Detecting language problems: a tech-assisted vision for earlier diagnosis

Charlotte Edwards 21 February 2019 (Last Updated February 20th, 2019 15:43)

A team at San Diego State University is using advanced computerised comprehension tasks to detect language problems in children. Using the program, the team has been able to spot problems a full two years earlier, a development which could lead to significant improvements for conditions that often hamper a child’s progress. San Diego State University psychology professor Margaret Friend explains the potentially life-changing technique.

Detecting language problems: a tech-assisted vision for earlier diagnosis
Children with undiagnosed language problems are more likely to find school challenging, and ultimately graduate and find employment at lower rates than children without these issues.

Studies in the UK have suggested that as many as one in ten children struggle with speech and language difficulties. Early speech and language problems are associated with reading and writing issues later on, as well as difficulties socialising and learning in general.

Children with undiagnosed language problems are more likely to find school challenging, and ultimately graduate and find employment at lower rates than children without these issues. Early diagnosis is beneficial so that a child’s development is affected as little as possible.

Researchers at San Diego State University in the US believe that most vocabulary assessments aimed at very young children are not designed to focus specifically on words that are the most important for predicting future language development, which makes early intervention difficult.

The researchers are taking part in an international, multi-site project, which aims to decode what babies comprehend before they start talking, and improve the tools to assess this early understanding. They have subsequently created a Computerized Comprehension Task (CCT) that could potentially indicate language problems two years earlier than current techniques.

San Diego State University psychology professor Margaret Friend explains more about this life-changing technology, how it was created and how it could potentially be adapted to help with other conditions such as stroke or cerebral palsy.

Charlotte Edwards: Can you explain how the program and the advanced CCT works?

Margaret Friend: The CCT is a forced-choice measure of vocabulary comprehension administered on a touchscreen designed for children between 16 and 24 months of age. There are 41 trials, one test word per trial, on which two pictures appear on the touch-sensitive screen. On each trial, a tester prompts a child to touch the picture that corresponds to a word. For example: “Where’s the car? Touch car.”

Children respond by touching one of the pictures on the screen, either a car or another picture, for example, an airplane. The task proceeds in this way until all 41 words have been tested. The number of children’s touches to correct pictures is taken as an estimate of their word comprehension relative to their peers of the same age.

CE: Why do you think it is more efficient than previous methods?

MF: The CCT is more efficient than previous methods in three ways. First, the procedure is easy to administer and takes only about ten minutes. Second, the procedure is easy to score. In about 15 minutes total, the CCT can be administered and scored to yield an estimate of early word comprehension. This contrasts with parent report measures that take longer to administer, up to 45 minutes, and looking time measures that are time-consuming to score.

Moreover, the CCT can be administered on a tablet, making it highly portable and ideal for assessing children in a variety of settings including integrating assessment into regular paediatric office visits.  More importantly, however, preliminary findings show that the CCT, in contrast to other measures, specifically taps into those words that children know really well. The number of words that children know really well, as opposed to superficially, is important to assess the maturity of their early language skill.

CE: What impact could your findings have on the lives of children?

MF: At two years of age, about 16% of children show a delay in spoken language. Most of these children improve with time. However, of those children who are also delayed in word comprehension, it is estimated that 75% will go on to have a persistent language difficulty. The CCT assesses word comprehension. Our recent research indicates that the CCT is the only measure that, as early as age two, predicts which children will exhibit language delay at age three. Other measures do not achieve comparable prediction until age four and epidemiological evidence indicates that approximately 15% of school-aged children are in need of services for language difficulties.

Furthermore, it is known that children who remain unidentified at school entry exhibit a number of difficulties including in language skills, academic performance, and behaviour problems. Imagine the frustration of a child who has difficulty expressing himself to teachers and classmates.

Ours is the first approach to show strong prospective sensitivity as early as two years of age with implications for the early assessment of developmental risk. The improvement in early identification could have a large impact on the healthy development of children. Early identification means that children could receive intervention a full year earlier. This maximises their chances of healthy outcomes in language and in school achievement.

CE: Could the program be adapted to help with any other conditions?

MF: In principle, the assessment could be adapted to any situation in which one wants to assess vocabulary in individuals who cannot respond verbally. For example, it might be possible to design an assessment to administer to stroke patients to determine the extent of their preserved vocabulary. Researchers in Amsterdam have used a comparable approach to assess word knowledge in children with cerebral palsy.

This allows children a means to express themselves that they did not have previously and provides a much more valid assessment of their development. It is important to note, given this example, that some individuals may be unable to execute a touch to the screen. In these cases, the program can be outfitted with a ‘jelly’: a kind of mouse that can be activated by other bodily movements.

CE: What’s next for the project and the researchers?

MF: The CCT has been developed and validated in English, Spanish and French. We are in the process of extending it to Danish, as well as to indigenous languages. We are also collecting data to evaluate prediction from early vocabulary to school achievement. A next step will be to complete automated scoring of the assessment and to provide percentiles for easy clinical application.