Using narrative language samples to identify DLD

The ATLAS Project logo

The ATLAS Project

About the project

Following the successful 2018 NIHR i4i Connect project, the team built a proof of concept of software to help evaluate children at risk of DLD (developmental language disorder). Now with a 2 year grant from NIHR, Therapy Box will work with North Bristol NHS Trust, University of Newcastle, and NHS sites in Hackney, Newcastle and Bristol. The software will help speech and language therapists to record a language sample, transcribe it and analyse it - saving up to 100 minutes per child. The goal is to allow more children with DLD to be identified more quickly. DLD affects 8% of the population and there are 800,000 children living with DLD in the UK.

NHS Funded Research

ATLAS is funded by the research arm of the NHS.

The project has been successfully funded for an additional 2 year project. This enables the team to build the product, evaluate it and bring it to market. An exciting aspect is a national citizen science project where we will gather language samples from 600 children. In 2018 the project’s proof of concept was funded by NIHR. It won a number of awards including the Virgin Media VOOM Award and was listed in the HealthTech Innovation of the Year category in the Digital Leaders 100.

Project Objective

Using machine learning to screen children for developmental language disorder

Approximately half of all children who have developmental language disorder (DLD) go undetected. This is partly because the NHS does have enough resources to screen all children for signs of DLD. The aim of the ATLAS project is to create an app that can screen children using machine learning. This will dramatically reduce the cost of screening and ultimately make it possible for all children to be screened.

Project Outcome

Automated analysis of language

Using speech recognition, transcription tools and language analysis, ATLAS will help clinicians to assess children more rapidly. We are hoping in 2020 to have an evaluation across three clinical sites, to determine how parents, children and clinicians find the app; how much time it saves and how it aligns with the gold standard assessment and the clinical decision making of a speech and language therapist.

Saving Time

ATLAS will save NHS speech and language therapists up to 100 minutes per patient.

The goal of ATLAS is ultimately save the NHS time and money - which will allow speech and language therapists to see more children and reduce waiting lists. The detailed analysis will also help with planning therapy more effectively.

Machine Learning

Using machine learning to speed up assessment of developmental language disorder

Two children in every classroom have DLD, and so it is essential that speech and language therapists can rapidly assess children to identify who needs intervention; tailor their plans for therapy and keep outcome measures. ATLAS will dramatically reduce the time needed and also provide a detailed analysis of language features, to help clinicians to identify areas of language competency and arease a child may need more help.

Standard Measures

As well as machine learning, ATLAS also includes industry standard measures

The ATLAS framework will offer both fixed analysis using industry standard measures as well as more in depth analysis of the language structure and content. In parallel the machine learning framework will run to either consolidate or supersede the outcome of the standard measures.