allows a quantitative (and qualitative) assessment of gross and subtle structural brain changes that might be more sensitive than cognitive measures for predicting POCD. In the present study, we expect that pre-operative atrophy in several specific brain regions would be associated with POD/POCD risk.

is a newly established magnetic resonance imaging method which allows directly imaging of the brain blood flow without the use of a contrast agent. ASL can easily be combined with conventional brain MRI to show the relationship between the cerebral perfusion territories, vascular anatomy/brain infarcts and other pathology. Since vascular hypoperfusion, is known to be associated with POD/POCD, hypoperfusion as measured with ASL would be expected in critical brain regions possibly before structural changes emerge.

During the past few years, technological advances have made it possible to conduct simultaneous fMRI/EEG experiments in the MR-scanner. Simultaneous fMRI/EEG has now evolved into a mature, generally accessible and in principle easily applicable technique. It is expected that pre-operative EEG-informed fMRI of critical brain regions will predict POD/POCD with high sensitivity,. Post-operative EEG-informed fMRI changes (compared to pre-operative EEG/fMRI) are thus anticipated to predict POCD with very high sensitivity.

No MRS studies have yet been conducted to probe the question whether glutamate and/or GABA concentrations are changed in POD/POCD. However, there are several lines of evidence suggesting that this might be the case.

No Gadolinium MRI studies have yet been conducted to probe the question whether there is blood-brain barrier (BBB) impairement in surgical patients who subsequently develop POD/POCD, or whether surgical interventions per se lead to impairment of the BBB , which in turn then leads to POD/POCD via (aseptic) postoperative inflammation of the brain.

Imaging will be conducted in all patients pre- and postoperatively (3-Tesla). Spectroscopy and Gadolinium MRI will be limited to a subset of patients (7-Tesla).

On the basis of available literature, a number of specific plasma markers can be tested for their ability to predict POD/POCD. Candidates include inflammatory and metabolic plasma proteins/molecules, cholinergic markers, markers associated with AD plaque formation and atherosclerosis, oxidative stress markers and inflammatory signatures as well as markers that indicate impaired blood-brain barrier (BBB) function.

Compared to plasma markers, cerebrospinal fluid (CSF-)based markers reflect brain pathology more closely as demonstrated in studies on AD. CSF (obtained during spinal anesthesia) is therefore ideal for validating possible plasma markers thought to predict POD/POCD (and dementia). For validation purposes, we will also collect post mortem brain tissue in a limited study subgroup.

Major cholinergic-nicotinic genes and mRNA, with their association with cognitive performance, are obvious candidates to study. In part, evidence for this has been obtained within the framework of our DFG-funded national priority program on the central effects of nicotine in CNS (www.nicotine-research.de). MiRNA will be screened using array-based technology.

In recent years, multivariate analysis algorithms have become a popular tool to diagnose neurological or psychiatric conditions based on structural or functional imaging data. The main challenge is the identification of features, which provide most information about the particular condition (so-called ‘disease signatures’). Features used in previous neuroimaging studies include local or global intensity patterns as well as geometric and surface-based features. An additional challenge is the integration of multiple measures such as neuroimaging and molecular biomarker to achieve greater power in diagnosis and prognosis. We will combine multiple features derived from brain imaging and other biomarkers.

are mechanistic molecular network-based modeling, useful for better understanding/interpreting particular molecular processes and their interactions (e.g. cholinergic-inflammatory response interaction). The complexity of molecular mechanisms underlying POD/POCD will be tackled with network-based approaches, which integrate analysis of many processes and pathways. These models would allow predictions about outcome (clinical, abnormal brain structure and function) and will be validated on the basis of the neuroimaging and biomarker data. A knowledge map will be constructed, based on iterative processes that elaborates and combine known relations between proteins or genes in question. This will be complemented by information from other public databases. The data generated by the consortium will constantly improve the knowledge map. The LCSB has developed large-scale text-mining pipelines based on full-text articles and is able to build semi-automatic knowledge maps and that will be the starting point of this task.