Neuroimaging is revealing more than ever how the brain changes over time and space. By combining information from different types of imaging that capture brain activity (functional magnetic resonance imaging, arterial spin labeling) and anatomy (T1-weighted and diffusion tensor imaging), we can begin to create individualized narratives of the brain across time. These brain narratives unfold from activation patterns of large populations of neurons, their synaptic integration by neurons, and the axonal pathways required to physically communicate the electrical signals that give rise to consciousness. It can be overwhelming to decipher brain imaging studies (even for the experts!), and this blog post is intended to demystify some fundamental principles of neuroimaging–especially of pain perception–so that you can critically evaluate this type of research.
Sculpting Brain Structure and Function
Different neuroimaging modalities are used to ask different types of biological questions about the brain. Suppose you want to know:
- What brain regions are active during pain perception?
- Is this pattern of activity unique to pain, or can other sensory experiences be mediated by the same network of brain areas?
- How do those brain regions naturally communicate with one another when you aren’t focused on pain?
- Does chronic pain cause neuronal atrophy in any of these regions?
- Are there predisposing differences in the anatomy (size or shape) of these regions that could make someone more susceptible to develop chronic pain?
- Over time, does the structure of the axonal highways that link these regions with the rest of the brain reflect chronic pain severity?
- By combining information about axonal pathways and communication between brain regions, can we identify new communication circuits that tell us something about an individual’s vulnerability or resilience to chronic pain?
These questions can be answered by using our knowledge of neuroplasticity from the basic science literature to guide the use of technology. Short-term changes in brain states are reflected in the fluctuations of BOLD activity over seconds to minutes, whereas longer-term changes are evident in gray (neuronal) and white (axonal) matter properties.
Functional magnetic resonance imaging (fMRI) captures snapshots of brain activity in 0.5- 3 second increments (depending on repetition time, or TR). Most fMRI studies are task-based experiments that detect changes local brain activity related to a specific task. For example, a task-based study that examines a person’s pain intensity ratings will use the ratings as discrete events that are used to extract relevant brain activity that strongly correlates with the rise and fall of perceived intensity. However, standard task-based analyses use event related designs that sample a larger period of time (approximately 3-5 seconds before and after a given TR). Therefore a single brain volume with a TR = 2.5 may actually reflect up to 12.5 seconds of activity. This poor temporal resolution isn’t necessarily a problem for capturing fluctuations in activity in regions like the somatosensory cortices, where natural fluctuations in blood oxygen level dependent (BOLD) activity take up to 20-25 seconds to peak and trough. On the other hand, this temporal resolution can be disastrous in capturing phasic activity in the nucleus accumbens (NAc). Brief phasic increases in the NAc appear to serve distinct biological purposes: a spike of NAc activity prior to a stimulus serves as a predictive signal, whereas the phasic burst after removal of a stimulus is likely a valuation signal (Baliki 2009).
Indeed, the natural fluctuations in brain regional activation suggest that, depending on when you happen to time your stimulus in comparison with an individual’s unique physiology as they lie in your scanner, you may fail to analyze data in a way that actually captures a region’s contribution to ANY 12.5 second event. Biyu He attempted to tackle this unique problem a few years ago by proposing algorithms to statistically account for these natural fluctuations (essentially by normalizing activity data to an individual’s basal BOLD responses for each region).
The anatomical features that emerge from large groups of neurons are captured in T1-anatomical imaging. Measurable increases and decreases in neocortical grey matter occur over weeks to months, and grey matter in the neocortex appears to correlate with clinical symptom severity across a broad variety of disease states. If we assume that symptoms are expressions of an underlying pathological process–which is a safe assumption–this means that grey matter changes may preferentially reflect consequences of a disease, rather than the disease itself.
The structural details of axon tracts (white matter) are measured with diffusion tensor imaging. At this point we know that measurable changes in axon structure emerge within 1-2 weeks complex motor task training (e.g., juggling), and that certain brain regions are more likely to show these changes than others (e.g., motor cortex). Many neurodegenerative diseases show a gradual but dramatic breakdown of axon tract structure, and this is thought to explain why information becomes far more difficult to integrate with diseases like Alzheimer’s, frontotemporal dementia, or multiple sclerosis. Very few diseases are associated with more coherent axon tract structure, and one of these conditions chronic bladder pain.
Given these basic ground rules, we can begin to piece together a narrative of the functional and structure interactions that promote disease development.
[CONTENT UNDER REVISIONS. PLEASE BE PATIENT!]