Combining Video‐Based Skeletal and Wearable Physiological Data for Early Detection of Agitation in Dementia Using Deep Learning
Bibliographic record
Abstract
BACKGROUND: Agitation and aggression (AA) in dementia pose significant challenges for patients, caregivers, and healthcare providers. Current detection methods rely on direct observation, delaying intervention and increasing inappropriate psychotropic medication use. Technological approaches, such as actigraphy and multimodal sensory wearables, have shown promise but remain limited in sensitivity and specificity. Advances in deep learning provide new opportunities to enhance early detection by integrating multimodal data sources. METHODS: This study employs a fusion of video-based skeletal keypoints and wearable-driven physiological data to improve AA prediction in individuals with advanced dementia in an inpatient setting. Physiological data were collected using the EmbracePlus wristband, which captures raw signals and digital biomarkers. Concurrently, privacy-compliant software extracted skeletal keypoints from video footage, ensuring patient anonymity. A standardized preprocessing pipeline aligned both data sources through sampling, normalization, and feature extraction. A deep learning framework incorporating recurrent neural networks (RNN) with gated recurrent units (GRU) and attention mechanisms was implemented to identify temporal patterns associated with agitation episodes. RESULTS: Preliminary analysis includes five patients (three females, two males; ages 63-85) with Alzheimer's or mixed pathology dementia and AA. Data collection per participant ranged from 36 to 95.5 hours. Model performance demonstrated an accuracy of 0.98 when integrating both video and wearable data, outperforming single-modality inputs (video: 0.94, wearable: 0.91). CONCLUSIONS: Integrating wearable physiological data with video-derived skeletal keypoints enhances the early detection of agitation in dementia care units. This approach is feasible, privacy-compliant, and improves predictive accuracy. Further research is underway to refine and scale this model for broader clinical application.
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 | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| 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 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".