This chapter describes a method for meta-research, based on image mining from neuroscientific publications. It extends earlier investigation to the study of a large scale data set. Using a framework for extraction and characterisation of reported fMRI images, based on their coordinates and colour profiles, we propose that significant information can be harvested automatically. The coordinates of the brain activity regions, in relation to a standard reference templates are estimated. We focus on the analysis of scientific reports of the default mode network. Both the commonalities and the differences of brain activity between control, Alzheimer and schizophrenic patients are identified.