Development of a machine learning approach for integrated medical diagnostic methodology using medical CT image analysis for liver diseases

About this project

The project aims to develop an integrated medical diagnostic methodology using machine learning to analyse medical CT images, identify abnormalities in organs, especially liver, and provide a consistent reporting system accordingly. Currently, medical imaging modalities are used to assist radiologists to make an accurate diagnosis for patient care and management of disease processes. While various machine learning tools are applied peripherally, the central outcome of image analysis in all existing research is focused on assisting diagnostic decision making. There is no doubt that it is critically important to have an accurate diagnosis in the first place. However, the practical application of such results for radiologists in the clinic requires to go one step further: fast and consistent reporting. The recent research conducted in UK shows that radiologists spent 40% of their time on reporting. Therefore, improving report turnaround time (TAT) is as significant as accurate diagnosis itself as key performance objectives of hospital radiologists. The industry partner in this project has recently developed a keyword-based search tool to address this issue by utilising the existing templates of similar cases stored in their database. Even though this can reduce TAT in some degrees, incorporating parameters identified in medical images is found to be necessary to provide faster and more accurate reporting. Therefore, this research will integrate two critical aspects of diagnostic system using medical images: analysis of images to make accurate diagnosis and data mining of image parameters for systematic reporting using a case study of CT images of upper body, especially liver. We expect that this result will provide a significant breakthrough in practice for radiologist in the clinic. This project is innovative since this is the first trial to integrate the diagnostic scheme to the consistent reporting system for the practical use by radiologists.

This project is closely related to Health, one of Science and Research priorities of Australian government. Accurate diagnosis and consistent reporting of health issues using medical images will greatly contribute to build healthy and resilient communities. The routinely measurement of parameters of essential organs, for example, reliable quantification of human organs, especially liver volumes, will be game changer for health assessment by allowing early medical intervention and much improved patient outcomes. It will provide a better health care and service model with improvement of diagnosis, identification and management of major health issues. The project will also lead to more efficient and more strategic allocation of time for radiologists by reducing their time for reporting. The project also fits well with UniSA’s strategic research stream, healthy futures. Automated and accurate detection of the early changes in every medical image as well as obtaining a documented baseline for future studies in the same individual to assess onset and progression of disease as well as prognosis after treatment of any related disease process would be very valuable across the whole patient cohort undergoing abdominal scans for any reason. It will provide a better and more accurate system for health care that can reduce potential health threats and provide more efficient and positive therapeutic outcomes. This is a project with global importance that should not be solely limited in the research environment. The project outcome will be outreached to and engaged with broader global medical industry (both government agencies and private companies) to be used in real world applications.

What you’ll do 

The industry partner of this project, Adelaide MRI, maintains a library about 20 million medical image set, catalogued with unique digital labels. Using this unique dataset, the project will produce the following main outcomes using a case study of CT images of upper body (about 200K images available): 

a) A methodology to segment multiple organs and identify necessary parameters: like shape, size, textures and volume for accurate diagnosis. A machine learning algorithm (potentially, deep neural network) is developed to automatically detect boundaries of target organs, especially liver, obtained from the CT images provided by Adelaide MRI. 3D volumetric measurements to calculate absolute organ volumes, which is largely ignored up to now due to time consuming and inaccurate nature of process, will also be included in the process. Our initial investigations confirmed that the boundaries of various organs are highly predictive to be segmented, even though it is still a difficult task. All CT images in the current database will be analysed and labelled according to the abnormalities identified by the proposed methodology. 

b) A systematic data mining algorithm for fast and systematic reporting: The current keyword-based search engine patented by the industry partner will be further developed to incorporate parameters found in step a). 

Existing templates of similar cases found in database by this algorithm will be retrieved and used with minor modification. This will provide more consistent and faster reporting, actually used by radiologists. This makes this project innovative and unique. The outcome of this project will be integrated to the current report system and utilised by medical experts in AdelaideMRI. This outcome will give huge advantage to save time to correctly assess patients and keep consistency in reporting to provide accurate diagnosis. This outcome can be easily extended to other radiologists in the fields.

Where you’ll be based 

Associate Professor Sang-Heon Lee has worked in the machine vision systems, machine learning and mechatronics for more than 20 years. He will be the principal supervisor of the project and provide valuable advice to develop a scheme to analyse the image data and develop a machine relearning algorithm. He has published about 170 peer-reviewed papers in journal and conferences in the related fields. 

Dr. Ivan Lee will also involve in the project as a co-supervisor of the project. His research focuses are machine learning, data mining and scholarly data analytics. He will provide critical advice to the development of reporting system and machine learning algorithm development. He has published about 100 journal or conference papers in the related fields. A/P. Ivan Lee has published about 15 journal or conference papers. 

Dr. Roger Davies is Associate Professor in Clinical Radiology at the Sydney University School of Medicine and also the CEO of the AdelaideMRI. He has more than 30 years of extensive experience as an interventional radiologist and 20 years of paediatric imaging. Over the past decades he has performed more than 100,000 radiological interventions for pain assessment and management. He will provide valuable information on the analysis of medical images.

Dr Hye Won Jung has worked in AdelaideMRI since 2021 as a professional medical researcher. His main research focuses are machine learning and image analysis. He will provide critical advice to the process of medical image analysis and machine learning algorithm development. He was a former PhD student of A/P Sang-Heon Lee and A/P Ivan Lee.

This strong supervisor team with consisting of industry and academic based members would result in not only a potential ARC Linkage project but more importantly contribution toward the practical applications to the real problems in the world. AdelaideMRI strongly supports potential ARC or NHMRC applications in the future by the research team. Financial Support 

This project is funded for reasonable research expenses. Additionally, a living allowance scholarship of $33,500 per annum is available to eligible applicants. Australian Aboriginal and/or Torres Strait Islander applicants will be eligible to receive an increased stipend rate of $46,653 per annum (2023 rates). A fee-offset or waiver for the standard term of the program is also included. For full terms and benefits of the scholarship please refer to our scholarship information for domestic students or international students.

Eligibility and Selection 

This project is open to application from both domestic and international applicants.

Applicants must meet the eligibility criteria for entrance into a PhD. Additionally, applicants must meet the projects selection criteria: 

•  Intermediate level of Understanding and research skills on machine learning 
• Research experiences in both digital image processing and machine learning, especially neural networks

All applications that meet the eligibility and selection criteria will be considered for this project. A merit selection process will be used to determine the successful candidate.

The successful applicant is expected to study full-time and to be based at our Mawson Lakes Campus in the north of Adelaide. Note that international students on a student visa will need to study full-time.

Essential Dates 

Applicants are expected to start in a timely fashion upon receipt of an offer.  Extended deferral periods are not available. Applications close on Friday, 27th of September.