Peter Attia MD363 ‒ A new frontier in neurosurgery: brain-computer interfaces, new hope for brain diseases, & more
CHAPTERS
- 0:00 – 10:58
From Cushing to Minimally Invasive Neurosurgery
Peter Attia asks for a history of neurosurgery, and Edward Chang describes Harvey Cushing and Wilder Penfield as the founders of the modern field. He explains how craniotomy has stayed similar while laser probes, focused ultrasound and catheter-based vascular treatments have changed much of the work.
- •Cushing is described as the father of modern neurosurgery, and Penfield as a pioneer of epilepsy surgery and awake brain surgery
- •The main categories are tumors, vascular, spine and functional neurosurgery
- •Roughly ninety percent of aneurysm procedures are now done through a catheter rather than open surgery
- •Clot retrieval is making stroke treatment more like heart-attack care
- •Surgery in general is moving toward less invasive approaches
- 10:58 – 17:39
Glioblastoma: Biology, Treatments and Delivery Challenges
Chang explains what glioblastoma is, why it is so lethal, and how genetic profiling and immune-based approaches may change treatment. They discuss the lack of known risk factors and the prospect of using focused ultrasound to open the blood-brain barrier.
- •Glioblastoma arises from glial support cells and grows quickly enough to outstrip its blood supply, leaving necrosis
- •Molecular profiling is allowing treatment to be tailored to tumor mutations
- •The tumor suppresses the immune system, so making it recognizable to immune cells is a promising strategy
- •More extensive resection prolongs survival but is not curative because microscopic cells remain
- •Focused ultrasound can open the blood-brain barrier for targeted drug delivery
- 17:39 – 24:22
Why Chang Chose Neurosurgery and the Rise of Brain Mapping
Chang recalls seeing an awake glioblastoma surgery as a medical student and being hooked. He describes brain mapping, from Penfield's electrical stimulation to recording from the brain to understand how neurons encode language.
- •A neuroanatomy professor took him to see an awake surgery, which inspired him
- •Brain mapping protects language and movement areas while maximizing tumor or seizure-zone removal
- •Penfield used electrical stimulation, which is still used today
- •Recording from the brain reveals how neurons convey information about words, consonants and vowels
- 24:22 – 30:10
How Awake Brain Surgery Works
Attia asks how surgery can be done without general anesthesia. Chang explains that the brain lacks pain receptors, that the scalp and dura are numbed locally, and that light sedation is lifted only for the mapping portion.
- •The brain and bone lack pain receptors, while the scalp and dura are numbed
- •Patients are not intubated and the head is fixed in a holder
- •Light propofol sedation is used, with a low dose compared to general anesthesia
- •The patient is awake for only an hour or two of a six-to-eight-hour surgery
- 30:10 – 35:43
Brain Redundancy, Plasticity and Resecting Brain Regions
Attia asks how surgeons can operate when every region seems precious. Chang discusses the ten-percent myth, redundancy in the frontal lobes, and plasticity through synaptic changes over time.
- •Only about ten to fifteen percent of the brain is critical for basic functions, though that is a lot more than the myth implies
- •Parts of the frontal lobe are redundant, and Chang has removed an entire frontal lobe on one side
- •Slow-growing lesions allow the brain to reorganize
- •Synaptic plasticity underlies learning and compensation
- 35:43 – 43:50
Corpus Callosotomy and the Question of Consciousness
Chang explains how severing the corpus callosum is used for drop-attack seizures, how it limits seizure spread, and what dissociation syndrome involves. Attia asks about the philosophical implications, and Chang says he thinks of consciousness mainly in clinical terms.
- •Callosotomy limits seizure propagation between hemispheres but does not cure seizures
- •Usually the anterior two-thirds is cut to reduce dissociation side effects
- •The surgery is delicate because of the superior sagittal sinus and the pericallosal arteries
- •Chang thinks about consciousness clinically, such as coma and brain injury, rather than philosophically
- 43:50 – 54:40
Neural Engineering, Individual Variation and Sensory Decline
Attia argues that traditional medicine has largely failed in neurodegenerative disease and asks about engineering approaches. Chang describes neural engineering, explains how thought representations are similar in sensory areas but individual further upstream, and discusses hearing loss and cognitive aging.
- •Therapy has mostly aimed at slowing progression rather than replacing function
- •The brain is an electrical organ, and neural engineering aims to decode and guide neural signaling
- •Early sensory processing is highly conserved across people, while higher processing is individualized
- •Hearing loss can reflect a perceptual problem in the brain and may accelerate age-related memory loss
- •Visual decline is thought to have less of this effect, though Chang is less certain
- 54:40 – 1:09:02
What a BCI Is and the Trade-offs of EEG, ECoG and Intracortical Electrodes
Chang defines a brain-computer interface and compares scalp EEG, surface ECoG and electrodes inserted into the brain. They cover resolution differences, glial scarring, signal drift, and why ECoG offers a safe and scalable balance.
