Skip to content
See how we’re building brain interfaces to enable silent speech

Engineering

Building Safe Implantable Devices

Our groundbreaking brain-computer interfaces (BCIs) have the potential to restore movement, enhance communication, and expand the human experience. We are achieving this through our implant, the Link, which directly interfaces with the brain to enable high-fidelity recording and modulation of neurons. The challenge then is, how do we ensure the safety of our technology? By prioritizing a meticulous approach to safety in our design and leveraging advanced assessment techniques, we have successfully launched clinical trials, providing our intracortical BCI implant to individuals living with paralysis and helping them regain their connection with the digital and physical worlds. 

This blog post highlights our approach to safety assessment and some of our findings, which demonstrate a substantial improvement in safety and biocompatibility compared to prior technologies.

Why Implantable BCI?

Neuralink is building a high-bandwidth, generalized brain interface to restore autonomy to individuals with unmet medical needs today, and to unlock superhuman capabilities across many people in the future. These goals fundamentally require the ability to interact with large areas of the brain with exceptional precision. Methods for recording neural activity from outside of the brain, such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI), or from the surface of the brain, such as electrocorticography (ECoG), are not suitable for our objectives. These approaches either lack neuron-level resolution or are impractical for continuous daily use.

A close-up image of Neuralink’s medical implant, sitting next to a U.S. quarter. The implant has a circular enclosure, comparable in size to the quarter, and thin, flexible electrode threads extending from it.
(Fig 01)
Our implant is roughly the diameter of a U.S. quarter and records neural activity with over 1,000 electrodes distributed across many thin, flexible threads.

Our approach requires a procedure to place electrodes adjacent to neurons of interest. To do this, a small disk of skull, fit to the implant’s diameter, and dura (protective layer around the brain) are surgically removed. Next, microscopic threads extending from the implant (Fig 01), each one-tenth the thickness of a human hair (Fig 02), are inserted into the brain by a surgical robot. The robot is designed to further minimize tissue disruption by using precise insertion parameters and vessel avoidance, all overseen by board-certified neurosurgeons. Following thread insertions, the implant is mounted into the opening in the skull and then the skin is fully closed over it. While this approach requires neurosurgery, it is designed from the ground up to be minimally disruptive, leaving the patient with no visible change and minimal disturbance to cortical tissue.

An illustrative diagram comparing the sizes of a thread from Neuralink’s implant, a needle from Neuralink’s surgical robot, a human hair, red blood cells, and a neuron. A scale bar indicates 100 micrometers.
(Fig 02)
In the implantation process, the surgical robot's needle inserts the implant's threads into the brain; both are thinner than a human hair and similar in thickness to a neuron.

Once implanted, the device must remain functioning and safe for a long time. One particular challenge is that the human body typically responds to foreign materials by encapsulating them in scar tissue. A significant reaction to our inserted threads could degrade the implant’s ability to detect and decode neural signals and potentially compromise the safety of our patients. We've designed safety and durability into the core architecture of the product, and through extensive preclinical studies, we've demonstrated highly favorable tissue response.

Evaluating Safety

At Neuralink, we approach preclinical safety from two perspectives: biocompatibility and procedural safety. Biocompatibility relates to how the body responds to the implanted material and relies on both in vitro and in vivo methods. Procedural safety encompasses the effects of the surgical process, including the surgical approach and implantation of the threads.

After developing confidence through extensive in vitro research, in vivo methods confirm these two aspects of safety. To do so, we rely on well-designed preclinical studies entailing both in-life and post-mortem analyses. During the in-life phase, animals are continuously monitored through daily veterinary physical and neurological examinations. Ultimately, safety must be demonstrated by inspection of the implanted tissue. Thus, after every study endpoint, a comprehensive analysis is performed to scrutinize the tissue response. 

In addition to routine, post-mortem evaluations like organ weighing and macroscopic inspection, we use several advanced high-resolution imaging techniques to study the interface between host tissue and our device. One key tool, microcomputed tomography (micro-CT), helps us create detailed 3D images of the implant and surrounding tissue. This imaging method allows us to see features as small as 1 micrometer (the size of a typical bacterium), providing valuable insights into how the Link interfaces with the adjacent bone and how the implanted threads are situated within the brain tissue.

Alongside micro-CT, we collect and process thin slices of implanted tissue to inspect different biological features. Historically, the key limitation of BCIs has been inevitable encapsulation by scar tissue due a process known as glial encapsulation (Fig 03).

