Living with Dementia October 2008

By Caroline Bradley

Scientists in London have pioneered the development of a computer programme to diagnose Alzheimer's disease.

New technique to identify disease
MRI (magnetic resonance imaging) scans can provide valuable information about the physical health of our brains. However, it can be difficult for the human eye to detect small changes on scans when analysing results.

Now, scientists in London have developed a programme to enable computers to interpret the scans. The programme can identify the typical pattern of damage to the brain caused by Alzheimer's disease that shows up on MRI scans.

Professor Nick Fox is part of a team based at University College London that has pioneered the idea to develop the new technology. He says,

'Historically we've tended to use scans to rule out other causes of symptoms rather than to rule dementia in. But it's increasingly clear that when radiologists recognise a pattern of loss of brain tissue, it can help to make a diagnosis in diseases such as Alzheimer's disease or fronto-temporal dementia.' 

Preliminary test results of the programme have proved positive, with Alzheimer's disease correctly identified in 96 per cent of cases. The team is hopeful that this technology could be used to support staff in making a clinical diagnosis at an early stage of the disease.

Exquisite detail
Professor Fox says,

'MRI scans provide an exquisitely detailed view of the brain. Each scan contains one million three-dimensional points of information (known as voxels). They allow us to view features in the brain which are as thin as a thumbnail.'

When a person develops dementia their brain loses brain cells, and specific patterns in the brain will indicate this loss. A radiologist viewing a scan looks out for particular signs of damage. For example, shrinking of the hippocampus, the area of the brain responsible for memory, is an indication of Alzheimer's disease.

But Alzheimer's disease and other dementias can affect many different areas of the brain, often quite subtly. It is difficult for the human eye to detect such small changes on MRI scans. Computers, on the other hand, are good at systematic analysis and can carry out a thorough evaluation of every area of an MRI scan.

Spotting the difference
To enable the computer programme to identify the damage typical of Alzheimer's disease, the researchers had to 'teach' it to spot the difference between scans of brains with the disease and those without. The programme analysed MRI scans from people who had the disease (confirmed at post-mortem examination) and 'learned' the difference between the patterns of damage they had analysed and scans of people who did not have dementia.

Professor Fox says,

'One of the strengths of this study is that we have been able to use MRI scans that we know for certain depict Alzheimer's disease. This is thanks to the generosity of people who agree to donate their brain or the brain of their family member to a brain bank.'

Impressive accuracy
Once the programme had been loaded with information on how to spot dementia, the team tested how well it performed. Running on a standard computer, the programme achieved an impressive rate of 96 per cent accuracy when it analysed the scans of 34 people who had Alzheimer's disease and 34 people of the same age who did not.

The rate dipped to 89 per cent when diagnosing people with dementia at an earlier stage, but all of the results were the same or better than current diagnosis rates in the best clinics. The programme also distinguished between fronto-temporal dementia and Alzheimer's disease with 89 per cent accuracy.

What's next?
Scientists now plan to focus on:

• testing the potential of this technology to diagnose dementia at an earlier stage than is currently possible.
• Establishing if it can track and measure the effects that therapies have on the progress of dementia.

Professor Fox says,

'We need to explore whether the programme will be useful when people are more mildly affected by dementia. We need a strong predictive value, somewhere up in the ninety per cents, for it to be useful in a clinic.'

If the programme can track and measure the progress of dementia in patients, it could be used to speed up and reduce the costs of clinical trials by automatically analysing the effects of new drugs on the structure of the brain.

In other research
At the International Conference on Alzheimer's Disease in July, researchers reported the discovery of a compound that might make it possible to identify Alzheimer's disease at an early stage. The compound attaches itself to the damaging protein amyloid and shows up on a special brain scan known as a PET (Positron Emission Tomography) scan. PET scans reveal chemical functions rather than physical structures. Researchers are working out whether this compound is stable enough to give reliable results.