Fluorescence spectroscopy and machine learning methods for detection of Alzheimer's disease from circulating white blood cells.

BackgroundAlzheimer's disease (AD) is the most common cause of dementia whose prevalence is projected to increase significantly in the coming decades. The recent advent of disease modifying therapies is a welcome development; however, it is also now apparent that early treatment maximizes the benefits of these drugs. Therefore, it is important to develop reliable methods of disease detection, preferably from an easily accessible matrix such as blood.ObjectiveTo develop a method for detecting AD from circulating white blood cells using spectral confocal microscopy.MethodsUsing K114-stained wild type and 5xFAD transgenic mouse cortical sections as proof-of-principle, spectral imaging of K114 fluorescence coupled with a signal processing/machine learning pipeline (spectral wavelet decomposition, dimensionality reduction, support vector machine classifier) can reliably distinguish non-plaque background parenchyma in the two strains. We then performed immunoprecipitation of Aβ from peripheral blood mononuclear cells (PBMCs) obtained from non-neurological controls and histopathologically-proven AD cases. We spectrally imaged the immunobeads labeled with K114, then used similar machine learning methods to classify control versus AD samples.ResultsNormal-appearing non-plaque 5xFAD background was reliably distinguished from wild type mouse brain. We could also classify AD with a high degree of reliability (area under the receiver operating curve = 0.95, p = 6.1e-5) and predict neuropathological scores from these blood elements (R = 0.89).ConclusionsOur spectral imaging method, together with automated machine learning analysis of spectral micrographs, using readily obtainable PBMCs from blood, represents a potentially useful approach for detection of AD in living subjects.