Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Leveraging Self-Training and Variational Autoencoder for Agitation Detection in People with Dementia Using Wearable Sensors

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that a variational autoencoder for feature compression plus a self-training loop for pseudo-labeling significantly improves agitation and aggression detection in people with dementia from wristband sensor data, with…

desk verdict The 90.18% balanced accuracy claim is undermined by a test set that appears to include the model's own pseudo-labels, though the new dataset and clinical angle are worth a look. read the letter →

arxiv 2412.19254 v1 pith:CGBROUQX submitted 2024-12-26 cs.AI

Abeer Badawi , Somayya Elmoghazy , Samira Choudhury , Khalid Elgazzar , Amer Burhan This is my paper · ORCID
classification cs.AI
keywords dementiaagitationdetectionwearablesensorsvariationalautoencoderself-trainingsemi-supervisedlearningXGBoostphysiologicalsignals
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that a semi-supervised pipeline can detect agitation and aggression episodes in people with dementia from wristband physiology when only a fraction of the data has clinical labels. The proposed system first compresses 182 engineered features into a 100-dimensional representation with a variational autoencoder, then runs self-training to generate pseudo-labels for unlabeled minutes and retrains three classifiers. The authors report that this combination beats fully supervised baselines, with XGBoost reaching a balanced accuracy of 90.18% and an AUC-ROC of 99.6%. If the result holds, continuous monitoring from routinely collected wearable data becomes a practical route to early agitation alerts without requiring exhaustive manual labeling.

What carries the argument

The load-bearing mechanism is a two-stage pipeline. First, a variational autoencoder, a probabilistic encoder-decoder trained to reconstruct input while regularizing a latent distribution, maps 182 hand-crafted features from heart rate, acceleration, electrodermal activity, and temperature into a 100-dimensional latent code per one-minute window. Second, a self-training loop trains a classifier on the labeled windows, predicts pseudo-labels for the 37,463 unlabeled windows, accepts those with predicted probability above a 0.7 threshold, and retrains, up to 100 iterations. The classifiers compared are Random Forest, Extra Trees, and XGBoost, with balanced accuracy as the headline metric because the agitation class is heavily outnumbered.

What would settle it

Train the identical pipeline but evaluate with leave-one-participant-out on the five fully labeled participants, and audit the 30% test windows to confirm they carry nurse-confirmed clinical labels rather than model-generated pseudo-labels; also reconcile the abstract's 90.16% figure with the 90.18% reported in Table 5. If balanced accuracy drops materially below 90% under these conditions, the central claim would not survive.

Watch

Extended reading notes

Core claim

The central claim is that combining variational autoencoder representation learning with self-training materially improves agitation and aggression classification compared with supervised learning on the labeled portion alone. On a dataset of 57,742 one-minute windows from 14 participants (5 fully labeled and 9 unlabeled), XGBoost trained on VAE-encoded features with self-training achieved a balanced accuracy of 90.18%, precision, recall, and F1-score of 99.0%, and an AUC-ROC of 99.6%, versus 85.03% balanced accuracy for the best fully supervised baseline. The paper interprets this as evidence that the VAE's compact latent representation, combined with confident pseudo-labels from self-training, lets models exploit unlabeled data rather than being limited by scarce clinical labels.

Load-bearing premise

The reported accuracy assumes that the 70/30 split of one-minute windows creates independent training and test sets whose test labels are clinically confirmed rather than pseudo-labels, and that windows from the same participant do not appear in both sets.

