According to a study recently published in PNAS, a total of 40,000 studies carried out over 15 years could be wrong due to errors in the software of the functional magnetic resonance (fMRI), one of the most useful imaging techniques to study brain activity.
In the study, data were taken at rest from 499 patients around the world and compared the results of their magnetic resonances to obtain 3 million random comparisons. 70% were false positives. Experts evaluated the three most popular software packages for functional magnetic resonance analysis: SPM, FSL and AFNI.
How is it possible that there are 40,000 poorly conducted studies and no one has noticed? The main reason is that performing MRIs is expensive, which has traditionally forced researchers to conduct studies with few patients and, most importantly, the experiments have not been repeated in many cases, to check if the results were the same.
In addition, until recently this problem joined the slowness of these computer programs. As he explains Anders Eklund, from the University of Linköping, in Sweden, and lead author of the study:
Despite the popularity of functional magnetic resonance imaging as a tool for the study of brain function, the statistical methods used have rarely been validated with real data.
Fortunately, we have made considerable progress in this field, and what was previously processed in 10 years can now be done in 20 days, and the results are available online for the study of other scientists.
This finding, therefore, does not invalidate all the research conducted with fMRI, but yes there could be many mistakes, and that all those conclusions about what our brains do while doing the exercise, we play, we practice sex or we experience addiction to drugs maybe they are partially wrong.
A blow to which we must add that last year it was found that less than half of the findings in psychology studies were reproducible.