Copyright Interesting Engineering
Engineers solving neuroprosthetic puzzles today face a veritable chain of bottlenecks: building electrodes and packaging that survive in the brain, capturing µV‑level spikes with low-noise electronics, decoding signals in real time, ensuring safety and regulatory approval, and finally scaling up production. Translating the brain’s faint electrical whispers into reliable control signals remains an uphill battle. Brain activity is measured in units as low as microvolts. It is buried in noise, and the body relentlessly reacts to foreign implants. As the University of Michigan’s Cynthia Chestek puts it, noninvasive measurements are “still like listening to the stadium a block away”—you can tell something’s happening, but you wouldn’t overhear a conversation. Durability at the neural interface The first hurdle is physical: putting enough contacts in the brain without provoking rejection or damage. Implantable electrodes must penetrate or sit on the cortex long-term while withstanding the harsh environment of the body. Invasive options like Blackrock Neurotech’s Utah Array use 96 silicon needles (~1–1.5 mm long) that literally stab into the motor cortex. (The Utah Array was launched in 2008 and is still used in research.) Other teams use softer arrays. For instance, Neuralink’s threads (∼4–6 µm wide) are ultrafine polymer probes that a robot inserts. As Precision Neuroscience co-founder Craig Rapoport warns, deep implants risk tissue damage: the Neuralink system uses “penetrating micro-electrodes, which cause damage when they’re inserted into the brain”. Indeed, the body often walls off electrodes with scar tissue or reacts with inflammation, degrading signals over weeks or months. Vasso Giagka at TU Delft notes that “long-term stability in the body is a major concern” for miniaturized implants. In one recent study, even bare silicon chips left in salty solutions or animals degraded significantly. In contrast, chips coated in a silicone (PDMS) polymer remained stable and “operated reliably in the body for months”. To increase longevity, companies are choosing proven materials and sealing methods. Paradromics, for example, stresses that “everything interfacing with the brain must be built from materials that last for decades”. Their Connexus implant uses platinum–iridium electrodes and aerospace-style hermetic enclosures similar to spacecraft, rather than soft plastics that deteriorate over time. Paradromics even points out that Neuralink’s polymer threads are expected to last under two years, whereas rigid metals and ceramics endure much longer. Likewise, growing evidence suggests stiff electrodes must be ultra-thin or flexible to avoid motion damage: Chestek notes that researchers are “finding ways of… creating electrodes that are smaller than neurons so they can go into the nerves and not do damage”. Still, failure modes abound. Tiny feedthroughs and wires can fracture, and moisture can seep through imperfect seals. Even high-quality ceramics or glass have tiny leaks. Each hermetic package (such as a titanium enclosure with glass-to-metal seals for hundreds of pins) is complex and must be individually tested to 10^-9 atm-cc/s leak rates—a slow, yield-limiting step. In practice, long-term studies of Utah Arrays show neural recordings “can last at least two years, with the possibility for arrays to last the better part of a decade”, but such longevity requires optimized design and rigorous testing. In summary, material choice and packaging dictate whether an implant can endure. Safety, surgery, and the path to patients Even a well-engineered device must survive a long testing, approval, and surgery process. One crucial aspect is safety interlocks and hermeticity. Brain implants must isolate tissue from electronics: typically, the implant body is sealed (ceramics, titanium) with tiny feedthroughs for each wire. Any breach invites fluid ingress and device failure. Designs often borrow pacemaker or cochlear implant techniques: for example, SCHOTT or CeramTec hermetic feedthroughs are used to carry up to hundreds of channels through glass-metal seals. According to Delft’s Vasso Giagka, addressing these packaging challenges “opens new doors” for implant longevity. Surgically, minimizing risk is vital. As Synchron’s Tom Oxley notes, his stent-based device intentionally uses veins rather than arteries so that if a blockage occurs, it’s “much more forgiving” than an arterial stroke – and in fact, “none [of our] ten users implanted … has had that issue”. Such design choices (implanting via blood vessels or using burr-hole surgery instead of full craniotomy) aim to shorten operating time and recovery. Still, any neurosurgery carries the risk of infection and/or bleeding. FDA clearance requires exhaustive safety testing. Synchron’s ongoing US pivotal trial