New di­ag­nos­tic tool could has­ten ovar­ian can­cer treat­ment

Vancouver Sun - - FRONT PAGE - SHAWN CON­NER

A new type of di­ag­nos­tic sys­tem us­ing so­phis­ti­cated com­puter soft­ware ca­pa­ble of an­a­lyz­ing and com­par­ing can­cer­ous tis­sue against a vast databank of dig­i­tal im­ages of can­cer sam­ples may speed up treat­ment for ovar­ian can­cer.

The sys­tem, be­ing de­signed by Aicha Ben­Taieb, a Si­mon Fraser Univer­sity com­put­ing science PhD stu­dent, aims to au­to­mate the iden­ti­fi­ca­tion of ovar­ian car­ci­no­mas for a faster and more re­li­able di­ag­no­sis.

Ovar­ian can­cer is the fifth most com­mon can­cer for women. It is es­ti­mated that this year in Canada, 2,800 women will be di­ag­nosed, and that 1,750 women will die from it.

The out­comes haven’t sig­nif­i­cantly im­proved in over 50 years. Its causes are un­known and there is a lim­ited un­der­stand­ing of its pro­gres­sion.

What is known is that there are five main sub­types. Ef­fec­tive treat­ment de­pends on iden­ti­fy­ing the sub­type as soon as pos­si­ble. But cur­rent meth­ods are sub­jec­tive, time-con­sum­ing and prone to er­ror.

Us­ing in­for­ma­tion col­lated via com­put­ers, Ben­Taieb be­lieves she has found a bet­ter way to iden­tify these sub­types.

Each sub­type shows in­di­vid­ual struc­tural and cel­lu­lar char­ac­ter­is­tics. Cur­rently, pathol­o­gists an­a­lyze tis­sue sam­ples us­ing a mi­cro­scope, dig­i­tal scan­ner and com­puter soft­ware.

How­ever, iden­ti­fi­ca­tion can easily be im­paired by tech­ni­cal fac­tors such as light­ing and the pathol­o­gist’s ex­pe­ri­ence. The anal­y­sis can also be time­con­sum­ing.

“What hap­pens now is a pathol­o­gist is of­ten un­sure, so he’s most likely to ask for a sec­ond opin­ion,” Ben­Taieb said.

“Depend­ing on where you are, in Van­cou­ver or the Yukon, that sec­ond opin­ion is go­ing to take more or less time. And if you don’t have ac­cess to an ex­pert, of­ten the start of the treat­ment is de­layed.”

The pathol­o­gist might also per­form ex­tra tests that are costly and not al­ways avail­able in ev­ery pathol­ogy cen­tre.

With Ben­Taieb’s method, an ar­ti­fi­cial in­tel­li­gence fea­ture is in­te­grated into the soft­ware that helps the pathol­o­gist an­a­lyze the tis­sue sam­ple. This fea­ture is trained, through a large data set of ex­pert-an­no­tated slides, to au­to­mat­i­cally iden­tify the char­ac­ter­is­tic vis­ual pat­terns for each sub­type of car­ci­noma.

“We’re look­ing at the whole im­age, in dif­fer­ent re­gions, in a more ef­fi­cient way, us­ing com­put­ers to ex­tract im­age-based fea­tures,” Ben­Taieb said.

Ben­Taieb is de­sign­ing the sys­tem at SFU’s Med­i­cal Im­age Anal­y­sis Lab un­der the su­per­vi­sion of com­put­ing science pro­fes­sor Ghas­san Ha­marneh, and in col­lab­o­ra­tion with pathol­o­gists Dr. David Hunts­man and Dr. Hec­tor Li Chang from the BC Can­cer Agency.

The tech­nol­ogy is still in its early days, and im­ple­men­ta­tion in op­er­at­ing rooms and pathol­ogy cen­tres may take months, if not years.

But the re­sults have been promis­ing. The sys­tem is al­ready able to cor­rectly clas­sify the sub­type in more than 92 per cent of cases.

Ideally, the sys­tem will even­tu­ally be in place in op­er­at­ing rooms, and take place in real time.

“A clin­i­cian or sur­geon would cut into the pa­tient,” Ben­Taieb said, “take some tis­sue, look at a sam­ple un­der a mi­cro­scope, and then the soft­ware would au­to­mat­i­cally say, ‘ it’s this sub­type.’”

Si­mon Fraser Univer­sity PhD stu­dent Aicha Ben­Taieb is de­vel­op­ing a di­ag­nos­tic tool that can au­to­mat­i­cally iden­tify ovar­ian can­cers.

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.