Tech­nol­ogy spots good em­bryos bet­ter

The Star Early Edition - - HEALTH -

LON­DON: Sci­en­tists are us­ing ar­ti­fi­cial in­tel­li­gence (AI) to help pre­dict which em­bryos will re­sult in IVF suc­cess.

In a new study, AI was found to be more ac­cu­rate than em­bry­ol­o­gists at pin­point­ing which em­bryos had the po­ten­tial to re­sult in the birth of a healthy baby.

Ex­perts from Sao Paulo State Uni­ver­sity in Brazil teamed up with Bos­ton Place Clinic in Lon­don to de­velop the tech­nol­ogy in col­lab­o­ra­tion with Dr Cristina Hick­man, sci­en­tific ad­viser to the Bri­tish Fer­til­ity So­ci­ety.

They believe the in­ex­pen­sive tech­nique has the po­ten­tial to trans­form care for pa­tients and help women achieve preg­nancy sooner.

Dur­ing the process, AI was “trained” in what a good em­bryo looks like from a se­ries of im­ages.

AI is able to recog­nise and quan­tify 24 image char­ac­ter­is­tics of em­bryos that are in­vis­i­ble to the hu­man eye.

These in­clude the size of the em­bryo, tex­ture of the image and bi­o­log­i­cal char­ac­ter­is­tics such as the num­ber and ho­mo­gene­ity of cells.

Dur­ing the study, which used cat­tle em­bryos, 48 im­ages were eval­u­ated three times each by em­bry­ol­o­gists and by the AI sys­tem.

The em­bry­ol­o­gists could not agree on their find­ings across the three im­ages, but AI led to com­plete agree­ment.

Stu­art Lav­ery, di­rec­tor of the Bos­ton Place Clinic, said the tech­nol­ogy would not re­place ex­am­in­ing chro­mo­somes in de­tail, which is thought to be a key fac­tor in de­ter­min­ing which em­bryos are “nor­mal” or “ab­nor­mal”.

“Look­ing at chro­mo­somes does work, but it is ex­pen­sive and it is in­va­sive to the em­bryo.

“What we are look­ing for here is some­thing that can be univer­sal. In­stead of a hu­man look­ing at thou­sands of im­ages, ac­tu­ally a piece of soft­ware looks at them and is ca­pa­ble of learn­ing all the time.

“As we get data about which em­bryos pro­duce a baby, that data will be fed back into the com­puter, and the com­puter will learn.

“What we have found is that the tech­nique is much more con­sis­tent than an em­bry­ol­o­gist, it is more re­li­able. It can also look for things that the hu­man eye can’t see.”

He said work was un­der way to look back at im­ages from par­ents who had ge­netic screen­ing and be­came preg­nant.

Ap­ply­ing AI to those im­ages would help the com­puter learn, he said.

“This is an in­no­va­tive and ex­cit­ing project com­bin­ing state-of-the-art em­bry­ol­ogy with new ad­vances in com­puter mod­el­ling, all with the aim of se­lect­ing the best pos­si­ble em­bryo for trans­fer to give all our pa­tients the best pos­si­ble chance of hav­ing a baby.

“Al­though fur­ther work is needed to op­ti­mise the tech­nique, we hope that a sys­tem will be avail­able shortly for use in a clin­i­cal set­ting,” Lav­ery added.

Newspapers in English

Newspapers from South Africa

© PressReader. All rights reserved.