# Prediction of dementia in older adults using nonlinear EEG features

> **NIH NIH K99** · MASSACHUSETTS GENERAL HOSPITAL · 2024 · $134,730

## Abstract

PROJECT SUMMARY
Early prediction of dementia can provide a unique opportunity for interventions to prevent or delay the onset of
dementia. There is evidence that alterations in many physiological functions precede the clinical onset of
Alzheimer’s disease and related dementia. The hypothesis behind this proposal is that brain activities assessed
by electroencephalogram (EEG) signals can offer prognostic value for persons at risk for dementia. The goal of
this project is to develop EEG biomarkers (EDBs) to predict future risk of dementia and monitor the
progress of cognitive decline. To achieve the goal, the PI and his team propose to investigate/assess the
complex EEG fluctuations using both traditional and novel methods derived from nonlinear dynamics theory.
Deep learning (DL) — the recent advance in the field of artificial intelligence — will be used to extract EEG
features from multiple EEG measures and develop EDBs for the risk for dementia. Using the existing datasets
of four cohorts in the National Sleep Research Resource (NSRR) that together provide >8,000 older adults with
assessments of overnight EEG, cognitive status and genetic data, four aims will be addressed. In Aim 1, deep
convolutional neuron network models will be built to construct EDBs using EEG measures at different wake/sleep
stages together (Aim 1A) or separately (Aim 1B). The effects of sex and ethnicity on EDBs and their performance
in predicting dementia will also be examined (Aim 1C). In Aim 2, the longitudinal changes in EDBs and their
associations with longitudinal change in cognition will be determined. In Aim 3, genetic factors related to the risk
for Alzheimer’s disease (Aim 3A, 3B) and sleep/circadian (Aim 3C) will be combined with derived EDBs to
improve the prediction of incident dementia. In Exploratory Aim 4, the links of EDBs to non-cognitive functional
outcomes such as sleep/circadian disturbances and risk for disability will be explored. Aim 1 will be completed
in the mentored phase and Aims 2-4 will be completed in the independent phase of this award. This research
project may provide a cost-efficient, non-invasive tool for the early prediction and monitoring of dementia as well
as for the evaluation of treatments of the disease. The project will be performed at Brigham & Women’s Hospital
and Harvard Medical School. The PI is well suited to accomplish this interdisciplinary research project with his
established expertise in biomedical engineering, data science, and biostatistics as well as the strong
mentoring/advisory team that will help the PI to receive advanced training in deep learning, human clinical
studies, and sleep/circadian physiology, gain new expertise in dementia and aging research, and learn about
genetic analysis. Fulfilling the proposed research and training activities will prepare the PI for launching his own
independent research program in the interdisciplinary field of data science and remote medicine.

## Key facts

- **NIH application ID:** 10807562
- **Project number:** 1K99AG083234-01A1
- **Recipient organization:** MASSACHUSETTS GENERAL HOSPITAL
- **Principal Investigator:** Shahab Haghayegh
- **Activity code:** K99 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $134,730
- **Award type:** 1
- **Project period:** 2024-07-15 → 2026-06-30

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10807562

## Citation

> US National Institutes of Health, RePORTER application 10807562, Prediction of dementia in older adults using nonlinear EEG features (1K99AG083234-01A1). Retrieved via AI Analytics 2026-05-25 from https://api.ai-analytics.org/grant/nih/10807562. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*