Can autism really be detected by voice alone?

The Lena System, a home device that records 16 hours of audio from a toddler's shirt pocket, appears to predict autism in children as young as 24 months with 91 percent accuracy.

(Credit: Lena Foundation)

The Lena Foundation, whose new autism-screening tool hit the market in September, claims that parents who use the Lena System are now able to determine with 91 percent accuracy whether their child is developing normally, has autism, or has unassociated language delays.

The home kit, which includes a digital audio recorder, an outfit to hold the recorder, and a questionnaire about the child's development thus far, costs $699. (The one-time language and autism screen, on the other hand, is $200.) The foundation, which develops technology for the screening of several types of language delays and disorders, says the kit works for children as young as 24 months.

"Roughly speaking, autistic children vocalize differently from other children," Dongxin Xu, manager of software and language engineering at the Lena Foundation, tells MIT's Technology Review.

Analyzing a child's vocal patterns to screen for autism isn't new. The three factors that seem to set the Lena System apart from traditional screening methods are portability (the recording device is small); amount of data (16-hour recordings); and the software Lena uses to analyze the recordings parents mail in dutifully each month.

According to Jeffrey Richards, a statistician and database technician for the Lena Foundation, the software first categorizes the 16-hour audio stream into sound types, such as child, parent, or television. The child clips are then further dissected, and analyzed for the phonological composition of each sound, as well as how it is clustered and paired. The resulting data is then compared with the data compiled on children who are considered normal, autistic, or delayed.

The 91 percent accuracy is high, and while Lena researchers continue to fine-tune their software to push that rate even higher, I remain somewhat skeptical that voice alone can determine whether a child is autistic. It is often suggested that Einstein didn't speak until he was at least 3, if not 4 or 5; I have to wonder how a 16-hour recording of Einstein at 24 months would be interpreted by Lena software.

[danielwsmithee]

You need to read up on Autism. <br /><br />"as many as 30% of autistic people have some sort of savant-like capability in areas such as calculation or music. " http://www.economist.com/sciencetechnology/displayStory.cfm?story_id=13489714<br /><br />It is quite possible that Einstein was himself Autistic.

[LancerEQ]

Researchers speculate that he did, in fact, have a type of autism:<br /><br />http://news.bbc.co.uk/2/hi/health/2988647.stm

[viper396]

...and the point that should be made by both of your comments is that autism is not always a handicap nor will autism by any definition prevent a person from having a meaningful and successful life.

[mike_ekim]

As an engineer I work with many people who I (and others) suspect are on the autism spectrum. Possibly I am on the spectrum as well, but I do just fine in life so I can't be bothered to be 'diagnosed'. I didn't take your comment personally - many people with asperger's wouldn't, after all. :) However, I would like to point out that the last paragraph of the article suggests that Einstein could not have been autistic because of his renouned intellect. I have a nephew who has been brillliant at math since he was 4, learned to speak late, and will likely go through life never combing his hair. Yes, he was diagnosed with autism, and one day he may very well develop practical cold fusion. <br /> <br />One lecturer on Autism once joked, "When I give a talk at a college I look around the audience and try to spot the people who are autistic. When I give a speach at an engineering college, I try to spot the people who aren't."

[mike_ekim]

My apologies, my previous comment was in reference to the article, not another commenter.

[ballmerisanape]

First you have to come up with a real definition of Autism.

[elinormills]

This is fascinating!

[dascha1]

Yes it is. I also think "sensory integration" is as well.

[rjmaldon]

Not sure that a misdiagnosis at such an early age would hurt the todler. Or, better stated, this should be used as a screening tool, not a treatment one.

[galaxyblur]

Autism is not a "thing", it's a spectrum classification based on certain behaviors and traits. We are all on that spectrum somewhere. That's why we'll never find a singular cause of autism -- there are likely many causes of these behaviors.<br /><br />danielwsmithee was right, Einstein could've been diagnosed if he was born today... and we may have hindered him for it. ballmerisanape is right also... you can't really define autism as anything, but rather group people together as having similar traits.<br /><br />An autistic person may have a vocal pattern that is similar to other people with autism. But there is certainly an element of over-diagnosis these days. The most important thing one eventually realizes is that there is nothing wrong with being autistic; it may not fit our society or lifestyle, but everyone is different.

[mbenedict]

I think you're mixing up Autism with Autism Spectrum Disorders (ASD). Autism is a "thing" in the sense that it's a very specific disorder within the spectrum of ASD. Two other ASD disorders are Asperger's Syndrome and PDD-NOS.<br /><br />It's incorrect to say that we "are all on [the autism spectrum] somewhere". The three ASD disorders are narrowly defined within the class of Pervasive Developmental Disorders (PDD). Each have specific medical diagnosis. The vast majority of people in the world do not have PDD, and therefore cannot be "in the spectrum" somewhere.<br /><br />If Einstein had PDD at all, he might be diagnosed with Asperger's Syndrome. He would NOT be diagnosed with Autism by modern standards.

