Using AI to transform speech therapy

Researchers from the Centre for Language and Communication Science Research at City St George’s, University of London have launched a new project that aims to use machine learning to improve clinical practices in speech and language therapy for people with aphasia, a condition that affects communication after a stroke.

Time-consuming manual processes

In particular, the ‘Machine Analysis and Reporting of Spoken personal stories with aphasia’ (MARS) project, supported by £470,000 in funding from Barts Charity, aims to address a significant barrier in NHS speech and language therapy: the time-consuming manual analysis of speech, which was highlighted in a previous paper by the team. Over the next three years, the MARS project team will work in partnership with the technology company Therapy Box to improve how clinicians, patients, and researchers assess and analyse spoken discourse in individuals with aphasia.

Aphasia, a communication disorder that affects a person’s ability to process and produce language, is a long-term condition that can occur following a stroke. Approximately one third of stroke survivors each year in the UK develop aphasia as a consequence of their stroke. World Stroke Day is observed on the 29 October, and the day aims to underscore the serious nature and high rates of stroke, raise awareness of the prevention and treatment of the condition, and ensure better care and support for survivors.

Currently, speech and language therapists (SLTs) rely on manual transcription and analysis of patients’ speech, a process that is time-consuming, resulting in many clinicians abandoning therapeutic work in this area. However, the MARS project – which involves Professor Lucy Dipper, Professor Madeline Cruice, Dr Niamh Devane and Dr Rebecca Moss at City St George’s – seeks to harness artificial intelligence (AI) to streamline these processes.

Developing AI tools

Working with Therapy Box, led by Rebecca Bright MBE and Swapnil Gadgil, the project will develop AI tools to transcribe, analyse, and report on speech from patients with aphasia. Therapy Box has already created innovative speech and language therapy tools, and its collaboration with City St George’s researchers will tailor these technologies specifically for patients with diverse speech patterns due to aphasia and related communication disorders.

The MARS project was inspired by a previous initiative, LUNA, funded by the Stroke Association, which highlighted the need for more accessible discourse analysis tools for clinicians.

Speaking about the MARS project, Professor Dipper said: “SLTs want to assess and treat everyday talking in their patients, but due to time constraints, current clinical practices are often limited to words and sentences. By automating speech transcription and analysis, MARS could enable therapists to spend more time working on discourse, and this could improve patient outcomes.”

Participatory approach to research

A major component of the MARS project is its participatory research approach, involving NHS SLTs, people with aphasia, and their families in the software’s development. The team first took this approach when preparing for the research grant, and they have since continued to work collaboratively in this way for the MARS project. This collaborative method ensures that the resulting AI tools will not only meet clinical standards for accuracy and reliability but will also be user-friendly for both clinicians and patients. The project will also investigate how AI can handle the challenges of diverse speech patterns, including accents, speech disorders such as dysarthria, and non-native English speakers.

The City St George’s team will soon begin working with Therapy Box on the software’s development. This collaboration aims to deliver fast and accurate analysis, allowing clinicians to personalise and improve treatment plans more effectively. The project is expected to benefit not only clinicians and researchers but also stroke survivors, who will receive more frequent and detailed assessments of their communication abilities.

Professor Madeline Cruice added: “With MARS, we are not only addressing a gap in speech therapy, but we also hope to empower patients and therapists by making discourse analysis more accessible, efficient, and inclusive. This could significantly improve patient care and support more personalised, effective treatments.”

Follow the MARS project on X (formerly Twitter) @MarsAphasia