Measuring Brain Activity: Design, Acquisition, and Analysis of fMRI data

By Paola Tine, student of the

Master in Applied Neuroscience.

What is fMRI

Functional Magnetic Resonance Imaging (fMRI) is the most advanced and commonly used non-invasive methodology for exploring the human brain while performing activities and experiencing mental states, including sensation, perception, attention, learning, memory, reasoning, decision making and so on, making it a crucial tool in investigating the relation between psychological processes and brain function (Poldrack, 2011). Differently from Magnetic Resonance Imaging (MRI), which only shows the brain in one given moment, fMRI allows researchers to acquire multiple images in a sequence obtained over time and can show how brain functions can change over time. fMRI is a great tool for enhancing studies about the human brain and the mind, especially when different disciplines such as neuroscience, medicine, sociology, anthropology, and psychology, among others, collaborate with each other. Using fMRI techniques is a very complex process that includes the careful design of experiments (directed by a certain research goal) followed by data acquisition, analysis and inference. Each of these phases represent an elaborate process in which physical, biological and mathematical knowledge, as well as statistical analysis are fundamental. This is why fMRI research should be carried out by a multidisciplinary team.

fMRI scans measure the consequences (or haemodynamic response) of the electrical impulses of neural activity. Neural communication – which involves the work of action potentials (electric impulses) and synapses releasing neurotransmitters to communicate with other synapses – produces the so called ‘phenomena of mind’. As neural activity is accompanied by changes in local blood flow, the observation of the presence of oxygenated blood in certain areas of the brain (blood-oxygen-level-dependent or BOLD response) is a crucial part of the fMRI technique. This is because when neurons are active, blood is needed to fuel them, consequently increasing the supply of oxygen. Trying to understand how the electric processes occurring at the level of the brain are linked to the human experience of emotions, thoughts and feelings is the main challenge of fMRI.

However, this link is not simple and the haemodynamic response can often be linked to synaptic events rather than action potentials (Stokes 2015). This is what Bennet, Miller and Wolford (2019) showed in a satirical, yet eye-opening, experiment, showing the brain responses of a dead salmon to which they displayed pictures of people in various social and emotional states. The fact that the dead salmon responded with brain activity to the images can be understood by looking at the second aspect of how fMRI works. The scan is a machine conventionally dividing the brain into 40,000 cubic units that can be seen in tridimensional resolution in a computer interface. Since each of these 2-3 mm cubes, called voxels (volumetric pixels), contains around six-hundred thousand neurons, some are likely to be activated even when not responding to an external stimulus. The experiment by Bennet et al., (2019), pointed out the “multiple comparisons problem”, meaning that when a large number of statistical tests is conducted, some results might occur that are misleading. These issues reveal the danger of applying fMRI techniques to detecting lies in a legal setting or communicating with people who are unable to communicate (Stix, 2008). However, important advancements have still been made by fMRI studies regarding general aspects of human thought (Ouellette, 2020; Cohen, 2020).

It is widely known that the mature brain has many specialized areas of function, and neurons that differ in structure and connections. The hippocampus, for example, which is a brain region that plays an important role in memory and spatial navigation, alone has at least 27 different types of neurons

What is not widely known is that the incredible diversity of neurons in the brain results from regulated neurogenesis during embryonic development. During the process, neural stem cells differentiate that is, they become any one of a number of specialized cell types, at specific times and regions in the brain

Stem cells can divide indefinitely to produce more stem cells, or differentiate to give rise to more specialized cells, such as neural progenitor cells. These progenitor cells themselves differentiate into specific types of neurons.

In 1928, Santiago Ramón y Cajal, the father of modern neuroscience, proclaimed that the brains of adult humans never make new neurons. “Once development was ended,” he wrote, “the founts of growth and regeneration … dried up irrevocably. In the adult centers the nerve paths are something fixed, ended and immutable. Everything must die, nothing may be regenerated.”

But from the 1980s onward, this dogma started to falter. Researchers showed that neurogenesis does occur in the brains of various adult animals, and eventually found signs of newly formed neurons in the adult human brain. Hundreds of these cells are supposedly added every day to the hippocampus—a comma-shaped structure involved in learning and memory. The concept of adult neurogenesis is now so widely accepted that you can find diets and exercise regimens that purportedly boost it.

