Early detection of autism spectrum disorder is vital for giving children the best start in life, but diagnosis is often delayed due to lengthy and resource-intensive assessments. A new study has developed a computer-based method that could make the process faster and more accurate, offering real hope to families waiting for answers. The findings are published in the journal SN Computer Science.
Researchers from the University of Hyderabad tested a novel algorithm that applies advanced mathematical techniques to screening data for autism. By reducing the size and complexity of the data while preserving its most important features, the approach allows quicker identification of patterns linked to the condition. Unlike traditional clinical methods, which rely heavily on expert observation and can take many hours, this technique is designed to deliver results in a fraction of the time.
The method, called low-rank binary matrix approximation using singular value decomposition, converts raw screening information into a simplified binary form. This makes it easier to cluster data points and classify whether someone is likely to show signs of autism. In practice, this means information gathered from common autism screening tools can be processed more efficiently, potentially making it easier to identify children at risk in community or primary care settings.
The researchers applied the algorithm to four datasets covering toddlers, children, adolescents, and adults. Results showed a marked improvement compared with existing machine learning methods, with accuracy rates reaching more than 65% for adults and close to 50% for toddlers. While these numbers may not appear high by everyday standards, they represent a significant leap over older techniques, some of which achieved less than 20% accuracy in certain groups. The new method also outperformed others in speed, completing analyses in seconds where older approaches required far longer.
Autism spectrum disorder affects communication, social interaction, and behaviour, and signs are usually present in early childhood. Early intervention has been shown to improve language skills, social development, and long-term outcomes, but timely diagnosis remains a challenge worldwide. Screening tools such as the Autism Diagnostic Observation Schedule or the Childhood Autism Rating Scale are widely used, yet they require trained clinicians and careful interpretation. A faster, computer-assisted approach could therefore play a vital supporting role.
Beyond autism, the researchers suggest that their algorithm could be adapted for other conditions where large sets of behavioural or medical data need to be analysed quickly. Its ability to cut through noise and highlight key patterns means it could contribute to more efficient healthcare systems, especially in resource-limited settings where clinical expertise is scarce.
The team note that their work does not replace traditional clinical assessments but rather complements them. The algorithm offers a way to screen more people in less time, flagging those most in need of detailed follow-up by specialists. As digital health tools continue to expand, approaches like this one may help bridge the gap between demand for assessments and the availability of trained professionals.
Meta-excerpt: A new algorithm speeds up autism screening by simplifying data, offering faster and more accurate early detection.