Details of Grant 

EPSRC Reference:	GR/J48887/01
Title:	INTELLIGENT DEFORMABLE MODELS FOR 2D AND 3D MEDICAL IMAGE SEGMENTATION
Principal Investigator:	Boyle, Professor R
Other Investigators:	Hogg, Professor D
Researcher Co-Investigators:
Project Partners:
Department:	Sch of Computing
Organisation:	University of Leeds
Scheme:	Standard Research (Pre-FEC)
Starts:	01 January 1994	Ends:	31 December 1996	Value (£):	146,233
EPSRC Research Topic Classifications:	Image & Vision Computing
EPSRC Industrial Sector Classifications:
Related Grants:
Panel History:
Summary on Grant Application Form
To develop state-of-the-art autonomous deformable models (contours, surfaces & volumes) for medical imaging applications; to improve the robustness and reliability of these models through the extensive use of internal and external constraints on behaviour; to incorporate practical implementations of these models into software tools likely to be of use in clinical practice.Progress:Initial work focused on the acquisition of 3D medical imaging data (primarily MR datasets of the human head) and on development and implementation of various 3D preprocessing techniques. Collectively, these techniques generate information which drives deformation of the models that we are developing.The first model we have developed is an extension into three dimensions of the 2D 'greedy' active contour model of Williams & Shah. Our model is a triangular mesh which can be initialised as a plane, a cylinder or an ellipsoid. The potential gradient that drives mesh deformation is derived from a 3D Euclidean distance transform of salient image structure, in combination with grey level gradient information.The model has a self-optimising topology: it can refine itself by adding nodes to the mesh to ensure that areas of high detail are represented accurately, yet it can also minimise redundancy in areas of low detail through a process of decimation. These procedures operate automatically whenever variables such as local surface curvature and mesh element size lie outside predefined limits. These same variables are also used in the automatic local adjustment of mesh elasticity and stiffness. Constraints on, e.g., smoothness can thus be relaxed as the mesh converges on regions of high detail and large local curvature. This approach yields more detailed surfaces than traditional methods. The fitted mesh has been used in a global surface matching algorithm that takes as its input values of local curvature and Koenderink's shape index computed at each mesh node. Development of the model is proceeding in several directions. We are currently exploring strategies for improving the results obtained for complex, convoluted objects such as the brains cortical surface. We are also experimenting with coupled mesh models which can simultaneously generate surface data for the inner and outer walls of, e.g., the left ventricle of the heart. A third area of study involves modification of the model such that deformation is influenced by the grey level statistics of voxels enclosed by the mesh. The implementation of several different parametric deformable models is underway; our intention is to compare their performance with our self-optimising, physically-deformable mesh.
Key Findings
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Potential use in non-academic contexts
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Impacts
Description This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk Summary Date Materialised
Sectors submitted by the Researcher
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Project URL:
Further Information:
Organisation Website:	http://www.leeds.ac.uk