Robust and interpretable AI-guided marker for early dementia prediction in real-world clinical settings
Bibliographic record
Abstract
Background: Predicting dementia early has major implications for clinical management and patient outcomes. Yet, we still lack sensitive tools for stratifying patients early, resulting in patients being undiagnosed or wrongly diagnosed. Despite rapid expansion in machine learning models for dementia prediction, limited model interpretability and generalizability impede translation to the clinic. Methods: We build a robust and interpretable predictive prognostic model (PPM) and validate its clinical utility using real-world, routinely-collected, non-invasive, and low-cost (cognitive tests, structural MRI) patient data. To enhance scalability and generalizability to the clinic, we: 1) train the PPM with clinically-relevant predictors (cognitive tests, grey matter atrophy) that are common across research and clinical cohorts, 2) test PPM predictions with independent multicenter real-world data from memory clinics across countries (UK, Singapore). Findings: PPM robustly predicts (accuracy: 81.66%, AUC: 0.84, sensitivity: 82.38%, specificity: 80.94%) whether patients at early disease stages (MCI) will remain stable or progress to Alzheimer's Disease (AD). PPM generalizes from research to real-world patient data across memory clinics and its predictions are validated against longitudinal clinical outcomes. PPM allows us to derive an individualized AI-guided multimodal marker (i.e. predictive prognostic index) that predicts progression to AD more precisely than standard clinical markers (grey matter atrophy, cognitive scores; PPM-derived marker: hazard ratio = 3.42, p = 0.01) or clinical diagnosis (PPM-derived marker: hazard ratio = 2.84, p < 0.01), reducing misdiagnosis. Interpretation: Our results provide evidence for a robust and explainable clinical AI-guided marker for early dementia prediction that is validated against longitudinal, multicenter patient data across countries, and has strong potential for adoption in clinical practice. Funding: Wellcome Trust, Royal Society, Alzheimer's Research UK, Alzheimer's Drug Discovery Foundation Diagnostics Accelerator, Alan Turing Institute.
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 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 | 0.005 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 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 | 0.001 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, 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".