Editorial extensions

If this is right

  • A semi-supervised pipeline can improve agitation detection over supervised baselines when labeled data are scarce.
  • The VAE dimension reduction from 182 to 100 features does not hurt and appears to help downstream classification.
  • The self-training label expansion roughly doubles the number of positive agitation windows in the training set before final classification.
  • XGBoost is the best-performing classifier in this setting, while Extra Trees is the fastest, indicating a practical speed-accuracy trade-off.
  • The system could support continuous monitoring in long-term care by alerting caregivers to probable agitation episodes.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A reader should verify the composition of the 30% test set: Table 3 reports label counts, such as 55,214 normal and 2,523 agitation minutes, that sum to nearly the full 57,742-minute dataset even in the fully supervised condition, which suggests some rows may already include pseudo-labeled or unlabeled windows; if so, the reported balanced accuracy is not a clean measure of clinical generalization
  • The natural next experiment is leave-one-participant-out evaluation, because the random window split can place windows from the same participants in both training and test, inflating performance relative to unseen patients.
  • The abstract's 90.16% and Section 4.4's 90.18% differ slightly; reconciling the exact reported number and the model version behind it would be a useful check.
  • The same VAE-plus-self-training recipe is a plausible template for other wearable monitoring tasks with sparse labels, such as stress or seizure detection, but the transfer depends on the pseudo-label threshold and class-imbalance handling being re-tuned per task.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper develops a semi-supervised pipeline for detecting agitation/aggression (AA) in people with dementia from Empatica E4 physiological data. The dataset includes 57,742 one-minute windows from 14 participants across three studies, of which 18,804 minutes are labeled normal, 1,475 are labeled AA, and 37,463 are unlabeled. The proposed system extracts features with the Flirt toolkit, learns a compact representation with a variational autoencoder (VAE), and then applies self-training with three classifiers (Random Forest, Extra Trees, XGBoost). The authors report that XGBoost with VAE and self-training achieves a balanced accuracy of 90.18% and an AUC ROC of 99.6%, and they conclude that combining VAE with self-training significantly improves AA detection.

Significance. The application is practically relevant, and the multi-site partially labeled dataset is a potentially useful resource. If the reported results were trustworthy, the paper would offer evidence that self-training can mitigate the labeling bottleneck in wearable-based agitation detection. However, the evaluation protocol does not support the central claim: the test set appears to contain pseudo-labeled windows, the split is at the window rather than participant level, and the supervised baseline in Table 3 is internally inconsistent. The quantitative headline is therefore not established, although the underlying idea and dataset may be salvageable with a rigorous re-analysis.

major comments (5)
  1. [§3.3, §3.2.3, Tables 4–5] The manuscript does not state that pseudo-labeled samples are excluded from the 30% test set. Section 3.2.3 describes adding pseudo-labeled samples to the training set, and Section 3.3 describes only a random 70/30 split. After self-training, the Normal/AA counts in Table 4 (e.g., 55,214/2,523 for Random Forest) and Table 5 (e.g., 54,712/3,026 for XGBoost) sum to nearly the full 57,742-minute dataset, so a 30% slice of the augmented data necessarily contains many windows whose labels were produced by the classifier itself. Unless the authors document an explicit exclusion of all pseudo-labeled windows and all unlabeled windows from the test partition, balanced accuracy and AUC measure agreement with the model's own pseudo-labels rather than clinical AA detection.
  2. [§3.3] The split is performed at the level of one-minute windows, not participants. With only five fully labeled participants, windows from the same participant are almost certainly present in both training and testing, so the reported metrics do not measure generalization to new patients or sessions. A participant-level split, or at least leave-one-participant-out evaluation, is required before any claim about detecting AA in unseen individuals can be made.
  3. [§4.2, Table 3] Table 3 is titled 'Supervised Learning and Variational Autoencoder' and Section 4.2 says the model was trained on 18,804 normal and 1,475 AA labeled minutes, but the table's Normal/AA row reports counts (55,214/2,523, 55,648/2,092, 54,768/2,974) identical to the self-training counts in Table 4. Either the row is mislabeled or the 'supervised' experiment already includes pseudo-labeled data; as written, the comparison in Section 4.4 between 'fully supervised learning with VAE' and the proposed system is not interpretable.
  4. [§4.4, Abstract] The claim that self-training and VAE 'significantly improve' performance is not supported by any significance test, confidence interval, or repeated-split analysis; the single 70/30 split with one random seed is insufficient. Additionally, the abstract reports an accuracy of 90.16% while Section 4.4 and Section 5 report 90.18%, and the test-set size for the augmented datasets is never given.
  5. [§3.2.2] The VAE's training procedure does not state whether the VAE was fit before or after the 70/30 split. If the encoder was trained on the entire dataset, test-window features are influenced by test data, which is a further form of leakage. The VAE should be trained on the training partition only, and the manuscript should specify this explicitly.
minor comments (6)
  1. [Abstract vs. §4.4] The abstract reports an accuracy of 90.16% while Section 4.4 and Section 5 report 90.18%; these numbers should be reconciled.
  2. [Equations (2) and (3)] Equations (2) and (3) contain the citation marker '[44]' inside the displayed formulas; the citations should be moved to the surrounding text.
  3. [§3.2.2] The sentence 'as shown in 4.4' appears to refer to Equation (4), not to a numbered section; the cross-reference should be corrected.
  4. [Tables 2–5] The meaning of the 'Normal/AA Labels' row is ambiguous: it should state clearly whether the counts refer to the training set, the full augmented dataset, or the test set.
  5. [§3.2.2] The text says the authors 'choose 100 features from the 182 original features' but also that the VAE produces a 100-dimensional latent representation; the relationship between the selected features and the latent dimension should be clarified.
  6. [Table 6] The comparison in Table 6 is across different datasets, different participants, and different feature sets; the caption should more strongly warn the reader that the AUC values are not directly comparable.

