Reliability and modeling of digital histological measurements in Alzheimer's disease neuropathologic change and Lewy body disease.

Abstract

Digital histology offers a more objective, continuous definition of neuropathological severity than traditional staging systems, but its reliability remains underexplored. We calculated regional percentage areas occupied by phosphorylated tau (p-Tau, AT8), amyloid-β (Aβ, NAB228), and phosphorylated α-synuclein (p-αSyn, 81A) pathology in 24 autopsied cases with varying degrees of Alzheimer disease neuropathological change and Lewy body disease (LBD) using manual and automated immunostaining methods to investigate variability across protocols. We then compared natural log-transformed percent area occupied values (ln%AO) to blinded ordinal severity scores, Braak stages, Thal phases, and McKeith LBD stages. p-Tau ln%AO from methodologically similar runs had the highest correlations (R2 = 0.91-0.95, β  =  0.95-0.97 for manual and automated methods, respectively); p-αSyn ln%AO from disparate immunostaining methods had the lowest (R2 = 0.16-0.34 β  =  0.40-0.59). p-Tau and Aβ ln%AO increased regionally with higher Braak and Thal stages (p-Tau: z = 2.06 P = .04. Aβ: z = 3.70 P < .001). Regional p-αSyn ln%AO increased from limbic to neocortical stages (z = 5.86 P < .001); amygdala-predominant type LBD cases peaked in the amygdala and dropped in other limbic regions. These findings show the potential to quantify differences in p-Tau, Aβ, and p-αSyn pathologies using digital histological methods in single-center studies.