[geeksyndrom]

mbenedict:<br />Please read more thoroughly the DSM-IV or any serious litterature on the subject. Or browse the screening tests developped by Simon Baron-Cohen at Cambridge. It's hard to see any "narrowly defined" concepts there.<br /><br />As the father of a teenager recently diagnosed with Asperger, I can testify that diagnosing this disorder is certainly not "hard fact" science. Furthermore, traits related to autism, like repetitive movements and eye contact avoidance, tend to disappear when Asperger or high-functioning autists without retardation grow up.<br /><br />As a matter of fact, most of the traits that lead to my son's diagnosis are shared by most men in my family, over 3 generations. Nobody has a diagnosis. Most didn't need it. You need a diagnosis when you need help. This is certainly the case of profound autists - a tragedy for the kid and the parents, but I do believe that most Aspergers can adapt to the society and live happy without much help.<br /><br />About Einstein, unfortunately we can't administer him any test. But from what I know of his bio, it's very plausible that he could have been diagnosed with some disorder on the autism spectrum. Being slow to learn language, though, is more typical of high-functioning autism (a la Temple Grandin) than Asperger. Some Asperger kids learn to speak very early (my son). Unfortunately the off button could be hard to find.<br /><br />About the article and the screening or diagnosis kit, I'd like to emphasize these points:<br /><br />(1) A voice with a lack of inflections can characterize other disorders, like schizophrenia and some mood disorders. I don't know if early onsets of these disorders include voice characteristics, though.<br />(2) A false positive or a diagnosis for a mild condition can be worse than no diagnosis at all. It can induce distress for the relatives and the kid himself where there wasn't.<br />(3) Throwing a statistic like 91% (why not 90,5%?) for such ill-defined disorders shows some lack of seriousness.<br /><br />So I remain skeptical.

[heathertml]

Why would you end this article so ....stupidly? <br /><br />91% accuracy, but you're skeptical? It's 91% percent flippin' accuracy already. <br /><br />The Einstein mind is one out of many thousands, and might well be missed. Who cares...91% chance of correct early dx of autism is amazing.

[ddesy]

You take the 91% accuracy as being a definite fact, and that's a problem. Don't you want proof?

[babyfacemagee]

Aspergers is a high functioning form of autism or at least on the autistic spectrum and such notables as Steven Spielberg and Mozart are thought to have had Aspergers. I agree with the others that it is a collection of traits that vary in each case...no two are exactly the same. Some have affected speech, others do not. Some have difficulty looking someone in the eye and interpreting facial expressions and body language and others less so. Some are very non-talkative and some are very verbose. It's really a set of genetic traits that seems to be in some famlies and affects mostly boys, but each differently.

[fear-teagaisg]

It certainly is autism silly-season. The only reason that Einstein (or Spielberg or Mozart) might be diagnosed as having autism nowadays is that the term "autism" no longer means anything in popular usage and is rapid losing its specificity as more professionals imagine themselves to be autism experts. The "spectrum" is now so broad that anyone can have autism, including the growing number of people who clearly do not have any real social, cognitive, or language impairments, but have books to sell, speaking tours to do, or blogs to write. As for Einstein, Spielberg, and Mozart, the actual facts of their lives are in no way consistent with a bona fide diagnosis of autism. A few years ago it seemed that everyone who was creative and productive had a touch of bipolar disorder; before that it was dyslexia; before that it was characteristics of schizophrenia. It is sometimes possible for high achievers to not have any diagnosable condition at all. <br /><br />More of the autism silly season is evidenced by the notion that autism can be detected by a voice analysis. There are no accepted measures of validity of this technique or anything like it. And, the 91% accuracy claim means nothing. It is a puzzle how such a number could be arrived at. What is the standard? The really important statistic, the false positive rate, is missing--although I'm sure there's a proprietary unpublished study done by the developers. If we accept that autism actually occurs in 1 in 150 kids, then a false positive rate of even just 2% means that more kids will be labeled with autism with than actually have it. That is, out of every 300 kids, 2 will have autism and 6 will be false positives. Even a tiny 0.5% false-positive rate means that over one-third of the sample are false positives. It would be a statistical miracle for a test such as this, even if it worked, to have a false positive rate in the 0.5-2% range. Thus, we're actually talking about a test that would label many times more kids with autism than would actually have it. <br /><br />Who sells something like this without testing its effectiveness? If it wasn't autism, where anything goes, doing such a thing would not be tolerated.

