RSS Feeds

Email a Friend Printer Friendly

New way to help diagnose dementia

August 01, 2000

A new way of interpreting 3D images of the brain has opened up the possibility of doctors being able to distinguish between Alzheimer’s disease and other degenerative brain diseases. Doctors need to be able to diagnose the correct disease accurately and as early as possible to implement the most appropriate treatment.

The work is being carried out by Professor Chris Taylor and his team at Manchester University’s Division of Imaging Science and Biomedical Engineering, with funding from the Engineering and Physical Sciences Research Council.

There are three main types of dementia: Alzheimer’s disease, fronto-temporal dementia and vascular dementia. Although the symptoms of each disease are similar, each has a different underlying cause and each requires different treatment.

All three diseases are characterised by a loss of brain tissue, but the pattern of loss is subtly different in each case and it is this fact which is the focus of the research at Manchester: is it possible to develop a way to interpret images of the brain that can automatically assess which parts of the brain are being lost and therefore help doctors both to diagnose the condition and to monitor the progress of the disease?

Currently, to distinguish conclusively between the three diseases a comprehensive battery of tests is needed, including brain scans, psychometric examination and monitoring how the disease develops.

The new work relies on analysis of three-dimensional images of the brain obtained from magnetic resonance scans.

“We use a model-based approach,” says Professor Taylor. “The main features of the brain and their natural variability are captured in a computer model. When a new image is presented to the computer we fit the model to the data and record how much and in what way it differs from the ‘average’ brain.”

The computer model is ‘trained’ by being shown many images of three-dimensional brain scans. The deep folds in the external surface of the brain - called sulci - are used as the main landmarks or ‘route map’ to define the shape of the brain. “Based on a large number of examples, the computer can learn to recognise the main features common to all brains,” says Professor Taylor. “The natural variation in shape can also be captured by recognising distinctive patterns of departure from the ‘average’ brain. The precise configuration of a particular brain can be expressed in terms of a relatively small number of parameters that record the extent to which each pattern of variation is present.” Only a few tens of parameters are required to describe the whole range of variability. “Things are not as complex as one might have expected,”
says Professor Taylor.

When the computer is presented with a new image, it chooses a set of parameters that adjusts the model to fit the new image. The values of the parameters give a compact description of brain shape and can be used to distinguish between normal variation and that due to disease. Normal brains and the three types of dementia are expected to be characterised by particular combinations of parameter values.

“Ultimately if we can measure more directly the pattern of loss of brain matter we could have a more specific indication of the type of disease and how the degeneration is progressing and thus be in a better position to manage the treatment of patients,” says Professor Taylor.

The main principles underlying the new technique are applicable not only to brain scans but to many problems where images need to be interpreted. “They have already been applied in areas as diverse as printed circuit board inspection for the electronics industry and face recognition”, adds Professor Taylor .


- ENDS -





Engineering and Physical Sciences Research Council (EPSRC)