My primary research interests are in
applying imaging and signal processing methodologies, including pattern
recognition and image registration, to problems within the medical and life sciences. In particular, I am interested
in extracting the maximum diagnostic and prognostic information from clinical images.
Currently I am actively involved in projects on
(i) The texture of clinical images. The potential of quantitative
texture measures to discriminate between normal
and abnormal status in medical imaging has long been recognized. I have investigated the clinical applicability
of a number of computationally tractable texture indices and have identified fractal signature and lacunarity as the
most promising discriminators for diagnosing compromised bone quality in osteoporosis. The relationship
between lacunarity and fractal dimension, the applicability of grayscale lacunarity, and the clinical sensitivity of the
indices are being investigated further.
(ii) Tortuosity (of blood vessels, and the spine). Increasing vessel
curvature or tortuosity with age may be a risk
factor in the development of atherosclerosis: I am investigating the quantitative measurement of arterial tortuosity
using CT and MRI images, based on the minimum curvature of approximating piece-wise splines to the mid-line.
I am keen to investigate its validity as an indicator of changes in morphology by applying it to a variety of vascular
systems (including retinal blood vessels) and to the curvature of the spine (scoliosis).
(iii) The curvature of the universe. We are exploring the feasibility of using lacunarity
analysis as a new method
of determining the Friedmann curvature of the Universe using theoretical cosmological models and observational
redshift data from the Sloan Digital Sky Survey. The immediate impact will be to provide the astrophysics
community with an alternative measure to fractal analysis of the nature and extent of the large scale distribution
(iv) Biometrics. The application of pattern recognition strategies to fingerprint and face recognition.