Prof. Sowmya from CSE is a collaborator on the recently successful ARDC Platform grant entitled `Australian Cancer Data Network: distributed learning from clinical data’ led by Associate Professor Lois Holloway, UNSW and Ingham Institute for Applied Medical Research.
Professor Arcot Sowmya from School of Computer Science and Engineering is a collaborator on the recently successful ARDC Platform grant entitled `Australian Cancer Data Network: distributed learning from clinical data’ led by Associate Professor Lois Holloway, University of New South Wales and Ingham Institute for Applied Medical Research.
The project establishes a nationally agreed capability to link regular treatment (clinical practice) and clinical trial data, for machine learning analysis with international links. It will improve accessibility and governance structures to support data users including clinicians, data scientists, governments and policy makers. It will leverage the scale and accessibility of an existing network in which data is decentralized and the analysis is taken to the data. It will streamline the administrative, ethical and political aspects of research, enabling learning across datasets that are otherwise difficult to merge due to size or ethical and governance challenges. It will use tools, knowledge and expertise from ARDC 2019 platform projects, and ensure cohesion with the Australian clinical trials data platform and the ARDC health studies data program. The value of the network will be demonstrated in radiation oncology, linking national and international centres and registry datasets.
This project will enable learning from locally stored patient level clinical practice data, linked with clinical trial data and international data in a distributed network. Learning across jurisdictions from higher level datasets (e.g. across NSW, VIC, SA, WA and QLD) will also become possible. It is well recognised that outcomes from clinical trials and clinical practice are different, however it is challenging to determine the influencing factors. Large variation in clinical practice is also well recognised. This distributed network enables these questions to be addressed, highlighting clinical practice variation and providing additional clinical evidence and health services analysis to inform service improvement.
This distributed learning approach enables a unique opportunity to learn from data across Australia and internationally. We facilitate the ability to learn from other international datasets and importantly international researchers can learn from Australian datasets. Australia gains from a broad international collaboration of experts who will strengthen the Australian data platform and strengthen our understanding of Australian data with likely gains in Australian clinical outcomes, health service support and data science expertise. This platform will enable a large range of projects to be undertaken by researchers, clinicians and policy makers from across Australia using national data. The outcomes of those projects will have significant social impact, potentially influencing the treatment of 65,000 Australians annually.
Professor Sowmya will be on the Technical Advisory Board as well as a Key user of the developed platform. For the latter, she will work on two different case studies:
(i) Multimodal learning from imaging and clinical datasets: This case study will draw upon distributed radiotherapy datasets for H&N cancer and lung cancer on the OzCAT network to learn a multimodal classification model for detection, staging and other relevant outcomes such as survival and recurrence.
(ii) Machine Learning based survival analysis from multimodal heterogeneous datasets: This case study with utilize multimodal heterogeneous radiotherapy datasets for H&N cancer and lung cancer on the OzCAT network to develop machine learning based survival analysis models. It will also analyse factors that differentiate between trial and clinical datasets, and thereby develop more robust models for clinical data that bootstraps from the trial datasets.