Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Abeer Badawi, Somayya Elmoghazy, Samira Choudhury, Khalid Elgazzar, Amer Burhan

TL;DR

This work tackles AA detection in PwD using wearable sensor data from the Empatica E4, addressing the scarcity of labeled data by integrating a variational autoencoder (VAE) for feature learning with a self-training semi-supervised loop. The methodology processes multi-site data, reduces features to a compact latent representation, and leverages unlabeled data to generate pseudo-labels for improved classification by Random Forest, Extra Trees, and XGBoost, achieving the best performance with XGBoost at about $90.18\%$ balanced accuracy and high AUC. The study demonstrates that combining VAE-based representations with self-training yields superior performance over fully supervised baselines across three Canadian hospital datasets, highlighting practical potential for continuous, real-world AA monitoring in dementia care. This approach offers a scalable path to leverage abundant unlabeled wearable data to enhance early detection and timely intervention for agitation in PwD, potentially improving patient outcomes and caregiver well-being.

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.

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

TL;DR

This work tackles AA detection in PwD using wearable sensor data from the Empatica E4, addressing the scarcity of labeled data by integrating a variational autoencoder (VAE) for feature learning with a self-training semi-supervised loop. The methodology processes multi-site data, reduces features to a compact latent representation, and leverages unlabeled data to generate pseudo-labels for improved classification by Random Forest, Extra Trees, and XGBoost, achieving the best performance with XGBoost at about balanced accuracy and high AUC. The study demonstrates that combining VAE-based representations with self-training yields superior performance over fully supervised baselines across three Canadian hospital datasets, highlighting practical potential for continuous, real-world AA monitoring in dementia care. This approach offers a scalable path to leverage abundant unlabeled wearable data to enhance early detection and timely intervention for agitation in PwD, potentially improving patient outcomes and caregiver well-being.

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.
Paper Structure (16 sections, 5 equations, 4 figures, 6 tables)

This paper contains 16 sections, 5 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Research Methodology
  4. Dataset Description
  5. Proposed System Architecture
  6. Data Pre-processing and Feature Extraction
  7. Variational Autoencoder
  8. Self-training
  9. Performance Evaluation Metrics
  10. Results and Discussion
  11. Agitation Detection using Supervised Learning
  12. Agitation Detection using Supervised Learning and VAE
  13. Agitation Detection using Self-training
  14. Agitation Detection using Self-training and VAE
  15. Discussion
  16. ...and 1 more sections

Figures (4)

  • Figure 1: A Block Diagram of the proposed System Architecture.
  • Figure 2: The Data Pre-processing and Feature Extraction Workflow Derived from Empatica E4 Wristband Signals.
  • Figure 3: Proposed System Architecture to Classify AA in PwD using VAE and Self-training.
  • Figure 4: AUC ROC and Precision-Recall Curve for XGBoost Classifier using VAE and Semi-Supervised Learning.