Chi­nese AI gives near­sighted chil­dren a glimpse of the fu­ture

China Daily - - CHINA - By YUAN QUAN and JING HUAIQIAO

If you are con­cerned that your child will be near­sighted, a new ar­ti­fi­cial in­tel­li­gence model de­vel­oped from mil­lions of eye­sight records could help pre­dict whether your off­spring will need glasses.

My­opia is the most com­mon vis­ual im­pair­ment in chil­dren, and China has an un­prece­dented rate of near­sight­ed­ness. A re­cent World Health Or­ga­ni­za­tion re­port showed that about 600 mil­lion Chi­nese, al­most half the pop­u­la­tion, are near­sighted, in­clud­ing more than 70 per­cent of high school and col­lege stu­dents, and 40 per­cent of pri­mary school chil­dren.

Cur­rent ap­proaches to curb­ing vi­sion loss in­clude eye­drops, glasses, con­tact lenses and surgery. How­ever, while these can be ef­fec­tive, they have side ef­fects, such as higher rates of re­cur­rence, eye in­fec­tions and other ail­ments.

If near­sight­ed­ness could be pre­dicted, med­i­cal pro­fes­sion­als could in­ter­vene with ap­pro­pri­ate treat­ments to help re­duce the risk of high my­opia, which is mea­sured by a fo­cus­ing power of -6 diopters, a mea­sure­ment of the op­ti­cal power of a lens, and higher.

After an­a­lyz­ing 1.25 mil­lion eye­sight records over three years, re­searchers from Zhong­shan Oph­thalmic Cen­ter at Sun Yat-sen Univer­sity in Guang­dong prov­ince have iden­ti­fied my­opia de­vel­op­ment rules, and built an AI model to pre­dict the con­di­tion in chil­dren and teenagers.

The study, pub­lished in the in­ter­na­tional jour­nal PLOS Medicine, in­volved chil­dren ages 5 to 18 who had eye­sight checks from 2005 to 2015 in eight of the largest oph­thalmic cen­ters in the south­ern prov­ince.

The re­searchers dis­cov­ered that near­sight­ed­ness usu­ally oc­curs at age 7, and rapidly de­vel­ops be­fore age 10. It can grow to -3 diopters dur­ing the teenage years and up to -6 diopters in the 20s.

There were few cases of high my­opia among school-age chil­dren, and re­searchers did not find the on­set or de­vel­op­ment age of high my­opia.

The re­searchers used age, the diopter and an­nual my­opia pro­gres­sion rates as the main vari­ables to de­velop an al­go­rithm to pre­dict de­grees of my­opia over 10 years and the pos­si­bil­ity of high my­opia be­fore 18 years.

To test the model, the de­vel­op­ers fed it about 687,000 eye­sight records of more than 129,000 peo­ple.

The di­ag­nos­tic ac­cu­racy was 90 per­cent within three years, and 80 per­cent within 10 years. It can also pre­dict high my­opia eight years in ad­vance, pro­vid­ing a sci­en­tific ba­sis for in­ter­ven­tion, study leader Liu Yizhi said.

The re­search team re­cently made the AI model pub­lic and demon­strated how it works by us­ing the case of a 5-year-old boy who was near­sighted and be­gan wear­ing glasses at -1 diopter last year.

A re­searcher in­putted two my­opia records taken at least 12 months apart, and within sec­onds the model showed that the boy’s diopters might reach -3 after 10 years, but he had lit­tle risk of high my­opia.

High my­opia usu­ally pro­gresses rapidly, and can cause blind­ness or other se­vere eye con­di­tions, said Lin Hao­tian, the lead au­thor of the study. The con­di­tion can also be as­so­ci­ated with ge­net­ics.

The risk of chil­dren de­vel­op­ing high my­opia is a great con­cern for Chi­nese par­ents, with thou­sands of stu­dents seek­ing care at hos­pi­tals and oph­thalmic clin­ics dur­ing school hol­i­days.

The rise in my­opia is partly driven by chil­dren spend­ing more time read­ing, study­ing, or glued to com­puter and smart­phone screens.

Re­search in­di­cates that an ef­fec­tive way of curb­ing wors­en­ing my­opia is to spend more time out­doors.

“Stud­ies have proved that if chil­dren have an ad­di­tional 40 min­utes out­side ev­ery day, the my­opia rate will fall by 23 per­cent in three years,” Liu said.

The AI model could also help com­bat the lim­ited num­ber of spe­cial­ists. By the end of 2014, China had just 36,000 oph­thal­mol­o­gists, with 70 per­cent of them work­ing in big cities.

Many spend most of their time treat­ing se­vere eye con­di­tions, but lit­tle on the preven­tion and con­trol of my­opia.

“The AI model will help ease the work­load for oph­thal­mol­o­gists and im­prove di­ag­nos­tic ac­cu­racy,” Liu said.

Last year, Liu’s team and re­searchers from other uni­ver­si­ties un­veiled an AI sys­tem that di­ag­noses cataracts with a high de­gree of ac­cu­racy.

The my­opia pre­dic­tion model will be put into clin­i­cal use soon.

Newspapers in English

Newspapers from China

© PressReader. All rights reserved.