Wear­ables take on autism

A new wear­able that mea­sures bio­met­rics could help to shine a light on liv­ing with autism.

Focus-Science and Technology - - CONTENTS -

“The frus­tra­tion of not be­ing able to com­mu­ni­cate their predica­ment to oth­ers, or reg­u­late the re­sult­ing emo­tional dis­tress, can lead to a melt­down”

Seven years ago, I had one of the worst con­ver­sa­tions of my life. A pae­di­a­tri­cian de­liv­ered the news to me and my part­ner that our 18-month-old son, Callum, had a se­vere devel­op­men­tal de­lay and was pre­sent­ing symp­toms con­sis­tent with autism spec­trum dis­or­der (ASD). The di­ag­no­sis of ASD was later con­firmed. Callum has since at­tended a spe­cial­ist preschool and now goes to a spe­cial ed­u­ca­tional needs pri­mary. His progress has been good, but his con­di­tion re­mains mod­er­ate to se­vere, and it’s likely he’ll re­quire con­tin­ued sup­port into adult­hood. Know­ing that your child prob­a­bly won’t achieve in­de­pen­dence in later life throws up un­pleas­ant ques­tions. What’s go­ing to hap­pen when I’m gone? Who will care for him? Who will love him? But, as we were to find out, that wasn’t the worst of it.

Autism is a life­long devel­op­men­tal con­di­tion char­ac­terised by dif­fi­cul­ties with lan­guage and so­cial in­ter­ac­tion, and a ten­dency for repet­i­tive be­hav­iours. The dis­or­der of­ten man­i­fests it­self with other men­tal health is­sues, in­clud­ing learn­ing dis­abil­i­ties, de­pres­sion and anx­i­ety. It is a spec­trum con­di­tion, mean­ing that its symp­toms and their sever­ity vary greatly from one in­di­vid­ual to the next. Those who ex­pe­ri­ence autism range from the high func­tion­ing, such as nat­u­ral­ist and tele­vi­sion pre­sen­ter Chris Pack­ham, through to peo­ple for whom it’s a pro­found dis­abil­ity, pre­clud­ing the pos­si­bil­ity of an in­de­pen­dent life.

The US Cen­ters for Dis­ease Con­trol and Preven­tion es­ti­mate the autism preva­lence to be 1 in 59 chil­dren, with ap­prox­i­mately five times more males be­ing di­ag­nosed than fe­males. In the UK, the rate is thought to be around 1 in 100. The pre­cise cause of autism isn’t fully un­der­stood. Re­search points to a com­bi­na­tion of en­vi­ron­men­tal fac­tors, such as air pol­lu­tion and con­di­tions in the womb, and ge­net­ics – al­though, de­spite many genes be­ing im­pli­cated al­ready, we’re only just be­gin­ning to iden­tify all of those that con­trib­ute to autism risk, and the com­plex in­ter­ac­tions be­tween them.

It has also emerged that many autis­tic peo­ple process sen­sory in­for­ma­tion dif­fer­ently – to the point that some sen­sa­tions, loud sounds, for ex­am­ple, can cause pain. The frus­tra­tion of not be­ing able to com­mu­ni­cate their predica­ment to oth­ers, or to reg­u­late the re­sult­ing emo­tional dis­tress, can lead to a state of ex­treme anx­i­ety, known col­lo­qui­ally as a melt­down. It’s not naugh­ti­ness and it’s not a tantrum. It’s a fight-or-flight re­sponse to a state of se­vere dis­tress – the same dis­tress you or I might ex­pe­ri­ence if our lives were in dan­ger.


My part­ner and I have wit­nessed count­less melt­downs, dur­ing which Callum will scream as if he were be­ing tor­tured, sob un­con­trol­lably, break any­thing within reach, hit and bite us, and self-harm – punch­ing him­self and bang­ing his head into walls and floors. It’s one thing when these episodes oc­cur in the pri­vacy of home, but when they take place in pub­lic – as they in­evitably will – the con­se­quences can be dire. There have been times when, in fury, he’s al­most run out into traf­fic. Vis­its to the bar­ber’s or the den­tist be­came im­pos­si­ble. Pubs, restau­rants – any­where noisy and in close prox­im­ity to other peo­ple

– equally so. And then there’s the spec­ta­cle of the full-on pub­lic melt­down – and the dis­ap­prov­ing looks and the mut­tered re­marks from those who don’t un­der­stand, and who write us off as in­ept par­ents of a dis­obe­di­ent child. What­ever the long-term out­look, when day-to-day life be­comes such an un­re­lent­ing strug­gle, it does drive you to the prover­bial depths.

