Spoken language-based automatic cognitive assessment of stroke survivors
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Stroke survivors (SSs) often experience cognitive decline following their initial stroke, necessitating repeat post-stroke cognitive assessments. Current methods of assessment, such as the pen-and-paper-based Montreal Cognitive Assessment (MoCA), is time-consuming and often reliant on seeing skilled clinicians in person. This is at a time when patients have a lot of often diverse rehabilitation needs. To address these challenges, our paper introduces the first system of its kind to be used for this cohort. CognoSpeak is an automated cognitive assessment system that people can use initially on the ward immediately post-stroke (baseline) and subsequently at home (follow-ups). CognoSpeak assesses cognitive decline by asking users to engage with a virtual agent by answering questions and completing clinically-motivated tasks and cognitive tests. The system then uses AI to extract and process speech, language, and interactional cues for cognitive decline. The system was originally developed for dementia; here, we show that it can successfully predict MoCA scores (regression) and identify cognitive decline predicated on a MoCA-based threshold (classification) in the stroke survivor cohort. We explore an extensive set of acoustic- and text-based features as well as different machine learning models. Leveraging a unique dataset of 55 SS CognoSpeak interactions, our findings show excellent performance for both regression and classification style prediction with the best regression result (Normalised Root Mean Squared Error (N-RMSE)) of 0.092. In addition, we show that direct classification of the MoCA score cutoff of 26 yields an F1-score of 0.74 (Specificity: 0.73, Sensitivity: 0.75) using a Logistic Regression Classifier. This demonstrates the first evidence of the system’s robustness and clinical potential.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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 it