some thoughts on resting state activity in the brain

I figured I would start writing with a small post on the subject that I have been interested in lately - resting state activity in the brain (thanks to reading this paper by Laumann and Snyder).

Some historical information first, if only to organize my thoughts: even though electrophysiology started with a discovery of spontaneous electrical activity in the brain (Hans Berger recording alpha waves all the way back in 1924 or so), neuroscientists and cognitive scientists weren’t much concerned with what the brain does when it’s not doing anything at first, focusing on task-related activity.

Studies have shown that the metabolic demands of the brain do not actually change all that much between resting state and task-related activity, which might mean that there might be some significant amount of background processing going on there. What do you do, when you’re just sitting there staring into space quietly?. According to the Amsterdam Resting-state Questionnaire, there are seven dimensions of the resting state: discontinuity of mind (rapidly switching thoughts), theory of mind (a.k.a. mentalizing, fancy terms for “putting yourself in other people’s shoes”), self, planning, sleepiness, comfort and somatic awareness (thinking about the state of your body).

There are interesting correlations of the prevalence of these dimensions with other cognitive tests and clinical variables, which is also a running theme in resting state activity research in general. Providing a total list of biomarkers people have attempted to pull out of the resting state would put anyone to sleep, but resting state activity differences in EEG, MEG and fMRI have been shown in neurological disorders, mental illnesses, brain fingerprinting studies, aging and so on.

Methods applied to this activity range from examining power in certain frequency bands, simple correlations between the activity of brain regions to complex graph theory metrics quantifying the relationship between “nodes” in the human brain. Criticism of these methods notwithstanding, resting state activity seems to be a promising and accessible venue of research - no complicated tasks needed, only a valid enough method and 5 to 10 minutes of brain activity recording and we can determine if you are at risk of disease or provide precision treatment (such as transcranial magnetic stimulation, prominent in Parkinson’s disease research, recently used in OCD).

Given how little we still know of the neurobiological underpinnings of mental illnesses in particular, it seems that developing biomarkers of disease and potential treatment response would be a welcome addition to the somewhat outdated tools of using the DSM and trying medications until something sticks. Some recent papers have shown advances in this field, albeit with potential replication issues. Data-driven approaches, utlizing previously unavailable huge datasets and machine learning models could additionally be combined with model-based approaches (such as reinforcement learning theories) in computational psychiatry to determine potential loci of pathology/intervention.

However, some caution is needed. Ethics in artificial intelligence has been a growing concern with the increasing prevalence of these methods and indications of potential gender, race and other biases, so ensuring fair and unbiased training data and interpretations is vital. Additionally, the infrastructural cost of precision medicine is something that is not often addressed, such as the costs of keeping terabytes of individualized data, running complex machine learning methods on multiple GPUs and communicating with scientists in the field to ensure state-of-the-art approaches are used.

To close with some food for thought, some interesting observations cast doubt on the content of resting state activity in general, from the paper I started this post with. Evidence suggests similar structure of resting state activity between wake and sleep, wake and anaesthesia, subjects in a population, and even across mammalian species. This can either mean that resting state activity does not amount to much outside of physiological confounds (e.g. breathing or heart rate artifacts in the data) or, on the contrary, that it captures some universal aspects of brain architecture, processing and cognition.