Fa­cial recog­ni­tion soft­ware mea­sures pain

Modern Healthcare - - INNOVATIONS - By Adam Ruben­fire

The ex­pe­ri­ence of pain is in­her­ently sub­jec­tive. One pa­tient’s rat­ing of pain as 5 on a 10-point scale could be another pa­tient’s 10.

It’s dif­fi­cult enough for providers to as­sess pain lev­els in adult pa­tients with the abil­ity to de­scribe their pain and com­pare it with past ex­pe­ri­ences. It’s much harder in treat­ing chil­dren or adults with de­men­tia, who of­ten are not able to ver­bally de­scribe their pain. Now some re­searchers are seek­ing a tech­no­log­i­cal way to ob­jec­tively mea­sure pain.

Dr. Jean­nie Huang, a pe­di­atric gas­troen­terol­o­gist at Rady Chil­dren’s Hos­pi­tal-San Diego, said she finds it dis­tress­ing not be­ing able to ac­cu­rately as­sess her young pa­tients’ pain. Some of her pa­tients have stayed in the hos­pi­tal longer be­cause clin­i­cians did not ac­cu­rately as­sess their pain, iden­tify the source and treat the prob­lem ef­fec­tively. Huang wants to find an un­bi­ased way to as­sess pain in pa­tients who can’t tell her.

A 2008 study in the Jour­nal of Ad­vanced Nurs­ing found that nurses’ es­ti­mates of pe­di­atric pa­tients’ pain were only mod­er­ately cor­re­lated with pa­tients’ self-re­ported pain rat­ing.

“The bot­tom line is (that) what the pa­tient is ex­pe­ri­enc­ing is what’s most im­por­tant,” said Bob Twill­man, ex­ec­u­tive di­rec­tor of the Amer­i­can Academy of Pain Man­age­ment. “To the ex­tent that we’re able to elicit that de­ter­mines whether we’ll be able to give (pa­tients) the re­lief they need.”

Huang, a pro­fes­sor of pe­di­atrics at the Univer­sity of Cal­i­for­nia, San Diego, so­licited the help of Mar­ian Bartlett, a re­search pro­fes­sor at UCSD’s In­sti­tute for Neu­ral Com­pu­ta­tion. Bartlett was us­ing fa­cial recog­ni­tion soft­ware to teach autis­tic chil­dren how to rec­og­nize and in­ter­pret fa­cial ex­pres­sions. Huang won­dered whether the same soft­ware could be used to mea­sure pain.

In 2012, Huang, Bartlett and a team of re­searchers launched a five-year Na­tional In­sti­tutes of Health-funded pro­ject to de­ter­mine whether pe­di­atric pa­tients’ pain could be ac­cu­rately mea­sured by fa­cial recog­ni­tion soft­ware de­vel­oped by Emo­tient, a San Diego com­pany co­founded by Bartlett.

The pro­to­type soft­ware was pro­grammed to rec­og­nize 20 fa­cial mus­cle move­ments known to in­di­cate pain. The ap­pli­ca­tion is based on data from prior soft­ware de­vel­oped by Emo­tient that uses com­puter vi­sion tech­niques to an­a­lyze fa­cial ex­pres­sions.

Through a process of ma­chine learn­ing, the soft­ware is “taught” to mea­sure pain based on video im­ages of pa­tients in pain. Then it per­forms a re­gres­sion anal­y­sis of the lev­els of in­ten­sity in those pa­tients’ dis­played signs of pain. The soft­ware ap­plies the data to mea­sure pain in other sub­jects, us­ing a 0-to-10 scale.

In a study to be pub­lished in the July 2015 is­sue of Pe­di­atrics, the soft­ware was taught and tested on 50 chil­dren, ages 5 to 18, who had un­der­gone la­paro­scopic ap­pen­dec­tomies at Rady Chil­dren’s Hos­pi­tal. Video im­ages of the chil­dren’s faces were cap­tured within 24 hours af­ter their ap­pen­dec­tomy, one day later, and at a fol­low-up visit two to four weeks af­ter surgery.

The ap­pli­ca­tion’s pain rat­ings cor­re­lated strongly with pa­tients’ self-re­ported pain lev­els. The soft­ware’s as­sess­ment of pain came closer than nurses’ assess­ments to those self-re­ported lev­els. And the soft­ware per­formed as well as the pa­tients’ par­ents in es­ti­mat­ing pain sever­ity. That’s sig­nif­i­cant be­cause par­ents’ es­ti­mates have been shown to be closer than clin­i­cians’ es­ti­mates to chil­dren’s self-re­ported pain.

Bartlett said Emo­tient even­tu­ally hopes to mar­ket the pain recog­ni­tion soft­ware. It’s not clear whether the de­vice would re­quire Food and Drug Ad­min­is­tra­tion ap­proval. The com­pany still needs to re­fine the soft­ware us­ing more video of pa­tients ex­pe­ri­enc­ing var­i­ous lev­els of pain.

Dr. Lynn Web­ster, a past pres­i­dent of the Amer­i­can Academy of Pain Medicine, said there’s prom­ise in the Emo­tient ap­pli­ca­tion’s real-time abil­ity to help clin­i­cians as­sess pain and bet­ter cal­i­brate their med­i­cal in­ter­ven­tions. It could help physi­cians re­duce or in­crease pain med­i­ca­tions. “We’re al­ways seek­ing biomark­ers that can help us treat chronic ill­nesses like pain,” Web­ster said. “Fa­cial recog­ni­tion could be very help­ful in pro­vid­ing use­ful in­for­ma­tion.”

Bartlett said that en­abling providers to track pa­tients’ pain is what could make the Emo­tient ap­pli­ca­tion com­mer­cially vi­able. The com­pany is in talks with sev­eral telemedicine providers in­ter­ested in the soft­ware and would like to bring it to mar­ket in 2016, she said.

Soft­ware is be­ing de­vel­oped to rec­og­nize fa­cial mus­cle move­ments as they re­late to pain.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.