Toshiba Medical Visualization Systems Europe, Ltd. (2007-11)

Title: Medical image analysis from multiple data sources
Research Engineer: Sean Murphy
Sponsor: Toshiba Medical Visualization Systems Europe, Ltd.
Academic Supervision: Dr Yvan Petillot, Heriot-Watt University
Industrial Supervision: Dr Ian Poole, Toshiba Medical Visualization Systems

Medical datasets arise from a variety of imaging modalities, including X-ray computed tomography (CT), magnetic resonance (MR), positron emission tomography (PET). Further variations arise from the variety of patient preparations protocols used, e.g., in CT and MR a contrast agent can be injected into the blood stream just prior to the scan, enhancing the vasculature and any perfusing tissues. Modern multi-detector CT scanners (64-slice acquisition is in common use, with 256 and 320 slice recently introduced) can acquire 3D datasets sufficiently rapidly to, e.g., capture phases of the beating heart or trace the perfusion of blood through the liver. MR is a highly flexible modality which can be tuned to image a wide range of tissue characteristics. PET is used to image a short-lived radioactive tracer, and depending on the molecule into which it is incorporated can be used to visualise a variety of metabolic processes - tumour growth being of particular interest.

As a patient progresses through screening, diagnosis, treatment and follow-up, image data may be acquired in several modalities, with varying protocols either simultaneously (there are combined PET/CT scanners), in close succession, or separated by months or years. Image analysis methods hitherto have mainly focused on extracting diagnostic information from single datasets, but there is a recognised need to integrate data across modalities and/or time to support diagnosis or treatment planning.

This broad goal gives rise to many technical challenges, such as inter-modal registration, time series analysis arising from perfusion studies, or the identification / segmentation of structures based on multiple sources of evidence. These challenges form the basis of this project.