Seminar Series: Technology Trends 2005
Automatic segmentation of anatomical structures
from medical images.
Dr Sébastien Ourselin, Project leader, Biomedical
Imaging, CSIRO ICT Centre
In this talk we will describe some of the past and current work
in BioMedIA Lab at the CSIRO ICT Centre, with a focus on medical image
segmentation. Segmentation is the process of delineating and
assigning a label to a biological structure. In the last two
decades, a wide variety of segmentation techniques have been
proposed but no single method can be applied successfully for every
medical imaging applications. We will illustrate our work through
different clinical applications.
First, we will describe an application we have developed for the
planning of surgery for abdominal aortic aneurysm (AAA). In the
planning stage surgeons often need to mentally evaluate multiple CT
slices in order to formulate a 3D representation of the vessels.
This must be taken into account when obtaining critical measurements
and planning the surgery. We proposed a semi-automatic approach to
fully segment and accurately design a patient specific AAA model
using knowledge-based vessel identification. Then, we will describe
some of our current work on the segmentation of soft tissues,
especially for the cartilages of the knee. Soft tissues are highly
variable in shape, their boundaries poorly defined and as a result
usually difficult to accurately segment using existing techniques.
Having an accurate estimation of the size and shape of the
cartilages is critical for a large number of clinical applications,
such as drug effect trials, cartilage grafting, and osteoarthritis
study. The aim of our work is to develop automatic segmentation
based on statistical shape model (SSM) of the different part of its
articular anatomy. We will then embed a priori knowledge of the
anatomical environment into the SSM to drive our segmentation. We
will also present our research in the accurate estimation of the
thickness of the cortex, and its applications to the study of
neurodegenerative disease, with a current clinical focus on early
stage of schizophrenia.
All these different tools have been developed in a common
software platform called MILX (TM), based on ITK and VTK. We will
give a brief overview of our software platform, and highlight the
need of a robust and stable software environment, critical for
successful clinical transfer of our technology.
Back to Technology Trends page
|
