Statistical Meta-Analyses in Imaging Neuroscience

The great increase in PET and fMRI imaging research has provided a large amount of data about the localization of motor, cognitive and affective processes in the human brain. However, there is an obvious discrepancy between the large number of studies available and the limited amount of knowledge that can be gained from each individual experiment. Quantitative coordinate-based meta-analyses have thus emerged as a principled way to integrate this large amount of findings. These approaches, in particular activation likelihood estimation (ALE) offer the opportunity to investigate convergent effects in large samples of healthy subjects and patients that could not be assessed in any individual study. It thus offers substantial advantages over the otherwise more subjective inferences about the relevance of individually observed effects. Ultimately, model-based analyses testing explicit hypotheses using quantitative meta-analysis approaches should offer an important step forward in the investigation of the physiological and pathological localization within the human brain. The development and application of neuroinformatics approaches to quantitative, data-driven meta-analysis may thus become an important aspect in the future strive to understand the healthy and diseased brain.