Background: Amyloid beta (Abeta) and Tau, the proteins thought to cause Alzheimer's disease (AD), can be readily measured in cerebrospinal fluid (CSF) and are promising AD laboratory biomarkers. Hippocampal atrophy is the most established AD imaging biomarker. Little is known about the relationship between CSF Abeta and Tau and hippocampal atrophy. Methods: We used an automated machine-learning algorithm, based on adaptive boosting, to segment 3D surface models of the hippocampi in baseline 3D T1-weighted brain MRI scans of 282 ADNI subjects (83 cognitively normal (NC), 140 mild cognitive impairment (MCI) and 59 AD) with baseline CSF Abeta measurements. Following 3D radial distance atrophy mapping, we applied linear regression at each hippocampal surface point to reveal associations between the structural and laboratory AD biomarkers in 3D. We then used false discovery rate based cumulative distribution plots (CDF) to rank the strength of the association between each CSF biomaker and hippocampal atrophy. Results: The CDF plot (Figure 1, top left) shows that Tau/Abeta ratio (in yellow) has the strongest association with hippocampal atrophy, followed by phosphorylated Tau (pTau)/Abeta ratio, Abeta, Tau and finally pTau. The 3D hippocampal significance maps (Figure 1, top right) show the areas with significant associations between Tau/Abeta ratio and hippocampal radial distance. The corrected p-values for the significance maps associating each CSF measure with hippocampal radial distance are shown in the Table (Figure 1, bottom). Conclusions: We demonstrated the hypothesized associations between hippocampal atrophy, the most established structural imaging AD biomarker, and several established CSF AD biomarkers. Strongest correlations were observed for Abeta, Tau/Abeta and pTau/Abeta ratios. Combinations of biomarkers, or biosignatures, such as the ones suggested by our data, may show even greater power for preclinical diagnosis, prognosis and as biomarkers for clinical trials.
FULL TEXT
