Our research aims to fill a fundamental gap of knowledge about the timing, location, and causal importance of specific neuronal populations in the brain that work together in the millisecond scale to subserve a given brain function.
We record directly from inside the brain in neurosurgical patients that are implanted with multiple electrodes across different anatomical and functional systems. We also apply direct electrical current to specific populations of neurons to alter their function while testing the effect of such perturbation on the human participant subjective feelings or task performance.
Our research is beneficial to each individual patient who volunteers to participate in our cognitive and behavioral experiments because we map the location of functional units within each patient’s brain and share this information with clinicians to make more precise and safer surgical plans and prevent major cognitive deficits after surgery. We also map the location of pathological activity and use the data to locate the source of seizures and the pathways for their propagation in each individual patient’s brain.
Our work is also relevant to public health and has societal impact. We strive to collect novel information about the functional architecture of the human brain, and improve our understanding of how the brain works. This will be vital for our understanding of the pathophysiology of neurological and psychiatric disorders that affect higher level cognitive functions and cause major problems for afflicted individuals and their families and the society.
We study human brain function at multiple levels of cognitive and behavioral processing. We study brain activity from the very early sensory input to very late decision making in even social or emotional domains. We do not focus on a specific area of the brain or on a narrow field of cognitive neuroscience. As documented by our published work, every level of human cognition, every stage of human brain function, and every regions of the human brain - are of interest to us – as we want to understand how different areas of the brain work together across different experimental tasks. For instance, we have studied the prefrontal cortex (PFC) as we have recorded directly from the human periaqueductal gray (PAG) and we have electrically stimulated the human default network as we have stimulated the human hypothalamus. Our goal is to acquire a universal understanding of the functional architecture of the human brain with millisecond and millimeter precision.
In our studies, we use a combination of intracranial electroencephalography (iEEG), functional MRI (fMRI), and intracranial electrical stimulation (iES). Here we explain each method very briefly:
Intracranial EEG (iEEG)
When EEG recordings are obtained from inside the skull with intracranial electrodes, we refer to it as intracranial EEG (iEEG). If strips or grids of electrodes are implanted over the bare cortex in the subdural space, we use the term electrocorticography or ECoG. If electrodes are inserted as thin wires of intracranial electrodes through small burr holes in pre-defined stereotactic coordinates of the brain without opening the skull bone, we refer to it as stereotaxic-EEG or simply stereo-EEG or sEEG.
Human iEEG can complement other methods of neuroscience beyond simply replicating what is already known, or can be known, from noninvasive lines of research in humans or from invasive recordings in nonhuman mammalian brains. For more information see: Parvizi J, Kastner S. Promises and limitations of human intracranial electroencephalography. Nature Neuroscience. 2018
Functional MRI or fMRI
This is an MRI scan of the brain that reveals where the brain is more activated or less activated during a task. If it is done during rest, it is called resting state fMRI or rs-fMRI to reveal which regions of the brain have synchronous fluctuation of activity (hence they might be working in unison and may be part of a so-called “resting state or intrinsic functional network”.
In each patient before surgical procedure, we obtain task or resting state fMRI. After the electrodes are placed, we can superimpose their location on the fMRI space and identify electrodes that are located in a specific functional region or in a specific intrinsic resting state network.
In this figure you see location of electrodes shown in the resting state fMRI space. Each color denotes a specific resting state network.
As another example, you see here how a patient’s two electrodes (1 and 2) are located exactly at the center of fMRI defined region of the brain that is specialized in processing faces.
Intracranial Electrical Stimulation (iES)
Direct electrical stimulation of the brain provides unique insights about the causal importance of a given brain region and its anatomical network in a specific set of functions. During this procedure, we inject electrical current through implanted electrodes when the patient is either relaxing or doing an experiment. We probe how their subjective state changes with real or sham electrical stimulations or how their performance in a given experiment is altered due to electrical perturbation of neurons in a given region of their brain.
