Researchers took pictures of children's retinas and screened them using deep learning artificial intelligence algorithms, and they were surprised to find that autism was diagnosed with 100 percent accuracy. The findings support the use of artificial intelligence as an objective screening tool for early diagnosis, especially when child psychiatry specialists have limited staff.
learned an artificial intelligence algorithm to screen for it, and they were surprised to find that it was 100% accurate in diagnosing autism. The findings support the use of artificial intelligence as an objective screening tool for early diagnosis, especially when child psychiatry specialists have limited staff.
At the back of the eye, the retina and optic nerve connect at the optic disc. The optic disc is an extension of the central nervous system, a window into the brain, and researchers have begun to exploit their ability to easily and non-invasively access this part of the body to obtain important brain-related information.
Recently, British researchers created a non-invasive method to quickly diagnose concussions by shining an eye-safe laser onto the retina. Now, researchers at Yonsei University School of Medicine in South Korea have developed a way to diagnose autism spectrum disorder (ASD) and symptom severity in children using retinal images screened by artificial intelligence algorithms.
The researchers recruited 958 participants with an average age of 7.8 years and photographed their retinas, resulting in a total of 1,890 images. Half of the participants were diagnosed with autism, and the other half were age- and gender-matched controls. Autism symptom severity was assessed using the Autism Diagnostic Observation Schedule-Second Edition (ADOS-2) calibrated severity score and the Social Responsiveness Scale-Second Edition (SRS-2) score.
trained a convolutional neural network (a deep learning algorithm) using 85% of retinal images and symptom severity test scores to build a model for screening ASD and ASD symptom severity. The remaining 15% of images are reserved for testing.
When screening for ASD on a test image set, the AI was able to single out children diagnosed with ASD with an average area under the receiver operating characteristic curve (AUROC) of 1.00. AUROC ranges from 0 to 1. A model that predicts 100% of the time incorrectly has an AUROC value of 0.0; a model that predicts 100% of the time correctly has an AUROC value of 1.0. Even when 95% of the least important regions in the image (excluding the optic disc) were removed, there was no significant decrease in average AUROC.
Researchers said: "Our models perform well in distinguishing ASD from TD (children with typical development) using retinal photographs, which means that retinal changes in ASD may have potential biomarker value. Interestingly, these models Using only 10% of the images containing the optic disc, an average AUROC value of 1.00 was retained, indicating that this region is critical for distinguishing ASD from TD."
The average AUROC value for symptom severity was 0.74, with an AUROC of 0.7 to 0.8 Values are "Acceptable" and AUROC values of 0.8 to 0.9 are "Excellent".
Researchers said: "Our results suggest that retinal photographs provide additional information about symptom severity. We found that only the ADOS-2 score allowed for feasible classification, but not the SRS-2 score. This may be because the ADOS- 2 is conducted by trained professionals with ample assessment time, while the SRS-2 is typically completed by caregivers within tens of minutes; therefore, the former will reflect a person's severity more accurately than the latter."
Study Participation The youngest is only four years old. The researchers say that based on their findings, their AI-based model could serve as an objective screening tool starting in this age group. Because the neonatal retina continues to grow until the age of four years, further research is needed to determine whether this tool can be accurately used on participants younger than four years old. "While future research is needed to determine generalizability, our study represents a noteworthy step toward the development of objective screening tools for ASD, which may help address pressing questions such as Specialized child psychiatric evaluation is not available due to limited resources."
The study was published in JAMA Network Open.
