Major assumption of fMRI likely correct

Functional Magnetic Resonance Imaging, more commonly known as fMRI, has revolutionized the way researchers study the brain. By using a magnet that is tens of thousands times stronger than the Earth’s magnetic field, scientists can peer into the deep recesses of the brain and see how it works.

Despite its widespread use, the interpretation of fMRI data relies on assumptions, some of which remain untested. Recently, researchers at Western University have shown that one important assumption in modern fMRI analysis is valid. Assumptions in science are things we accept as true, even if there is no concrete evidence for them—if an assumption is wrong, it might mean that our scientific conclusions are wrong.

Researchers must make assumptions when using fMRI because it does not directly measure brain activity. Rather, it measures  the oxygen content of the blood in the brain, which is related to brain activity. The idea is that more oxygen will be required by active brain areas during an fMRI experiment. This advanced technology presents some potential problems.

            “There is some disassociation between what the neurons are actually doing and what we’re actually measuring,” says Spencer Arbuckle, a PhD student in neuroscience and lead author of this research study that was published last year in the journal NeuroImage.

In the early days of fMRI experiments in the 90s, and even now, researchers would examine the average activity within a brain region and observe if there were any overall increases or decreases of activity in response to some task. Results of this method proved limited, leading to modern fMRI analysis, which focuses on the differences between activity patterns within the brain regions. Think of this way: if you averaged all the colours of the rainbow, you would get white light, and you would miss the beauty of each colour. In much the same way, averaging brain activity in fMRI misses the fine details of activity within each region.

Today, researchers do not average, but instead decode the pattern of activity within a region. They look at all the colours, so to speak. “We think that this pattern represents something, some information. They key word is represents,” says Arbuckle.

To understand what these patterns represent, scientists examine how activity patterns change between conditions, for example, when patients are doing different activities. But new assumptions come with these new techniques. One important assumption that is commonly made is that these activity patterns are the same, regardless of the average activity in a region. In other words, scientists assume that the same pattern of activity is measured whether the brain region is only moderately active versus very active. If we consider the rainbow again, imagine we wanted to measure the colours at daytime versus at night. Scientists have assumed that the patterns we measure are the same. But what if they are not? What if it’s like comparing apples to oranges? This assumption is critical to test if, for example, one wants to make inferences between patient populations and healthy controls, who may have differences in overall activity.

Arbuckle and his colleagues aimed to make sure that this assumption is a reasonable one to make. To do this, they looked at the brain activity patterns of people pressing buttons.

Arbuckle explains it as such: when you move one finger, you have a certain pattern of activity in the motor cortex in the brain. These patterns are largely unique to each finger. If you move the same finger twice within the same time-frame, the overall activity increases, but ideally the same fine-grained activity pattern is still present. Therefore, if the assumption about the stability of fMRI activity patterns is valid, you would expect to see similar patterns for the same finger whether it was pressed at slow speeds or fast speeds.

Through a series of analyses, they found that the patterns of the finger presses were measurably quite stable as the pressing speed increased. This provides strong evidence in support of this assumption.

This is good news for a large body of studies and analysis approaches that rely on measuring brain patterns for their fMRI analysis.

“We are a bit more confident,” Arbuckle says. “We think that these representations actually, truly do represent something meaningful and informative.”

Original Research Article: Arbuckle SA, Yokoi A, Pruszynski JA, Diedrichsen J. Stability of representational geometry across a wide range of fMRI activity levels. Neuroimage [Internet]. 2019 Feb 1 [cited 2019 Dec 20];186:155–63. Available from: https://www.sciencedirect.com/science/article/abs/pii/S105381191832072X