Remote Sensing for ADRD-Specific Activities Identification in Older Adults
试验速览
- 阶段
- 不适用
- 状态
- 招募中
- 入组人数
- 16
- 试验地点
- 1
- 主要终点
- Classification accuracy of ambient sensor-based daily activity models
研究概览
简要总结
The investigators aim to use smart-home sensors and artificial intelligence (AI) to monitor and detect Alzheimer's Disease and Related Dementias (ADRD)-specific daily activities among older adults, with the goal of early symptom detection and personalized support. Dementia, which impacts memory and cognition, remains a global concern. In the United States, more than 6.7 million individuals aged 65 and older are living with ADRD, and projected annual healthcare costs are expected to reach $1 trillion by 2050. This underscores the need for deeper understanding and innovative support. To address the unique challenges associated with ADRD, such as cognitive decline, personalized strategies that promote independent well-being are essential. Smart-home sensors can support older adults with ADRD as they continue to live in their homes. These sensors provide real-time data on health and daily activities, offering insights into their daily lives. However, adoption of these technologies is low, and the practical application of AI remains limited. This highlights the need for further research to make these devices more accessible to this population. The investigators' aims include:
Conducting focus groups with individuals with and without ADRD and their caregivers to identify daily activities that can be measured using in-home sensors; Collecting in-home sensor data from older adults with and without ADRD; and Using AI to develop a tool for recognizing daily activities. The integration of smart-home sensors with advanced data-analysis techniques holds significant potential for transforming the support and care provided to individuals with ADRD. Ultimately, the investigators' findings will contribute to improving the quality of life for affected individuals and alleviating the burden on caregivers and healthcare systems.
详细描述
Dementia, which impacts memory and cognitive abilities, constitutes a global concern that intensifies with the aging population. In the United States, 6.7 million individuals aged 65 and older live with Alzheimer's Disease and Related Dementias (ADRD), and projected annual healthcare costs are expected to reach $1 trillion by 2050. This underscores the urgent need for enhanced understanding and innovative support. Individuals with ADRD face unique challenges, including behavioral changes and cognitive decline, necessitating tailored strategies for their well-being. Aligning with the National Institute on Aging's research goals, the investigators' study explores a promising avenue: the use of smart-home sensors to monitor and assist ADRD patients while they reside in their homes. These sensors provide real-time insights into health, activity, and environmental factors. However, adoption of these technologies among people with ADRD is low, and the practical application of artificial intelligence (AI) in this context remains limited. This underscores the need for further research to make these devices more accessible to this population.
The investigators aim to utilize a fully modular smart-home sensor system, combined with AI-based data-analysis methods, to monitor and analyze activities specific to individuals with ADRD. Remote sensor installations have been deployed across Missouri to facilitate the seamless delivery of sensor data to the investigators' interdisciplinary team, known as the Age-friendly Smart, Sustainable, and Equitable Technologies for Access intervention research team. The investigators' approach involves applying AI with causal inference to gain a nuanced understanding of the daily activities and behavioral patterns of those with ADRD. The investigators hypothesize that incorporating modeled causal features into the AI process will 1) enable identification of ADRD-specific daily activities, and 2) enhance the AI's ability to recognize these activities.
The investigators' aims include:
Conducting focus groups with individuals with and without ADRD and their caregivers to identify daily activities measurable with in-home sensors; Collecting in-home sensor data from older adults with and without ADRD; and Developing an AI system using machine-learning (ML) models for ADRD-specific daily activity recognition. Aim 3 will encompass three key elements: identification of causal features associated with ADRD-specific daily activities, development and refinement of ML models for recognizing these activities informed by the causal features, and creation of personalized ML models for individuals diagnosed with ADRD.
The integration of smart-home sensors with advanced data-analysis techniques holds significant potential for transforming the support and care provided to individuals with ADRD. Ultimately, the investigators' findings will contribute to improving the quality of life for affected individuals and alleviating the burden on caregivers and healthcare systems.
研究设计
- 研究类型
- Interventional
- 分配方式
- Non Randomized
- 干预模型
- Parallel
- 主要目的
- Supportive Care
- 盲法
- Single (Participant)
入排标准
- 年龄范围
- 50 Years 至 —（Adult, Older Adult）
- 性别
- All
- 接受健康志愿者
- 是
入选标准
- •Community-dwelling, English-speaking adults aged ≥ 50 years
- •Clinical diagnosis of mild cognitive impairment or mild dementia (ADRD)
- •Diagnosis established by a neuropsychologist, neurologist, or geriatrician within the University of Missouri Healthcare System
- •Diagnosis confirmed using the latest consensus criteria and verified through record review
- •No restriction on the etiology of the cognitive disorder (e.g., Alzheimer's disease, vascular dementia, mixed dementia)
排除标准
- •Clinical Dementia Rating (CDR) global score > 1 (moderate or severe dementia)
- •Cognitive or functional impairments that would preclude meaningful participation in daily activities
研究组 & 干预措施
Participants w/o ADRD
干预措施: Remote Ambient Sensor System (Other)
Pariticpants w/ ADRD
干预措施: Remote Ambient Sensor System (Other)
结局指标
主要结局
Classification accuracy of ambient sensor-based daily activity models
时间窗: Weeks 1-4 after enrollment
Percent of daily activity labels correctly predicted by SVM, XGBoost, LSTM and Transformer models, trained and tested on ambient motion and environmental sensor data collected during weeks 1-4 from participants with and without early-stage ADRD.
F1 score of ambient sensor-based daily activity models
时间窗: Weeks 1-4 after enrollment
Harmonic mean of precision and recall for SVM, XGBoost, LSTM and Transformer models, evaluated on held-out portions of the week 1-4 ambient sensor data.
Area under the ROC curve of ambient sensor-based daily activity models
时间窗: Weeks 1-4 after enrollment
AUC of ROC curves for SVM, XGBoost, LSTM and Transformer models distinguishing among daily activity classes, based on training and testing using weeks 1-4 ambient sensor readings.
次要结局
未报告次要终点
研究者
Knoo Lee
Ast Professor
University of Missouri-Columbia