A multi-panel illustrative graphic showing the longitudinal development of glial encapsulation around a single electrode. In the top left is a graphic of naive tissue, showing blood vessels descending into the tissue and neurons throughout the tissue. The panels to the right of the naive panel show progressive increase and maturation of the glial capsule at 12 hours, 1 week, 1 month, and 3 months. Three additional illustrations, in a horizontal viewing plane, are located below the 1-week, 1-month and 3-month panels. The horizontal views highlight the two-part ring of cells that form around the probe, composed of an inner ring of microglia and outer ring of astrocytes.
(Fig 03)
An overview of the principal challenge faced by brain-computer interfaces: glial encapsulation. After implantation in the brain, microglia and astrocytes become activated and, over the course of a few months, mature into a capsule around the probe.

One way we assess the presence or absence of microglial response is by using a stain that dyes scar tissue blue—Masson’s Trichrome. Compared to published data on the response to other microelectrode arrays, we have consistently witnessed no encapsulation around our threads (Fig 04).

Two side-by-side photomicrographs of cortical tissue. (Left) Tissue that has been implanted with a non-Neuralink microelectrode array, which is depicted in an inset image. The photomicrograph includes labels for leptomeningeal fibrosis, fibrotic scar, and white matter necrosis, along with a scale bar of 1 millimeter. (Right) Tissue that has been implanted with the Neuralink implant, which is depicted in an inset image, along with a scale bar of 500 micrometers. Another inset image shows a magnified view of a region containing a thread cross-section, along with a scale bar of 100 micrometers.
(Fig 04)
Photomicrographs of implanted cortical tissue stained with Masson’s Trichrome; (Left) tissue response due to implantation with a non-Neuralink microelectrode array and (right) the response due to implantation with Neuralink threads. The arrowhead in both images indicates a location of implantation—many such locations are visible in both images.

Another technique, immunohistochemistry, allows us to visualize key brain cells—neurons, microglia, and astrocytes—within the implanted tissue (Fig 05). This helps us understand the presence and health of these cells, offering a clearer picture of how our technology interacts with tissue on a cellular level.

Two side-by-side photomicrographs of cortical tissue with implanted Neuralink threads, 6 months after implantation. (Left) A low-magnification image of grey matter with a region of interest indicated by a box which expands out to the right panel. A scale bar of 300 micrometers is provided. (Right) High-magnification image of the region boxed in the left panel, showing neurons (brown), astroglia (purple), and microglia (red), along with a scale bar of 30 micrometers. A cross-section of a thread is also visible at the center of the image.
(Fig 05)
Photomicrographs of cortical tissue with implanted threads from the Neuralink implant, 6 months after implantation. (Left) A low magnification image of the grey matter which contains several threads. (Right) High-magnification regions of interest highlighting various cell types of interest and the cellular response around the thread.

By applying cell detection image processing algorithms to photomicrographs of cortical tissue, we can count the number of neurons next to each thread and compare this number to unimplanted control tissue. Contrasting historic literature describing the response to other microelectrode arrays, our studies have shown minimal tissue response, maintaining about 98% of neurons at the thread interface.

A side-by-side of a photomicrograph and a bar graph, overviewing a quantitative analysis of neuronal density around Neuralink threads and the probes of a non-Neuralink microelectrode array. (Left) An example photomicrograph of tissue around a Neuralink thread containing annotated neurons, a thread, and a scale bar corresponding to 50 micrometers. (Right) Quantification of neurons around Neuralink threads and probes of a non-Neuralink microelectrode array (MEA). The average neuron density around Neuralink threads is 98% and the average neuron density around the microelectrode array probe, as described in Patel et al. 2023, is 37%. The methods outlined by Patel et al. were replicated for Neuralink’s analysis.
(Fig 06)
Quantitative analysis of neuronal density. (Left) An example photomicrograph containing annotated neurons and a thread (Right) Quantification of neurons around Neuralink threads and probes of a non-Neuralink microelectrode array (MEA). The methods outlined by Patel et al. were replicated for Neuralink’s analysis.

As we continue to refine and enhance our systems, we invite forward-thinking experts to join our mission. If you’re excited to tackle the complex challenges in building safe, effective neural interfaces, explore our open roles and apply. Join our pioneering team to help shape the future of neurotechnology.

Acknowledgment

At Neuralink, we value our animal participants and honor their contributions by constantly improving their care and enrichment. Our engineering and preclinical teams systematically refine surgical techniques and device designs through a range of benchtop and preclinical studies. All studies involving animals are conducted under strict ethical oversight, adhering to and exceeding federal regulations while upholding the principles of the Three Rs.