Paul Hayton

Analysis of dynamic MR breast images using a model of contrast enhancement

We describe a model of dynamic contrast enhancement in breast MRI designed to aid the radiologist in cases for which X-ray mammography is ineffective. The breasts are segmented from the image slices by a dynamic programming algorithm after morphological opening. A pharmacokinetic model has been derived to fit the rise in intensities after injection of a contrast agent, in a way that facilitates investigation of the effects of different models of bolus injection. The pharmacokinetic model is used in a modified Horn-Schunck algorithm to correct for motion effects during the seven minute acquisition period. The results show significant localization of tumours and enable discrimination of cancerous tissue. In particular, we illustrate the approach with an image that shows a carcinoma, whose appearance and localization are greatly improved by the registration algorithm.

Static and dynamic novelty detection methods for jet engine health monitoring.

Novelty detection requires models of normality to be learnt from training data known to be normal. The first model considered in this paper is a static model trained to detect novel events associated with changes in the vibration spectra recorded from a jet engine. We describe how the distribution of energy across the harmonics of a rotating shaft can be learnt by a support vector machine model of normality. The second model is a dynamic model partially learnt from data using an expectation–maximization-based method. This model uses a Kalman filter to fuse performance data in order to characterize normal engine behaviour. Deviations from normal operation are detected using the normalized innovations squared from the Kalman filter.

A non-rigid registration algorithm for dynamic breast MR images

Abstract Magnetic resonance image analysis is a promising technique for diagnosing breast cancer, particularly in women for whom X-ray mammography is ineffective. If breast motion is not corrected for, diagnostic accuracy is significantly reduced. In this paper, we analyse the kinds of motion that arise during image formation and we describe a model based non-rigid registration algorithm to estimate and correct for breast motion. Registration of breast MR images is complicated by the use of a contrast agent which results in a non-uniform increase in intensity across the image. The work described here forms part of an implemented breast MR analysis system which allows automatic detection and segmentation of regions of focal enhancement and non-rigid image registration.