Callum was non-ver­bal un­til the age of six. He’s now eight and, hap­pily, his melt­downs have sub­sided some­what as his com­mu­ni­ca­tion skills have im­proved. But while they lasted, they kept us un­der ef­fec­tive house ar­rest much of the time. And even when we did go out it would rarely be to the shops, or the cin­ema, or a res­tau­rant – but usu­ally on a long walk, away from other peo­ple.

The thing was that we never quite knew when a melt­down was go­ing to strike. They were of­ten brought on by the most seem­ingly in­nocu­ous trig­gers, like the sound of a dog bark­ing or even birds singing. Some­times there was no ob­vi­ous trig­ger at all. And this left us vir­tu­ally in­ca­pable of do­ing any­thing to stop them.

But imag­ine if par­ents, or care­givers in gen­eral, could re­ceive a no­ti­fi­ca­tion to their mo­bile phone the in­stant their child’s anx­i­ety lev­els be­gin to rise. Re­searchers at North­east­ern Univer­sity, Maine Med­i­cal Cen­tre and the Univer­sity of Pitts­burgh are de­vel­op­ing just such a sys­tem. It works us­ing a wrist­band, rather like a sports watch, that mon­i­tors bio­met­ric data (literally mean­ing ‘body mea­sure­ments’) – specif­i­cally, the wearer’s heart­beat, skin tem­per­a­ture, sweat lev­els and ac­cel­er­a­tion. The lat­ter is im­por­tant in autis­tic peo­ple, who of­ten flap their arms as a way to emo­tion­ally reg­u­late them­selves (one of a group of be­hav­iours known col­lec­tively as ‘stim­ming’).

In a study con­ducted in the Devel­op­men­tal Dis­or­ders Unit at Spring Harbor Hos­pi­tal in Port­land, Maine, and pub­lished ear­lier this year, the team fit­ted the wrist­band to 20 non-ver­bal autis­tic in-pa­tients, aged be­tween 6 and 17 years. Data from the bands was trans­mit­ted by Blue­tooth to a server where it was times­tamped and stored. Con­cur­rently, car­ers kept a times­tamped log of each pa­tient’s be­hav­iour. The study gen­er­ated 87 hours of data, in which the re­searchers ob­served a to­tal of 548 ag­gres­sions – full melt­downs as well as iso­lated out­bursts. 2

2 The re­searchers chopped this data up into in­ter­vals, each 15 sec­onds in length. For each in­ter­val they cal­cu­lated sum­mary statis­tics – the av­er­age, min­i­mum, max­i­mum and the vari­abil­ity within the in­ter­val – for each of the biomark­ers that the wrist­band records. Fi­nally, they added a la­bel – es­sen­tially a yes/no in­di­ca­tor – from the car­ers’ notes, to show whether or not there was a melt­down episode in each in­ter­val.

Next, they looked for pat­terns in the bio­met­ric marker data that pre­saged the on­set of the ag­gres­sive episodes recorded by the la­bels. They did this by crunching the pro­cessed num­bers through a ma­chine learn­ing al­go­rithm – a piece of com­puter soft­ware that uses sta­tis­ti­cal meth­ods to ex­tract knowl­edge and in­sights from an ab­stract mass of data. In this case, it was try­ing to con­struct a model that could fore­cast whether or not a melt­down was go­ing to oc­cur in any given 15-sec­ond in­ter­val, based on the bio­met­ric read­ings in the pre­ced­ing in­ter­vals. In fact, they built a num­ber of dif­fer­ent mod­els – a per­son-de­pen­dent model for each pa­tient, us­ing just that in­di­vid­ual’s data, plus one ‘global’ model pool­ing to­gether the data from every­one.

“If we used three min­utes of past data, that gave us the high­est ac­cu­racy of pre­dic­tion one minute into the fu­ture,” says team mem­ber Prof Matthew Good­win, of North­east­ern Univer­sity. “For the global model, that ac­cu­racy was 71 per cent. Out of ev­ery 10 times that you would make a pre­dic­tion that an ag­gres­sion is go­ing to hap­pen in the next minute, then roughly seven times it will come to that.”

