Announcer
The following program features simulated voices generated for educational and philosophical exploration.
Vera Castellanos
Good afternoon. I'm Vera Castellanos.
Ryan Nakamura
And I'm Ryan Nakamura. Welcome to Simulectics Radio.
Vera Castellanos
Yesterday we discussed longevity interventions and the distinction between extending lifespan and healthspan. Today we turn to brain-computer interfaces, specifically the pursuit of high-bandwidth neural recording without invasive surgery. We're joined by Dr. Michel Maharbiz, an electrical engineer at UC Berkeley whose work on neural dust represents a fundamentally different approach to reading brain activity.
Ryan Nakamura
Dr. Maharbiz, welcome. Before we dive into neural dust specifically, can you contextualize the problem? Why is reading neural activity so technically challenging, and what are the current limitations of existing approaches?
Dr. Michel Maharbiz
Thanks for having me. The fundamental challenge is one of scale and access. The human brain contains roughly eighty-six billion neurons, each capable of firing at hundreds of times per second, organized into circuits spanning millimeters to centimeters. Current technologies fall into two categories. Non-invasive methods like EEG or fMRI capture aggregate activity from outside the skull but lack cellular resolution and temporal precision. Invasive methods like Utah arrays provide single-neuron resolution but require opening the skull, inserting electrode arrays, and accepting all the risks that entails—infection, immune response, tissue damage, signal degradation over time as scar tissue forms.
Vera Castellanos
So we have a resolution-invasiveness tradeoff. You either get coarse signals safely or precise signals dangerously. Neural dust aims to thread that needle how exactly?
Dr. Michel Maharbiz
The concept is to create wireless sensors small enough to be injected or implanted minimally invasively—think devices the size of a grain of sand or smaller—that can record neural activity and transmit data wirelessly. They'd be powered externally via ultrasound, eliminating the need for batteries, wires, or transcutaneous connectors. You could potentially scatter hundreds or thousands of these sensors throughout neural tissue through minimally invasive procedures, achieving high spatial resolution without the risks of traditional electrode arrays.
Ryan Nakamura
That sounds almost too elegant. What are the technical barriers? Why isn't this already deployed clinically?
Dr. Michel Maharbiz
Several interconnected challenges. First, power and communication. These devices need sufficient energy to sense neural signals and transmit data, but they're too small for batteries. We use ultrasound for power transfer because it penetrates tissue better than electromagnetic waves at the frequencies we need. But converting ultrasound to electrical energy efficiently at sub-millimeter scales is non-trivial. Second, biocompatibility. Any foreign object in neural tissue triggers immune response. We need materials and form factors that minimize inflammation and encapsulation. Third, data bandwidth. If you have thousands of sensors transmitting simultaneously, managing that communication without interference requires sophisticated protocols.
Vera Castellanos
Let's unpack the biocompatibility issue. The brain is immunologically privileged but not immune to foreign body response. What happens when you introduce even tiny devices into neural tissue?
Dr. Michel Maharbiz
You get microglial activation, astrocyte proliferation, eventually glial scar formation around the device. This is the same process that limits the longevity of conventional electrodes. The scar tissue electrically isolates the device from neurons, degrading signal quality over months to years. Our approach is to make devices small enough and use materials inert enough that the immune response is minimal. There's theoretical and some experimental evidence that devices below critical size thresholds—roughly ten to fifty microns depending on geometry—provoke much less scarring.
Ryan Nakamura
Is there a lower bound? Can you make these so small they're essentially invisible to the immune system?
Dr. Michel Maharbiz
You hit fundamental limits. The device needs an antenna for communication, sensors for neural signals, circuitry for processing and transmission. Miniaturizing further means reduced sensitivity, shorter range, lower data rates. There's an optimization space where you balance size, performance, and biocompatibility. Current prototypes are in the millimeter to sub-millimeter range, but we're working toward smaller.
Vera Castellanos
You mentioned ultrasound for power. That introduces an external dependency—some kind of wearable or implanted ultrasound source. Doesn't that compromise the minimally invasive premise?
Dr. Michel Maharbiz
It's a tradeoff. Yes, you need an external ultrasound transducer, likely worn on the head or implanted subdermally. But that's far less invasive than penetrating the skull with electrode arrays. The transducer could be a thin patch, integrated into headwear, or a small subdermal unit. The key is that the interface with the brain itself—the neural dust particles—requires minimal disruption.
