*** This page is archived and provided for reference purposes only ***

Skip Over Navigation Links

NIH Research Matters

March 3, 2008

Computers Detect Alzheimer’s Disease in Brain Scans

Computers can be trained to detect early signs of Alzheimer's disease in MRI brain scans, according to a new report. The finding could help doctors diagnose the disease earlier and more accurately than they can now, so treatment can begin earlier.

Image of brain with scattered, small areas of color.

Computer-based diagnostics show brain areas that help to classify patients. The blue and green areas increase the likelihood of “normal” classification. Red and yellow increase the likelihood of an Alzheimer’s disease diagnosis. Image courtesy of the journal Brain (PubMed Citation).

Physicians today can diagnose Alzheimer's disease only after symptoms appear. But earlier detection could provide new opportunities for slowing progression of the disease or preventing it altogether. Several lines of research funded by NIH have been aiming to reliably diagnose Alzheimer's disease as early in the disease process as possible.

In the past few years, researchers have shown that brain scans using MRI could help doctors detect Alzheimer's disease. However, analyzing these MRI scans can be time-consuming and difficult. A research team led by Dr. Richard Frackowiak at the Wellcome Trust Centre for Neuroimaging at University College London became interested in a technique that uses computers called support vector machines to do the hard work of analyzing brain images. Their work was supported by NIH's National Institute on Aging (NIA), the Wellcome Trust, the Mayo Clinic and others.

The researchers used support vector machines to compare MRI scans from proven Alzheimer's disease patients with scans from cognitively normal elderly people. The whole-brain images came from 2 facilities with different scanning equipment. As reported in the March 2008 issue of Brain, up to 96% of the verified Alzheimer's disease patients were correctly classified using the new technique. They also found that data from one facility could be used to train a support vector machine from another facility to perform the task with different subjects and different scanning equipment.

The researchers next tested the ability of the support vector machines to differentiate scans between patients who appeared to have mild Alzheimer's disease and age/sex matched controls. The computers correctly identified the scans in 89% of cases, comparable with results in the best clinical centers.

The researchers also tested to see if the method could differentiate between different forms of dementia. The computer analyzed scans from people with Alzheimer's disease and a disease with similar symptoms called frontotemporal lobar degeneration. Post-mortem diagnoses confirmed that the computer correctly identified 89% of the patients.

This study showed that a computer-based diagnostic method could successfully distinguish patients with Alzheimer's disease from healthy aging subjects and from those with frontotemporal lobar degeneration. The technique proved effective in different facilities with scans from different equipment.

“The advantage of using computers is that they may prove to be cheaper, faster and more accurate than the current method of diagnosis,” Frackowiak explained. “This will be particularly attractive for areas of the world where there is a shortage of trained clinicians and when a standardized reliable diagnosis is needed, for example in drug trials.”

—by Harrison Wein, Ph.D.

Related Links:

Contact Us

E-mail: nihresearchmatters@od.nih.gov

Mailing Address:
NIH Research Matters
Bldg. 31, Rm. 5B64A, MSC 2094
Bethesda, MD 20892-2094

About NIH Research Matters

Editor: Harrison Wein, Ph.D.
Assistant Editors: Vicki Contie, Carol Torgan, Ph.D.

NIH Research Matters is a weekly update of NIH research highlights from the Office of Communications and Public Liaison, Office of the Director, National Institutes of Health.

This page last reviewed on December 3, 2012

Social Media Links