Neurosurgeon Dr Casey Halpern discusses deep brain stimulation, compulsive behavior, and emerging brain therapies
Andrew Huberman interviews Dr Casey Halpern, a neurosurgeon and chief of stereotactic functional neurosurgery at Penn Medicine.
Summary
Andrew Huberman interviews Dr. Casey Halpern, chief of stereotactic functional neurosurgery at Penn Medicine, about deep brain stimulation (DBS) and its applications for compulsive and impulsive disorders. Dr. Halpern explains how DBS — originally developed for movement disorders like Parkinson's disease — has revealed unexpected psychiatric benefits, leading researchers to investigate its use for OCD, binge eating disorder, addiction, and depression. He describes his laboratory's work identifying "craving cells" in the nucleus accumbens and using intraoperative recordings to find precise neural targets for stimulation. The conversation also covers stereo-EEG, an invasive electrode-recording technique borrowed from epilepsy care that is being applied to psychiatric disorders such as depression (at Baylor and UCSF) and OCD, as a pathway to defining new targets for non-invasive therapies. Non-invasive approaches such as transcranial magnetic stimulation (TMS) and MRI-guided focused ultrasound are discussed, as is the potential for machine learning to detect impulsive or compulsive episodes before they occur.
Key Takeaways
FULL TRANSCRIPT
Introduction to Neurosurgery and Deep Brain Stimulation
Andrew Huberman: Casey — I should say Dr. Halpern — welcome. You're a neurosurgeon, which I consider the astronauts of neuroscience. For those who aren't familiar with the differences between neurosurgery, neurology, and psychiatry, could you educate us a bit? What does a neurosurgeon do, and how do you think about and conceptualize the brain?
Dr Casey Halpern: The scope of neurosurgery is quite broad. We take out brain tumors, we clip aneurysms in the brain, we take care of patients who have had traumatic brain injury, concussion, spine surgeries — probably 90% of what neurosurgeons do around the country, taking care of herniated discs and lumbar fusions. The scope covers the entire central nervous system, including the peripheral nervous system. We take care of patients with carpal tunnel syndrome and nerve disorders. Historically, neurosurgeons did everything in that domain, but now we subspecialize. I'm lucky to be at Penn Medicine, where we can focus on one of these areas. I'm chief of stereotactic functional neurosurgery. All I do is deep brain stimulation surgery, and a complement to that is focused ultrasound — or transcranial focused ultrasound — which is a non-invasive way to perform an ablation in the brain. It's recently FDA approved, currently for tremor.
Deep brain stimulation is a procedure where we place a very thin, insulated wire deep into a part of the brain involved in Parkinson's disease, for example. But that's actually not the therapy itself. The therapy is delivering electrical stimulation through the tip of that wire — or one of the tips, as there are actually multiple contacts at the bottom of the wire. They're very small. It's a bit more like I have to implant a tool to deliver a medication, but that medication is going to be in the form of electricity, delivered into a very small region of the brain.
I'm very privileged to be able to interact with the human brain in this way. It's always with the goal of trying to provide somebody with a meaningful therapy. When we deliver electrical stimulation, these electrodes — while they might be sitting in a very small region of the brain — are near regions within a few millimeters that, if stimulated, could cause a temporary, very brief side effect: a moment of laughter, or a moment of panic. We can just shut that electrode off. But often these side effects could actually be therapeutic. That's how we have discovered ways to use deep brain stimulation not just for movement disorders like Parkinson's disease, but for patients with Parkinson's who have a psychiatric comorbidity like depression or obsessive-compulsive disorder. A lot of these patients are highly compulsive and impulsive. Sometimes these problems actually melt away while we're trying to help their tremor — patients tell us their gambling issue has gotten better, or their mood has improved.
Why is that? Well, there's probably more than one reason. You can help somebody's mood by making their tremor go away, of course. But we also see laughter in the clinic sometimes. That's because we're stimulating parts of the brain that are not just involved in motor circuits, but also in what we call the limbic circuit — the part of the brain involved in emotion. If we learn how to modulate those areas therapeutically, step by step, we can actually develop these therapies for other indications like depression.