will involve dozens of patients to demonstrate safety and efficacy. As neurosurgeon Elad Levy notes, even selecting who can implant matters – for example, the FDA asked Synchron to test stroke patients for residual brain activity first, warning that “if limited to quadriplegia, the market is way too small to be sustainable”. Despite these hurdles, progress continues. As Blackrock CEO Marcus Gerhardt notes, their Utah-array-based systems are already helping patients “to accomplish things… that were unimaginable 10 years ago”. Early trials have enabled tetraplegic users to move cursors and robotic limbs or even attempt speech. However, before a BCI becomes routine, companies must prove they can manufacture and support these devices at scale, which is the final challenge. From prototype to production Building a few dozen implants in academic labs is one thing; mass-producing thousands of sterile, hermetic neurodevices is another. Several factors drive cost and feasibility: Wafer/yield: Modern BCIs use bespoke silicon or MEMS. Fabricating hundreds of tiny electrodes on silicon wafers inevitably yields defects. Multi-step micromachining (electrode patterning, passivation layers, releasing structures) might yield only 20–50% viable devices unless highly optimized. This low yield dramatically raises per-unit cost until volume scaling and process control improve. By contrast, large-scale foundries achieve >90% yields on logic chips, but implant geometries and materials are often outside standard CMOS processes. Hermetic packaging: Sealing electronics with dozens of wires is expensive. Specialty processes (glass-frit sealing, brazing, laser welding) often involve manual steps or small batches. Paradromics touts “air-tight packaging methods similar to those used in spacecraft” to ensure longevity, but such aerospace-grade methods can cost ~$10k–$50k per device in low volumes. In short, packaging strongly influences the cost of goods. Sterilization and testing: Each implant must endure a sterilization cycle (typically EtO gas for electronics, which is slow and requires aeration). Many chip and connector materials must pass biocompatibility tests (ISO 10993) and burn-in. Testing each unit’s functionality is time-consuming: every channel, ADC, and safety switch must be verified. These steps add both time and labor per device. Supply chain and team: Specialized components (medical-grade titanium, custom ASICs, biocompatible polymers) have limited suppliers. Early companies often integrate vertically—Synchron even took equity in a medical-component manufacturer to boost capacity. A supporting infrastructure for repairs and updates is also needed. These factors keep BCI devices very expensive right now. One analysis notes that Neuralink’s first human implant cost is estimated at ~$10.5k in parts and labor, and insurers might be charged $50k or more. Given the specialized surgery and follow-up, the “landed” price could be well into five figures. For comparison, even cochlear implants (simpler electronics) can cost tens of thousands. Engineers often identify near-insurmountable cliffs in BCI projects. In the electrode domain, minimizing tissue damage and inflammation is the top priority; materials and geometry trade chronic robustness for acuity. In electronics, the power-noise-bandwidth triangle is paramount. In software, decoding methods must balance accuracy with latency and robustness. Safety and regulatory compliance, though not “engineering” in the lab sense, impose design constraints (such as required circuitry for electrical isolation and watchdog timers) and prolonged development cycles. Finally, manufacturing scale ties them all together: an implant design that works in 10 patients may encounter yield and reliability issues when pushed to 1,000 units. In summary, while dramatic demonstrations (paralyzed patients typing via thought, or monkeys texting with mind control) have grabbed headlines, BCI engineering remains a balance of trade-offs. Teams learn from each company’s experience: for instance, Paradromics’ choice of metal electrodes and robust packaging stresses longevity, Blackrock’s Utah Array provides a proven baseline (though with limited channel count), and Neuralink’s thin polymer threads push channel count (at the cost of uncertain chronic durability). Even Synchron’s creative stentrode uses a vein to sidestep some risks, but it still must meet the same implant regs. As Precision’s Rapoport observes, only by building a “safe and effective” system for one group can developers then ask, “are there ways to scale it beyond that original use case?”. The field’s momentum suggests answers are coming. Still, the final reckoning will come when these devices leave the lab and enter millions of lives, which will depend squarely on overcoming manufacturability and economic hurdles.
