Metabolic Imaging to Improve Clinical Workflows

Metabolic Imaging to Improve Clinical Workflows

The work in these projects aims to mine the large amounts of data contained by medical images in a way that gives clinicians the most relevant information for the decisions they need to make.

Aim

Metabolic images contain a wealth of information that is seldom used, due to our inability to package it effectively for time-poor clinical specialists. For instance, the way PET images are recorded allows them to display the dynamics of radio-active tracers as they are taken up by cells and then flushed out over time. In the video link below an example of a dynamic PET scan using a tracer called FDOPA is shown. The tracer is taken up by various structures in the brain over a 75 minute period. In the clinic however, PET images are typically viewed as a single snapshot in time.

Fig 1. Single frame from a maximum intensity projection of a dynamic FDOPA PET image

Fig 1. Single frame from a maximum intensity projection of a dynamic FDOPA PET image

The work in these projects aims to mine the large amounts of data contained by medical images in a way that gives clinicians the most relevant information for the decisions they need to make. This can reduce the time needed for reporting and enable better decision making, thereby helping to contain the currently unsustainable increases in health costs. The project is following several avenues to achieve its aim, working closely with clinical collaborators in the departments of nuclear medicine, radiology, surgery, and radiation therapy at the Royal Brisbane and Women’s Hospital, the University of Queensland and the University of South Australia. Some individual areas of research are briefly highlighted below.

Physiologically-informed kinetic analysis models

Dr Charles Baker has developed a physiologically based kinetic analysis model for high blood perfusion organs. In this work, which won the Canon Extreme Imaging Competition in 2014, the model distributes tracer between the blood and intracellular compartments in more physiologically correct proportions than the standard model, producing parametric images with improved contrast between lesions and normal tissues as shown in the image and video below. A publication has just been accepted in the Journal of Theoretical Biology.

Fig 2. Single frame from a dynamic FDG PET image highlighting colon cancer metastases within the liver C Baker, N Dowson, P Thomas, S Rose, “Modelling of FDG metabolism in liver voxels,” Journal of Theoretical Biology, in press 2014.

Fig 2. Single frame from a dynamic FDG PET image highlighting colon cancer metastases within the liver
C Baker, N Dowson, P Thomas, S Rose, “Modelling of FDG metabolism in liver voxels,” Journal of Theoretical Biology, in press 2014.

Demultiplexing PET tracers

Christopher Bell, a PhD student, is exploring ways in which we can de-multiplex pairs of PET tracers, because in some circumstances several biological factors could be relevant to during the planning of therapy: e.g. using FDOPA to localise metabolically active neoplasm, and FMISO to locate regions of hypoxia. However for clinical use it is important that the logistical impact of such approaches is minimised. Methods have been developed to help clinicians to design protocols to best trade-off the use of PET scanners and the image fidelity:

C Bell, S Rose, S Puttick, A Pagnozzi, C Poole, Y Gal, P Thomas, M Fay, R Jeffree, N Dowson, “Dual acquisition of 18F-FMISO and 18F-FDOPA,” Physics in medicine and biology 59(14): 3925, Jul 2014.

Accelerate and improve accuracy of kinetic analysis

The dynamic analysis of images remains computationally expensive, which impacts upon the feasibility of its use in the clinical setting. We have developed methods to speed up kinetic analysis by an order and improve its accuracy at the same time. A publication is forthcoming, with previous work in this area made available below.

N Dowson, P Bourgeat, S Rose, M Daglish, J Smith, M Fay, A Coulthard, C Winter, D MacFarlane, P Thomas, and O Salvado “Joint Factor and Kinetic Analysis of Dynamic FDOPA PET Scans of Brain Cancer Patients,” Medical Image Computing and Computer-Assisted Intervention, pages 185-192, 2010 Springer

Quantify cerebral palsy injury

Alex Pagnozzi, a PhD student, is developing tools to support clinicians when quantifying the extent of cerebral palsy in children in collaboration with researchers at the Stella Maris Institute in Pisa, Italy. A link between the extent of injuries and the observed symptoms has been established, but the reporting on MR images remains laborious. Hence, new approaches have been developed to reduce the complexity of existing reporting tools. A publication is forthcoming.

Vascular measures for planning endovascular surgery

The repair of endovascular aneurysms is an increasingly common but complex surgical procedure. Extensive planning before surgery is required and it is important to assess how much vessels twist, because this increases the difficulty of the procedure. However surgeons typically assess tortuosity by eye, and accurate computational methods for this purpose are not readily available. Hence, in collaboration with clinicians and researchers at the University of Adelaide, we are developing a computational approach to assess tortuosity. Recent work has had a focus on the particular question of how much the presence of tortuosity increases the chances of an adverse event during or after surgery, to assist patients in avoiding procedures with a greater likelihood of futility:

N Dowson, M Boult, P Cowled, T De Loryn, R Fitridge, “Development of an Automated Measure of Iliac Artery Tortuosity that Successfully Predicts Early Graft-Related Complications Associated with Endovascular Aneurysm Repair,” European Journal of Vascular and Endovascular Surgery, 48(2):153-160, Aug 2014.

Early markers for treatment response

In collaboration with Queensland Health including clinicians in Specialised PET Services, Queensland, Neurosurgery, Radiology and Radiation Oncology and the University of Queensland, our group has developed tools to analyse cancer at a local, intra-tumour scale. These tools can reveal the patterns of treatment failure and hence identify better treatment strategies. Such tools are enabling the development of biomarkers that can assist in establishing treatment response at an earlier stage, with the potential of playing a role in treatment planning.

Fig. 3. Image showing example of FDOPA PET's ability to accurately reveal regions of active tumour at an early stage

Fig. 3. Image showing example of FDOPA PET's ability to accurately reveal regions of active tumour at an early stage

Some of our recent research has found that focal metabolic changes in PET uptake have a correlation with patient prognosis:

N Dowson, P Thomas, M Fay, R Jeffree, Y Gal, P Bourgeat, J Smith, C Winter, A Coulthard, O Salvado, S Crozier, S Rose, “Early prediction of treatment response in advanced gliomas with 18F-DOPA positron-emission tomography,” Current Oncology 21(1):e172, Feb 2014.

Y Gal, N Dowson (co-first author), Pierrick Bourgeat, Olivier Salvado, Paul Thomas, Michael Fay, Stephen Rose, R Jeffery, Amir Fazlollahi, Stuart Crozier, “Amorphous regions-of-interest projection method for simplified longitudinal comparison of dynamic regions in cancer imaging,” IEEE T. Biomedical Engineering 61(2): 264-272, Feb 2014.

N Dowson, P Thomas, Y Gal (co first-author), M Fay, R Jeffree, C Winter, A Coulthard, J Smith, P Bourgeat, O Salvado, S Crozier, S Rose, “Assessing local outcomes in heterogeneous gliomas,” Journal of Physics: Conference Series, 489(1), Mar 2014.

N Dowson, P Thomas, Y Gal, M Fay, R Jeffree, C Winter, A Coulthard, J Smith, P Bourgeat, O Salvado, S Crozier, S Rose, “Contribution of FDOPA PET to radiotherapy planning for advanced glioma,” Journal of Physics: Conference Series, 489(1), Mar 2014.

Acknowledgements

This work has been supported by the National Health and Medical Research Council, Cancer Council Queensland the Queensland Government Smart State scheme and the Australian eHealth Research Centre.