Ivo Dinov was interviewed by NOVA/PBS to discuss a recent PNAS Report identifying significant potential shortfalls of Big Data functional magnetic resonance imaging (fMRI) studies, of which there may be over 35,000 in the past 25 years.
The article used 500 normal control subjects (null data) to generate 3 million simulation studies where every experiment included randomly chosen subjects, either resting state or task activation fMRI, and found false-positive discoveries (significant grouping effects where there were none) in up to 70 percent of the simulations.
Although this does not discredit any specific previously published fMRI findings, the investigations suggests the need for novel Big Data analytics methods, and scalable software tools, that can reduce the false-positive rate.