- •A BCI records brain signals and uses a computer to interpret them and act on them
- •ECoG sits on the brain surface under the dura and avoids injury from penetrating electrodes
- •ECoG offers roughly a thousandfold better resolution than scalp EEG
- •Single-neuron recordings are hard to keep stable because of immune scarring and micromotion
- •The trial used a percutaneous port with infection risk, and Chang expects a fully implantable wireless version in about a year
- 1:09:02 – 1:09:15
Anne's Story and the Bravo Trial Setup
Chang describes Anne, who had a brainstem stroke from a vertebral artery injury in her twenties and became quadriplegic and unable to speak. He explains the Bravo trial, the 253-electrode ECoG array, and the uncertainty over whether the speech areas would still function after so long.
- •Anne's stroke affected the brainstem via the basilar artery, which has no backup supply
- •She waited until her daughter graduated before joining the trial
- •The trial used existing medical materials, which helped win FDA approval
- •A credit-card-sized array of 253 electrodes was placed over speech motor cortex
- •The biggest unknown was whether those brain areas would still work after years of silence
- 1:09:15 – 1:35:38
Training the Decoder: NATO Words, Phonemes and AI
Chang explains how Anne trained the system by trying to say prompted words, starting with NATO code words and reaching near-perfect accuracy within a week. He then explains the AI decoding approach, which uses short windows of brain activity, speech units from a speech-recognition model, and a language model.
- •The decoder reads volitional intent to speak, not reading or inner monologue
- •Accuracy on the NATO words rose from about 50% on day one to the 95-100% range in about a week
- •Neural activity is decoded in 10-20 millisecond windows into speech units, then assembled with a language model
- •The approach draws on about 36 participants' earlier data for the neural code and on open-source speech recognition
- •Constant decoding also seemed to strengthen Anne's facial muscles, suggesting a rehabilitation role
- 1:35:38 – 1:41:08
Speed, Latency and the Engineering Path to Scalable BCIs
Chang reports that Anne reached about 80 words per minute versus roughly 150 to 160 for normal speech, and that a newer streaming approach cut latency below a second with the same hardware. They discuss the move to a wireless implant with more channels, treating the remaining work as engineering.
- •Anne averaged about 80 words per minute, versus roughly 150-160 in normal conversation
- •A 2025 study decoded in a streaming way with less than a second of latency using a new algorithm on the same hardware
- •Next-generation arrays would have about four times as many sensors and wireless connectivity
- •Chang says the proof of concept was the hard part and the rest is now largely optimization
- •Recent injuries would likely make decoding easier because brain function is better preserved
- 1:41:08 – 1:47:57
BCIs Beyond Speech: ALS, Breathing and Biological Engineering
Attia asks whether BCIs could restore breathing and movement in ALS. Chang describes functional electrical stimulation paired with decoding, the need for multidisciplinary teams, and a future in which biological engineering such as organoids complements electrodes.
- •Restoring breathing or movement would need further engineering steps, including stimulating muscles directly
- •Functional electrical stimulation could couple a BCI to muscle stimulation
- •Progress requires engineers, neurosurgeons, neurologists and neuroscientists working together
- •Electrodes can record from only thousands of cells versus 86 billion neurons, so biology may help address scale
- •Organoids may eventually interface with BCIs, though not in the near term
- 1:47:57 – 1:52:36
The 2030 Outlook and the BCI Industry Landscape
Chang's goal for 2030 is for BCI systems to be available to many patients with conditions such as ALS, spinal cord injury, stroke and MS. He discusses companies such as Neuralink, his own focus on customized ECoG, and why he expects devices to become less invasive but not scalp-based.
- •The 2030 goal is to move from research proof of concept to practical devices for varied conditions
- •Neuralink uses a robot to insert electrodes at fine resolution, with notable technical challenges
- •Chang's group is pursuing a higher-resolution, fully implantable ECoG system
- •He expects less invasive devices over time, but not a scalp-based solution because of the physics
- •Devices may eventually move toward enhancement, raising ethical questions
- 1:52:36 – 1:57:54
Stem Cell Therapies and Regenerating Lost Brain Function
Attia asks about regeneration after injuries like Anne's. Chang says earlier stem cell work gave modest results but cell-based therapies are returning, with Parkinson's disease the most promising focal target, and he discusses past fetal grafts and immunosuppression.
- •Stem cell results about ten to fifteen years ago were modest
- •Focal replacement of lost cells, as in Parkinson's substantia nigra, shows promise
- •Earlier fetal grafts helped some patients but could cause side effects such as dyskinesias from excess dopamine
- •New therapies aim at better cell models, dopamine control and delivery
- •Patients would initially need immunosuppression, and engineering aims to make cells less immunogenic
- 1:57:54 – 2:09:13
2040 Outlook, Chiropractic Risk and Cushing's Perspective
Chang predicts that glioblastoma could become a chronic condition and that Parkinson's and early Alzheimer's detection will see progress. Attia asks about vertebral artery dissection risk from neck manipulation, and they close with how Cushing might view the field and a call for young people to enter medicine.
- •Chang hopes glioblastoma becomes chronic or curable by targeting known genetic mechanisms and the immune system
- •He is more optimistic about Parkinson's, which has a focal target, than Alzheimer's, which involves multiple systems
- •Certain chiropractic neck movements can cause vertebral artery dissection, though the incidence is very low
- •Cushing would recognize craniotomy but be amazed by clot retrieval and BCIs
- •Neurosurgery is shifting from avoiding injury to understanding and tapping into brain function
