Artificial Intelligence Used to Improve Formal Autism Diagnosis ProcessCategory: Science & Technology
Posted: April 11, 2012 02:31PM
For the first time, machine learning techniques have been used to greatly simplify a medical classification tool. Researchers at Harvard Medical School examined the Autism Diagnostic Interview, Revised (ADI-R) and Autism Diagnostic Observation Schedule (ADOS) tools which are used in the diagnosis of autism. One or both of these tests are used to make a diagnosis, but both require a trained clinician to administer them, which has caused a wait of more than a year between the warning signs of autism in a child and when they can receive a formal diagnosis. The new test reduces the time needed for a diagnosis by 95% without sacrificing any accuracy.
With the artificial intelligence technique the researchers looked though hundreds of the completed surveys and tested to see what questions were redundant. The 93 questions of the ADI-R were reduced to just seven questions while maintaining near 100% accuracy. The ADOS was shortened from 29 steps to just eight and again, accuracy was not compromised.
The survey and some videos are now up online for the public to look at, as a means to evaluate the effectiveness of the new tests further. The researchers hope to mobilize the tests even further, so the entire population has access to them.