Understanding how analysis of fMRI data can influence the reproducibility of studies

New paper published in NeuroImage by Ludovica Griffanti et al.

fMRI reproducibility paper

Functional magnetic resonance imaging (fMRI) studies allow researchers to measure brain activity by looking at changes in blood flow to different areas of the brain: the greater the blood flow in a certain area of the brain, the greater the amount of brain activity in that area. This information can tell researchers which areas of the brain are functionally connected to one another, in other words, which areas of the brain are active at the same time. This gives clues as to which areas of the brain need to work together in order to enable a certain function, such as the ability to move. In some individuals these connections may be abnormal, and thus certain functions, such as movement, may be disrupted. Using fMRI to identify abnormal patterns of brain activity can provide an early indication (or biomarker) that an individual may go on to develop a disease, such as Parkinson’s.

A recent fMRI study found that connections in the basal ganglia (a region of the brain which is known to degenerate in Parkinson’s) are disrupted in patients in the early stages of Parkinson’s disease. Looking at the brain activity of patients allowed researchers to differentiate them from healthy individuals, which indicates the potential of fMRI to provide an early diagnosis of Parkinson’s disease. However, one huge issue in science is the ability to reproduce data – if the same changes in connectivity in the basal ganglia are not seen repeatedly in patients with Parkinson’s disease, then a diagnosis of Parkinson’s with fMRI cannot be trusted. Reproducibility is particularly difficult to achieve in fMRI studies due to the large variety of ways in which researchers can analyse their data. The recent OPDC study by Ludovica Griffanti et al. found that when fMRI data is analysed with a certain method, data is increasingly reproducible. The OPDC researchers found that the method of artefact removal (how noisy data is ‘cleaned up’) and the choice of template against which to compare patient data have a large influence on results, and thus recommend that researchers publish their method of analysis so that different studies can be compared effectively. Understanding how different analysis settings can influence results is an important step in improving the reproducibility of fMRI studies, and thus brings us one step closer to defining a reliable imaging biomarker of Parkinson’s disease.

Read the full paper here