LENA Language and Autism Screen Q & A

Any parent of a child age 24-48 months should use the screen service, regardless of whether they are concerned about their child’s development. The American Academy of Pediatrics (AAP) recommends that children be screened for autism twice before the age of 2, but only 5-7 percent of pediatricians formally screen for autism spectrum disorders (ASDs). Parents who raise concerns are typically told to “wait and see.” The average age of diagnosis is 5.7 years, but research has shown that children who receive intervention before the age of four have better prognoses. The current “wait and see” system is not working—the foundation is trying to take action to empower parents and professionals to decrease the average age of diagnosis; 5.7 years is unacceptable when we know that treatment needs to start earlier.

The language and autism screen is an automatic screening service that uses acoustic information in the speech signal of the child to determine whether the child is at risk for autism. We use speech recognition technology to categorize the phones or sounds produced by the child, and then apply a statistical cluster-based method to look at the distribution, or pattern, of those sounds. Based on comparisons to the sound pattern of children with autism, we are able to determine with 89 percent accuracy whether a child is at risk.

The language and autism screen was developed on a sample of children age 24-48 months, including both typically developing children and children diagnosed with autism, as well as children with language delays not including autism. Our speech recognition engineers use acoustic modeling techniques to detect the voice of the child wearing the system. They use a unique combination of a phone-based and cluster-based approach to automatically analyze the speech of the child and determine whether the child is at risk for autism.

The language and autism screen is different from traditional screening methods because it is based on data obtained from the child’s natural environment over the course of an entire day. Traditional screening instruments typically involve a parent questionnaire or an observational session where the child is in an unfamiliar environment with a stranger. Until this technology was developed professionals did not have objective information about the child’s behavior in the natural home environment—data obtained in a completely unobtrusive manner with no demands made of the child.

The language and autism screen is remarkably accurate given that it is a completely objective and automatic measure. We are continuing to conduct research on the screen at the foundation. In our sample of 190 children, the system was quite robust, with accuracy rates around 89 percent. In other words, it was able to successfully identify 89 percent of the children diagnosed with autism based on the distribution of sounds produced by the child on the recording day.

There is a crucial and unfortunate disparity in the situation for early autism screening: The average age of diagnosis is 5.7 years, but research has shown that intervention started before the age of four is more beneficial than intervention at older ages. We need to identify these children as early as possible in order to improve their developmental trajectory. The system was developed to help parents become aware when there is cause for concern. Similar to any screening system, it is considered a first step. If a child is identified as high risk, the parents are advised to take the information to a pediatrician and schedule a comprehensive evaluation.

Research has shown that children with autism vocalize differently from typically developing children. However, this behavior is usually not included as part of a comprehensive evaluation for autism because it requires extensive professional training and the time-consuming transcription of audio. However, our experts have taken advantage of advancements in speech recognition technology to produce algorithms that can analyze an audio recording of a child and automatically detect the acoustic anomalies specific to children with autism.

It is designed for children age 24-48 months. A parent or speech-language professional can order it through our website at www.lenababy.com. We send the parent a LENA Digital Language Processor (DLP) that is worn in the front pocket of the child’s clothing; the DLP records continuously for up to 16 hours. The parent sends the recording back to the foundation, where it is processed. The sounds produced by the child are identified, and further analyses are conducted to determine the probability that the child is at-risk for autism. Finally, we mail the parent or professional a confidential report on the probability that the child is at-risk, as well as information about his or her language skills and the quality of his or her language environment.

The LENA System software processes the audio recording into segments from several seconds to several minutes in duration, assigning a sound category (e.g., key child vocalizations, adult male speech, TV/electronic sound, or silence) to each segment based on previously developed acoustic models. Key child vocalization segments are further processed to determine the probability that the child’s vocal output is consistent with a pre-defined classification model for ASD.

It will provide you with a confidential report that includes an at-risk probability rating. We also provide percentile information about the child’s language environment in terms of adult word count and conversational turns, as well as a developmental age regarding the child’s language skills.

