Astronomical & Medical Imaging

One Day BMVA Technical Meeting in association with the IEE/E4 at the Royal Statistical Society, London, UK on 18th April 2001

Chairpersons: Seb Oliver (Sussex University); Lewis Griffin (King's College London)

www.bmva.ac.uk/meetings

The aim of this one-day meeting will be to explore common areas in Medical Imaging and Astronomy. Despite the radical difference in subject matter, it is clear that there are methodological issues common to these two disciplines. In both there is a need to acquire, process, interpret and store images of 2, 3 or more dimensions. In addition both are interested in temporally and spectrally resolved data that may be of a scalar, vectorial or tensorial nature

Programme

10:30 Registration and coffee

10:55 Introduction and welcome, Lewis Griffin (King's College), Seb Oliver (Sussex Uni)

11:00 Mammogram Analysis, Lance Miller, University of Oxford.

11:30 3D Texture, Maria Petrou, Surrey University

12:00 Adaptive Optics: From Astronomy to Ophthalmology, Chris Dainty, L Diaz S Haro, I Munro, C Paterson, Imperial College.

12:30      The role of the Transport Equation in Medical Imaging, Simon Arridge, University College London.

13:00 Lunch

14:00 Statistics on Manifolds, Xavier Pennec, INRIA Sophia.

14:30 Doppler Tomography, Tom Marsh, University of Southampton.

15:00 Tea

15:30 Data Fusion, Dave Hawkes, King's College London.

16:00 Astro-Medical Instrumentation, John Fordham, University College London.

16:30 Summary and discussion

16:40 Closing remarks and finish

REGISTRATION FORM: 18th April 2000

BMVA MEMBER: YES/NO
IEE MEMBER:        YES/NO

Author:        Simon Arridge

Institution:    University College, London

URL:            http://www.cs.ucl.ac.uk/staff/S.Arridge/

Title:            "The role of the Transport Equation in Medical Imaging"


Abstract:

The Radiative Transport Equation is a balance equation describing the propagation of particles in the presence of absorption and scattering. It can be described by a seven dimensional state space taking into account spatial position, direction of movement, time, and energy. It arises in many areas of medical imaging. The Radon Transform, and the attenuated Radon Transform are special cases. In Radiotherapy Treatment Planning it is used to predict the best beam placement to concentrate radiation on a tumour without affecting surrounding tissue. In Optical Tomography, when scattering is the dominant process the RTE can usually be replaced with a simpler hyperbolic or parabolic PDE. However in some cases, such as in the presence of void regions this is too simplistic.

In this talk I will discuss some methods for solving the RTE, in particular using the Pn approximations. A novel method, the radiosity diffusion method,
will be presented which combines methods from Computer Graphics with standard Finite Element Methods for solving 2nd order PDEs. Some results from the latter will be presented, of relevance to Optical Tomography

Authors:      Chris Dainty, Luis Diaz Santana Haro,Ian Munro and Carl Paterson

Institution:    Imperial College, London

URL:           http://op.ph.ic.ac.uk/

Title:            Adaptive Optics: From Astronomy to Ophthalmology

Abstract:

Adaptive optics is a technology in which optical wavefront imperfections are compensated so as to produce diffraction-limited images. All modern large telescopes are now being equipped with this technology, so that the existing limited angular resolution due to the effects of atmospheric turbulence are overcome. Similar technology can be applied to the human eye and in this talk we shall describe work going on in our Group and elsewhere on potential ophthalmic applications of adaptive optics.

Author:        John Fordham

Institution:    University College, London

URL:            http://www.phys.ucl.ac.uk/cgi-bin/people?Fordham

Title:    "Astro-medical Instrumentation"

Abstract:

The instrumentation demands of the scientist in many bio-medical applications overlap those required in astronomy. However, the approach to instrumentation in the two fields is, in general, very different. The astronomical community develops one-off instruments using state-of-the-art technology whereas the bio-medical scientist primarily relies upon commercial products that may not ideally match demands in terms of data quality. Cross-collaboration between the two communities can be mutually beneficial and can, broadly, be put into three categories: (1) transfer of technology, (2) transfer of know-how and (3) joint development of new technology. Here, I give examples of overlap areas and address the advantages and drawbacks associated with cross-collaboration.

Author:        David Hawkes

Institution:    King's College, London

URL:            http://www-ipg.umds.ac.uk/rad/staff/cisg_staff/davidhawkes_staffpage.htm

Title:            "Data Fusion"

There is a long history of technology transfer and cross fertilisation between medical imaging and astronomy. Fourier methods originally developed for astronomy are the mainstay of image reconstruction in CT scanning and nuclear medicine while gamma cameras and solid state x-ray detectors are technologies common to both disciplines. Data fusion is one area where the two communities may not be so aware of advances in each other's discipline.

This talk will describe the importance of relating information from one medical image to another image or to physical space in image guided interventions. The definition of spatial correspondence is crucial to this process. This talk will present advances in assessing similarity between images including the use of information theoretic measures such
as mutual information for fully automated image alignment. Image transformations can be rigid-body, affine or non-rigid. Recent advances in practical non-rigid registration algorithms will be described. The talk will be illustrated with examples in the diagnosis and treatment of cancer, the neurosciences and image guided interventions.

Author:        Tom Marsh

Institution:    University of Southampton

URL:            http://www.astro.soton.ac.uk/~trm/

Title:            "Doppler Tomography"

Abstract:
I describe the method of Doppler tomography used to image structures within close pairs of orbiting stars (binary stars). The method relies on Doppler shifts from high speed gas flows, but I will show that it is entirely analogous to medical X-ray tomography. I will show results from applications of this method and discuss the case of gas flows out of the orbital plane of the binary orbit which involves information in 3D.

Author:        Lance Miller

Institution:    University of Oxford

URL:            http://www-astro.physics.ox.ac.uk/~lam/

Title:            "Analysis of Mammaograms".

Abstract:
The detection of discrete objects within an image and classification based on observed features are operations common to both astronomical and medical image analysis. A number of projects have previously attempted to transport ideas between these two disciplines. I discuss here the similarities and the differences between the problems of astronomical and medical image analysis, illustrated by a recent programme which aimed to develop methods of detecting maligant masses in mammograms.

Author:        Maria Petrou

Institution:    University of Surrey

URL:             http://www.ee.surrey.ac.uk/Personal/M.Petrou/

Title:             "3D texture features"

Abstract:

Texture in 2D is easy to perceive visually. However, texture in 3D is very difficult to visualise. In the past few years we have been developing techniques for quantifying volume roughness and representing it in a form that makes it explicit to the human vision system. We have been applying our methodology to MRI data of Alzheimer's patients, to MRI data of schizophrenic patients, and to seismic data of the crust of the Earth. We find that we can identify features derived from the volume texture that correlate well with the stage of the condition in the case of Alzheimer's, with other indicators, like for example results of the MMSE test used for diagnosing the disease. In the case of schizophrenia, the features can discriminate between schizophrenics and normal controls in a statistically significant way. Our methodology is general and can be applied to any 3D volume distributed data.

Author:        Xavier Pennec

Institution:    INRIA Sophia

URL:            http://www-sop.inria.fr/epidaure/personnel/pennec/pennec.html

Title:            "Statistics on Manifolds"

Abstracts:

Probabilities and Statistics on Riemannian Manifolds - Basic tools for geometric measurements with application in medical image analysis and molecular biology