Protocol for detection and monitoring of post-stroke cognitive impairment through AI-powered speech analysis: a mixed methods pilot study
Bibliographic record
Abstract
Introduction: Post-stroke cognitive impairment (PSCI) affects up to 75% of stroke survivors but remains challenging to detect with traditional neuropsychological assessments. Recent advances in artificial intelligence and natural language processing have opened new avenues for cognitive screening through speech analysis, yet their application to PSCI remains largely unexplored. This study aims to characterize speech markers of PSCI in the first-year post-stroke and evaluate their utility for predicting cognitive outcomes in a Singapore cohort. Methods: This prospective mixed-methods study will recruit 30 stroke survivors from the Alexandra Hospital and National University Hospital in Singapore. Participants will be assessed at four timepoints: baseline (within 6 weeks of stroke onset), 3-, 6-, and 12-months post-stroke. At each visit, participants will complete the Montreal Cognitive Assessment (MoCA) and a standardized speech protocol comprising picture description and semi-structured conversation tasks. Speech recordings will be automatically transcribed using automated speech recognition (ASR) systems based on pretrained acoustic models, and comprehensive linguistic and acoustic features will be extracted. Machine learning models will be developed to predict MoCA-defined cognitive impairment. Statistical analysis will include correlation analysis between speech features and MoCA scores, as well as machine learning classification and regression models to predict cognitive impairment. Linear mixed-effects models will characterize trajectories of MoCA scores and speech features over time. Qualitative analysis will follow an inductive thematic approach to explore acceptability and usability of speech-based screening. Discussion: This study represents a critical step toward developing speech-based digital biomarkers for PSCI detection that are sensitive, culturally appropriate, and clinically feasible. If validated, this approach could transform current models of PSCI care by enabling remote, frequent, and naturalistic monitoring of cognitive health, potentially improving outcomes through earlier intervention.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.060 | 0.047 |
| Meta-epidemiology (narrow) | 0.003 | 0.002 |
| Meta-epidemiology (broad) | 0.003 | 0.003 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.003 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.004 | 0.004 |
| Insufficient payload (model declined to judge) | 0.037 | 0.016 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".