Using Artificial Intelligence to Diagnose ADHD


WASHINGTON, Nov. 10, 2021 (GLOBE NEWSWIRE) -- A recently published article in Experimental Biology and Medicine (Volume 246, Issue 21, November, 2021) investigates the use of artificial intelligence in diagnosing attention-deficit hyperactivity disorder. The study, led by Dr. Yichun Liu, in the Center for Applied Genomics (CAG) at Children's Hospital of Philadelphia in Philadelphia, PA (USA), reports that the accuracy for deep-learning methods (78%) is superior to traditional clustering methods (50%). 

Attention-deficit hyperactivity disorder (ADHD) is a common psychiatric disorder in children that causes lifelong impairments. Accurate diagnosis of ADHD is impaired by disease heterogeneity as well as the long delay between symptom onset and diagnosis. While new guidelines have improved mental disorder diagnosis, objective screening methodologies and lab tests are still lacking for ADHD. There is increasing evidence that copy number variation (CNV), a type of duplication or deletion event within the human genome, correlates with and influences the development of mental disorders, including ADHD. Artificial intelligence models have been used to analyze whole genome sequences from large numbers of patients and successfully predict drug resistance as well as classify tumors. However, these models have not been applied to the diagnosis of ADHD.  

In this study, Dr. Liu and colleagues performed whole genome sequencing on 524 African American individuals, including 116 ADHD patients and 408 healthy controls. When copy number variation (CNV) intensity was used as the predictive feature vector for the artificial intelligence (AI) model, the accuracy of identifying African American ADHD patients and healthy controls was ~78%, which represents a significant improvement when compared to traditional methods (50%). The accuracy for identifying ADHD and healthy control European American children was reduced but still above 70%. Another important finding is that the accuracy of CNVs in non-coding regions was comparable to CNVs in protein coding regions, suggesting that non-coding genomic regions may play a more important role than previously thought. Dr. Liu said, "Next generation sequencing and AI deep learning have achieved success in each of their fields in the past decade. Combining both together forms a powerful tool for the prediction and diagnosis of complex mental disorders, such as ADHD, especially for toddlers who are too young to be diagnosed. Dr. Hakonarson, a co-author for the study, added, "Besides a valuable tool for ADHD diagnosis, the AI model combined with whole genome sequencing technology also identified copy number variations (CNVs) and their corresponding genomic regions that warrant further study. These genomic regions, especially non-coding regions with structural variations, may impact both susceptibility to ADHD and how patients respond to therapeutic interventions."

Dr. Steven R. Goodman, Editor-in-Chief of Experimental Biology and Medicine, said, "Liu and colleagues have combined the powerful tools of next gen sequencing and AI/Deep Learning in the diagnosis of ADHD in African American and European American children. In both cases, the accuracy was superior to traditional clustering methodology."

Experimental Biology and Medicine is a global journal dedicated to the publication of multidisciplinary and interdisciplinary research in the biomedical sciences. The journal was first established in 1903. Experimental Biology and Medicine is the journal of the Society of Experimental Biology and Medicine. To learn about the benefits of society membership, visit www.sebm.org. If you are interested in publishing in the journal, please visit http://ebm.sagepub.com/.

For more information, please contact ebm@sebm.org.

Related Images






Image 1



This content was issued through the press release distribution service at Newswire.com.

Attachment


Featured Image for Experimental Biology and Medicine

Tags