The main goal of our studies has been to use MRI, FDG-PET, and CSF biomarkers to identify in cognitively normal elderly (NL) subjects and in patients with mild cognitive impairment (MCI), the earliest clinically detectable evidence for brain changes due to Alzheimer's disease (AD). A second goal has been to describe the cross-sectional and longitudinal interrelationships amongst anatomical, CSF and cognition measures in these patient groups. It is now well known that MRI-determined hippocampal atrophy predicts the conversion from MCI to AD. In our summarized studies, we show that the conversion of NL subjects to MCI can also be predicted by reduced entorhinal cortex (EC) glucose metabolism, and by the rate of medial temporal lobe atrophy as determined by a semi-automated regional boundary shift analysis (BSA-R). However, whilst atrophy rates are predictive under research conditions, they are not specific for AD and cannot be used as primary evidence for AD. Consequently, we will also review our effort to improve the diagnostic specificity by evaluating the use of CSF biomarkers and to evaluate their performance in combination with neuroimaging. Neuropathology studies of normal ageing and MCI identify the hippocampal formation as an early locus of neuronal damage, tau protein pathology, elevated isoprostane levels, and deposition of amyloid beta 1-42 (Aβ42). Many CSF studies of MCI and AD report elevated T-tau levels (a marker of neuronal damage) and reduced Aβ42 levels (possibly due to increased plaque sequestration). However, CSF T-tau and Aβ42 level elevations may not be specific to AD. Elevated isoprostane levels are also reported in AD and MCI but these too are not specific for AD. Importantly, it has been recently observed that CSF levels of P-tau, tau hyperphosphorylated at threonine 231 (P-tau231) are uniquely elevated in AD and elevations found in MCI are useful in predicting the conversion to AD. In our current MCI studies, we are examining the hypothesis that elevations in P-tau231 are accurate and specific indicators of AD-related changes in brain and cognition. In cross-section and longitudinally, our results show that evaluations of the P-tau231 level are highly correlated with reductions in the MRI hippocampal volume and by using CSF and MRI measures together one improves the separation of NL and MCI. The data suggests that by combining MRI and CSF measures, an early (sensitive) and more specific diagnosis of AD is at hand. Numerous studies show that neither T-tau nor P-tauX (X refers to all hyper-phosphorylation site assays) levels are sensitive to the longitudinal progression of AD. The explanation for the failure to observe longitudinal changes is not known. One possibility is that brain-derived proteins are diluted in the CSF compartment. We recently used MRI to estimate ventricular CSF volume and demonstrated that an MRI-based adjustment for CSF volume dilution enables detection of a diagnostically useful longitudinal P-tau231 elevation. Curiously, our most recent data show that the CSF isoprostane level does show significant longitudinal elevations in MCI in the absence of dilution correction. In summary, we conclude that the combined use of MRI and CSF incrementally contributes to the early diagnosis of AD and to monitor the course of AD. The interim results also suggest that a panel of CSF biomarkers can provide measures both sensitive to longitudinal change as well as measures that lend specificity to the AD diagnosis.