Circularity Check

2 steps flagged · score 7.0 of 10

Evaluation against the model's own pseudo-labels makes the headline 90.18% balanced accuracy a self-consistency result, not a clinical detection result.

  1. fitted input called prediction [Section 3.3 (Performance Evaluation Metrics)]
    "We use three classification models: Extra Trees, Random Forest, and XGBoost classification models to classify the Self-training model-generated labels. ... We then split the dataset into training and testing sets, with 70% used for training and 30% for testing."

    The split is introduced as part of classifying 'Self-training model-generated labels,' and no step excludes pseudo-labeled windows from the test partition. Table 5 shows the XGBoost run ends with 54,712 normal + 3,026 AA = 57,738 labeled minutes, essentially the whole 57,742-minute dataset. A 30% test slice taken from this augmented set therefore contains many labels produced by the self-training model itself. Those pseudo-labels were admitted only when classifier confidence exceeded 0.7, so test predictions agree with them by construction; the reported 90.18% balanced accuracy and 99.6% AUC measure self-consistency, not clinical AA detection.

  2. fitted input called prediction [Section 4.2 and Table 3]
    "For the supervised learning scenario shown in Table 3, the model is trained and evaluated using labeled data, consisting of 18,804 minutes labeled as normal and 1,475 minutes labeled as AA. (Table 3 lists "Normal/AA Labels 55214/2523".)"

    The text says only original clinical labels (18,804 normal, 1,475 AA) are used, but Table 3 reports Normal/AA counts of 55,214/2,523, which sum to 57,737 and can only be reached by adding self-training pseudo-labels. This contradiction shows the evaluation tables include model-generated labels as ground truth even in the 'fully supervised' comparison. The balanced-accuracy improvements are therefore at least partly the classifier agreeing with its own high-confidence pseudo-labels rather than detecting clinically confirmed agitation.

full rationale

The paper's central claim is that VAE-based feature extraction plus self-training improves agitation detection, with XGBoost reaching 90.18% balanced accuracy and 99.6% AUC (Table 5). The load-bearing condition for that claim is a test set containing only genuine clinician-provided labels on data not used to generate pseudo-labels. The manuscript does not establish this: Section 3.3 describes the 70/30 split immediately after saying the models 'classify the Self-training model-generated labels,' and nowhere excludes pseudo-labeled samples from the test partition. The label counts in Tables 3-5 sum to nearly the full 57,742 minutes, which is only possible if the augmented self-training labels are present in the evaluation set. Since pseudo-labels are accepted only above a 0.7 confidence threshold, scoring the model on those same pseudo-labels is scoring it on its own outputs; the reported metrics are largely self-consistency. The VAE feature dimension (100) and pseudo-label threshold (0.7) are also selected on the same data, and the split is at the window rather than participant level, so even original-label windows from the same participant appear in both training and test partitions. Self-citations [41,42] are used to justify preprocessing and model choice, but they are not the main circularity and do not independently raise the score. The method itself is not inherently circular, but the evaluation as reported does not provide an independent test of AA detection.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The empirical performance claim depends on several hand-set hyperparameters, domain labeling assumptions, and an independence assumption for the data split. No new physical or algorithmic entities are introduced.