In a series of experiments, we have probed the causal importance of specific brain areas (and their interconnected networks) in specific functions. Here are some examples:
A global brain map of stimulation effects. Unique responses are observed with stimulation of each specific region of the brain
Fox KCR, Shi L, Baek S, Raccah O, Foster BL, Saha S, Margulies DS, Kucyi A, Parvizi J. Intrinsic network architecture predicts the effects elicited by intracranial electrical stimulation of the human brain. Nat Hum Behav. 2020. Epub 2020/07/08. doi: 10.1038/s41562-020-0910-1. PubMed PMID: 32632334.
What happens when you stimulate the fusiform face area
Watch 2 examples below:
Parvizi J, Jacques C, Foster BL, Witthoft N, Rangarajan V, Weiner KS, Grill-Spector K. Electrical stimulation of human fusiform face-selective regions distorts face perception. J Neurosci. 2012;32(43):14915-20. Epub 2012/10/27. doi: 10.1523/JNEUROSCI.2609-12.2012. PubMed PMID: 23100414; PMCID: PMC3517886.
Schrouff J, Raccah O, Baek S, Rangarajan V, Salehi S, Mourao-Miranda J, Helili Z, Daitch AL, Parvizi J. Fast temporal dynamics and causal relevance of face processing in the human temporal cortex. Nature communications. 2020;11(1):656. Epub 2020/02/02. doi: 10.1038/s41467-020-14432-8. PubMed PMID: 32005819; PMCID: PMC6994602.
Like every other method our approach has its own limitations and this is how we deal with them:
The iEEG signal is obtained in clinical populations with medication resistant epilepsies.
We actively minimize the confounding effect of epilepsy on the acquired data by the following measures: 1) we exclude patients with diffuse brain disease; 2) we exclude electrodes with epileptiform discharges (more than 80% of recording sites are usually void of epileptic activity); 3) we exclude trials coinciding with epileptic discharges; 4) we obtain data several hours outside the window of seizures; and most importantly, 5) we confirm that findings are anatomically and functionally consistent across a number of patients - each with different source and type of seizures. When possible, we also show that findings in our clinical population are akin to findings reported in non-invasive studies of healthy subjects.
Electrodes cannot be implanted wherever we want to have recordings from.
The implanted electrodes are inserted only where there is a clinical (and not research) need for recording. Patients with epilepsy decide to go through the risky procedure of invasive recordings in order to gain seizure freedom. Hence, the location of electrodes is only decided by clinicians and solely for clinical monitoring purposes. We have to live with this limitation and we have no ways around it.
The recording sites are sparse and may vary across patients.
We mitigate this problem by collecting data from a relatively large number of patients. Each patient is usually implanted with about 100-200 electrode contacts. Pairs of regions of interest will always be covered in a sufficient number of patients.
The iEEG signal captures averaged signal and not firing rate of individual neurons. The iEEG method is not for deciphering computations at the local circuit level. However, it is suited for mesoscale level studies of timing and causal importance of a population of neurons within millimeters of the brain as well as their relationship with other remote regions.
Recently, we confirmed that non-lesional epileptic tissue has normal physiological responses to relevant cognitive stimuli, but their responses are “seized” by ongoing spontaneous epileptic activity in a window of -1050ms to +200ms around the stimulus onset.
Liu S, Parvizi J. Cognitive refractory state caused by spontaneous epileptic high frequency oscillations in the human brain. Science Translational Medicine. 2019
Highlights of Our Published Work:
Recordings with ECoG and FMRI in the same subjects show that a given population of neurons within a specific coordinate of the primary visual cortex respond to visual stimuli in a specific visual field. Stimulating the same neurons will give rise to phosphenes in exactly the same location of the visual field. We can even predict how much of the cortex a specific dose of electrical current will activate and how the person’s subjective experience will change.
Numbers in the Brain: Where in the brain 2 and 2 are added together
In 2013, we discovered a special area within the posterior inferior temporal gyrus (pITG)
Shum J, Hermes D, Foster BL, et al. A brain area for visual numerals. Journal of Neuroscience. 2013
In a group of subjects with simultaneous recordings in the lateral parietal and inferior temporal regions we tracked neuronal responses as subjects were doing simple additions (e.g. “2”, “+”, “2”, “=”, “4”). Non-selective responses in visual areas (LOG) and medial fusiform gyrus (mFG) were nicely contrasted with the selective responses to numerals in the posterior inferior temporal gyrus (pITG, red) and anterior intraparietal sulcus region (aIPS, upper left panel).
