Neural NoiseBCI Annual Review — 2009
1 January–31 December 2009
Introduction
If 2008 marked the fracturing of the first commercial BCI venture, 2009 was the year in which the field’s institutional architecture was rebuilt on more durable, federally-supported foundations. The most consequential structural event was the formal launch in June 2009 of the BrainGate2 pilot clinical trial at Massachusetts General Hospital, under the direction of Leigh Hochberg with John Donoghue, Michael Black, and Arto Nurmikko at Brown University and the Providence VA Medical Center. The new Investigational Device Exemption — submitted to the FDA in 2008 and approved for recruitment in 2009 — replaced the defunct Cyberkinetics-sponsored trial and was designed as a multi-site, multi-year investigation of intracortical BCI safety and feasibility. The transition from a commercial to an academic model would prove critical, providing the stability that allowed BrainGate participants to be followed for years and ultimately generating the high-impact publications of 2011–2012.
In the neuroscience of BMI, 2009 produced a landmark paper from Jose Carmena’s laboratory at UC Berkeley: Ganguly and Carmena, “Emergence of a stable cortical map for neuroprosthetic control,” published in PLoS Biology (7, e1000153). Using non-human primates performing two-dimensional cursor BMI control with a fixed linear decoder over multiple weeks, they demonstrated that motor cortex neural ensembles underwent gradual plastic reorganization, converging on a stable “cortical map” optimized for prosthetic control — a map that could be recalled across days and resisted interference from other learned decoders. This was the first rigorous demonstration that brains could form dedicated motor memories for BMI control, with deep implications for long-term clinical neural prosthetics. Donoghue also published his Annual Review of Neuroscience article that year, providing a comprehensive synthesis of the field.
On the cochlear implant front, Cochlear launched the Nucleus 5 sound processor in September 2009, featuring the world’s smallest BTE processor with SmartSound 2 and AutoPhone technology, building on the Freedom platform. The global cochlear implant installed base was expanding rapidly, with over 150,000 total recipients by year’s end — representing, by unit count, the most widely deployed active neural prosthesis in clinical use. The retinal prosthesis field was accumulating longer-term safety and efficacy data from the Phase I/II Argus II trial (NCT00407602), which had enrolled thirty subjects across US and European centers since 2007.
Timelines
January–March. The early months of 2009 were dominated by the aftermath of Cyberkinetics’ final asset sales. In February, NeuroMetrix acquired the rights to the Andara oscillating field stimulator technology for $350,000 and appointed Cyberkinetics’ former CEO Timothy Surgenor to its board. The BrainGate IP portfolio was sold in April 2009 to privately held BrainGate Co. (acquired by Jeffrey Stibel for under $1 million), covering clinical application patents. Blackrock Microsystems, which had acquired the microelectrode array manufacturing operations in 2008, began establishing itself as the primary supplier of Utah arrays (NeuroPort system) to academic researchers. Within the scientific community, Kim, Simeral, Hochberg et al. continued accumulating BrainGate participant data, and early-stage publications from these sessions characterized the statistical structure of neural population coding in human motor cortex.
April–June. The pivotal institutional event of 2009 came in June, when Brown University and Massachusetts General Hospital announced the launch of BrainGate2 (NCT00912041). The new trial was funded by the NIH (NINDS, NIDCD) and the Department of Veterans Affairs, with no commercial sponsor. The trial’s IDE had been submitted in 2008 following Donoghue’s resignation from Cyberkinetics; the new protocol expanded enrollment criteria to include ALS and brainstem stroke in addition to spinal cord injury, and established research sites at MGH, the Providence VA, and Stanford. The press release from Brown on June 10, 2009 articulated the four goals of BrainGate2: (1) reliable point-and-click computer cursor control, (2) robotic or prosthetic limb control, (3) functional electrical stimulation of paralyzed limbs, and (4) characterization of motor cortex neuroscience in humans with tetraplegia. This framing would guide the trial’s research agenda for the next decade.
July–September. The Ganguly and Carmena PLoS Biology paper on stable cortical maps was published in July 2009, immediately drawing wide attention. Using two macaques with chronically implanted 32-channel Utah arrays, they showed that practice with a fixed decoder over two to three weeks led to the stabilization of directional tuning curves, the emergence of a reproducible population-level cortical map, and the capacity to learn multiple distinct decoders without mutual interference. The implications for clinical BCIs were direct: a fixed, stable decoder could potentially be implanted and used long-term without daily recalibration, reducing the clinical burden of intracortical systems. Cochlear launched the Nucleus 5 in September 2009, offering the smallest BTE processor with enhanced SmartSound technology and an upgrade path for Freedom implant recipients — an industry milestone given it was the first cochlear system to offer immediate backward compatibility.