[ps_martin]

This screen isn't a diagnosis. It is simply a diagnostic tool. If you check their website, they clearly state, "If a problem is detected, we recommend that you take the reports and information to your pediatrician or speech-language pathologist for more immediate intervention and therapy."<br /><br />The people most likely to use this are people who already have concerns about their child's development. Perhaps there is a lack of 'typical' social interaction or maybe the child's prosody is 'off' or any number of other red flags of developmental issues or delay. The point is, this test is simply a diagnostic tool to help figure out what special assistance the child need, if any.<br /><br />As for your concerns about the validity of their screen or the results, they provide PDFs that seem to address them: http://www.lenababy.com/DownloadFile.aspx/pdf/Automated_Detection_of_ASD &#38; http://www.lenababy.com/DownloadFile.aspx/pdf/Accuracy_For_ASD_Screens_vs_LAS .

[dawgosaur]

Having an autistic (actually Asperger's) son, and having interacted with others like him at his schools, they do speak differently, generally lacking any inflection in the voice. There is a lack of excitement in the voice, almost robotic. I have meet kids who are totally strangers, and correctly guessed they were autistic by the voice. My wife has noticed this as well.

[mike_ekim]

Agreed, I have heard this in people including my nephew. I'm not saying that all people with autism sound the same, but when someone 'talks like an autistic person' it is unmistakeable.

[miamoe]

As an employee of the not-for-profit LENA Foundation, I wanted to clarify a couple of misunderstandings. First the Language and Autism Screen is available for $200. The LENA Home System which monitors the effectiveness of intervention therapies and is used for several years is $699. Accuracy of the autism screen is determined using the same methods as other professional and widely used tools such as the M-Chat and ASQ. 91% accuracy refers to the sensitivity at which the algorithmic model is set - in other words, the children in sample with a diagnosis of ASD ? the percentage that the screen identified as autistic. Sensitivity can be increased, however, you get a decrease in specificity - meaning, you increase your false positives. Arguably, this may be considered a good thing - as catching more children with autism and accepting some false positives is an overall good, versus missing children with autism and not getting them into treatment at an early enough age. Please view both our websites: www.lenababy.com or www.lenafoundation.org for more information on our research and the list of universities and children's hospitals currently using our technology.

[SergeM256]

Sounds like totally useless invention. There is no treatment for autism and similar disorders and there is no benefit of early diagnosing of this disorder. If your child has autism - you will know this sooner or later without spending $700 on software and equipment.

[ps_martin]

This is, most likely, the most ignorant post I'll read on the Internet today and that is saying a lot.

[dawgosaur]

Apparently you know nothing about autism, or how early intervention can help. My son is as success as he is because he was diagnosed at 3 years old. A specialized education started at a very young age (3 yrs) has helped him become a young adult, with a chance of being a productive member of society. With your way of thinking, we should just let them fall to the wayside and institutionalize them when they get in the way. Let's not be proactive at all.

[mike_ekim]

I agree with the other responders. Serge, if you're trying to validate your own inaction, or the negligence of someone you know who failed to help an autistic person, you're doing it with the wrong crowd.

[fear-teagaisg]

Miamoe tell us here are some "misunderstandings" about the Lena Foundation System--which seem to be actual "understandings"--then seems to offer the astounding suggestion that a test that would seem to label several times more kids with autism than actually have it is a good thing. Getting lots of false positives is helpful, especially if one of the purposes of the system might be to drum up attention and business for Lena's parent organization, Renaissance Learning Inc., which is the internet tells us is owned by the same people who created the non-profit Lena Foundation. It is not so good for those children who are labelled and directed into inappropriate or unneeded services, which would seem to be most of the kids caught by the Lena autism assessment system. <br /><br />But never mind that. We all know that it has become perfectly OK to do anything whatsoever to families affected by autism--especially if it involves separating desperate parents from their money for "revolutionary" new methods that turn out to be ineffective but expensive larks at best, like swimming with dolphins, or dangerous shams like facilitated communication and holding therapy. The market is now flooded with assessments as well, most of which only seems to work because they are selectively applied to kids who are already exhibiting lots of odd behavior. Are the people sold this expensive system being told that the results would not be accepted by a competent therapist or treatment center, and that they would have to go and pay even more money for a well-researched, standardized autism assessment? We also know that if the condition is autism, it is perfectly OK to rush a treatment or assessment onto the market before it has any general acceptance in the field, and before it has been subjected to real, independent testing. It was not long ago that such behavior was not considered professional or ethical. Nowadays, as we see here, it is standard practice.

[Michichael]

Hey, think of how the world would be a better place of us Autistic folks ran it. &gt;:)

[CharlieGill3]

Interesting article right up to the "I remain somewhat skeptical ...[as] It is often suggested..."<br /><br />I might suggest an evening course in journalism.

[hayesjohn]

Personally, I find it all to hard to believe that it is possible to monitor the effectiveness of how well an intervention is working with the system. I just don't believe that testing ( which starts at 24 months ) will be able to tell the difference month by month . <br />Doesn't that imply the ability to tell the age of a child by voice? Where are the standards that reflect month by month development of communication skills ? Is it assumed to be straight line growth for every child? Don't children do everything in spurts ? Good months the intervention is working,bad months it isn't ?<br /><br />Sounds more like magic wrapped up in technology .