I would say the most impressive and consistent effect we have is when a patient with tremor who has been tremoring for the past 20 years receives stimulation through that electrode in the clinic — we get immediate relief of tremor. That is the effect that inspired me to become a neurosurgeon when I was in college. I've never really wanted to do anything else except help develop that type of therapeutic for another kind of symptom.
OCD: Definition, Brain Circuits, and Current Treatments
Andrew Huberman: I'd love to learn more from you about OCD. Could you tell us what OCD is, what brain areas are involved, what the current range of treatments is, and what's the difference between someone who is obsessive and somebody who has true OCD?
Dr Casey Halpern: My perspective on OCD may be a little different from a psychiatrist who lives and breathes OCD and sees patients every single day. I probably take care of three to five patients a year with deep brain stimulation for obsessive-compulsive disorder. So I don't see these patients as routinely, but my laboratory as a researcher is very focused on trying to improve outcomes of deep brain stimulation for OCD. I do feel I have expertise and a perspective to share.
As a neurosurgeon, I feel obligated to better understand where the obsessions in the brain come from and how we can interrupt them to stop the associated compulsion — better than we're actually doing it now. I've been leading an endeavor with a number of collaborators around the country to better understand these circuits in the brain and study them in humans both invasively and non-invasively. That would be with an electrode-based surgery, similar to what we do in epilepsy to understand where seizures come from — we want to understand better where obsessions come from. But we're also working with imaging experts and geneticists to understand OCD at a broader level.
I consider OCD to be a spectrum disorder in a way. I apologize to those who might feel I'm using that term incorrectly — I'm using it to describe patients who have obsessions and even some related compulsions that might not meet criteria for OCD. As a neurosurgeon, I'm really obsessive about safety and compulsive about my surgical procedures. Some aspect of OCD — which we often joke about but should consider seriously, because people do suffer from this — some aspect of it helps us. There are famous CEOs that probably have some level of OCD, as well as surgeons and scientists alike. Perhaps if it can be controlled, it's an asset. But if it goes awry and is uncontrollable, it becomes obsessive-compulsive disorder. I tend to see the patients who are the most severe — those who have failed medication.
There are multiple medications worth trying for OCD, and some can actually be very helpful.
Andrew Huberman: Which neurotransmitter systems do they tend to target?
Dr Casey Halpern: SSRIs are sort of the first line for OCD, but tricyclics can also be helpful. So this is still the serotonin system. But as we know, the serotonin system interacts with the noradrenergic system and the dopamine system, so it's hard to be specific to one of these things. I think that's also why it's hard to predict how these medications are going to work for these kinds of patients. But tricyclics and SSRIs can be very helpful and are definitely first line, and there are others.
Exposure and response prevention is probably the most effective option — it's a kind of cognitive behavioral therapy, but these are distinct and offered by psychologists. There's a whole field and a whole clinic at my institution focused on this, started by Edna Foa at Penn. What they do for these patients is offer these types of cognitive therapies: exposure to the stressor, trying to get patients to habituate to whatever it is that stresses them and causes these compulsions, to help these patients live every day and function.
These are all fabulously helpful therapies for a variety of patients, but there's still about 30% of patients who still suffer from OCD, and some of them have severe OCD. Sometimes it's moderate to severe, and those are the patients I'm really motivated to try to help. Our therapies for those patients right now I would say are worth pursuing but not optimal. It's one of those things we have to balance as a researcher, because when you see patients like this you want to do everything you can to help them.
It's important to educate patients on the risks and benefits — this is deep brain stimulation surgery, but also capsulotomy, which is more of an ablation approach. Rather than delivering stimulation through an electrode, you can actually heat the tissue and even destroy it. Some would say this part of the brain is very safe to destroy — it's kind of like an appendix. Others would say it's safer to modulate. I have seen patients do very well with these ablations. Sometimes the lack of effect is what's so amazing — you can actually traverse parts of the brain without having any adverse effects on a patient's function, at least that you can test. You can also destroy small parts of the brain — we're talking 3 or 4 millimeters in size. These little ablations can be really helpful for patients and have no obvious side effects that we can tell, perhaps after a short recovery from surgery.
