM. Software is often an excellent tool for demos at school and public outreach events. It can be used to stimulate curiosity and engage students in scientific and technological disciplines.
What Australian computing research infrastructure do you use for your research?
We use National Computational Infrastructure (NCI)’s Gadi supercomputer, accessed via the National Computational Merit Allocation Scheme (NCMAS) grant. NCMAS is Australia’s premiere grant scheme for access to High-Performance Computing (HPC) resources. The two Tier-1 HPC facilities in Australia, the National Computational Infrastructure (NCI) and the Pawsey Supercomputing Research Centre, together offer hundreds of millions of hours of computing time to meritorious researchers.
What communities are you part of and do you recommend?
I am a member of the Australian Institute of Physics (AIP). Anyone with a tertiary education in physics from an AIP-accredited institution can join. I highly recommend becoming a member of this national institute as it offers many opportunities to engage with other physicists, with the aim of collaborating in research, education and public outreach.
If working with Geant4
, it is recommended to follow and participate in the Geant4 Forum, which is a space dedicated to Q&A between Geant4
users and developers on technical matters.
Keep In Touch
You can connect with Susanna via LinkedIn and her UOW profile.
If you’d like to be part of a growing community of RSEs in Australia, become a member of RSE-AUNZ – it’s free!
Research Software Awards Open
The ARDC is proud to sponsor awards for research software and research software engineers in all stages of their careers. The goal of the awards is to strengthen the recognition of research software and those who develop and maintain it as being vital to modern research.
The Astronomical Society of Australia’s Emerging Leaders in Astronomy Software Development Prize, sponsored by the ARDC, is now open for submissions for early-career researchers (ECRs) who have produced or contributed to new open-source astronomy software. Applications close 14 February – learn more and apply now.
The ARDC continues to sponsor a wide range of research software awards for 2025.
The ARDC is funded through the National Collaborative Research Infrastructure Strategy (NCRIS) to support national digital research infrastructure for Australian researchers.
Note
- Read the papers:
- Mentzel F, Paino J, Barnes M, Cameron M, Corde S, Engels E, Kröninger K, Lerch M, Nackenhorst O, Rosenfeld A, Tehei M, Tsoi AC, Vogel S, Weingarten J, Hagenbuchner M, Guatelli S. Accurate and Fast Deep Learning Dose Prediction for a Preclinical Microbeam Radiation Therapy Study Using Low-Statistics Monte Carlo Simulations. Cancers (Basel). 2023 Apr 4;15(7):2137. doi: 10.3390/cancers15072137. PMID: 37046798; PMCID: PMC10093595.
- Mentzel F, Kröninger K, Lerch M, Nackenhorst O, Rosenfeld A, Tsoi AC, Weingarten J, Hagenbuchner M, Guatelli S. Small beams, fast predictions: a comparison of machine learning dose prediction models for proton minibeam therapy. Med Phys. 2022 Dec;49(12):7791-7801. doi: 10.1002/mp.16066. Epub 2022 Nov 10. PMID: 36309820.
- Mentzel F, Kröninger K, Lerch M, Nackenhorst O, Paino J, Rosenfeld A, Saraswati A, Tsoi AC, Weingarten J, Hagenbuchner M, Guatelli S. Fast and accurate dose predictions for novel radiotherapy treatments in heterogeneous phantoms using conditional 3D-UNet generative adversarial networks. Med Phys. 2022 May;49(5):3389-3404. doi: 10.1002/mp.15555. Epub 2022 Mar 3. PMID: 35184310.
Author
Reviewed by
Categories
Research Topic
Related Program
Related Projects
Related News & Events
- Shaping Research Software: An Interview with Anthony Truskinger
- Shaping Research Software: An Interview with Ben Foley
- Shaping Research Software: An Interview with Catherine Bromhead
- Shaping Research Software: An Interview with Cynthia Huang
- Shaping Research Software: An Interview with Dr Adam Sparks
- Shaping Research Software: An Interview with Dr Emily Kahl
- Shaping Research Software: An Interview with Dr Juan Nunez-Iglesias
- Shaping Research Software: An Interview with Dr Juergen Knauer
- Shaping Research Software: An Interview with Dr Kate Harborne
- Shaping Research Software: An Interview with Dr Marcel Keller
- Shaping Research Software: An Interview with Dr Robert Smith
- Shaping Research Software: An Interview with Dr Saras Windecker
- Shaping Research Software: An Interview with Fonti Kar
- Shaping Research Software: An Interview with Johanna Bayer
- Shaping Research Software: An Interview with Lydia Lucchesi and Sam Nelson
- Shaping Research Software: An Interview with Manodeep Sinha
- Shaping Research Software: An Interview with Matthew Sainsbury-Dale and Andrew Zammit-Mangion
- Shaping Research Software: An Interview with Melina Vidoni
- Shaping Research Software: An Interview with Michael Roach
- Shaping Research Software: An Interview with Roozbeh Valavi
- Shaping Research Software: An Interview with Ryan Wick
- Shaping Research Software: An Interview with Sharon Tickell
- Shaping Research Software: An Interview with Varvara Efremova
- Shaping Research Software: An Interview with the MiniZinc Team
- Shaping Research Software: An Interview with the predictNMB Team