Neuroimaging and spatial resolution

Introduction

Neuroimaging has been at the forefront of brain discovery ever since the first ever images of the brain were recorded in 1919 by Walter Dandy, using a technique called pneumoencephalography (PET). 

Fast-forward over a decade and neuroimaging is more than just blurry singular images. Modern techniques allow us to observe real time changes in brain activity with millisecond resolution, leading to breakthroughs in scientific discovery that would not be possible without it. Memory is a great example - with functional magnetic resonance imaging (fMRI) techniques we have been able to demonstrate that more recent long-term memories are stored and retrieved with brain activity in the hippocampus, but as memories become more in the distant past, they are transferred to the medial temporal lobe.

While neuroimaging techniques keep the doors open for new and exciting discoveries, spatial limitations leave many questions unanswered, especially at a cellular and circuit level. For example - within the hippocampus, is each memory encoded via complete distinct neural circuits? Or do similar memories share similar neural pathways?

Within just a millimetre cubed of brain tissue we could have up to 57,000 cells (most of them neurons), all of which may have different properties, be part of different circuits, and produce different outcomes. This almost makes revolutionary techniques such as fMRI, with almost unparalleled image quality, seem pointless.

To truly understand how neural circuits work, we have to dig as deep as possible to record the smallest regions possible. So that begs the question, how small can we actually record in the human brain?

EEG

2024 marks a decade since the first recorded electroencephalography (also known as EEG) scan by Hans Berger in Germany. This technique involves placing electrodes all around the scalp to record activity throughout the whole outer surface of the brain (Figure 1). Unlike the methods we see later on, EEG scans provide a direct measure of activity in the brain, by measuring electrical activity when the brain is active.

However, because electrodes are only placed across the scalp, EEG scans are only able to pick up activity from the outer cortex, missing important activity in deeper parts of the brain. In our memory example, this means it would completely miss any activity in the hippocampus. EEG resolution is also quite underwhelming, typically being able to resolve activity with a few centimetres’ resolution - not great for mapping behaviours to specific structures in the brain. 

EEG scans are used in a medical environment to measure overall activity levels, assisting with epilepsy diagnosis. Let's look at what we can use to dig deeper into the brain and locate signals of activity…

PET

Position emission tomography (PET) scans offer a chance to record activity throughout the whole brain by ingesting a radioactive tracer, typically glucose labelled with a mildly radioactive substance. This tracer is tracked and uptake in specific parts of the brain is a sign for greater metabolic activity, indicating a higher signalling rate.

PET scans already offer a resolution far beyond the capacities of EEG scans, distinguishing activity between areas with a resolution of up to 4mm. With the use of different radioactive labels, we can also detect activity of specific populations of neurons  such as dopamine neurons to diagnose Parkinson's disease. In fact, many studies have reliably demonstrated the ability of PET scans to detect the root cause of Parkinson's disease, which is a reduced number of dopamine neurons in the basal ganglia, before symptoms become too extreme.

As impressive as it sounds, a 4mm resolution can locate activity in large areas of the cortex, but is limited in its resolving power for discrete cortical layers. Take the human motor cortex for example - all 6 layers have an average width of only 2.79mm. A PET scan would not be powerful enough to determine which layer is most active, so we need to dig a little deeper…

fMRI

Since its inception in the early 90's, fMRI has gained the reputation of becoming the gold standard for human neuroimaging, thanks to its non-invasiveness, lack of artefacts, and reliable signalling.

fMRI uses Nuclear Magnetic Resonance to measure changes in oxygenated blood flow, which is correlative of neural activity, known as BOLD signals. In comparison to EEG, measuring blood oxygen levels cannot reach a highly impressive temporal resolution, and is also not a direct measure of neural activity. fMRI makes up for this with its superior spatial resolution, resolving spaces as small as 1mm apart. Using our human motor cortex example, this would allow us to resolve activity between every 2-3 layers - not a bad return considering it doesn’t even leave a scar.

PET, and especially EEG, pales in comparison to the capabilities of fMRI that has since been used for a wide range of neuroimaging research. Most notably, structural MRI has been used to support the idea of hippocampal involvement during spatial navigation from memory tasks (figure 2). Its resolving power and highly precise images also make it suitable to be used for mapping surgical procedures.

Conclusion

With a resolution of up to 1mm, fMRI takes the crown as the human neuroimaging technique with the best spatial resolution! Table 1 shows a brief summary of each neuroimaging method.

Unfortunately though, there is still so much more we need to do to look at individual circuits and connections. As mentioned before, even within a millimetre cubed of brain, we have 5 figures worth of cells, making the number of neurons that make up the whole brain impossible to comprehend. To observe the activity of a single neuron, we would need an imaging technique with the power of viewing cells in the 10’s of micrometre range. So what can we do to get to the resolution we desire while still being suitable for humans?

Maybe there isn't a solution. Instead, maybe if we want to record singular neuron activity, we have to take inspiration from invasive animal techniques such as microelectrode recordings. Typically used in rats and mice, these can achieve single-cell resolution to look at neuroscience from the smallest of components.

It would be unethical to stick an electrode into a healthy human's brain and record activity, but perhaps in the future a non-invasive form of electrode recording could be developed?

The current neuroscience field is foggy and shrouded in mystery. Most of these mysteries simply cannot be solved with the current research techniques we have at our disposal. But this is what makes neuroscience exciting - there is still so much to explore! Who knows when we will be able to map behaviours to neural circuits with single-cell precision, but with how quickly imaging techniques are being enhanced and fine-tuned, I wouldn't be surprised if it's sooner than we think.

Written by Ramim Rahman

REFERENCES

Hoeffner, E.G. et al. (2011) ‘Neuroradiology back to the future: Brain Imaging’, American Journal of Neuroradiology, 33(1), pp. 5–11. doi:10.3174/ajnr.a2936.

Maguire, E.A. and Frith, C.D. (2003) ‘Lateral asymmetry in the hippocampal response to the remoteness of autobiographical memories’, The Journal of Neuroscience, 23(12), pp. 5302–5307. doi:10.1523/jneurosci.23-12-05302.2003.

Wong, C. (2024) ‘Cubic millimetre of brain mapped in spectacular detail’, Nature, 629(8013), pp. 739–740. doi:10.1038/d41586-024-01387-9.

Butman, J. A., & Floeter, M. K. (2007). Decreased thickness of primary motor cortex in primary lateral sclerosis. AJNR. American journal of neuroradiology, 28(1), 87–91.

Loane, C., & Politis, M. (2011). Positron emission tomography neuroimaging in Parkinson's disease. American journal of translational research, 3(4), 323–341.

Maguire, E.A. et al. (2000) ‘Navigation-related structural change in the hippocampi of taxi drivers’, Proceedings of the National Academy of Sciences, 97(8), pp. 4398–4403. doi:10.1073/pnas.070039597.

[Figure 1] EEG (electroencephalogram) (2024) Mayo Clinic. Available at: https://www.mayoclinic.org/tests-procedures/eeg/about/pac-20393875 (Accessed: 18 October 2024). 

[Figure 2] Boccia, M. et al. (2016) ‘Direct and indirect parieto-medial temporal pathways for spatial navigation in humans: Evidence from resting-state functional connectivity’, Brain Structure and Function, 222(4), pp. 1945–1957. doi:10.1007/s00429-016-1318-6.