Nonetheless, despite how safe they might be, these surgical procedures are still surgical procedures, and patients are hesitant to proceed — especially when they know that their chance of a transformative effect is quite low. We can generally achieve a responder rate of about 50%, and responders still have symptomatic OCD. So I'm really inspired to find a way to deliver these therapies in a more disease-specific or symptom-specific way.
Brain Targets: Cortex, Basal Ganglia, and the Nucleus Accumbens
Andrew Huberman: Were one to come into your clinic for this sort of work of ablations or stimulation, where would you first start to probe in the brain?
Dr Casey Halpern: This is a disorder of both cortex and subcortex. We find that areas in the cortex like the prefrontal and orbitofrontal cortex are not functioning the way they would in a non-OCD patient. They're often hyperfunctioning, and we need to find a way to try to normalize their function. Then there are projections to the subcortex — this is the basal ganglia, the caudate-putamen, or the dorsal striatum. These are interconnected with the ventral striatum. This is an area of the brain that I focus a lot of my energy on — the ventral striatum, which is not limited to but includes the nucleus accumbens.
This is an area of the brain that we know to be involved in gating reward-seeking behavior. When it's perturbed, it seems to gate compulsive behavior — meaning a rat will pursue a reward despite punishment, despite foot shock, for example. That can be similar to an OCD patient. They will check their home for safety until 3:00 a.m. and not sleep that night — doing something because of the urge, but despite the risk.
When our judgment consistently puts us at risk, that's where we have something like OCD. Contamination behavior, for example: if they feel contaminated, they will wash their hands for hours repeatedly, or if they drop their toothbrush on the floor, this will lead to a compulsive behavior of cleaning the toothbrush or brushing their teeth repeatedly. Very common symptoms that patients report to us or that we observe.
Patients with eating disorders — if they have binge eating disorder, they'll overeat; if they have bulimia, they might purge, despite the risks of these things. Addiction is similar — we tend to drug-seek if we're addicted. We'll pay off a dealer in order to get our fix despite the risk. That type of urge despite the risk is something I've always been really interested in, and it's a common denominator to all of these problems. These are some of the most common conditions in our society today. The nucleus accumbens and the cortical areas that send projections to these areas are probably at least one of the main circuits involved in these kinds of things.
Andrew Huberman: What is the nucleus accumbens? What roles does it play in healthy brain behavior and in pathology?
Dr Casey Halpern: The nucleus accumbens is a part of the brain, part of our reward circuits. It has a lot of functions and interconnects with many parts of the brain. When I started getting interested in reward and what I could do as a surgeon to try to improve how we manage rewards — what I mean by that specifically is: if you have an urge for a reward, that's a normal phenomenon. That's not something we're trying to stop. The issue is if you have an urge for a reward that either puts you or somebody else at risk — it's probably a reward we shouldn't have. If you're a drug addict and you use heroin or opiates, that opiate might make you feel better because life is stressful. But the risk of doing those things is really high — in fact, potentially lethal. If you have OCD and you can't sleep at night because you're so nervous that you didn't lock the door and you've checked 30 times, that's an urge we've got to treat. Eating disorder is the same.
This problem can be ameliorated or improved upon by a better understanding and a tailored treatment targeting the nucleus accumbens. Specifically, it seems that repeated exposure to something like a drug of abuse — or any type of very strong reward — can in a way hijack normal functioning of the nucleus accumbens. The goal is to disrupt what is kind of habitual, or at least this kind of recurring problem that is happening.
Identifying "Craving Cells" in the Operating Room
People with binge eating disorder, at least at a severe level, tend to binge about once a day. What we decided to do in the operating room was to leverage a tool that we use all the time when we take care of patients with Parkinson's. With Parkinson's, a lot of these patients — not all — have tremor. When we place an electrode into the motor structure to try to improve their movement disorder, we can often hear tremor cells. We convert their electrical signal to an audible signal, so we can actually hear it, and it sounds like the tremor looks — the frequency of the signal is the same as the hand shaking.
Andrew Huberman: So you're poking around in a dedicated, careful way — one poke at a time.
Dr Casey Halpern: One poke at a time, with a very fine wire — a set of wires — listening to the electrical activity until you encounter some cells that are sending out electrical activity at a similar frequency.
Andrew Huberman: And then you can stimulate them or quiet them and see if the tremor goes away.