A parent using the autism screen can view it as a first step in investigating at-risk behaviors for autism. It will provide them with a private and confidential analysis of their child’s vocal activity and refer them to a professional if there is cause for concern. This system was not intended and should not be used as a replacement for a professional evaluation. It was produced to give parents immediate access to crucial information about at-risk behavior so they can take the steps necessary to make the right decisions for their child.

The screen was tested on a sample of 190 children age 24-48 months. There were 75 children with ASD in the sample, which is large compared to the sample size used to test other screening systems. There were 81 typically developing children in the sample, as well as 34 children with language delays not including autism.

Autism is typically associated with three broad categories: 1) language delay, 2) impaired social behavior (e.g., lack of eye contact) and 3) restrictive and repetitive behaviors.

They can differ in several dimensions. For example, children with autism are sometimes atypical with respect to the pitch quality and rhythmicity of speech, or vocalization frequency and duration.

The screening algorithms are sensitive to all of the sounds that come out of the child’s mouth, including grunts, squeals, babbles and words. The screening algorithms have been shown to work reliably and accurately with children who are not yet talking. However, if the child is completely mute there will be no data to work with. If a child is completely mute (i.e., not babbling or producing any sounds) at any age, then he or she should receive a comprehensive evaluation from a professional.

This is a great question and a concern we had during development—we needed to make sure that we were distinguishing children with autism from children with language delays. The autism screen component does not detect language delay, it detects subtle anomalies in the vocal production of the child. The power of our processing algorithm can perform statistical analyses beyond what can be deciphered by the untrained human ear. For example, autism is typically related to a number of different behaviors. One of them is repetition. Not all children with autism engage in repetition, but many of them will repeat the same thing over and over or will produce sentences that are markedly different in terms of prosody. With enough audio data, the screening algorithm can use this statistical information, in conjunction with several other features, to determine a child’s at risk probability.

It is important to note that our test sample includes children with language delays who passed the screen. The screening system is specific to autism spectrum disorders.

If a parent has strong evidence that their child is autistic, then they should get the child into treatment as soon as possible. This is the foundation’s mission and the purpose of the screening service. Many parents have suspicions and are told by the pediatrician to “wait and see.” If parents have cause for concern but cannot get the referral, then the $250 spent on the screen will give them validating information in a private and confidential manner. We are a not-for-profit organization so the price of the screen covers our costs for transportation and labor. Our mission is to help children and empower parents.

We are empowering parents with information. The current situation for early autism screening in this country is unacceptable, and we are taking action to offer a solution. The average age of diagnosis for autism is 5.7 years. However, every autism researcher and clinician in the country knows that children who receive intervention before the age of four have better developmental trajectories. There is a problem with this system, and we are working with parents to try to fix it. Our mission is to empower parents with information that will allow them to provide their child with maximum opportunities to reach their full developmental potential.

The language and autism screen is not a treatment, it is a screening service designed to provide parents with information about at-risk behavior. Parents and professionals can use the LENA technology to monitor their child’s progress and response to intervention.

There are a number of different therapy types available, many of which focus on language input and behavior modification. We encourage parents to visit the Autism Society of America (ASA) and Autism Speaks websites to learn about treatment types so they can make an informed decision about the best options available.

If a child is identified by the screen as being at-risk for autism, the parent is instructed to take the information to a pediatrician and obtain a referral for a comprehensive evaluation. However, the typical wait for a professional evaluation is 3-6 months, or up to a year in rural areas. Since we know that time is of the essence with respect to early identification, we encourage parents to take action by seeking out treatment programs as early as possible.

The language and autism screen is not considered a medical device so it does not fall under the purview of the FDA. It is simply a measurement instrument; digital recorders do not need FDA approval.

The language and autism screen is not currently available for purchase outside of the United States and Canada. The screening algorithms were developed and tested with monolingual American English-speaking families in the United States. The screen has not been tested with accented English or with other languages. The foundation would like to collaborate with researchers worldwide to test the system cross linguistically; we encourage interested researchers to contact us.