Brain tau network derived biomarker improves the diagnostic accuracy of Alzheimer's disease.

BACKGROUND

To develop and validate brain tau network based biomarker for the early diagnosis and prognostication of Alzheimer's disease (AD). A total of 456 subjects, including 226 healthy controls (HC), 166 mild cognitive impairment (MCI), and 64 AD from the Alzheimer's Disease Neuroimaging Initiative (ADNI), were included in this study. Preprocessed 18F-AV1451 PET images and clinical assessments were used. Regional standardized uptake value ratios (SUVR) relative to the cerebellum were calculated using partial volume corrected PET images. For graphical analysis of brain network at individual level, thirty regional 18F-AV1451 SUVRs sampled for whole brain volume were used to compute network characteristics, including connectivity, nodal and global efficiency. Group differences in network characteristics were examined and four classification models based on different predictors were constructed. Cross-sectional and longitudinal verification were performed to assess the efficiency of the network characteristics.

RESULTS

The cross-sectional and longitudinal results showed decreasing trends in connectivity, nodal and global efficiency from HC to MCI and AD. The results of the network-based model significantly enhanced the classification accuracy compared to the model only using SUVR as predictors. Notably, the apolipoprotein E4 (apoE4) positive participants exhibited reduced connectivity, nodal efficiency, and global efficiency compared to the negative.

CONCLUSIONS

The network characteristics calculated at the individual level, particularly the global efficiency of the brain tau network, are quantitative biomarkers for monitoring AD progression and evaluating therapeutic responses.