Most importantly, we found that neuronal populations that are just a 3-5 millimeters apart from each other show clearly different profiles of responses. Note the most selective responses to numbers in the number form area (NFA) that has previously been reported in a different set of subjects.
Daitch AL, Foster BL, Schrouff J, et al. Mapping human temporal and parietal neuronal population activity and functional coupling during mathematical cognition. Proceedings of National Academy of Sciences U S A. 2016
We successfully invented a method to decode the brain activity during daily life events. We showed that the parietal region that is activated during experimental task of numerical cognition is also activated when subjects freely utter words that denote magnitude, time, or distance. At the single word level, we could see activations in the intraparietal sulcus regions.
Dastjerdi M, Ozker M, Foster BL, Rangarajan V, Parvizi J. Numerical processing in the human parietal cortex during experimental and natural conditions. Nature communications. 2013
Hermes D, Rangarajan V, Foster BL, et al. Electrophysiological Responses in the Ventral Temporal Cortex During Reading of Numerals and Calculation. Cereb Cortex. 2017;27(1):567-575.
Baek S, Daitch AL, Pinheiro-Chagas P, Parvizi J. Neuronal Population Responses in the Human Ventral Temporal and Lateral Parietal Cortex during Arithmetic Processing with Digits and Number Words. Journal of Cognitive Neuroscience. 2018
How faces are processed in the brain
A specific patch of the ventral temporal cortex known as the fusiform face area (mFUS and pFUS) show increased fMRI activity when a human subject views faces. The same area when recorded with ECoG shows highly selective responses to faces only. If you stimulate the same area in the same subject’s brain, you can cause distortion of face perception.
Parvizi J, Jacques C, Foster BL, Withoft N, Rangarajan V, Weiner KS, Grill-Spector K. Electrical stimulation of human fusiform face-selective regions distorts face perception. Journal of Neuroscience 2012
In another study we replicated prior findings and recorded responses in the fusiform face areas to human faces, monkey faces, bird faces, and marine (fish or dolphin) faces. We found sites with selective responses to faces, and using Multi Kernel Learning Method (Schrouff et al 2016), confirmed once again that selectivity of responses is best measured in high frequency broadband (HFB) instead of other narrowband alpha, beta, theta, or gamma rhythms.
Our findings clearly demonstrated that activations in the fusiform face areas are highest for human faces (pink), then monkey faces (blue) and bird faces (orange) and are lowest to marine (fish or dolphin) faces (green). Face selective areas (averaged across different sites) respond weakly to non-face stimuli (gray).
We showed that face related electrophysiological responses in the ventral temporal cortex (VTC) were seen earliest in the most face selective areas and then in less selective areas (that are more anteriorly located) indicating that the face information is only distributed in time: early modular and then late distributed.
We then used machine learning tools to prove that information in non-selective sites is redundant.
Schrouff J, Raccah O, Baek S, et al. Fast temporal dynamics and causal relevance of face processing in the human temporal cortex. Nature communications. 2020
Human Brain at Rest:
We have shown that a part of the brain called default network is activated immediately after transition to rest (within ~200-300ms) and that this rest activation is not due to mind wandering but that it makes mind wandering possible.
Dastjerdi M, Foster BL, Nasrullah S, et al. Differential electrophysiological response during rest, self-referential, and non-self-referential tasks in human posteromedial cortex. Proc Natl Acad Sci U S A. 2011
In another study we showed that the resting state connectivity in the brain exists between neuronal populations that are activated together during task. This connectivity persists between functionally co-activated neuronal ensembles even during deep sleep.
Furthermore, we showed that the resting connectivity in the brain extends to the connectivity between the brain and the body. Using simultaneous pupillometry and intracranial recording from the human insula, we showed that spontaneous activations occur within the same neuronal populations in the human insula and with the same spectrotemporal profiles as task-evoked activations. Interestingly, the spontaneous insular activations are time locked to pupillary responses in a manner similar to task-evoked brain and pupillary responses.
Foster BL, Rangarajan V, Shirer WR, Parvizi J. Intrinsic and task-dependent coupling of neuronal population activity in human parietal cortex. Neuron. 2015.
For more information see list of Publications.