Search Contact information
University of Cambridge Home Department of Engineering
University of Cambridge > Engineering Department > Machine Intelligence Lab

Abstract for syn_thesis

PhD thesis, University of Cambridge


Michael Syn

May 1996

A 3D freehand ultrasound system augments a conventional clinical scanner with a position sensor on the hand-held probe. Such systems are safe, cheap, portable, and allow clinicians to scan using conventional techniques. Unfortunately the resulting freehand images are non-parallel, sometimes self-intersecting, and retain the noisy image artefacts inherent in conventional 2D ultrasound.

This dissertation proposes two model-based strategies for interpreting such images: an organ shape model is used for geometric reconstruction of scattered organ landmarks in the images, and the Gompertz growth model is used to register organ shape models to each other in a coherent and biologically justified way.

Both strategies are robust to noise and inaccuracies in the organ model meshes, and are intended to complement future work on the detection of tissue boundaries in ultrasound images. So a model-based framework to organise sparse and noisy cues about tissue boundaries, is a key element in any attempt at fully-automated interpretation of 3D freehand ultrasound images.

A biological model of organ growth is first developed using Oster-Murray mechanisms, whose eigenmodes describe the organ's modes of shape variation. An iterative procedure allows these idealised modes to be refined from organ examples. 3D freehand ultrasound images are then segmented by such organ models, for the purpose of organ volume estimation. However, an organ model can only be refined from the segmented organ shape if they both share a common shape parameterisation.

They are therefore registered to each other using their eigenmodes, which are proposed to represent homologous (`biologically corresponding') landmarks. The choice of registration solutions is restricted to biologically plausible ones using the Gompertz metric. Bayesian combination of the likelihood of eigenmode homology, with the prior constraint of Gompertzian growth, results in a posterior measure of homology which must be minimised for an optimal registration. The minimisation is efficiently performed using a multi-resolution implementation of the highest confidence first algorithm.

Keywords: medical imaging, 3D freehand ultrasound, volume estimation, registration

[6.7 MBytes compressed using gzip]

(ftp:) (http:)
PDF (automatically generated from original PostScript document - may be badly aliased on screen):
  (ftp:) syn_thesis.pdf | (http:) syn_thesis.pdf

If you have difficulty viewing files that end '.gz', which are gzip compressed, then you may be able to find tools to uncompress them at the gzip web site.

If you have difficulty viewing files that are in PostScript, (ending '.ps' or '.ps.gz'), then you may be able to find tools to view them at the gsview web site.

We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.

© 2005 Cambridge University Engineering Dept
Information provided by milab-maintainer