Scan software

is claiming a breakthrough in medical diagnostics after interpreting positron emission tomography (PET) brain scan images to identify various forms of dementia.

The Hamburg-based project is claimed to enhance one of the most effective techniques, called FDG-PET, for diagnosing early-stage dementia, which is performed using a PET brain scan with the tracer fluorodeoxyglucose (FDG).

It arose from work involving Philips and doctors at the

, who had observed that while FDG-PET images give high-grade diagnostic information, clinicians need considerable expertise in nuclear medicine imagery and brain anatomy to interpret them.

In an attempt to help, Philips researchers combined their capabilities in decision support software, imaging technology and image analysis software.

Philips scientist Stewart Young said: 'The system contains a database of scans of people with healthy brains, and patients with particular verified diseases — Alzheimer's, frontotemporal dementia or Lewy body dementia. Following a scan, the software compares the patient's image to this database by aligning the images, then computing a simple set of numbers which indicates the similarity of patterns in the patient's brain to the patterns in the database.'

The software presents two results to the user. The first is a picture of the brain overlaid in blue with regions where disease has been detected. The second is a set of figures giving an estimate of how close the patterns seen in the image are to the disease pattern.

Young said this capability was the key advantage of the system over current methods. 'It gives the ability not only to say this is a pattern most similar to a pattern of, say, Alzheimer's disease, but also to attach a quantifiable number to that to give a percentage of similarity to a typical pattern.'

Another key feature is that the system allows the integration of FDG-PET information with magnetic resonance imaging (MRI), which gives a much more detailed image of brain anatomy.

'FDG-PET does not show so much detail, but gives much more of the functional information — not just where the tissue is but what it is doing. The combination of these two different types of information is another powerful feature of the system,' said Young.

Further study is needed to show whether the system can give earlier diagnosis than current methods, but Young argued that the trials showed it will be able to support people who do not have the high level of expertise now required. Medical institutions can also continually add to the information in the databases and make the information available to a larger potential base of users, which could eventually have some impact on early detection.

Philips recently completed a study with the University Medical Centre in Hamburg on the feasibility and usability of the system. It showed that using FDG-PET scans from 83 previously diagnosed patients, the software achieved better than 98 per cent correspondence with doctors' interpretation of Alzheimer's and frontotemporal lobe dementia characteristic.

The next stage will be to perform comparative prospective studies on un-diagnosed patients alongside established methods with clinical partners.