Aphasia researchers and computer scientists from universities, hospitals, and research centers in Montréal and Paris have collaborated on a scoping review of the uses of artificial intelligence in service of aphasia rehabilitation during the years 1990 to 2023. Their study objectives were – firstly – to describe how AI is being used to advance the field of aphasia rehabilitation broadly, and – secondly – to determine if and how AI is being incorporated into augmentative and alternative communication devices or applications for the benefit of persons with aphasia.
The researchers first formulated their research question: “In what way is artificial intelligence (AI) used in language rehabilitation for people with aphasia (PWA)?” They subsequently conducted a search of six key databases using the search terms ‘artificial intelligence’, ‘aphasia’, ‘AAC’, ‘rehabilitation’ and others. They winnowed the 663 returned candidate studies using inclusionary and exclusionary criteria to yield 28 studies that met their scoping review criteria, charted, collated, and summarized their contents in order to generate four main themes, namely: (1) AI used for the classification or diagnosis of aphasia per se or diagnostic categories of aphasia; (2) AI used as a supplement to existing aphasia therapies; (3) AI used to create performance models that simulate lexical issues observed in aphasia; and (4) AI used to create performance models that simulate paraphasic issues documented in aphasia.
A large majority of articles reviewed – 17 of 26 (65.4%) employed AI to process clinical and demographic data in order to diagnose aphasia, to classify aphasia diagnostic syndromes, or to diagnose primary progressive aphasia and its variants. These articles were clearly motivated by very successful AI applications in the diagnosis and classification of other medical conditions. Seven other studies – or 26.9% – focused on AI as a supplementary technology that could improve interventions with existing therapies. Le and colleagues, for example, used AI to provide automatic feedback on verbal output during practice at home, which was effective but deemed impractical in current implementation, as their approach requires hand-coding of all patient utterances. Other studies suggested potential application of automatic speech recognition in assessing verbal output, but only when aphasic impairments are mild. As of the date of submission of the article – in late 2022 – no AI was being incorporated directly into AAC devices or apps in the context of aphasia rehabilitation, but given recent advances in generative conversational AI technologies such as ChatGTP, it is likely that situation will change as developers exploit such new capabilities.
For further reading: N. Azevedo, E. Kehayia, G. Jarema et al., 2024,
How artificial intelligence (AI) is used in aphasia rehabilitation: A scoping review. Aphasiology, 38(2):305-336. https://doi.org/10.1080/02687038.2023.2189513