Neuroimaging research produces very large amounts of data. This data is then often processed with sophisticated algorithms that take a lot of computing power. With all this data storage and computation, comes a cost in terms of pollution. A common way to do some analysis of fMRI images is a tool called fMRIPrep. A lot of neuroimaging researchers use this to get their data ready for final analysis steps. We were interested in how much carbon pollution this tool creates and in how it could be optimised to reduce this pollution.

To do this, we ran fMRIPrep on data from people who did a task in the MRI scanner. We ran it with different options turned on and off to see how this affected the amount of energy, and therefore carbon pollution, that was required. We discovered that turning off some options in the software could reduce the energy use considerably. Importantly, we also showed that changing these settings did not affect the quality of the data and that the results from subsequent analyses were the same as with more polluting options. We recommend that people turn off these options.

Reference:

Souter, N.E., Bhagwat, N., Racey, C., Wilkinson, R., Duncan, N.W., Samuel, G., Lannelongue, L., Selvan, R., Rae, C.L., 2024. Measuring and reducing the carbon footprint of fMRI preprocessing in fMRIPrep. Human Brain Mapping 45, e70003. https://doi.org/10.1002/hbm.70003