One of medical science’s indispensable diagnostic and research tools over the past quarter-century has been functional magnetic resonance imaging technology, or fMRI. This is a non-invasive imaging technique that makes use of strong magnetic fields and radio waves to produce three-dimensional images of the interior of the human body, and has revolutionized research into brain activity, using increased blood flow to indicate corresponding increases in neural activity. However, a new study has called the accuracy of the device’s software into question, after discovering a bug in commonly-used MRI interpretation software packages — a bug that may very well call the results of over 40,000 medical research studies into question.
The study, conducted at Linköping University in Sweden, came about because of the realization that the images produced by fMRI software have never been verified against real data — the software uses statistical methods to extrapolate what is actually going on inside the subject. The team made use of 499 healthy control subjects to scan for resting-state data, when the subjects were not performing a task, with their brains at rest. The researchers were expecting a false-positive rate of about 5 percent, but instead found that some of the more commonly-used software packages used would produce false-positive rates of up to 70 percent — basically, many of the scans showed brain activity while the subject wasn’t actually doing anything.
While this poses an obvious problem for previous diagnoses of potential neurological problems in patients, the team found that the statistical bug in the software has been present for fifteen years, and may call into question the findings of over 40,000 medical studies that made use of fMRI imaging. This bug was corrected in May of last year, so at the very least future studies won’t be affected in a similar manner. As well, the research team points out that while it’s impossible to know for certain how many of the studies that have made use of fMRI imaging in the past might actually be affected, a growing body of MRI results are being made available to researchers, to give them a broader range of material to compare and validate their findings against.