Alzheimers AI 10.1.17

Machine Learning Breakthrough Delivers Early Alzheimer’s Diagnosis

Italian researchers have devised a machine learning algorithm that uses MRI brain scans to identify signs of Alzheimer’s disease far earlier than current non-invasive methods.

The algorithm, developed by a team at Italy’s University of Bari, can detect changes in the brain that signal the onset of the disease nearly 10 years before the appearance of clinical symptoms that are noticeable to doctors. This machine learning technique can distinguish between a healthy brain and one afflicted by Alzheimer’s with an accuracy rate of 86 percent, according to New Scientist magazine.

Using the more than 100 MRI scans, the team trained the algorithm to differentiate between healthy and diseased subjects.

Each scan was divided into small segments and the neuronal connectivity between them was analyzed. The researchers determined that the algorithm identified Alzheimer’s signs most accurately when the brain segments being compared ranged in size from 2,250 to 3,200 cubic millimeters. This corresponds to the approximate size of brain structures affected by the disease, including the amygdala and hippocampus, according to Marianna La Rocca at the University of Bari.

Early diagnosis is critical for the effective treatment of Alzheimer’s, increasing the chances that drug therapy or lifestyle changes will yield positive results. However, current methods of early detection –such as brain imaging employing radioactive tracers—are invasive, costly and not widely available, La Rocca said.

MRI scans are comparatively inexpensive and accessible, making the machine-learning algorithm an attractive alternative to existing approaches.

The Italian researchers said their technology could also be used to diagnose Parkinson’s disease.
The University of Bari’s work represents just one area where AI technology is being used to fight Alzheimer’s.

Canadian firm WinterLight Labs has devised a robot that employs voice analysis to identify signs of Alzheimer’s by listening to patients’ speech patterns. WinterLight said its system can detect the condition with an accuracy rate ranging from 82 percent and 100 percent.

Improved diagnosis of Alzheimer’s using AI technologies could deliver major benefits for patients and for society in general.

More than 5 million Americans now live with the disease, according to the Alzheimer’s Association. As the U.S. population ages, this total is expected to rise dramatically, reaching a total of more than 16 million by 2050.

The cost of treating all these patients is astronomical. Total medical payments in 2017 for all individuals with Alzheimer’s and other forms of dementia are expected to reach $259 billion, the association reports.

Tyler Schulze is vice president, strategy & development at Veritone.  He serves as general manager for developer partnerships, cognitive engine ecosystem, and media ingestion for the Veritone platform. Learn more about our platform and join the Veritone developer ecosystem today.