High density EEG biomarkers of sporadic and familial neurodegeneration: A comparison with cognitive and neuroimaging markers in multicultural samples

The worldwide growth of neurodegenerative diseases (NDs)1-5 calls for an urgent development of sensitive markers allowing the identification and differentiation of various NDs. In this sense, assessment of high density EEG cently emerged as a highly promising framework for establishing both transdiagnostic and disease-specific markers. This approach stands out as it consists in highly affordable, non-fatiguing, non-invasive measures which can reveal early deficits traceable to well-established neurophysiological process affected across NDs. Moreover, across samples with mild variants, multi-ethnic profiles, or low education, the combination of EEG and sensitive tasks can capture key neurocognitive deficits that usually escape the cognitive screenings used in international harmonized initiatives from national institutes like NIH-NIA. Our team have a strong expertise in EEG markers, whichoffers different sets of measures including ERPs, oscillations, connectivity measures, source space analysis, decoding and machine learning approaches, for both active tasks and resting state recordings6-39. Such markers may be critical for early discrimination and characterization of Alzheimer¿s disease (AD), and frontotemporal dementia (FTD). Of note, these markers might hold great relevance in Latin-American health institutions, most of which lack neuroimaging and genetic expertise, so that ND diagnosis relies predominantly on coarse-grained, unspecific clinical tests. Crucially, to prevent advocating sub-standard alternatives for future clinical trials, EEG markers need to prove at least as sensitive as other gold-standard markers across NDs. Against this background, our long-term goal is to establish the sensitivity of multimodal EEG markers vis-à-vis cognitive (standardized coarse-grained cognitive tests), and neuroimaging ones (i.e., structural MRI, functional connectivity derived from fMRI) for discriminating among NDs subtypes (Ad vd FTD), status (familial vs sporadic) and predicting disease progression. We will implement a multi-feature analytical framework to characterize and predict the subtype, status, and clinical severity of ND conditions via a novel multimodal electrophysiological markers (derived from EEG) neuroimaging pipeline (high density EEG, MRI and fMRI) and a set of sensitive behavioral (cognitive) measures. We aim to identify the optimal set of features affording accurate characterization of each patient¿s condition (AD, FTD); its familial presentation (sporadic vs familial) and disease severity. To this end, we will establish a cohort totaling 210 participants (with 80 controls, 30 sporadic AD patients, 30 familial AD patients, 30 sporadic FTD patients, 30 familial FTD patients). In previous ND reports40-47, we found robust effects with smaller samples using MRI48-51and hdEEG22,41,42,46,52 alone or in combination10,41,42,46,53-55, joint with classification methods12,40,56-58, and even when considering four groups of patient and controls41,42. We will couple clinical assessments and sensitive cognitive tasks with innovative analytical techniques to account for heterogeneity in these populations. With this framework, we will test the postulate that the characterization of ND subtype, status, and disease severity will be optimized through the combination of EEG with multiple features obtained from both sensitive tasks and multimodal imaging. We will adopt a three-fold approach, including EEG-focused studies (to establish the most sensitive EEG features), comparative studies (to test the robustness of those features relative to clinical and neuroimaging ones), and multidimensional studies (to examine whether differentiation among NDs in terms of subtype, status and severity prediction is boosted when EEGfeatures are included in multidimensional classifiers involving clinical,neuroimaging, and socio-environmental data). Additionally, we will factoring in the role of socio-environmental determinants (e.g., socioeconomic status [SES and social determinants of health [SDH). Through a combination of inferential statistics and machine-learning analyses (e.g., random forest, support-vector machines, progressive feature elimination), we aim to detect the optimal set of EEGfeatures affording accurate discrimination of ND subtype (AD, FTD), familial status (sporadic vs. familial) and severity. Moreover, we will tune our methodological approaches with datasets already obtained by our team. We will profit from our ongoing international project to apply our protocol in large relevant samples. First, our study will benefit from a strategic integration with the project ¿Multi-partner consortium to expand dementia research in Latin America¿ (ReDLat, multi-funds from NIH-NIA-R01 AG057234 / Alzheimer¿s Association / Tau Consortium / Global Brain Health Institute) to recruit participants with AD, FTD, and controls from 10 sites in Argentina, Brazil, Chile, Colombia, Mexico, Peru, and the US(but only a subset of 4 selected sites will be included in this application). Our multi-partner grant will cover the main costs for capacity building, diagnosis, genetic studies and MRI recordings. With these multicultural samples, we will test the underlying hypothesis that, in discriminating among NDs and tracking disease severity and familial status, potential EEG markers will prove at least as sensitive as cognitive and neuroimaging markers, irrespective of the patients¿ SES-SDH. Moreover, we will examine the complementary hypothesis that overall prediction of disease subtype, status and severity, based on clinical, imaging, and cognitive data, will be boosted by the inclusion of EEGmarkers.In testing both hypotheses, multimodal measures from controls will be used for disease-specific normalization of patient data and between-group comparisons. Furthermore, the use of data-driven machine-learning analysis will offer a cutting-edge platform to tackle the many sources of heterogeneity across culturally diverse samples. Crucially, ensuing results will be contrasted, validated, and harmonized against multicenter databases.