Premium
[P4–249]: TAU BURDEN CORRELATES WITH LONGITUDINAL STRUCTURAL CHANGE IN REGIONS AFFECTED BY EARLIEST NEUROFIBRILLARY TANGLE PATHOLOGY IN AMYLOID‐β‐POSITIVE ADNI SUBJECTS
Author(s) -
Xie Long,
Das Sandhitsu R.,
Wisse Laura E.M.,
Ittyerah Ranjit,
Yushkevich Paul A.,
Wolk David A.
Publication year - 2017
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2017.06.2117
Subject(s) - entorhinal cortex , neurofibrillary tangle , atrophy , temporal lobe , alzheimer's disease , nuclear medicine , correlation , tangle , pathology , psychology , medicine , hippocampus , neuroscience , disease , senile plaques , epilepsy , mathematics , pure mathematics , geometry
Figure 1. Scatter plots of age-adjusted AV1451 uptake in right medial temporal lobe ROI vs. annual longitudinal thickness change in right BA35 (a). Their correlation (b) remains highly significant after removing the two outliers in (a). Positive change in thickness indicates cortical thinning. Background:During the last years clinical multi-institutional studies are becoming more popular, since they offer neuroscientists the possibility of pooling larger datasets at the cost of strict data standardization and quality assessment rules. This work presents a set of rules and strategies, developed for the DELCODE study, designed to ensure data homogeneity across sites. DELCODE stands for: DZNE longitudinal cognitive impairment and dementia study. Starting in 2014, it is conducted over ten sites across Germany and currently contains about 1.350 scanning sessions from about 825 subjects. It aims to scan 1.000 subjects annually over a period of five years to establish the progression of dementia in the pre-clinical phase. Methods: Based on standard operating procedures designed for the study, every scan undergoes a quality check. Acquisition parameters, sequenceand protocol completeness are assessed automatically. Artifacts and coverage of the brain are evaluated manually, spatial orientation is inspected semi-automatically. The results are used as feedback in each center to minimize errors. Data is classified into three categories: A) Complete datasets without deviations from the protocol, B) deviations from the protocol but complete acquisition of scans e.g. suboptimal FoV positioning; and C) incomplete datasets e.g. only structural scans. Results: In 2014, of 84 scans, 52,4% were of type A, 27,4% were of type B and 20,2% were classified as type C. In 2015, of 562 scans, 66,2% were type A, 17,3% were type B and 16,5% were type C. In 2016, of 683 scans, 68,8% were type A, 18,2% were type B and 13% were type C. This shows an increase of quality in data during the first year of feedback, followed by a steady quality assurance in the passing years with increased data acquisition. Conclusions:We were able to establish a harmonized scanning protocol for a multi-center study, assuring a steady monitoring of MRI data quality. Optimization is an ongoing process. The numbers above prove the usefulness of the proposed QA rules and effectiveness of regular feedback to sites. In the future, the currently manual/ semi-automatic quality check information will be used to train an algorithm to assess the image quality automatically.