free parameters (6)
  • VAE latent dimension = 100
    Chosen 'after several trials and investigations of the parameters and features set' (Section 3.2.2); no held-out validation or sensitivity analysis is reported.
  • Pseudo-label confidence threshold = 0.7
    Set in Section 3.2.3 to filter pseudo-labels; no analysis of how this threshold changes performance or which samples are added.
  • Self-training maximum iterations = 100 (actual iterations: RF 10, ET 3, XGB 7)
    The maximum is set to 100, but reported iteration counts vary per model, indicating tuning on the evaluation set.
  • Window size = 1 minute, no overlap
    Sliding-window choice in Section 3.2.1; determines the feature granularity and creates strong temporal correlation between consecutive windows.
  • VAE training epochs and batch size = 50 epochs, 128 batch
    Reported in Section 3.2.2 without a validation-based selection procedure.
  • Selected feature count = 182 of 198 features
    Features yielding NaN or infinity values were dropped (Section 3.2.1); this data-dependent filtering is a modeling choice.
assumptions (5)
  • domain assumption Nurses' shift notes and Dementia Observation System records provide accurate minute-level ground truth for agitation events.
    Section 3.1 states AA episodes were labeled from nurses' notes and DOS records; no inter-rater reliability or validation of label timing is reported.
  • domain assumption One-minute windows are independent, so a random 70/30 split is unbiased.
    Section 3.3 splits the dataset without grouping by participant, ignoring temporal autocorrelation within continuous 48-72 hour sessions.
  • domain assumption Classifier confidence above 0.7 is a reliable proxy for pseudo-label correctness.
    Section 3.2.3 uses this threshold to admit unlabeled samples; no calibration, manual verification, or error analysis of pseudo-labels is given.
  • domain assumption VAE latent features retain agitation-relevant information despite being trained only for reconstruction and KL regularization.
    Section 3.2.2 defines the VAE loss without any downstream validation that the compressed features preserve the signal needed for AA classification.
  • standard math Standard VAE reparameterization and self-training convergence behavior are valid background mathematics.
    Equations 1 through 8 in Section 3.2.2 use standard VAE machinery; this is accepted background.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Self-Training and Variational Autoencoder for Agitation Detection in People with Dementia Using Wearable Sensors." pith.science (2026). https://pith.science/paper/CGBROUQX

@misc{pith2026241219254,
  author       = {Pith},
  title        = {Pith review of: Leveraging Self-Training and Variational Autoencoder for Agitation Detection in People with Dementia Using Wearable Sensors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CGBROUQX}},
  note         = {Machine review of arXiv:2412.19254}
}
read the original abstract

Dementia is a neurodegenerative disorder that has been growing among elder people over the past decades. This growth profoundly impacts the quality of life for patients and caregivers due to the symptoms arising from it. Agitation and aggression (AA) are some of the symptoms of people with severe dementia (PwD) in long-term care or hospitals. AA not only causes discomfort but also puts the patients or others at potential risk. Existing monitoring solutions utilizing different wearable sensors integrated with Artificial Intelligence (AI) offer a way to detect AA early enough for timely and adequate medical intervention. However, most studies are limited by the availability of accurately labeled datasets, which significantly affects the efficacy of such solutions in real-world scenarios. This study presents a novel comprehensive approach to detect AA in PwD using physiological data from the Empatica E4 wristbands. The research creates a diverse dataset, consisting of three distinct datasets gathered from 14 participants across multiple hospitals in Canada. These datasets have not been extensively explored due to their limited labeling. We propose a novel approach employing self-training and a variational autoencoder (VAE) to detect AA in PwD effectively. The proposed approach aims to learn the representation of the features extracted using the VAE and then uses a semi-supervised block to generate labels, classify events, and detect AA. We demonstrate that combining Self-Training and Variational Autoencoder mechanism significantly improves model performance in classifying AA in PwD. Among the tested techniques, the XGBoost classifier achieved the highest accuracy of 90.16\%. By effectively addressing the challenge of limited labeled data, the proposed system not only learns new labels but also proves its superiority in detecting AA.

