A new AI can detect and diagnose childhood autism just by scanning photos of their eyes. The system is comprised of a deep-learning AI algorithm, which diagnosed childhood autism with 100% accuracy during a test phase. The findings surrounding the new detection system could allow for better screening tools and early diagnosis.
The screening relies on a new method that scientists came up with, which allows them to access information about the brain by looking at the back of the eye, where the retina and optic nerve connect to the optic disc. This method has allowed scientists to develop a non-invasive means of rapidly diagnosing concussions by shining eye-safe lasers onto the retina.
The AI autism detection system works similarly, and thus far, it appears to work out exceptionally well. To test the algorithms, the researchers recruited 958 participants, all with a mean age of 7.8 years. They then photographed their retinas for a total of 1,890 images.
The study notes that half of the participants had already been diagnosed with autism spectrum disorder (ADS) and that half were age- and sex-matched control participants. The AI system was then trained on 85 percent of the retina images, along with symptom severity scores, to help it construct models on which to base its diagnosis. The other 15 percent were kept for testing.
The researchers found that the AI was able to detect children with autism within a mean area under the receiver operating characteristic (AUROC) curve of 1.00, showcasing that the AI was able to detect the correct diagnosis 100 percent of the time.
This model, of course, is exceptionally promising, as it shows that AI could help with the detection of childhood autism much faster, especially in places where a specialty child psychiatrist isn’t readily available. Of course, it still has a long way to go before it can replace human-driven testing entirely, but it is a notable accomplishment that continues to show the efficiency and helpfulness AI systems could bring to the medical industry.
A new AI can detect and diagnose childhood autism just by scanning photos of their eyes. The system is comprised of a deep-learning AI algorithm, which diagnosed childhood autism with 100% accuracy during a test phase. The findings surrounding the new detection system could allow for better screening tools and early diagnosis.
The screening relies on a new method that scientists came up with, which allows them to access information about the brain by looking at the back of the eye, where the retina and optic nerve connect to the optic disc. This method has allowed scientists to develop a non-invasive means of rapidly diagnosing concussions by shining eye-safe lasers onto the retina.
The AI autism detection system works similarly, and thus far, it appears to work out exceptionally well. To test the algorithms, the researchers recruited 958 participants, all with a mean age of 7.8 years. They then photographed their retinas for a total of 1,890 images.
The study notes that half of the participants had already been diagnosed with autism spectrum disorder (ADS) and that half were age- and sex-matched control participants. The AI system was then trained on 85 percent of the retina images, along with symptom severity scores, to help it construct models on which to base its diagnosis. The other 15 percent were kept for testing.
The researchers found that the AI was able to detect children with autism within a mean area under the receiver operating characteristic (AUROC) curve of 1.00, showcasing that the AI was able to detect the correct diagnosis 100 percent of the time.
This model, of course, is exceptionally promising, as it shows that AI could help with the detection of childhood autism much faster, especially in places where a specialty child psychiatrist isn’t readily available. Of course, it still has a long way to go before it can replace human-driven testing entirely, but it is a notable accomplishment that continues to show the efficiency and helpfulness AI systems could bring to the medical industry.