Automatic retinal image analysis enables autism screening

Experts at the Chinese University of Hong Kong (CUHK) have developed an automatic retinal image analysis (ARIA) technology for screening of autism spectrum disorder (ASD).

“The ARIA technology uses a machine learning approach to analyze retinal images, including those obtained from very young children, for objective risk assessment of ASD that is noninvasive, fully automatic and relatively convenient,” said Professor Benny Zee of the Jockey Club School of Public Health and Primary Care (JCSPHPC), CUHK, who led the study.

The study included 46 participants aged 7–20 years with ASD from three Hong Chi Schools, a government-funded school system for students with ASD and mild-to-moderate intellectual disabilities, and 24 control participants from the community optometry clinic. Within the cohort, 23 participants with ASD were age- and gender-matched to 23 control participants. All colour retinal images of the 46 participants (average age, 13.17 years; male, 73.9 percent) were captured using a nonmydriatic fundus camera without dilation. Machine learning and deep learning techniques were used to develop a classification model for ASD based on information of the retina. [EClinicalMedicine, doi: 10.1016/j.eclinm.2020.100588]

Study results showed that participants with vs without ASD had significantly increased nipping (p<0.001), haemorrhage (p=0.032 and p=0.055 for left and right eyes, respectively), exudates (p=0.008 and p=0.042 for left and right eyes, respectively), optic disc diameter (p=0.034), and optic cup diameter (p=0.050). The sensitivity and specificity of the classification model were 95.7 percent and 91.3 percent, respectively.

“The importance of the study lies not only in scientific and technological advancement, but also in the positive social impact it creates in improving social capital through close collaboration among different institutions and team members,” said Ms Maria Lai of JCSPHPC, CUHK.

In Hong Kong, the ARIA technology has been applied to risk assessment of stroke and presence of white matter hyperintensities as an early risk factor of dementia. The research team has recently received the Outstanding Social Capital Partnership Award and the Social Capital Builder Logo Award from the Hong Kong government as recognition of their contribution with the ARIA technology in improving social capital for health promotion.

“Every student’s need is unique, especially for students with ASD and special educational needs. Through the development of such an innovative technology and partnership with optometrists in the local community, it is hoped that ASD screening and appropriate intervention can be provided to at-risk children at an early stage in the community setting, and public awareness about the needs of children with ASD and special education needs can also be enhanced, thus improving social acceptance and inclusion of these children to the society,” said Ms Sally Chiu of Hong Chi Association.