Figures

Figures reproduced from arXiv: 2412.19254 by the authors.

Figure 1
Figure 1. A Block Diagram of the proposed System Architecture. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The Data Pre-processing and Feature Extraction Workflow Derived from Empatica E4 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Proposed System Architecture to Classify AA in PwD using VAE and Self-training. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: AUC ROC and Precision-Recall Curve for XGBoost Classifier using VAE and Semi [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Benchmarking Early Agitation Prediction in Community-Dwelling People with Dementia Using Multimodal Sensors and Machine Learning

    eess.SP 2025-05 conditional novelty 5.0 of 10

    A LightGBM model using sensor features plus time-of-day and current agitation status predicts next-6-hour agitation with AUC-ROC 0.972 and AUC-PR 0.432 on the TIHM dataset, but random-fold evaluation and label-derived...

Reference graph

Works this paper leans on

46 extracted references · 42 canonical work pages · cited by 1 Pith paper

  1. [32]

    Semi-supervised learning for emotion recognition in dementia care: A review,

    J. Doe et al. , “Semi-supervised learning for emotion recognition in dementia care: A review,” International Journal of Dementia Research , vol. 7, pp. 65–80, 2021

    work page 2021

  2. [20]

    A pilot study to detect agitation in people living with dementia using multi-modal sensors,

    S. Spasojevic, J. Nogas, A. Iaboni, B. Ye, A. Mihailidis, A. Wang, S. J. Li, L. S. Martin, K. New- man, and S. S. Khan, “A pilot study to detect agitation in people living with dementia using multi-modal sensors,” Journal of Healthcare Informatics Research , vol. 5, no. 3, pp. 342–358, 2021

    work page 2021

  3. [1]

    Dementia,

    W. H. Organization, “Dementia,” https://www.who.int/news-room/fact-sheets/detail/dementia, 2023, [Accessed December 30, 2024]

    work page 2023

  4. [2]

    Dementia,

    S. A. Gale, D. Acar, and K. R. Daffner, “Dementia,” Am J Med , vol. 131, no. 10, pp. 1161–1169, 2018

    work page 2018

  5. [3]

    Dementia: An overview,

    N. Bansal and M. Parle, “Dementia: An overview,” Journal of Pharmaceutical Technology, Re- search and Management , vol. 2, pp. 29–45, 2014

    work page 2014

  6. [4]

    Dementia: the bare essentials,

    M. I. Kester and P. Scheltens, “Dementia: the bare essentials,” Practical Neurology, vol. 9, no. 4, pp. 241–251, 2009

    work page 2009

  7. [5]

    Preva- lence of neuropsychiatric symptoms in dementia and mild cognitive impairment: results from the cardiovascular health study,

    C. G. Lyketsos, O. Lopez, B. Jones, A. L. Fitzpatrick, J. Breitner, and S. DeKosky, “Preva- lence of neuropsychiatric symptoms in dementia and mild cognitive impairment: results from the cardiovascular health study,” Jama, vol. 288, no. 12, pp. 1475–1483, Sep 25 2002, (in eng)

    work page 2002

  8. [6]

    S13: Technology enabled care for neuropsychiatric symptoms of dementia: implementation at the point of care,