October–December. The final quarter of 2009 saw Cyberkinetics complete its wind-down; the company’s remaining shell was acquired by Blackrock Microsystems. The four participants who had been enrolled in the first-generation BrainGate trial between 2004 and 2009 either transitioned to BrainGate2 enrollment (participant S3, the long-duration brainstem stroke participant, transferred to the new protocol) or concluded their participation. In the EEG-BCI literature, the BCI Competition IV review article was being prepared by Blankertz, Tangermann, and colleagues and would appear in Frontiers in Neuroscience in 2012, serving as a comprehensive post-mortem on the state of the art. DBS programming consensus guidelines were being developed, with a JAMA Neurology consensus workshop leading to publication in 2010.
Trends
The Academic Turn in Intracortical BCI
The transition from Cyberkinetics to BrainGate2 in 2009 exemplified a broader structural shift in the field: the recognition that early-stage neural interface technology required the patient capital and institutional accountability of academic medical centers rather than venture-backed startups. The BrainGate2 model — multi-PI, multi-site, federally funded, with explicit neuroscientific as well as clinical feasibility objectives — would become the template for subsequent intracortical trials. Critically, the academic model allowed participants to remain in the trial for years beyond the formal safety endpoints, generating the long-duration datasets (700, 1000, and eventually 2000+ days of implantation) that would define the field’s understanding of signal longevity.
Neural Plasticity and Prosthetic Motor Memory
Ganguly and Carmena’s 2009 PLoS Biology paper introduced the concept of “prosthetic motor memory” — the idea that the brain forms a dedicated neural representation of the decoder transformation, analogous to natural motor skill acquisition. This was not merely a curiosity for systems neuroscience but had direct clinical implications: it suggested that consistent practice with a stable BCI system could lead to rapid, robust recall across sessions and resistance to interference, reducing the training burden for users. The paper generated significant follow-on work on closed-loop decoder adaptation (CLDA) algorithms that could accelerate the stabilization process and manage the non-stationarity of neural recordings.
Cochlear Implants as a Clinical Benchmark
With over 150,000 recipients worldwide and growing, cochlear implants in 2009 represented the gold standard of clinical neural prosthetics: regulatory-approved, reimbursed by insurers, manufactured by multiple competing companies (Cochlear, Advanced Bionics, MED-EL), and achieving speech understanding outcomes that transformed quality of life for recipients. The Nucleus 5 release demonstrated the industry’s ability to iterate processor technology while maintaining implant compatibility. For the BCI field, cochlear implants provided both inspiration and a daunting comparison: despite decades of effort, no motor BCI had achieved the clinical deployment scale or outcomes consistency of cochlear technology. The gap reflected not only engineering challenges but also the greater complexity of motor cortical signal processing compared with spiral ganglion stimulation.
Retinal Prosthetics Accumulate Clinical Evidence
The Argus II multi-center trial, running since 2007, was yielding its first meaningful follow-up data on visual function restoration in retinitis pigmentosa patients. The 30-electrode stimulation array, implanted on the epiretinal surface and driven by a glasses-mounted camera, could restore rudimentary shape perception, edge detection, and some motion detection. Results were encouraging for the basic safety profile and showed statistically significant improvement on standardized visual function tests (square localization, direction of motion discrimination, grating acuity) compared with device-off baselines. Second Sight Medical Products was preparing for the CE Mark application in Europe, which would come in 2011.
ECoG as an Intermediate-Invasiveness Platform
A growing body of work, spurred partly by the BCI Competition IV ECoG dataset and partly by clinical recording opportunities in epilepsy patients undergoing pre-surgical monitoring, was establishing electrocorticography as a practical middle ground between EEG (low invasiveness, limited bandwidth) and intracortical microelectrodes (high resolution, high invasiveness). Schalk and colleagues had shown two-dimensional ECoG cursor control comparable to single-unit systems in 2008; work from the Leuthardt group at Washington University was extending this to speech and language paradigms. The ECoG modality attracted particular interest because of its potentially more stable long-term recording profile — the cortical surface, unlike microelectrode penetration sites, did not generate the inflammatory gliosis response believed to degrade intracortical recordings.
Suggested Titles
- BrainGate2: Rebuilding the Neural Interface on Federal Ground
- The Stable Map: Neural Plasticity Rewrites the BMI Learning Curve
- Cochlear at Scale: The Benchmark No Motor BCI Has Reached
- From Microelectrodes to Cortical Surface: ECoG Finds Its Place
- After Cyberkinetics: The Commercial Lessons of a Pioneering Failure