Research Project
10301875
PROJECT SUMMARY/ABSTRACT Presently, no established biomarker exists to robustly predict the clinical manifestation of cognitive symptoms in persons with Alzheimer?s disease (AD) and AD-related dementias. PET and MRI brain biomarkers are costly and invasive, thus there is a critical need for a noninvasive, inexpensive, and portable AD screening tool that can be easily deployed in-home or in residential communities. While non-brain signals characterizing biological and behavioral traits may prove valuable, two types of brain signal also hold strong promise as digital biomarkers of early stages of AD: epileptogenic activity (EA), and aberrant functional brain networks. Importantly, both biomarkers can be collected affordably and reliably with the latest dry-electrode ambulatory electroencephalography (EEG) technology. AD patients have a tenfold higher seizure prevalence compared to the general population (Pandis and Scarmeas, 2012); however, the use of EA as an AD digital biomarker is largely unexplored. It is also well known that amnestic MCI (aMCI) and AD patients show subtle functional network disruptions that are promising predictors of AD, as shown by our group (e.g Pusil et al., 2019) and others, but there is no previous research assessing the joint impact between EA and functional networks. The scientific premise of this proposal is two-fold: (i) a combinatorial EA and functional network biomarker will predict conversion from aMCI to AD more robustly than a single signal in isolation, and (ii) a novel deep learning model that performs multimodal (MEG and EEG) learning to find shared signatures of AD, but ultimately yields a model that needs affordable EEG-only data, will yield a powerful biomarker. This proposal will pursue three specific aims. 1) Identify specific features of EA that prognosticate aMCI conversion; 2) Design a digital biomarker that predicts aMCI conversion from EA features and functional brain networks; 3) Extend the digital biomarker to ambulatory EEG with dry electrode technology. To achieve these aims, we will collect MEG, wet-electrode EEG, and dry-electrode (ambulatory) EEG data from 200 aMCI patients, evaluate their signals with expert epileptologists, and monitor the patient?s yearly conversion rate to AD. We will then design and validate a deep learning model called Siamese Multiple Graph to Gauss (SMG2G), which performs multimodal learning on MEG and EEG network (graph) data but ultimately yields a model that needs EEG-only data to make predictions of aMCI conversion. The final product will be a dry-electrode ambulatory EEG digital biomarker that can be readily measured in home or in a residential facility. The research proposed in this application is innovative because it is the first to combine EA and functional network signals to design an AD biomarker and achieves this by cutting- edge machine learning. It is also significant because it will advance the field vertically both scientifically and clinically by enabling large-scale, early detection of AD. Our team is especially well-prepared to undertake this project, with clinical and engineering expertise, strong collaboration over the years, with preliminary data supporting the aims, and institutional support.
