Treatment Online

Machines Learn to Identify Alzheimer's Accurately
> 2/22/2008 3:05:23 PM

Alzheimer's disease (AD) has a devastating but gradual effect on the brain, making it extremely difficult to diagnose. While relatives might strongly suspect its presence because of forgotten birthdays or uncharacteristic confusion, and physicians may be able to use brain scans to make fairly accurate guesses, a definite diagnosis has only been possible by directly examining the brain during autopsy for the presence of neural plaques. A study in the most recent issue of Brain presents a new method for reliably diagnosing AD without the help of human interpretation.

Skilled clinicians can use MRIs to correctly identify AD in the majority of cases. As always though, the necessity of relying on expert interpretation opens up room for mistakes and levels of accuracy that vary depending on experience and training. A team of researchers led by Stefan Kl�ppel has successful trained support vector machines to diagnose AD. After studying thousands of whole brain images from a variety of patients, the machines learned to recognize the patterns most likely to end in a post-mortem confirmation of AD. Their accuracy ranged from 85.6% to 96.4%, depending on the group of images used in training. The vector machines were able to generalize their pattern recognition skills to other hospitals that used different equipment, making them useful in diverse settings across the country.

The machine diagnosis was not much better than the best clinical analysis under ideal conditions, but it was much less labour intensive. In addition, it was able to distinguish AD from frontotemporal lobar degeneration with much better accuracy. This suggests that vector machines may be able to differentiate a large array of neurological diseases with the proper training. If computer pattern recognition can aide in diagnosis, doctors will have more time to spend on treatment and patients will have more time to come to terms with their condition. The only thing worse than noticing that your mind going is not knowing where it is going or why.

Comments

No comments yet.