Linking Multidimensional Sleep Health to Cognitive Function in Older Adults Using Machine Learning
PI: Sayed Reza (Computer Science)
Dr. Reza plans to fund graduate student tuition through discussions with the School Director, SSET at Penn State Harrisburg. Additionally, an undergraduate student currently supported by Dr. Reza’s CRS grant will be contributing to the project in the Summer 2025, focusing on data preprocessing, model development, and interpretation of results. Dr. Roberts is supported 75% on NIH grants and 25% on discretionary funds of Dr. Buxton, with multiple grants pending funding or revie
Dr. Sayed Reza(PI), Dr. Daniel Roberts(Co-PI), Dr. Orfeu Buxton(Mentor)
A short statement (1 sentence/1 paragraph) explains the connection of the project to ICDS’s mission: This project aligns with ICDS’s mission by integrating advanced computational and data science methods(interpretable AI and time series analysis). The methods will be applicable to the domain of sleep and cognitive health, allowing interdisciplinary collaboration to address scientific and societal challenges related to healthy aging and cognitive decline.
Brief (1 page) description of the proposed project
This project will evaluate the relationship between sleep health and cognitive function in older adults by leveraging wearable device time series data and applying interpretable AI/ML techniques. Building on prior work[1], we will examine how multidimensional sleep metrics, spanning ~50 measures, relate to cognitive outcomes such as clinically assessed Mild Cognitive Impairment (MCI) and global cognition assessed via the Montreal Cognitive Assessment (MoCA), and momentary cognitive performance using Ecological Momentary Assessments on a smartphone implementation of the M2C2 (a NIH mobile cognitive assessment toolkit).
Objectives: The project’s primary objective is to apply observable AI/ML models to identify interpretable patterns linking sleep health to cognitive status using a large, diverse cohort of older adults. An additional objective of the project is to evaluate new ML classification approaches for sleep-wake classification[2].
Methodology: Sleep metrics (e.g. duration, timing, disruption) are extracted and aligned with next-day cognitive performance. The project will use sequential pattern analysis for mining the interpretable patterns linking sleep health to cognitive status. We will use AI/ML models (Unet, EfficientNet) to classify sleep-wake in wearable sensor data. Also, the project uses SHAP for explainable AI to ensure transparency and interpretability of findings on the relationship between sleep variables to cognitive performance.
Significance: The project will create new knowledge or understanding of how mean and daily variations in sleep affect clinical cognition measures and daily variation in cognitive performance. The result can help identify early markers of cognitive decline and inform personalized interventions. This project offers a unique opportunity to work at the intersection of aging, sleep science, and explainable AI.
Opportunities for Junior researcher: Dr. Sayed Reza directs the Data Driven Decision Making Lab (https://www.smreza.com/profile/d3m-lab/). A student in his lab will contribute to data preprocessing, model development, and interpretation of results to gain hands-on experience with wearable time series analysis and interpretable machine learning, while contributing to a high-impact, interdisciplinary project.
List of specific areas of computational and/or data science skills that you are particularly interested in recruiting: Time series analysis, machine learning, data integration, Python programming, and high-performance computing.
Any specific requirements or expectations of potential ICDS Junior Researchers: Potential ICDS Junior Researchers are expected to have strong analytical skills, a collaborative mindset, and a willingness to engage with interdisciplinary teams at the intersection of data science, aging, and health research.
List of specific objectives for work supported by this call: Here’s a list of specific objectives for work supported by this ICDS proposal call, tailored to the project:
1. Develop interpretable machine learning to uncover actionable patterns between multidimensional sleep health and cognitive function in older adults.
2. Develop scalable pipelines for integrating and analyzing multimodal data (e.g., actigraphy, EMA, MoCA) with support from the ICDS computational infrastructure and consulting.
3. Promote interdisciplinary collaboration between data scientists and health researchers, aligning with ICDS’s mission to continue data-driven discovery.
At least one medium to long-term goal: A medium to long-term goal of this project is to characterize sleep patterns as a non-invasive biomarker for early detection of Alzheimer’s and related dementias. Results are expected to lead to competitive external funding proposals (NIH; foundations) for observational or intervention-focused projects.
A paragraph summarizing the team member’s recent and/or planned engagement with ICDS: Our team has recently started collaborating with ICDS through consultations on AI methods. We plan to engage by connecting ICDS’s technical expertise and infrastructure to support scalable analysis and interdisciplinary collaboration throughout the project.
Collaborator Information
1. Dr. Sayed Reza, PhD (PI) Assistant Professor of Computer Science Penn State University Harrisburg
2. Dr. Daniel Roberts, PhD(Co-PI) Research Assistant Professor Pennsylvania State University
3. Orfeu M. Buxton, PhD(Mentor) Elizabeth Fenton Susman Professor of Biobehavioral Health Co-funded faculty, Associate Director: Social Science Research Institute (SSRI) Associate Director: Clinical and Translational Science Institute (CTSI) Director: Sleep, Health & Society Collaboratory Pennsylvania State University 2025 Pattishall Outstanding Research Achievement Award Editor in Chief, Sleep Health