TL;DR:
- Researchers employ AI to diagnose childhood autism with 100% accuracy through retinal imaging.
- The AI algorithm analyzes retinal images and outperforms human specialists in early diagnosis.
- This technology could serve as an objective screening tool when access to child psychiatrists is limited.
- Retinal alterations may hold potential as biomarkers for autism.
- The AI model shows promise in assessing symptom severity.
- Researchers suggest its applicability from age four onwards, with further validation needed for younger children.
Main AI News:
In a groundbreaking development, researchers have harnessed the power of artificial intelligence to diagnose childhood autism with unparalleled precision. Through the analysis of retinal images captured from children, a deep learning AI algorithm has achieved a remarkable 100% accuracy rate. This pioneering advancement underscores the potential of AI as a definitive screening tool for early autism diagnosis, particularly in situations where access to specialized child psychiatrists is limited.
At the core of this innovation lies the intricate structure of the eye, where the retina and the optic nerve intersect at the optic disc. Serving as an extension of the central nervous system, this delicate structure offers a unique window into the complexities of the brain. Researchers have increasingly recognized the value of non-invasive access to the retina in gathering crucial brain-related information.
Notably, British researchers have previously introduced a non-invasive method for swiftly diagnosing concussions by employing an eye-safe laser to examine the retina. Building upon this foundation, scientists from Yonsei University College of Medicine in South Korea have devised an ingenious technique for diagnosing autism spectrum disorder (ASD) and assessing symptom severity in children. This method relies on the examination of retinal images, analyzed by a sophisticated AI algorithm.
The research team meticulously recruited 958 participants, with an average age of 7.8 years, and captured retinal images, amassing a total of 1,890 images. Half of the participants had received an ASD diagnosis, while the other half consisted of age- and gender-matched controls. To gauge ASD symptom severity, the researchers employed the Autism Diagnostic Observation Schedule – Second Edition (ADOS-2) calibrated severity scores and Social Responsiveness Scale – Second Edition (SRS-2) scores.
Using 85% of the retinal images and the associated symptom severity scores, a convolutional neural network, a deep learning algorithm, was trained to formulate models for ASD screening and assessing ASD symptom severity. The remaining 15% of the images were held for testing purposes.
The AI demonstrated remarkable accuracy in identifying children with ASD in the test set, with a mean area under the receiver operating characteristic (AUROC) curve of 1.00. AUROC values range from 0 to 1, with 0 indicating complete incorrectness and 1 signifying absolute correctness. This means that the AI’s predictions in this study achieved a flawless 100% accuracy rate, even when 95% of the least critical areas of the image were removed, excluding the optic disc.
The researchers commented, “Our models exhibited promising performance in distinguishing between ASD and TD [typically developing] children using retinal photographs, suggesting that retinal changes in ASD may hold potential value as biomarkers. Interestingly, these models maintained a mean AUROC of 1.00 while utilizing only 10% of the image containing the optic disc, underscoring the critical role of this area in distinguishing ASD from TD.”
For assessing symptom severity, the mean AUROC value was 0.74. In the context of AUROC values, 0.7 to 0.8 is considered ‘acceptable,’ and 0.8 to 0.9 is deemed ‘excellent.’
The researchers noted, “Our findings imply that retinal photographs can provide supplementary insights into symptom severity. We observed that reliable classification was attainable only for ADOS-2 scores, not for SRS-2 scores. This divergence may stem from the fact that ADOS-2 assessments are conducted by trained professionals with ample time for evaluation, whereas SRS-2 assessments are typically completed by caregivers in a shorter timeframe, thereby potentially less accurately reflecting an individual’s severity status.”
Remarkably, study participants as young as four years old were included in the research. Based on their findings, the researchers propose that their AI-based model could serve as an objective screening tool from this early age onward. However, as the retina continues to develop until the age of four, further investigation is warranted to validate the tool’s accuracy for participants younger than that.
The researchers concluded, “Although additional studies are needed to establish the tool’s applicability across a broader population, our study marks a significant stride towards the development of objective screening tools for ASD. This advancement could play a pivotal role in addressing pressing issues, such as the limited accessibility of specialized child psychiatry assessments due to resource constraints.“
Conclusion:
The groundbreaking AI-powered diagnosis of childhood autism through retinal imaging represents a significant advancement in healthcare technology. This innovation has the potential to address the scarcity of specialized child psychiatry resources and provide more accessible and accurate early diagnoses. Moreover, the discovery of potential biomarkers in the retina opens doors for further research and development in the field of neurological disorders, promising both medical and commercial opportunities in the market.
