Stanford University expert Dennis Wall is developing an algorithm for detecting child autism in collaboration with doctors and data scientists.
The idea of the algorithm is to identify the disease by watching a few minutes of video of a child playing at home. Normally diagnosis of autism takes doctors up to ten hours of intensive observation and testing of child behaviour. This process is also expensive.
Machine learning algorithms will precisely identify the problem, and this solution will be much better at comparing expensive face-to-face research with highly-skilled professionals, moreover it will take a matter of minutes.
In accordance with a recent publication in PLOS journal, AI algorithms help to significantly reduce autism detection time.
Researchers have developed a web application, applying 8ML models. The algorithm was tested with 162 minutes of home videos of children with and without autism. It combines 30 behavioural features, and will identify autism within four minutes.
This algorithm will help to distinguish kids with autism from kids with atypical development, using only short home videos up to six minutes long. This models help to identify the risk of autism as well. Already in use the algorithm applied to mobile videos helps to develop a more accurate disease detection model. The autism authentication programme will make it possible to identify autism out of clinic, so patients can get help faster.
Autism has a huge impact in the US, and diagnosis has increased 700% since 1996. This mobile solution will decrease diagnostics time so patients in need will get help quicker.