Physician in the Loop
For youSign in
Speech-language pathology ·

AI app R-SPEAK builds speech for people with aphasia; deep learning automates ultrasound swallowing measurement

A communication tablet on a stand, its grid of picture buttons showing house, sun, cup, ball, apple and tree, the cup button lit amber, under a speech bubble.
THE SHORT READ

Researchers in the United Kingdom reported in the Journal of Medical Internet Research that people with expressive aphasia found R-SPEAK, an AI smartphone app that builds coherent speech from their utterances, acceptable. A deep learning framework automated scanline selection in ultrasound-based dysphagia assessment.

THE DETAILS
  • AI Oct 8, 2026
    AI app R-SPEAK, which builds coherent speech from utterances in expressive aphasia, rated acceptable in UK co-design study
    Researchers at Sheffield Hallam University, the University of Sheffield and the University of Nottingham co-designed R-SPEAK, a smartphone app meant to build upon the utterances of people with mild-to-moderate aphasia to produce coherent speech, according to a study published Oct. 8 in the Journal of Medical Internet Research. Four people with aphasia and one carer evaluated it in interviews; the four with aphasia gave it a mean System Usability Scale score of 75 (SD 9.19). The authors said the lightweight Qwen2.5:3B model performed strongest overall, within 3% of larger models on quality. They said it was acceptable to patients.
    Sources: Journal of Medical Internet Research, PubMed
  • AI Oct 7, 2026
    Deep learning picks suprahyoid ultrasound scanlines for dysphagia assessment with 72.3% agreement with expert consensus
    A deep learning framework segmented suprahyoid muscle ultrasound images and automatically chose the scanline used for quantitative dysphagia assessment, according to a study published online Oct. 7 in Medical & Biological Engineering & Computing. The study analyzed 371 ultrasound images from patients with normal, mild and severe dysphagia. The U-Net segmentation model achieved an average intersection-over-union of 0.9599, and automated scanline determination reached an overall agreement of 72.3% with expert consensus. The authors said the framework provides an objective and reproducible approach for ultrasound-based dysphagia assessment.
    Sources: Medical & Biological Engineering & Computing, PubMed

The weekly innovation roundup for speech-language pathology, published every Sunday for the week before; every item links to its sources. Latest speech-language pathology news · All specialties