Brain scans using functional magnetic resonance imaging (fMRI) have taken a battering. Most famously, the technique appeared to show brain and spinal cord activity when used on a dead salmon.
A later study found a bug in software most commonly used to analyse fMRI scan data.
fMRI works because increases in brain activity increases blood flow in the brain. It also churns out huge amounts of data. So pinpointing brain areas which are more active than others means processing lots of data. This is where errors emerge.
To solve this, researchers, often use a small sample size, which brings its own limitations. “The average statistical power of studies in the neurosciences is very low,” notes a study in Nature Reviews Neuroscience, “consequences include overestimates of effect.”
Now this question of how to analyse fMRI scan data is making scientists also question how to collect the data. Researchers are moving away from using scan data to test pre-existing theories and instead taking a data-driven approach to pull the interesting information directly from the data.
These new data-driven methods throw up more possibilities as to how neuroimaging data can unlock undiscovered brain mechanisms.
Researchers at Brunel University London and Denmark’s Aarhus University looked at fMRI data through a new lens called binarisation of consensus partition matrices (Bi-CoPaM). Bi-CoPaM merges results from several different ways of looking at data, to get stronger results that can be re-produced. This fresh approach finally may spark grow consensus on brain scan results.
They studied for brain scans from people hearing music clips to test Bi-CoPaM for the first time against older methods of analysing scan data.
“Bi-CoPaM allows us to find clusters including functionally and anatomically related neural networks consistently responding similarly to emotional music,” said Brunel’s Professor Asoke Nandi.
“The most important finding is that our proposed approach discovered a single cluster, including the anatomically connected subcortical and cortical structures of the reward circuit, responding selectively to liked, enjoyed music. This is one of few studies obtaining such a finding with a data-driven approach.”
Oh and that dead salmon reacting? Once the data underwent correction for multiple comparisons the false positives were eliminated – it really was a deceased fish!
Towards tunable consensus clustering for studying functional brain connectivity during affective processing, is published in International Journal of Neural Systems.