    A. M. Burhan, W. Sun, M. Chiu, S. Choudhury, A. Badawi, and K. Elgazzar, “S13: Technology enabled care for neuropsychiatric symptoms of dementia: implementation at the point of care,” International Psychogeriatrics, vol. 35, no. S1, pp. 40–40, 2023

    work page 2023

Show all 46 references
  1. [7]

    Agitation and aggression in people with alzheimer’s disease,

    C. Ballard and A. Corbett, “Agitation and aggression in people with alzheimer’s disease,” Curr Opin Psychiatry , vol. 26, no. 3, pp. 252–259, May 2013, (in eng)

    2013

  2. [8]

    Detecting agitation and aggression in people with dementia using sensors—a systematic review,

    S. S. Khan, B. Ye, B. Taati, and A. Mihailidis, “Detecting agitation and aggression in people with dementia using sensors—a systematic review,” Alzheimer’s & Dementia , vol. 14, no. 6, pp. 824–832, 2018. 13

    2018

  3. [9]

    Dementia and agitation in nursing home residents: How are they related?

    J. Cohen-Mansfield, M. S. Marx, and A. S. Rosenthal, “Dementia and agitation in nursing home residents: How are they related?” Psychology and Aging , vol. 5, no. 1, p. 3, 1990

    1990

  4. [10]

    Intelligent assistive technology applications to dementia care: current capabilities, limitations, and future challenges,

    A. J. Bharucha, V. Anand, J. Forlizzi, M. A. Dew, C. F. Reynolds III, S. Stevens, and H. Wactlar, “Intelligent assistive technology applications to dementia care: current capabilities, limitations, and future challenges,” The American journal of geriatric psychiatry , vol. 17,...

    2009

  5. [11]

    Artificial intelligence for alzheimer’s disease: promise or challenge?

    C. Fabrizio, A. Termine, C. Caltagirone, and G. Sancesario, “Artificial intelligence for alzheimer’s disease: promise or challenge?” Diagnostics, vol. 11, no. 8, p. 1473, 2021

    2021

  6. [12]

    Semi-supervised learning for identifying the likelihood of agitation in people with dementia,

    R. Rezvani et al. , “Semi-supervised learning for identifying the likelihood of agitation in people with dementia,” Journal of Healthcare Informatics Research , vol. 5, pp. 45–58, 2021

    2021

  7. [13]

    Automatic feature extraction from wearable sensor data by use of machine learnings,

    K. Sato, M. Chida, Y. Hayakawa, and N. M. Fujiki, “Automatic feature extraction from wearable sensor data by use of machine learnings,” Proceedings of The 7th International Conference on Intelligent Systems and Image Processing 2019 , 2019

    2019

  8. [14]

    Towards automatic feature extraction for activity recognition from wearable sensors: A deep learning approach,

    B. Chikhaoui and F. Gouineau, “Towards automatic feature extraction for activity recognition from wearable sensors: A deep learning approach,” 2017 IEEE International Conference on Data Mining Workshops (ICDMW) , pp. 693–702, 2017

    2017

  9. [15]

    Activity recognition with evolving data streams: A review,

    Z. S. Abdallah, M. M. Gaber, B. Srinivasan, and S. Krishnaswamy, “Activity recognition with evolving data streams: A review,” ACM Computing Surveys (CSUR) , vol. 51, no. 4, p. 71, 2018

    2018

  10. [16]

    Extraction and interpretation of deep autoencoder-based temporal features from wearables for forecasting personalized mood, health, and stress,

    B. Li and A. Sano, “Extraction and interpretation of deep autoencoder-based temporal features from wearables for forecasting personalized mood, health, and stress,” Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , vol. 4, pp. 1 – 26, 2020

    2020

  11. [17]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013

    2013 arXiv

  12. [18]

    Nabilone for agitation blinded intervention trial (nab-it),

    “Nabilone for agitation blinded intervention trial (nab-it),” ClinicalTrials.gov, 2021, [Accessed December 30, 2024]. [Online]. Available: https://clinicaltrials.gov/ct2/show/NCT04516057

    2021

  13. [19]

    Standardizing care for neuropsychiatric symptoms and quality of life in dementia (stan),

