Description:
Presenter: Prof Lois Holloway and Dr Fahim Alam
Date/Time: Tuesday 16 June at 12:30pm
Zoom Meeting ID: 81207801095
Location: Via Zoom
Title: From Retrospective Radiotherapy Data to Federated Learning: Building AI-based
Solution through ACDN
Abstract:
Radiotherapy generates rich clinical, imaging, treatment planning, and dosimetry data
that can support large-scale learning health systems. However, these data are often
distributed across hospitals, stored in heterogeneous clinical systems, and affected by
variations in local workflows, naming conventions, data quality, and governance
requirements. The Australian Cancer Data Network (ACDN) aims to address these
challenges by enabling structured, multi-centre radiotherapy data extraction,
standardisation, analysis, and privacy-preserving research through federated learning.
This talk will give an overview of our work in developing practical AI and data-mining
workflows within the AusCAT/ACDN ecosystem. This will cover how radiotherapy data
are extracted from oncology information systems, treatment planning systems, and
DICOM archives, anonymised, and transformed into structured research-ready datasets.
Several ongoing research projects, including automated standardisation of radiotherapy
structure nomenclature, cardiac contour comparison will be discussed.
The talk will also discuss our current efforts to support federated learning, where models
can be trained across multiple hospital datasets without requiring centralised sharing of
patient-level data. Overall, the talk will highlight how AusCAT/ACDN provides a platform
for connecting clinical radiotherapy data, machine learning methods, and collaborative
research to support evidence generation and future AI-enabled decision-making in
cancer care.
Bio:
Prof Lois Holloway
Conjoint Professor Lois Holloway leads the medical physics research group at the
Ingham Institute and Liverpool and Macarthur Cancer Therapy centres. She has over
200 peer reviewed publications, has been chief investigator on grants totalling over
$15M and has supervised over 15 PhD students to completion. She holds academic
appointments at University of New South Wales, University of Sydney and University of
Wollongong.
Her research interests are in the use of data and advanced analysis techniques to
provide evidence for patient and clinician decisions. This is particularly important where
evidence from clinical trials is limited due to trial eligibility criteria and/or there is a lack of
representation of particular patient groups within clinical trials. Related to this she has
strong interests in how we can use imaging to improve personalisation of patient care,
through improving the imaging approaches we use, particularly for radiotherapy and
improving the information we can obtain from images.
She leads the Australian Cancer Data Network, a federated data network where patient
data remains at local institutions but can still be learnt from as a combined cohort using
machine learning approaches. She is passionate about multi-disciplinary collaboration
and is always keen to hear from potential new collaborators, both technical and clinical.
Dr Fahim Alam
Dr Fahim Irfan Alam is a post-doctoral research fellow, working on artificial intelligence,
radiotherapy data, and multi-centre clinical data research. He has a background in
computer science and machine learning, with a PhD focused on deep learning-based
image analysis. His current work contributes to the Australian Cancer Data Network,
where he supports the development of infrastructure and analytical pipelines for
extracting, standardising, and analysing radiotherapy imaging and clinical data across
multiple hospitals.
His research focuses on building AI solutions, including automated standardisation of
radiotherapy structure nomenclature, cardiac dosimetry analysis, outcome prediction,
and federated learning approaches for privacy-preserving multi-centre research. He is
particularly interested in bridging the gap between retrospective clinical data, technical
AI development, and practical deployment within hospital-based research environments.
|
