Can computers be trained to detect Alzheimer's disease? According to a new international study, the high-tech concept looks promising.
"This technology in theory can create a library of diagnoses to one day make a positive diagnosis (of Alzheimer's disease)," said one of the leading authors of the study, Dr. Cynthia Stonnington, assistant professor of Psychiatry at the Mayo Clinic in Scottsdale.
The aim of the study, titled "Automatic Classification of MRI scans in Alzheimer's Disease," was to assess how successfully a computer could be "trained" to diagnose Alzheimer's or other causes of dementia based on MRI brain images it was fed.
Stonnington said the process was based on "a pattern recognition technique" using three controlled categories. Each consisted of scans of elderly patients with normal cognitive skills, those with a specific type of degeneration and those with pathologically verified Alzheimer's.
Alzheimer's disease, which is a neurodegenerative disorder, is the most common cause of dementia.
It's a fatal disease that more than 5 million people in the U.S. live with, according to the Alzheimer's Association.
Currently, a postmortem autopsy is the only definitive way to positively diagnose the disease.
Physicians currently base a probable diagnosis in living patients by identifying commonly associated clinical symptoms associated with Alzheimer's.
As the computer methodology in the study shows, that may change.
The study, conducted between July 2006 and July 2007, was led by a team that included physicians and researchers from the Mayo Clinic in Scottsdale, the Mayo Clinic in Rochester, Minn., and a number of European neuroimaging and neurological centers.
Those centers include the Institute of Neurology at University College London, where the study took place.
Stonnington said the study used two sets of MRI images (in the three categories) supplied by more than one MRI center.
She said the study found up to 96 percent of pathologically verified Alzheimer's scans were correctly classified using the brain images. Moreover, she said that images supplied from the different MRI centers achieved comparable results in separate analysis.
Stonnington said one of the ultimate goals is collect a broad range of brain scans - those including Alzheimer's and varying forms and degrees of dementia - and train the computer to read them and create a universal database that can aide physicians in forming a diagnosis.
For those afflicted with the disease, Stonnington said the earlier the detection, the better.
"More and more is happening in the way of slowing the progression of the disease," Stonnington said.
The best hope, she said, is to identify the illness before there is extensive brain deterioration, so people can take advantage of the latest science has to offer to combat it. Stonnington said the reality of using this technological method may become possible within the next decade.
The complete findings of the study are published in this month's issue of Brain (http://brain.oxfordjournals.org), a neurology journal published by Oxford University Press.
[CORRECTION: The original version of this story mistakenly gave Rochester, N.Y. as the location for one of the Mayo Clinic's branches]