    “Standardizing care for neuropsychiatric symptoms and quality of life in dementia (stan),” ClinicalTrials.gov, [Accessed December 30, 2024]. [Online]. Available: https: //clinicaltrials.gov/ct2/show/NCT03672201

    2024

  14. [21]

    15 feasibility of using wearable sensors to detect agitation in persons with dementia,

    C. Nesbitt, A. Gupta, K. Maly, H. R. Okhravi, and S. Jain, “15 feasibility of using wearable sensors to detect agitation in persons with dementia,” CNS Spectrums, vol. 24, no. 1, pp. 181–181, 2019

    2019

  15. [22]

    Agitation detection in people living with dementia using multimodal sensors,

    S. S. Khan, S. Spasojevic, J. Nogas, B. Ye, A. Mihailidis, A. Iaboni, A. Wang, L. S. Martin, and K. Newman, “Agitation detection in people living with dementia using multimodal sensors,” in 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biolo...

    2019

  16. [23]

    Deep learning in human activity recognition with wearable sensors: A review on advances,

    S. Zhang, Y. Li, S. Zhang, F. Shahabi, S. Xia, Y. Deng, and N. Alshurafa, “Deep learning in human activity recognition with wearable sensors: A review on advances,” Sensors, vol. 22, no. 4, p. 1476, 2022

    2022

  17. [24]

    Towards automatic feature extraction for activity recognition from wearable sensors: a deep learning approach,

    B. Chikhaoui and F. Gouineau, “Towards automatic feature extraction for activity recognition from wearable sensors: a deep learning approach,” in 2017 IEEE international conference on data mining workshops (ICDMW) . IEEE, 2017, pp. 693–702

    2017

  18. [25]

    Recognition of human activities using continuous autoencoders with wearable sensors,

    L. Wang, “Recognition of human activities using continuous autoencoders with wearable sensors,” Sensors, vol. 16, no. 2, p. 189, 2016. 14

    2016

  19. [26]

    Zhu and A

    X. Zhu and A. B. Goldberg, Introduction to semi-supervised learning. Springer Nature, 2022

    2022

  20. [27]

    Clinical document classification using labeled and unlabeled data across hospitals,

    H. Hassanzadeh, M. Kholghi, A. Nguyen, and K. Chu, “Clinical document classification using labeled and unlabeled data across hospitals,” in AMIA annual symposium proceedings, vol. 2018. American Medical Informatics Association, 2018, p. 545

    2018

  21. [28]

    Mean teachers are better role models: Weight-averaged consis- tency targets improve semi-supervised deep learning results,

    A. Tarvainen and H. Valpola, “Mean teachers are better role models: Weight-averaged consis- tency targets improve semi-supervised deep learning results,” in Advances in neural information processing systems, vol. 30, 2017, pp. 1195–1204

    2017

  22. [29]

    Self-labeled techniques for semi-supervised learning: Tax- onomy, software and empirical study,

    I. Triguero, S. Garc ´ ıa, and F. Herrera, “Self-labeled techniques for semi-supervised learning: Tax- onomy, software and empirical study,” Knowledge and Information Systems , vol. 42, no. 2, pp. 245–284, 2015

    2015

  23. [30]

    Self-supervised ecg representation learning for emotion recognition,

    P. Sarkar and A. Etemad, “Self-supervised ecg representation learning for emotion recognition,” arXiv preprint arXiv:2002.03898 , 2020

    2002 arXiv

  24. [31]

    Applying self-supervised representation learning for emotion recognition using physiological signals,

    K. G. M. Quispe et al., “Applying self-supervised representation learning for emotion recognition using physiological signals,” IEEE Transactions on Affective Computing , vol. 9, pp. 78–91, 2021

    2021

  25. [33]

    An end-to-end methodology for semi-supervised har data collection, labeling, and classification using a wristband,

    L. B. Hinkle et al., “An end-to-end methodology for semi-supervised har data collection, labeling, and classification using a wristband,” Journal of Ambient Intelligence and Humanized Computing , vol. 11, pp. 215–230, 2021