Ryan Nakamura
What about data resolution? If I'm interested in reading out motor intent for prosthetic control, or decoding speech, or eventually reading thoughts, what kind of signal quality do these devices provide compared to Utah arrays or other invasive methods?
Dr. Michel Maharbiz
Current prototypes measure local field potentials rather than single-neuron spikes. That's lower resolution than the best invasive arrays but higher than EEG. For many applications—motor control, seizure detection, closed-loop stimulation—local field potentials are sufficient. Single-neuron resolution would require further miniaturization and sensitivity improvements. But we're not trying to replace Utah arrays for research applications requiring spike-level precision. We're targeting clinical applications where the risk-benefit calculus of invasive surgery doesn't currently favor intervention.
Vera Castellanos
Which applications specifically? What conditions or use cases justify implanting even minimally invasive devices?
Dr. Michel Maharbiz
Epilepsy monitoring is a natural fit. Current epilepsy patients who don't respond to medication often need invasive monitoring to localize seizure foci before resective surgery. Neural dust could provide continuous, long-term monitoring less invasively. Paralysis and prosthetic control is another—reading motor cortex to control robotic limbs or restore communication in locked-in patients. Chronic pain, where you might do closed-loop stimulation based on neural signatures of pain perception. Depression and psychiatric conditions where existing therapies fail and deep brain stimulation is considered.
Ryan Nakamura
Those are all medical applications. What about enhancement? If this technology matures, do we envision healthy individuals using it for augmentation—direct brain-computer communication, cognitive enhancement, merged reality interfaces?
Dr. Michel Maharbiz
That's beyond my research focus, but I'll say that any invasive intervention in healthy individuals faces enormous regulatory and ethical barriers. Even minimal invasion carries risk. Until we have extraordinary safety data and compelling benefit, I don't see enhancement applications gaining approval. That said, if the technology becomes sufficiently safe and capabilities sufficiently compelling, the pressure for enhancement use will be intense.
Vera Castellanos
Ryan's point touches on something important. The medical-enhancement distinction is contextual and shifts over time. Cochlear implants started as medical devices but are arguably enhancements for congenitally deaf individuals who never had hearing to restore. If neural interfaces become safe enough, the line will blur.
Ryan Nakamura
Exactly. And there are intermediate cases. What about students using interfaces for accelerated learning? Professionals in high-stakes fields like surgery or aviation using them for enhanced focus and performance? These aren't traditionally medical but aren't purely cosmetic either.
Dr. Michel Maharbiz
Those scenarios require capabilities we don't yet have. Reading neural activity is one thing. Writing to the brain—stimulating in ways that produce specific cognitive effects—is harder and less well understood. Most enhancement scenarios people imagine require bidirectional interfaces with precision we're nowhere near achieving. We can stimulate broadly, as in deep brain stimulation, but programming specific memories or skills requires understanding neural coding we simply don't have.
Vera Castellanos
Let's discuss the write problem. You're focused on reading, but closed-loop systems require stimulation. How do you deliver stimulation through these miniaturized devices?
Dr. Michel Maharbiz
That's an active area of development. The same devices that sense can theoretically stimulate by reversing the transduction process—converting incoming ultrasound or electromagnetic energy into local electrical fields. But stimulation requires more power than sensing, which exacerbates the power budget problem. There are also safety concerns about local heating and unintended stimulation of adjacent tissue. The precision of stimulation is limited by device size and placement accuracy.
Ryan Nakamura
What about the placement question? How do you get these devices to specific locations in the brain? If you're injecting or implanting hundreds of sensors, how do you control distribution?
Dr. Michel Maharbiz
That's a major challenge. Current concepts involve either direct injection through fine needles, potentially guided by imaging, or vascular delivery where devices are small enough to traverse capillaries and lodge in targeted regions. Each approach has problems. Direct injection is more invasive and limited in coverage. Vascular delivery raises concerns about emboli, unintended lodging, and limited control over final distribution. We need better delivery mechanisms.
Vera Castellanos
Vascular delivery is particularly concerning. You're essentially creating intentional microemboli, betting they lodge harmlessly in capillary beds rather than causing ischemia. That seems extraordinarily risky.