Dr Casey Halpern: We are very confident that when we stimulate that area — in this case the subthalamic nucleus — we will disrupt that tremor circuit and that tremor will dissolve. And it does.
Andrew Huberman: So what is the analog to tremor in terms of appetite and the desire to binge?
Dr Casey Halpern: Craving. There are probably other terms we could use, but that's the term we've chosen for a number of reasons. One is because people relate to it. People with binge eating disorder or obesity — if you ask them if they crave, the answer will often be yes. If you ask them if they lose control or binge, they might not know what you mean, or they might not actually feel out of control even when they are. But the word craving is relatable.
So we set out to see if we could identify craving cells. In a patient with OCD — which is related; in fact, we target a very similar part of the brain — we tried to identify cells related to obsessions, and we believe we did. It was a single case study where we tried to optimize where our electrode was placed. So we had some proof of concept that we would be able to elicit a disease-specific symptom in the operating room, assuming the patient could tolerate being awake. Not everybody needs to be awake for this procedure, but at least for these first-in-human trials where we're trying to establish where in the brain we need to be, I think this type of approach is really critical.
Non-Invasive Brain Stimulation: TMS and Focused Ultrasound
Andrew Huberman: What is the status of non-invasive brain stimulation, ablation, and blocking activity in the brain? My understanding is that transcranial magnetic stimulation is being used to treat depression and a number of other brain syndromes non-invasively — no drilling through the skull. My understanding is that the spatial precision isn't that great. Ultrasound is something I hear a lot about these days, and my understanding is that ultrasound can allow researchers and clinicians to stimulate specific brain areas. What are your thoughts on these forms of non-invasive brain stimulation and blockade of brain activity?
Dr Casey Halpern: We need to embrace non-invasive approaches. Some of them are a little fluffy in that we don't understand how they work — though we don't necessarily understand exactly how deep brain stimulation works either. Because we don't know exactly how they work, they're not as precise as we would like them to be. We have work to do there, and I actually think that work is doable and is underway.
TMS — transcranial magnetic stimulation — is FDA approved for depression. It's also FDA approved for OCD and for nicotine addiction. We believe we can use TMS to define a circuit that, if modulated, improves OCD, albeit temporarily. In those patients, if it's temporary, they would be appropriate for an invasive study. That's something we're actively working on. I've always believed that neurosurgeons need to be part of the discussion with these non-invasive approaches. We don't need to do them ourselves, but I think we can help make them more precise and to probe non-invasively with purpose. Perhaps one day there will be a TMS target for anorexia and obesity. If we're scratching the surface with invasive approaches to these problems, we're doing even less with non-invasive brain stimulation. We have so much work to do there. Eating disorders and TMS have been so scarcely studied, and it is an area we need to work on.
Transcranial MRI-guided focused ultrasound is an FDA-approved method to deliver an ablation to the brain non-invasively. There are researchers — myself included — trying to use MRI-guided focused ultrasound in a modulatory way, not just as an ablation, but to drive neuronal activity or inhibit it. We're still learning how to do that. There are trials trying to understand if you can use ultrasound to open the blood-brain barrier so you can deliver a medication to a specific area — perhaps for a brain tumor or something like that.
It's a very exciting field, and it is FDA approved for tremor right now. I actually do it routinely for patients with tremor from Parkinson's or essential tremor, and I love doing it. It's often just kind of a miracle because there's no incision. I don't have to place an electrode into the brain to achieve a similar result. It's fabulously effective for these patients. It treats patients on one side, usually their dominant hand or their worse hand. It really speaks to the fact that you can deliver non-invasively an ablation to the brain in a hypothesized zone that we think is related to the problem at hand. At least with tremor, it works really well.
Could this be effective for psychiatric disease, obesity, eating disorders? Perhaps — actually, that would be the ideal. The problem is we don't know where to do the ablation. There is a trial we would like to do for OCD where we would deliver an ablation to the same area of the brain that we've been delivering ablations to for years for patients with OCD — it helps a bit, and that's called a capsulotomy. But the outcome is probably going to be about the same. It's a nice method because it's non-invasive, but we need to find a new target for these conditions.