Experiment Design

The most common experiment designs used to conduct fMRI studies are blocked and event-related designs. Block designs provide the participant to be examined with alternating blocks of stimuli or cognitive tasks to perform and these tests are generally used to explore functional areas and steady state processes (such as attention). During the stimulus, the haemodynamic response  is detected by the fMRI machine due to a change in neuronal activity, after a few seconds of delay, to allow enough time for the brain to obtain a supply of glucose, generally reaching the peak after 5 seconds (Huettel, Song & McCarthy, 2009), which is followed by a drop in the BOLD signal eventually returning to its starting point or baseline. Block designs have the advantage of providing a larger BOLD signal contrast in comparison with the baseline (Buxton et al., 1998). Differently, event designs provide the participant with different types of stimuli, over a short period of time (where each event is divided by intervals lasting a few seconds) and in a random order. The goal of event designs is to detect transient changes in brain activity, allowing for investigation of individual responses to trials (Schacter et al., 1997), with the advantage of being less sensitive to head motion (Birn et al., 1999).

All of these designs operate based on the ‘subtraction paradigm’, which means that the blood flow (the BOLD signal detecting neural activity) detected when the subject is experiencing the cognitive results of a given experimental task, is subtracted from the pattern of blood flow detected in a control, or resting state, to detect the blood flow of the studied cognitive process. However, the non-linear nature of brain processes and the fact that the brain is never at rest creates a grey area in this method. In When zero is not zero: The problem of ambiguous baseline conditions in fMRI, Stark and Squire (2001) discuss the problematicity of discerning baseline conditions from activity conditions. As they report in this study, in their experiments, participants’ medial temporal lobe appeared to be proportionally more active while at rest than during cognitive tasks. This demonstrates that cognitive activity can be significant in periods of rest and they pointed out the risk of using resting states as a baseline for cognition tasks.

Acquisition and Analysis of fMRI data.

fMRI is a non-invasive test that is performed with a patient/participant lying in a traditional MRI unit, a cylinder-shaped tube that is surrounded by a circular magnet. During the examination, the patient is asked to perform some simple tasks, such as tapping their fingers or viewing pictures on a screen located inside the scanning machine. When acquiring fMRI data, technical issues to consider are unwanted sources of ‘noise’, such as head motion notion,  which can alter the data by producing neural activity independently from the stimuli given by the experiment. These aspects cannot be controlled through mathematical modelling and have to be considered and potentially avoided through experimental design. To help address this issue, the first part of the analysis process, known as pre-processing, can make use of data smoothing and spatial filtering.

Another important, though problematic, aspect of fMRI analysis occurring in the interpretative phase, is the use of reverse inference, according to which researchers might wrongly base their interpretations of the studied cognitive processes on previous knowledge of the functioning of certain brain regions (Poldrak, 2011). This is a common issue when using imaging to address psychological theories. When inducing a psychological state (for example through the task of punishing someone), some brain activity is observed and if this is in an area or pattern of the brain associated with a specific mental state (for example pain, or pleasure) then false reverse inferences can be drawn. This can be visualised through the famous example of a logical fallacy that says: dogs like ice cream; Mary likes ice cream; therefore, Mary is a dog.  This example demonstrates that logic can only be used in very well-defined contexts to make meaningful conclusions. It is particularly important to know this when looking at brain functioning, as it is rare for a specific area of the brain to be activated by only one cognitive process.  Consequently, while reverse inference can serve as a hint to progress with deeper research, it should not be used as a direct means to interpret results (Poldrack, 2011). Differently, through ‘forward inference’ it is possible to design experiments that investigates what areas of the brain are associated with given tasks. However, there are also several issues related to this methodology, including the fact that certain brain patterns or areas might activate during specific cognitive tasks without being strictly related to the task performed or the relative emotional state (Poldrack, 2008; 2006). Care needs to be taken in interpreting the results of fMRI experiments as the data can be quite complex. Statistics play a crucial role in decoding this data and should be taken into account when designing experiments to ensure that they promote easily codified and relevant results for interpretation by neuroscientists (Landquist, 2008). This requires close communication between researchers and those designing and running the experiments at all levels of planning and interpretation.

Conclusion

fMRI is undoubtedly one of the fastest growing areas of research in the neuroscientific field and one of the most promising. Thanks to fMRI, researchers can now have access to brain activity in order to investigate human behaviour, perception and cognition. However, this methodology encounters several limitations that make it particularly vulnerable to mistakes when not handled properly. Ultimately, the main problem of fMRI is understanding what can be achieved by looking at brain patterns and what cannot, and above all, what questions should we strive to answer regarding human society, culture, and behaviour when looking at humans’ most intricate and intimate structure.

Bennett, CM, Miller, MB & Wolford, GL 2009, ‘Neural correlates of interspecies perspective taking in the post-mortem Atlantic Salmon: An argument for multiple comparisons correction, NeuroImage, vol. 47,  no. S125.

