New technologies may lead to improved diagnosis and treatment of disorders such as autism

Feb 28 2005

Researchers at OHSU's OGI School of Science & Engineering have launched a unique, new research project that will use advanced speech and language technologies to better understand childhood communication disorders.

These new technologies may lead to improved diagnosis and treatment of disorders such as autism in which communication problems are central. The research is being funded through a $2.9.million, 5-year grant provided by the National Institutes of Health.

"This newly-funded research will allow us to do something that has not been done before," explained Jan van Santen, Ph.D., Director of the Center for Spoken Language Understanding (CSLU) at OGI and Chair of the Department of Computer Science and Electrical Engineering, who leads the project. We will apply and adapt technologies created for non-medical purposes to further our understanding of autism and other neurodevelopmental disorders.

Lois Black, Ph.D., a Research Scientist at CSLU, and van Santen, conceived of a series of studies to analyze prosody and communication in autism.

"Prosody, which involves the melody, timing, and intonation of speech, refers to the 'how it is said,' not to the 'what is said,' of language," explained Black. "We use prosody to convey meaning, intent, and emotions. The ways in which we emphasize or express what we say is incredibly important in conveying the correct meaning of our statements. For example, the word 'really' can convey surprise, shock, skepticism and a whole host of other meanings depending on the way the word is expressed. Autism has to do, in part, with the inability to understand and connect with other minds. There is some evidence suggesting that difficulties with prosody may be strongly involved in this and, perhaps, a critical feature of autism."

"The problem," van Santen added, "is that research on prosody has been hampered by not being able to objectively measure it, and that is what is so unique and important about the new speech technologies that will be used."

The research will look at the ability to express and to understand prosody in a range of contexts, such as conversations, listening to and telling stories, or simply deciding whether a computer generated voice sounds angry or sad.

In one pilot study with preliminary findings, van Santen and Black had children with autism and a control group of typically developing children answer questions about the factual and social-emotional contents of stories. The stories were presented with either neutral prosody or with very enhanced prosody. It was found that both groups understood social-emotional information better with very enhanced prosody, even though children with autism did not reach quite the same performance levels as their typically developing peers.

"Children with autism are usually able to remember simple facts. For example, they can recall names, dates, and numbers," said Black, but they usually do poorly processing and remembering social information, such as motives for action and feelings. This preliminary finding using enhanced prosody has important treatment implications for children with autism, given their weakness in social cognition."

In another part of the study, van Santen and Black will use new technologies to look at what is called "prosodic attunement" or "reciprosody" during conversations, and investigate its relation to a child's social and communicative competence. A challenge in research on autism has been to identify who the children with autism are. Autism diagnoses are on the increase. There is some controversy around whether this escalation reflects an increase in incidence due to an interaction between genetic vulnerability and environmental triggers, an expanded awareness of the disorder, or more liberal application of diagnostic criteria. Despite this ongoing debate, there is a broader and more heterogeneous group of children being diagnosed with an autism spectrum disorder than ever before. The researchers believe a sharper understanding of the core issues and differences between children with different forms of communication disorders is needed along with a more precise understanding of the children who make up the autism spectrum group. The scientists hope to address this in their research by comparing children with autism to those with developmental language disorders and by giving an extensive battery of neuropsychological measures, so that they know exactly who the children are.

Besides Black, who is a child neuropsychologist and expert in neurodevelopmental disorders, and van Santen, who is an expert in speech synthesis and signal processing, the project also includes Peter Heeman, Ph.D., an expert in speech recognition and human-computer dialogue.

To assess prosodic abilities, the project will use speech technologies initially developed at CSLU for speech synthesis, signal processing, speech recognition, and human-computer dialogue. These technologies will be used to measure prosody in the children's speech, or to create speech stimuli to measure the children's abilities to understand prosodic cues.

"We are only now becoming aware of the potential of these technologies for studying communication disorders," said van Santen. "They were never developed for this purpose, but it turns out that we do not have to reinvent the wheel. Here is how these techniques make a critical difference. In the past, researchers would have to learn how to listen to prosody, and different listeners would never agree. But now, we have the quantitative models that allow us to make statements such as: 'This child's prosody is only 25 percent connected with meaning.'"

The researchers can also now manipulate speech with signal processing methods to very precisely control the speech stimuli used for measuring the ability to understand prosody. For example, in the story experiment, scientists were able to exactly match two versions of the story in terms of speaking rate, loudness, and pause durations. This is critical for methodologically clean experiments. It also allows the scientists to create diagnostically important stimuli that simply cannot be produced in any other way, such as videos of different emotions where the audio portions are swapped. These videos will be used to analyze the ability to infer emotion from the combination of facial and vocal prosody.

Towards the end of the research project a treatment program that uses computer games to improve communication and social skills will be set up.

"Obtaining an improved understanding of the communication difficulties associated with autism is very important," explained Robert Nickel, M.D., a professor of pediatrics at the OHSU School of Medicine and practitioner at the OHSU Child Development and Rehabilitation Center (CDRC). "OGI's use of computer technologies will hopefully result in new strategies to address autism-related communication problems and to improve patients' quality of life We are very enthusiastic about this project being awarded to OHSU, and view this as a wonderful opportunity for OHSU to become an important player in research on autism and communication disorders. CDRC will be working with CSLU to do whatever it can to help recruit participants for the studies."

The study will recruit children between the ages of 3 and 8 who are verbally fluent (at least 3 to 4 word sentences) and of "normal intelligence" (IQ over 70). In all, about 300 to 400 children will participate, including children with an autism spectrum disorder, developmental language disorder, and typically developing children. Half of the study's subjects will be seen at OHSU's Beaverton campus at the OGI School of Science and Engineering. The other half will be seen at the Yale Child Study Center, which is an internationally renowned center of excellence in autism.