Because of the common denominator of the urge despite the risk — that compulsion — perhaps it could be the same target. I don't know. But I would argue we need to do these modulatory experiments either with a device or with invasive recordings to better understand where these problems are coming from, to define where we should do an ultrasound treatment.
Stereo-EEG and Invasive Recording for Psychiatric Disorders
There has been a revolution in America — it was in Europe before it was in America — where we do stereo-EEG, which is basically like doing an EEG of patients with epilepsy but with invasive electrodes. We place tiny little wires, less than a millimeter in diameter, all throughout the brain into parts we believe are involved in seizures. We admit the patients to the hospital, figure out where the seizures are starting and propagating, and then we can stimulate these electrodes to see if there's a symptom that's important, and try to identify a region that we think we could either remove surgically, ablate with a laser, or put a stimulator in. That's commonplace now for epilepsy. It works extremely well and is very safe. Of course, it's still a brain procedure, but the complication rate is surprisingly low for the number of electrodes we place, and it's extremely well tolerated. Most of these patients leave the hospital and don't even feel like they've had surgery.
There's actually a lot of interest in using that procedure to study mental health disorders. We are trying to do it for patients with obsessive-compulsive disorder — we're awaiting an FDA decision on that. I credit our colleagues at Baylor and at UCSF for already studying this, bringing together the epilepsy technique and the psychiatry expertise to study how we could better target electrodes in depression. If they have a consistent target, perhaps there becomes an ultrasound target. Right now the approach is a bit more reversible because you can always shut that electrode off or even remove it if it's not in the optimal location to treat the depression. After a large volume of cases, perhaps they could pool that data to develop a new ultrasound target for depression. I think that would be fabulous, and it's probably their long-term goal — not to speak for them, but I'm sure it's on their radar.
You might ask why we aren't doing this for obesity right now in our study. The reason is that we've developed a target for obesity and binge eating disorder out of mouse models, and we believe it's relevant for the human state because you can model this problem in a mouse a bit better than you can model depression or OCD. We feel like we can rely on the pre-clinical studies more. Whereas with these more human mental health conditions that are hard to model in a mouse, you really have to study it in the human. You can perhaps start in an epileptic patient — a patient who has electrodes — and try to provoke a depressed state, or study epileptics who have comorbid depression. That can really validate this approach as well. But in the end, it's getting into the human brain in the disease specifically that will eventually lead to a non-invasive approach — either a lesion or modulatory approach. Modulatory would be like TMS; a lesion approach would be with ultrasound.
Awareness, Machine Learning, and Predicting Compulsive Episodes
Andrew Huberman: If people can be made to feel — or make themselves feel — just a little bit better, a little less anxious just prior to a craving episode or a binge episode, maybe even if people can become better at detecting their own internal states and when they're veering toward a binge, or veering toward using a drug, or maybe even veering toward suicidal thinking — that awareness seems like it might be among the best tools that people could develop.
Dr Casey Halpern: Yes, I've always thought that if we can improve awareness, we can improve outcomes. I think that's probably true for many of these patients. The problem comes down to the fact that some of these patients are so resistant to treatment. The patients we see as surgeons are the patients who have tried cognitive behavioral therapy, certainly have tried medications, have tried behavioral management — they're as aware as they could possibly be, and they still lose control.
We've studied this in the lab. We bring patients to the laboratory with the implanted device to try to provoke this electrographic signal that can be detected by the actual device that will stimulate them when they're at home. Before we actually initiate stimulation, we want to see: can this device detect this craving cell signal? It's going to be different from what we saw in the operating room, because that was a single cell. These devices, these electrodes, are about a millimeter in diameter instead of about a tenth of a millimeter, which is what we use in the operating room. So they're only detecting thousands of cells' responses.
We actually have a way to provoke binges — it's called a mood provocation. It's very well validated. It's a little bit like provoking seizures in the epilepsy monitoring unit, but here in the sort of psychiatric monitoring unit or the food monitoring unit, we have a psychiatrist and eating disorder specialist come in and induce a mood that is related to each patient's self-described binge episode.
Andrew Huberman: So the psychiatrist comes in and provokes a feeling that can evoke the negative behavior.
