ICE – Direct brain recordings
My primary research method is Intra-cranial Electrophysiology, or ICE, which is a rare method for studying human brain function at a fine-grained level not possible with other current techniques.
The method relies on patients who are undergoing brain surgery in order to cure epilepsy (we thank them sincerely for their magnanimity!). As a normal part of the procedure, neurosurgeons carefully place electrodes directly in contact with the patients’ brains. These electrodes are left in place for some days while the patients spend time in the hospital (the insertion sites are all covered and bandaged).
The surgeons wait for seizures to occur naturally, and then the power of the electrodes is unlocked: they sense the exact source of the seizures the way seismographs sense the source of an earthquake. Seizures may start in a small piece of unhealthy tissue deep in the brain. If the surgeons identify that tissue, and remove it, the patient’s epilepsy can be cured. The identification has to be impeccably precise, of course, and not all configurations of seizure-causing tissue can be treated by surgery, which is one reason they must use the electrodes to precisely guide their work.
Importantly, surgeons also use the electrodes to identify and map all the healthyband operational brain tissue around the surgical target. This allows the surgeons to design a trajectory to remove the epileptic tissue (which is usually deep in the center of the brain) while disrupting as little healthy tissue as possible – to avoid impairing their language or memory or other faculties.
Because the electrodes are placed broadly around the surgical target, to map and spare the normal functions, we can record directly from healthy tissue right where we hypothesize cognitive functions like language are computed. We of course never tell the surgeons where to place the electrodes – that decision is made for medical reasons only. Instead, we look at the surgical plans for each new patient, and only ask some of them if they will be willing to participate in the research – namely, those who will have electrodes in parts of the brain we study. We also use a set of strict further criteria to select patients for study here: for instance, that they must have left-hemisphere language (usually right-handed), no clinical language deficits, no psychiatric disorders, English as first and primary language, late onset of seizures within their life – usually after language faculty has largely developed, epilepsy focus far away from recording sites, and no seizures within several hours of testing. These criteria drastically reduce the number of patients on whom we can perform our research, but they also help assure that the data can be interpreted and can be trusted. I currently collaborate with five hospitals, in order to collect research data from enough patients.
If the patients give us consent to do research, we simply have them look at laptop computer screens and perform our research tasks in moments in the hospital room when they have nothing else going on. (For instance, they might see words on screen, and have to say them in the right grammatical form for a given sentence.) Patients often have many hours of such time. We greatly appreciate their donating some of it to science, for all that it can do to assist our understanding of the human mind and brain. We view the privilege to access such rare and high-fidelity data as a responsibility – to do all we can to get the most out of the data.
For each patient, several researchers collaborate to run many tasks (experiments), so that science as a whole can get as much important information as possible. Some of my collaborators study epilepsy itself, and aim to develop cures and therapies. I study the language system – how the brain is wired to allow us to speak. Language is central to the human condition – allowing us to teach and learn, to know others’ minds, to form groups and even nations, and to transmit the details of our thoughts and emotions from person to person and from generation to generation.
Language is also central to human cognition, encompassing many of the elemental talents of the human mind, such as vision and memory, and drawing on a larger territory of the brain than any other faculty. Understanding the biology of language, the mechanisms used by brain cells to translate between thoughts and words, should allow us to understand in general how the brain computes the mind — and how to fix it when something goes wrong.
Importantly, language cannot be studied in animals, because only humans have language. (Communication systems of other animals, such as parrots and whales, are not as complex as ours, lack the central talent for grammar that allows us to express ourselves infinitely, and in any case currently lack any means of translating across species.) Therefore, the biology of language can only be studied in humans, and the ICE method brings us closer than ever before to actual biological mechanisms of language.
ICE electrodes record the electrochemical activity of brain cells and groups of cells, with extremely high temporal and spatial resolution. The temporal resolution is limited only by the recording computers (and disk space we wish to use), which continue to improve (we record much data at 20,000 samples per second); and the spatial resolution is limited only by the size of the electrode. Micro-scale electrodes can record from single cells. ICE also affords great physiological resolution. We can determine what kinds of cells in what part of the cortex are activate in what ways (for instance we can differentiate between bursts or firing and periods of sub-threshold activity; and can detect when cells oscillate together). This means that we can get down to the detailed mechanistic level when investigating the biological seat of cognition, especially language.
In contrast, most of our current understanding of how the human brain works first of all came from patients with brain damage (from stroke, for instance, or also from wartime injuries): doctors associated specific brain damage with specific impairments in mental abilities. This was approximate, protracted work that revealed large-order regions of the brain crucial to a given faculty, but could not reveal all regions partially involved in the faculty, nor the timing as to when in a series of processing stages a region might come into play. In recent times, much of our understanding of the brain basis of cognition has come from non-invasive imaging methods such as fMRI and MEG. These methods ushered in a categorical leap in what aspects of cognition could be studied, and continue to yield insights that change the way we understand our inner (mental) world. However, they suffer from coarse resolution in time or space or both, and are loosely coupled to neurophysiological events so cannot distinguish among physiological processes. Therefore it is not currently possible to use them to assemble, for instance, a model of speaking that includes the timing and location of processing steps required, nor a model of which sets of brain cells connect with which other ones when, in order to share information. It is necessary for information to be shared, for instance among computational entities in charge of word meaning, word sound, and word structure – so that these can be brought together when we utter the final word.
The way a recording session works is this: While the patient does one of our tasks (or is in some resting state) the computers that normally record their brain signals continue doing so, and later we are able to pair their data with what they were doing at the time. We get hundreds of channels of ICE data directly from sites within the brain, plus measures like heart rate and eye movements.
Coming soon, a description of my favorite part: Data analysis!