Birn, RM, Bandettini, PA, Cox, RW & Shaker, R 1999, ‘Event-related fMRI of tasks involving brief motion’, Human Brain Mapping, vol. 7, pp. 106–114.

Brockway, JP 2000, ‘Two functional magnetic resonance imaging f(MRI) tasks that may replace the gold standard, Wada testing, for language lateralization while giving additional localization information’, Brain Cognition, vol. 43, pp. 57–59

Buxton, RB, Wong, EC & Frank, LR 1998, Dynamics of blood flow and oxygenation changes during brain activation: the balloon model, Magnetic Resonance in Medicine, vol. 39, pp. 855–864

Cohen, A 2020, ‘Duke University researchers say every brain activity study you’ve ever read is wrong’, Fast Company, 25 June 2020, viewed 10 September 2020.<fastcompany.com/90520750/duke-university-researchers-say-every-brain-activity-study-youve-ever-read-is-wrong#:~:text=The%20ones%20that%20reveal%20thought,structures%20involved%20in%20a%20task.>.

D’Esposito M, Zarahn E & Aguirre GK 1999, ‘Event-related functional MRI: implications for cognitive psychology’, Psychological Bulletin, vol. 125, pp. 155–164

Friston, KJ et al. 1998, ‘Event-related fMRI: characterizing differential responses’, NeuroImage, vol. 7, pp. 30–40

Huettel, SA, Song, AW & McCarthy, G 2009, Functional Magnetic Resonance Imaging, Sinauer, Massachusetts.

Lindquist, MA 2008, ‘The Statistical Analysis of fMRI Data’, Institute of Mathematical Statistics, vol. 23, no. 4, pp. 439–464.

Ouellette, J 2020, ‘Duke scientist questions his own research with new study faulting task fMRI’, Ars Technica, 8 July 2020, viewed 15 September 2020, <arstechnica.com/science/2020/07/study-task-fmri-cant-reliably-predict-a-single-persons-thoughts-and-feelings/>.

Poldrack, RA 2011, ‘Inferring mental states from neuroimaging data: from reverse inference to large-scale decoding’, Neuron, vol. 72, no. 5, pp. 692–697.

Poldrack, RA 2008, ‘The role of fMRI in cognitive neuroscience: where do we stand?’, Current Opinion in Neurobiology, vol. 18, no. 2, pp. 223–7. 

Poldrack, RA 2006, ‘Can cognitive processes be inferred from neuroimaging data?’, Trends in Cognitive Science, vol. 10, no. 2, pp. 59–63. 

Rombouts, SA et al. 1997, ‘Test-retest analysis with functional MR of the activated area in the human visual cortex’, Neuroradiology, vol. 18, pp. 1317–1322.

Schacter DL, et al. 1997, ‘Late onset of anterior prefrontal activity during true and false recognition: an event-related fMRI study’, NeuroImage, vol. 6, pp. 259–269

Stark CE & Squire LR 2001, ‘When zero is not zero: The problem of ambiguous baseline conditions in fMRI’, Proceedings of the National Academy of Sciences, vol. 98, no. 22, pp. 12760-12766.

Stix, G 2008, ‘Can fMRI Really Tell If You’re Lying?’, Scientific American, viewed 18 September 2020, <scientificamerican.com/article/new-lie-detector/>.

Stokes, M 2015, ‘What does fMRI measure?’, Scitable,  16 May, viewed 16 September 2020, <nature.com/scitable/blog/brain-metrics/what_does_fmri_measure/>.

Tie, Y et al. 2009, ‘Comparison of blocked and event-related fMRI designs for pre-surgical language mapping’, NeuroImage, vol. 47, no. 2, pp. T107–T115.

Kempermann, G., Gage, F. H., Aigner, L., Song, H., Curtis, M. A., Thuret, S., Kuhn, H. G., Jessberger, S., Frankland, P. W., Cameron, H. A., Gould, E., Hen, R., Abrous, D. N., Toni, N., Schinder, A. F., Zhao, X., Lucassen, P. J., & Frisén, J. (2018). Human Adult Neurogenesis: Evidence and Remaining Questions. Cell stem cell23(1), 25–30. https://doi.org/10.1016/j.stem.2018.04.004

Farmer, J., Zhao, X., van Praag, H., Wodtke, K., Gage, F. H., & Christie, B. R. (2004). Effects of voluntary exercise on synaptic plasticity and gene expression in the dentate gyrus of adult male Sprague-Dawley rats in vivo. Neuroscience124(1), 71–79.