Protocol for detection and monitoring of post-stroke cognitive impairment through AI-powered speech analysis: a mixed methods pilot study DOI Creative Commons
Ravi Shankar, Effie Chew, Anjali Bundele

et al.

Frontiers in Aging Neuroscience, Journal Year: 2025, Volume and Issue: 17

Published: May 1, 2025

Introduction Post-stroke cognitive impairment (PSCI) affects up to 75% of stroke survivors but remains challenging detect with traditional neuropsychological assessments. Recent advances in artificial intelligence and natural language processing have opened new avenues for screening through speech analysis, yet their application PSCI largely unexplored. This study aims characterize markers the first-year post-stroke evaluate utility predicting outcomes a Singapore cohort. Methods prospective mixed-methods will recruit 30 from Alexandra Hospital National University Singapore. Participants be assessed at four timepoints: baseline (within 6 weeks onset), 3-, 6-, 12-months post-stroke. At each visit, participants complete Montreal Cognitive Assessment (MoCA) standardized protocol comprising picture description semi-structured conversation tasks. Speech recordings automatically transcribed using automated recognition (ASR) systems based on pretrained acoustic models, comprehensive linguistic features extracted. Machine learning models developed predict MoCA-defined impairment. Statistical analysis include correlation between MoCA scores, as well machine classification regression Linear mixed-effects trajectories scores over time. Qualitative follow an inductive thematic approach explore acceptability usability speech-based screening. Discussion represents critical step toward developing digital biomarkers detection that are sensitive, culturally appropriate, clinically feasible. If validated, this could transform current care by enabling remote, frequent, naturalistic monitoring health, potentially improving earlier intervention.

Language: Английский

Protocol for detection and monitoring of post-stroke cognitive impairment through AI-powered speech analysis: a mixed methods pilot study DOI Creative Commons
Ravi Shankar, Effie Chew, Anjali Bundele

et al.

Frontiers in Aging Neuroscience, Journal Year: 2025, Volume and Issue: 17

Published: May 1, 2025

Introduction Post-stroke cognitive impairment (PSCI) affects up to 75% of stroke survivors but remains challenging detect with traditional neuropsychological assessments. Recent advances in artificial intelligence and natural language processing have opened new avenues for screening through speech analysis, yet their application PSCI largely unexplored. This study aims characterize markers the first-year post-stroke evaluate utility predicting outcomes a Singapore cohort. Methods prospective mixed-methods will recruit 30 from Alexandra Hospital National University Singapore. Participants be assessed at four timepoints: baseline (within 6 weeks onset), 3-, 6-, 12-months post-stroke. At each visit, participants complete Montreal Cognitive Assessment (MoCA) standardized protocol comprising picture description semi-structured conversation tasks. Speech recordings automatically transcribed using automated recognition (ASR) systems based on pretrained acoustic models, comprehensive linguistic features extracted. Machine learning models developed predict MoCA-defined impairment. Statistical analysis include correlation between MoCA scores, as well machine classification regression Linear mixed-effects trajectories scores over time. Qualitative follow an inductive thematic approach explore acceptability usability speech-based screening. Discussion represents critical step toward developing digital biomarkers detection that are sensitive, culturally appropriate, clinically feasible. If validated, this could transform current care by enabling remote, frequent, naturalistic monitoring health, potentially improving earlier intervention.

Language: Английский

Citations

0