AI and speech-based identification of cognitive impairment in maintenance haemodialysis-the ASPIRE study.

Authors

Anne Møller Lohmann, Maria Dolmer Knop, Laurine Dargaud, Abhista Partal, Dimitrios Salogiannis, Henrik Birn, Jeanette Finderup

Published in

Clinical kidney journal. Volume 19. Issue 10. Pages sfag316. Epub Sep 11, 2026.

Abstract

Cognitive impairment (CI) is common among individuals receiving maintenance haemodialysis and is associated with impaired treatment adherence, reduced quality of life, and mortality. However, CI often remains under-recognized in nephrology care because existing diagnostic approaches are resource-intensive. Speech-based artificial intelligence (AI) is a promising approach for identifying CI through analysis of speech characteristics. This proof-of-concept study investigated whether a speech-based AI model could identify Montreal Cognitive Assessment (MoCA)-based CI and distinguish individuals receiving haemodialysis from clinically relevant controls.
Adults receiving maintenance haemodialysis and controls completed the MoCA and speech tasks. Participants were categorized into low-, moderate-, and high-risk CI groups according to MoCA scores. Acoustic, linguistic, and multimodal speech features were extracted using machine-learning models. Model performance was evaluated.
Speech-based AI differentiated individuals receiving haemodialysis from controls (n = 40). Multimodal models achieved the highest performance, with accuracies ranging from 0.78 to 0.80 for uraemia classification. For MoCA-based cognitive risk classification (n = 63), verbal fluency tasks yielded the strongest results. The best-performing model was the two-task fusion, which identified the high-risk group with a weighted F1-score of 73%, area under the receiver operating characteristic curve (ROC AUC) of 79%, and Matthew's correlation coefficient (MCC) of 0.43. Moderate MoCA-based cognitive risk was more difficult to classify.
Speech-based AI suggested potential as a scalable, non-invasive, and low-burden approach for identifying CI. Verbal fluency tasks and multimodal speech analysis showed the strongest classification performance. Although larger multicentre studies are needed, these findings provide proof of concept that speech may serve as a clinically relevant biomarker of CI in nephrology and support further validation using comprehensive neuropsychological assessment as a reference standard.

PMID:
42859739
Bibliographic data and abstract were imported from PubMed on 11 Oct 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.