Of course, that also means that 3 times in 10 it’ll be a false alarm. How­ever, Good­win says that clin­i­cians – and, in­deed, most par­ents – would rather deal with these ‘false pos­i­tives’ than for gen­uine melt­downs to go un­de­tected. “They would rather have their at­ten­tion triaged and noth­ing oc­curs than have a false be­lief that they don’t have to at­tend and some­thing does.” He also be­lieves there’s room to im­prove the model’s ac­cu­racy, both in the way the data is pre-pro­cessed and by em­ploy­ing more so­phis­ti­cated ma­chine learn­ing al­go­rithms. And this work is on­go­ing. Al­ready, for the per­son-de­pen­dent model, the ac­cu­racy is higher, av­er­ag­ing around 84 per cent.

Good­win and his team are about to trial the tech­nol­ogy at a res­i­den­tial care fa­cil­ity for autis­tic peo­ple in the UK. The char­ity Autism To­gether operates the Raby Hall care home in the Wir­ral. From July they’ll be tri­alling bio­met­ric wrist­bands on a group of res­i­dents, the first time the tech­nol­ogy has been ap­plied in a res­i­den­tial care set­ting. But it won’t just be bio­met­ric sys­tems on test. Good­win’s team will also be in­stalling video and au­dio mon­i­tor­ing equip­ment, as well as de­vices to record light lev­els, am­bi­ent tem­per­a­ture, hu­mid­ity and baro­met­ric pres­sure.

The hope is that all this ex­tra data will permit not just pre­dic­tion of melt­downs, but also as­sist in un­der­stand­ing how an autis­tic per­son’s im­me­di­ate en­vi­ron­ment can ex­ac­er­bate their con­di­tion. And that could help ar­chi­tects de­sign new res­i­den­tial homes tai­lored to peo­ple on the autis­tic spec­trum, and to con­sider the needs of the autis­tic in­di­vid­ual when de­sign­ing other build­ings, such as shops, cinemas and restau­rants.

“The hope is that all this ex­tra data will as­sist in un­der­stand­ing how an autis­tic per­son’s im­me­di­ate en­vi­ron­ment can ex­ac­er­bate their con­di­tion”

Good­win imag­ines that, in the com­ing years, this tech­nol­ogy may com­bine with the In­ter­net of Things to en­able au­to­mated safe­guards in the care of those on the autism spec­trum. “Some­body starts to es­ca­late phys­i­o­log­i­cally,” says Good­win. “Might the lights in the room au­to­mat­i­cally dim down? Might some quiet mu­sic come up?” With so many peo­ple now re­ceiv­ing autism di­ag­noses, the de­mand for care is in­creas­ing – au­to­mated so­lu­tions such as this could help triage the at­ten­tion of hu­man car­ers more ef­fec­tively.


The tech­nol­ogy also rep­re­sents a po­ten­tial sea change in how we mea­sure some­one’s emo­tional state gen­er­ally. Other as­pects of health and de­vel­op­ment can be de­ter­mined quan­ti­ta­tively – by, for ex­am­ple, a brain scan, an X-ray or a ge­netic test. But as­sess­ing be­hav­iour, mood or over­all state of mind, is com­par­a­tively im­pre­cise. It might, for ex­am­ple, in­volve them fill­ing out a ques­tion­naire, the an­swers to which are nat­u­rally sub­jec­tive. And even if you could put pa­tients in some kind of lab­o­ra­tory-based ‘be­hav­iour scan­ner’, ob­tain­ing an ac­cu­rate pic­ture in such an un­nat­u­ral and po­ten­tially stress­ful en­vi­ron­ment seems un­likely.

Bio­met­rics, on the other hand, of­fers an un­ob­tru­sive win­dow to look in­side a pa­tient and quan­tify these neb­u­lous as­pects of hu­man health ob­jec­tively, in a way never be­fore pos­si­ble. And for peo­ple on the autism spec­trum – who are of­ten in­tel­lec­tu­ally im­paired, or who may lack the lan­guage skills nec­es­sary to ex­press how they’re feel­ing – the ben­e­fits could be even more pro­found. “These are the folks that we un­der­stand the least,” says Good­win. “These are the folks that we need to sup­port the most.” Dr Paul Par­sons is a sci­ence writer and data an­a­lyst based in Buck­ing­hamshire. He tweets from @NasaProPlus. His lat­est book, The Be­gin­ning And The End Of Ev­ery­thing (£16.99, Michael O’Mara Books), is out in Novem­ber.

Newspapers in English

Newspapers from UK

© PressReader. All rights reserved.