    2021

  26. [34]

    Semi-supervised learning and data augmentation for wearable-based health monitoring system in the wild,

    H. Yu and A. Sano, “Semi-supervised learning and data augmentation for wearable-based health monitoring system in the wild,” in NeurIPS 2022 Workshop on Learning from Time Series for Health, 2022

    2022

  27. [35]

    Data-driven forecasting of agitation for persons with dementia: A deep learning-based approach,

    S. HekmatiAthar et al. , “Data-driven forecasting of agitation for persons with dementia: A deep learning-based approach,” International Journal of Dementia Care , vol. 8, pp. 112–126, 2022

    2022

  28. [36]

    A. P. Association, Diagnostic and Statistical Manual of Mental Disorders , 5th ed. Arlington, V A, US: American Psychiatric Publishing, Inc., 2013

    2013

  29. [37]

    Agitation in cognitive disorders: International psychogeriatric association provisional consensus clinical and research definition,

    J. Cummings et al. , “Agitation in cognitive disorders: International psychogeriatric association provisional consensus clinical and research definition,” International Psychogeriatrics , vol. 27, no. 1, pp. 7–17, 2015

    2015

  30. [38]

    The mini-mental state examination (mmse),

    L. Kurlowicz and M. Wallace, “The mini-mental state examination (mmse),” pp. 8–9, 1999

    1999

  31. [39]

    Agitation in cognitive disorders: International psychogeriatric association provisional consensus clinical and research definition,

    J. Cummings, J. Mintzer, H. Brodaty, M. Sano, S. Banerjee, D. Devanand, S. Gauthier, R. Howard, K. Lanctˆ ot, C. G. Lyketsos et al. , “Agitation in cognitive disorders: International psychogeriatric association provisional consensus clinical and research definition,” Internati...

    2015

  32. [40]

    Empatica — Medical devices, AI and algorithms for remote patient monitoring,

    Empatica Inc., “Empatica — Medical devices, AI and algorithms for remote patient monitoring,” Empatica, [Accessed December 30, 2024]. [Online]. Available: https://www.empatica.com

    2024

  33. [41]

    Inves- tigating multimodal sensor features importance to detect agitation in people with dementia,

    A. Badawi, K. Elgazzar, B. Ye, K. Newman, A. Mihailidis, A. Iaboni, and S. S. Khan, “Inves- tigating multimodal sensor features importance to detect agitation in people with dementia,” in 2023 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE) , 2023, pp. 77–82

    2023

  34. [42]

    Artificial intelligence and features investigating to detect neuropsychiatric symptoms in patients with dementia: A pilot study,

    A. Badawi, S. Choudhury, K. Elgazzar, and A. M. Burhan, “Artificial intelligence and features investigating to detect neuropsychiatric symptoms in patients with dementia: A pilot study,” in 2023 IEEE Symposium Series on Computational Intelligence (SSCI) , 2023, pp. 741–746

    2023

  35. [43]

    Flirt: A feature generation toolkit for wearable data,

    S. F¨ oll, M. Maritsch, F. Spinola, V. Mishra, F. Barata, T. Kowatsch, E. Fleisch, and F. Wortmann, “Flirt: A feature generation toolkit for wearable data,” Computer Methods and Programs in Biomedicine, vol. 212, p. 106461, 2021. 15

    2021

  36. [44]

    An introduction to variational autoencoders,

    D. P. Kingma and M. Welling, “An introduction to variational autoencoders,” arXiv preprint arXiv:1906.02691, 2019

    1906 arXiv

  37. [45]

    Tutorial on variational autoencoders,

    C. Doersch, “Tutorial on variational autoencoders,” arXiv preprint arXiv:1606.05908 , 2016

    2016 arXiv

  38. [46]

    A survey on semi-supervised learning,

    J. E. Van Engelen and H. H. Hoos, “A survey on semi-supervised learning,” Machine learning , vol. 109, no. 2, pp. 373–440, 2020. 16

    2020

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.