Using Machine Learning to Predict Dementia from Neuropsychiatric Symptom and Neuroimaging Data
Bibliographic record
Abstract
BACKGROUND: Machine learning (ML) is a promising technique for patient-specific prediction of mild cognitive impairment (MCI) and dementia development. Neuropsychiatric symptoms (NPS) might improve the accuracy of ML models but have barely been used for this purpose. OBJECTIVES: To investigate if baseline mild behavioral impairment (MBI) status used for NPS quantification along with brain morphology features are predictive of follow-up diagnosis, median 40 months later in patients with normal cognition (NC) or MCI. METHOD: Baseline neuroimaging, neuropsychiatric, and clinical data from 102 individuals with NC and 239 with MCI were extracted from the Alzheimer's Disease Neuroimaging Initiative database. Neuropsychiatric inventory questionnaire items were transformed to MBI domains using a published algorithm. Diagnosis at latest follow-up was used as the outcome variable and ground truth classification. A logistic model tree classifier combined with information gain feature selection was trained to predict follow-up diagnosis. RESULTS: In the binary classification (NC versus MCI/AD), the optimal ML model required only two features from over 200, MBI total score and left hippocampal volume. These features correctly classified participants as remaining normal or developing cognitive impairment with 84.4% accuracy (area under the receiver operating characteristics curve [ROC-AUC] = 0.86). Seven features were selected for the three-class model (NC versus MCI versus dementia) achieving an accuracy of 58.8% (ROC-AUC=0.73). CONCLUSION: Baseline NPS, categorized for MBI domain and duration, have prognostic utility in addition to brain morphology measures for predicting diagnosis change using ML. MBI total score, followed by impulse dyscontrol and affective dysregulation were most predictive of future diagnosis.
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 | 0.003 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".