From diagnosis to intervention: EEG and MEG biomarkers in Alzheimer's disease

This thesis aimed to highlight the value of electroencephalography (EEG) and magnetoencephalography (MEG) biomarkers in diagnosing Alzheimer's disease (AD), advancing our understanding of its underlying mechanisms, and evaluating treatment effects on neuronal function in AD clinical trials. To this end, we investigated whether MEG biomarkers could serve as a supportive diagnostic tool in the memory clinic, explored the relationship between MEG biomarkers, amyloid-beta (Aβ) pathology and the excitation-inhibition (E-I) balance, and evaluated which EEG biomarkers are most sensitive to change over time. Additionally, this thesis included studies that contribute to the advancement of methods for analyzing EEG and MEG recordings. Our results indicated that a machine learning model trained on MEG measures of oscillatory power and functional connectivity holds promise to support diagnostic decision-making in the memory clinic, particularly for patients with dementia with Lewy bodies or dementia due to AD. This raised the question of whether even more sensitive markers of neuronal dysfunction could be derived from MEG recordings of AD patients by examining alternative neural characteristics (e.g. variability). We therefore introduced a novel MEG-based functional connectivity measure: the joint permutation entropy (JPE). The JPE was shown to be a sensitive biomarker for early-stage AD, with classification performance (mild cognitive impairment [MCI] due to AD versus controls) at least as good as that of relative theta power, an established biomarker for AD. In a replication study, we showed that the JPE, as well as another recently introduced MEG biomarker called the phase lag time (PLT), are reliable measures of functional connectivity that are sensitive to neuronal dysfunction in patients with dementia due to AD. The novel biomarkers performed as well as, or better than, the previously validated corrected amplitude envelope correlation (AEC-c) and phase lag index (PLI) in terms of sensitivity to group differences, replicability of findings, and classification performance (dementia due to AD versus controls). Given that the JPE was designed to assess rapid fluctuations in neuronal activity and connectivity, we hypothesized that the measure could serve as a surrogate marker for E-I balance. We have provided evidence for this hypothesis through two complementary approaches: (1) a computational modeling study, in which we showed that MEG measures of local signal variability and functional connectivity are sensitive to changes in the E-I ratio in the model; and (2) a longitudinal MEG-PET study in cognitively unimpaired adults, in which we found that disruptions in MEG oscillatory power and functional connectivity are associated with the emerging deposition of Aβ in this population. The results of the final study in this thesis indicate that EEG measures of oscillatory power, particularly recorded from temporal regions, are most sensitive to change over time. The disease stage of the patients under investigation influenced the rates of change of the EEG biomarkers, with patients with MCI due to AD exhibiting faster progression than patients with dementia due to AD. The reported effect sizes indicate that EEG biomarkers have potential to reduce the size, duration and thereby costs of AD clinical trials. In conclusion, this thesis provides support for the hypothesis that MEG biomarkers offer a non-invasive way to detect pathology-induced changes in neuronal excitability and E-I balance over the course of AD. EEG and MEG biomarkers furthermore hold significant potential to benefit patients, with key applications in diagnosis and the evaluation of treatment effects. With the growing interest in activity-targeting interventions and the introduction of OPM-MEG, a novel and highly sensitive technique for measuring large-scale neuronal activity, we anticipate that additional applications of EEG and MEG biomarkers in AD will become evident in the future.