Software packages that analyze the results of functional magnetic resonance imaging (fMRI) contain flaws that showed false brain activity on 60 percent of images, claim Swedish researchers. A reasonable percentage is five.
Researchers Anders Eklund and Hans Knutsson from Linköping University, and Thomas Nichols from the University of Warwick in Great Britain, found that the majority of fMRI studies use SPM, FSL or AFNI software packages based on “parametric statistical methods that depend on a variety of assumptions.” However, these methods have been validated only with simulated—as opposed to real—data. As a result, the researchers questioned whether these methods could potentially show brain activity in its absence, raising the issue of false positives.
Their claims would call into question the results of 40,000 fMRI studies going back 15 years.
Published in the journal PNAS, the study reports the researchers used resting-state fMRI data from 499 healthy controls to conduct three million-task group analyses.
"Using this null data with different experimental designs, we estimate the incidence of significant results. In theory, we should find 5 percent false positives, ... but instead we found that the most common software packages for fMRI analysis (SPM, FSL, AFNI) can result in false-positive rates of up to 70 percent," explain the researchers. “These results question the validity of some 40,000 fMRI studies and may have a large impact on the interpretation of neuroimaging results.”
FUNCTIONAL MAGNETIC RESONANCE IMAGING
Functional magnetic resonance imaging (fMRI) measures blood flow in the brain and shows which areas of the brain are most active when processing a stimulus, allowing researchers to study deep structures of the subject’s brain during exposure to advertising or when faced with purchase decisions. The technique doesn’t require the administration of radioactive contrast dye so that repeat observations can be made on the same subject, according to BCC Research analyst Joanna Sousa.
“The technology’s based on the fact that the level of oxygen in the blood changes in response to neural activity,” she explains. “Neurons need oxygen for their activity and each time a synapse discharges more oxygen is extracted from the arteries that feed the neuron.”
Differences in blood oxygen saturation correlate with neural activity. Sousa says that functional magnetic resonance imaging is a technique for measuring brain activity across the entire brain simultaneously. It works by detecting changes in blood oxygenation and flow that occur in response to neural activity. When a brain area is more active, it consumes additional oxygen. To meet this demand, blood flow increases to the active area.
“Functional magnetic resonance imaging identifies which brain areas are activated and modulated during specific tasks, hence the term ‘functional’ MRI,” says Sousa.
RESEARCHERS URGE COMMON STATISTICAL METHODS USE REAL FMRI DATA
Eklund proposed another method in which few assumptions are made and significantly more calculations--a thousand times more--are done, which yields fewer errors. He adds that more modern graphics cards would reduce the processing time for images, as well..
“Thanks to modern graphic cards, large calculations can be run,” Eklund says. “It would take 1,000 times longer to run the calculations using a normal computer, but thanks to the graphic cards, I reduced the processing time from 10 years to 20 days.”
Eklund applied his newer method on the same data that was analyzed on the older method and found that the new method yielded more accurate results with differences in five percent of cases.
Our results suggest that the principal cause of the invalid cluster inferences is spatial autocorrelation functions that do not follow the assumed Gaussian shape,” he concludes. “It is not feasible to redo 40,000 fMRI studies, and lamentable archiving and data-sharing practices mean most could not be reanalyzed either. Considering that it is now possible to evaluate common statistical methods using real fMRI data, the fMRI community should, in our opinion, focus on validation of existing methods.”