Dr. Michel Maharbiz
Agreed, which is why that approach is very speculative and would require extensive safety validation. The devices would need to be designed to avoid critical vessels, dissolve or clear over time if they don't lodge appropriately, and be trackable during deployment. It's not near-term.
Ryan Nakamura
Let's talk about data security and privacy. If we're reading neural signals wirelessly, how do you prevent interception or unauthorized access? Brain data is arguably the most intimate information possible.
Dr. Michel Maharbiz
Encryption is essential but technically challenging at this scale. These devices have minimal computational capacity, so implementing robust encryption is difficult. You could potentially do key exchange and encryption at the external receiver level, but the wireless link from neural dust to receiver is vulnerable. This is an area where the technology is currently inadequate for the privacy requirements that should apply.
Vera Castellanos
That's alarming. We're contemplating devices that read brain activity but can't guarantee the data won't be intercepted. What are the implications for clinical deployment or commercial development?
Dr. Michel Maharbiz
It means we need to solve the security problem before wide deployment. For early clinical applications in controlled settings—hospital monitoring, research studies—the risk is manageable with physical security and isolated networks. For take-home devices or anything approaching consumer use, we need cryptographic solutions that work within the power and computational constraints. That's an active research area but not solved.
Ryan Nakamura
What about removal? If someone wants these devices out, can they be extracted, or are they permanent once implanted?
Dr. Michel Maharbiz
That depends on size, location, and encapsulation. Millimeter-scale devices implanted superficially could potentially be located with imaging and removed surgically. Sub-millimeter devices distributed throughout brain tissue would be functionally irretrievable without causing more damage than leaving them in place. One approach is to design them to biodegrade after a certain period, turning off and dissolving into biocompatible components. But that creates other issues—you lose functionality after degradation, and you need to ensure degradation products are truly safe.
Vera Castellanos
So we might be creating permanent brain implants that can't be fully removed. That raises consent and autonomy issues. What happens if someone changes their mind or the technology is superseded by better approaches?
Dr. Michel Maharbiz
Those are valid concerns requiring careful consideration in clinical trial design and informed consent processes. Patients need to understand that these devices may be permanent. We might limit early applications to patients where the alternative is more invasive and equally irreversible interventions, like resective brain surgery for epilepsy.
Ryan Nakamura
Let's discuss the competitive landscape. Neuralink, Synchron, Paradromics—there are multiple approaches to brain interfaces. How does neural dust compare strategically?
Dr. Michel Maharbiz
Each approach optimizes different tradeoffs. Neuralink focuses on high-bandwidth invasive arrays, accepting surgical risk for maximal performance. Synchron uses vascular stents with electrodes, less invasive but lower channel count and resolution. Neural dust aims for distributed sensing with minimal invasion. The right technology depends on application requirements. For research needing single-neuron resolution across large populations, invasive arrays may remain superior. For clinical monitoring and stimulation where risk-benefit matters more than maximal performance, minimally invasive approaches have advantages.
Vera Castellanos
What's your timeline for clinical translation? When might we see neural dust in human trials, and what would those trials target?
Dr. Michel Maharbiz
We're in advanced preclinical development. First-in-human trials are probably three to five years away, targeting epilepsy monitoring as the initial indication. The path through regulatory approval is long—we need extensive safety data in large animal models, demonstration of clinical utility, manufacturing scalability. Full commercialization is a decade or more away.
Ryan Nakamura
Final question: what's the most important unsolved problem in neural interfaces generally that you think deserves more attention?
Dr. Michel Maharbiz
Understanding neural coding. We can record activity, but interpreting what it means—how populations of neurons represent information—is still primitive. We decode simple things like reaching movements, but complex cognition, emotion, subjective experience remain opaque. Better interfaces won't help if we don't understand the language the brain speaks. That requires sustained investment in basic neuroscience alongside device development.
Vera Castellanos
We're out of time. Dr. Maharbiz, thank you for this illuminating discussion of the technical realities behind the brain-interface hype.
Dr. Michel Maharbiz
Thank you for having me.
Ryan Nakamura
Tomorrow we'll examine xenotransplantation with Dr. Luhan Yang.
Vera Castellanos
Until then. Good afternoon.