Dr Casey Halpern: That's exactly right. We video the session and synchronize the video to the brain signal recordings. The patients all wear an eye tracker so we can see what they're eating at all times and what they're looking at specifically. That allows us to have the best temporal resolution possible to understand what is happening right before the bite.
Even under video surveillance, through a one-way mirror in a laboratory setting, when patients are very well aware that they're there to be studied and that they might binge — they still do. We believe they do because they just can't control it, as aware as they are of it. It's probably because they're the most severe.
I think if we can improve awareness — not just the societal awareness I was talking about earlier, but the patient's own awareness around their problem — that could be a powerful way to help so many of these patients. That's sort of the role of cognitive behavioral therapy. The limitation of it — and I actually don't have any problem with it; I think it's a wonderful treatment — is that if you stop it, many of these patients go back to their old behaviors. It's not necessarily lasting in the absence of continued cognitive behavioral therapy. Some people can benefit from it long term, but some can't. In the less severe patients, improving awareness is key. But in these really refractory patients, this is the disease — despite the awareness, they can't control themselves. What we're trying to restore is that improved ability to control their behavior.
Andrew Huberman: Do you think there's a role for machines and artificial intelligence here? There are a couple of laboratories up at the University of Washington that are using particular signature patterns within voice to try and help people who are suicidally depressed know when they're headed towards an episode before they can even consciously know. This gets right down to issues of free will and whether or not machines can be smarter than we are. One could argue that some of the search algorithms on Google and other search engines are actually more aware of our preferences than we are. Basically, these are devices that are listening to people talk all day, paying attention to patterns of breathing and how well people slept, integrating a huge number of cues, and then signaling somebody with a yellow light — "you're headed into a depressive episode" — and the person might say, "I feel fine, this is kind of baseline state for me," and the device says, "Uh-uh, this is where you were preceding the last episode that took you down a deep, dark trench and it took months to get out of." I wonder whether some of these devices could help with the sorts of things we're talking about today.
Dr Casey Halpern: I think so. I've always said we have to get in the brain before we get out of it. If we get in the brain and understand what these signals look like, we'll know what those non-invasive signals are. I think it's possible that we are scientifically sophisticated enough to use machine learning and this kind of bot technique to anticipate when somebody is going to be highly impulsive.
Suicide is the most dangerous impulse. It's something that is immensely a focus of the lab — impulsivity. We've talked mostly about compulsion. Compulsion being going after a reward, or the urge despite the risk. Impulsivity is similar but different. If you model impulsivity in a mouse, it's related to going after a food reward without the paired tone that the mouse is supposed to wait for. The mouse doesn't want to wait anymore — they just go after the food.
Andrew Huberman: I've been that mouse.
Dr Casey Halpern: We've all been that mouse — we can all relate with this to a certain extent. Again, it's a spectrum.
In any case, I certainly think that there is a way to use our own body's physiology to anticipate when these impulses are coming online. How best to do that — I think we're just scratching the surface. But these are the kinds of solutions we need. Some of these problems are of epidemic proportions: the largest public health problems in this country and in this world — obesity, the opiate crisis, depression, suicidality. That's like a third of our country, maybe more. We need scalable solutions.
I'm a neurosurgeon. I'm only going to be able to treat the most severe of patients with these problems. We've only done about 200,000 deep brain stimulation surgeries ever. The problem we're talking about here is 50 million Americans. There's no possibility that surgeons can address that problem. But we could help inspire an initiative to go after that kind of problem, or help make it more rigorous — because the last thing we need is some sort of wearable fancy tool that wastes people's money and time. We need real therapies for these things.
Not that these devices we're discussing lack promise — I think there's actually a lot of promise, and we use machine learning in the lab all the time. I'm not an electrical engineer or a computational neuroscientist doing this type of work; I just help develop the hypothesis around it and help fundraise around it. But I definitely think there's a future for it. I just suspect we're scratching the surface on how best to do it.
Andrew Huberman: I really appreciate you sharing those tools. A number of people out there might want to become neurosurgeons. I really believe that in hearing today's conversation, you will spark an interest in medicine and neurosurgery. On behalf of everybody, and myself as well, thank you so very much.
Dr Casey Halpern: I'm honored. Thank you so much for having me.