## Does Brainsmart's 256-Channel ECoG Array Signal a Turning Point for Semi-Invasive BCI in China?

Hong Kong-based Brainsmart is seeking approximately RMB 200 million in a Series A round — its first institutional raise after closing a USD 10 million angel round in Q1 2026 led by Bayland Capital and Lanchi Ventures, with HKIC as a limited partner. The company's 256-channel flexible [ECoG](https://bciintel.com/glossary/ecog) array has already cleared NMPA type-testing for clinical trial use and is claimed to carry the highest channel density among devices currently implanted in Chinese trials. Early animal experiments achieved approximately 90% accuracy reconstructing the pitch contour of a Cantonese song from neural signals. Human data from awake brain tumor surgeries has captured language-related signals for speech restoration research. GCP clinical trials are targeted for initiation in 2027, with NMPA type-testing of the company's fully implanted product targeted by year-end 2026. Founded in May 2025 by Professor Paddy Kwok Leung Chan of the University of Hong Kong, Brainsmart is one of the younger entrants in a field where credible clinical milestones typically take years longer to accumulate.

---

## What Is Brainsmart and Who Is Behind It?

Brainsmart was founded in May 2025 — making it barely 15 months old at Series A — by Professor Paddy Kwok Leung Chan, a mechanical engineering professor at the University of Hong Kong whose research spans organic semiconductors, transistors, sensors, and bioelectronics with a specific focus on neuroengineering. Chan serves as Managing Director and Co-director of the InnoHK Advanced Biomedical Instrumentation Centre.

His co-founder and CEO is Sean Xiao (Hai Xiao), described as a serial medtech entrepreneur and former Sinovation Ventures investor with prior experience at Danaher and GE Healthcare. The founding team also includes a researcher who contributed to what the company calls the world's first BCI neural foundation model — a generalized decoder designed to transfer across individuals, time points, and tasks — alongside R&D professionals from established medical device firms. An overseas team of algorithm specialists rounds out the technical bench.

This combination of academic pedigree, medtech commercialization experience, and algorithm depth is exactly the profile that has historically made Chinese BCI startups competitive at the clinical translation stage. The skeptic's caveat: the team is young as an operating unit, and executing GMP manufacturing, GCP trials, and NMPA submission simultaneously within an 18-month window is an aggressive operational ask.

---

## The Technology: 256-Channel Flexible ECoG as the Strategic Wedge

The core product is a thin, flexible [electrocorticography](https://bciintel.com/glossary/electrocorticography) [electrode array](https://bciintel.com/glossary/electrode-array) designed for semi-invasive implantation. "Semi-invasive" here means subdural or epidural placement through small burr holes — avoiding the large craniotomy required for traditional ECoG grids and the deeper tissue penetration of fully intracortical arrays like the Utah array or SEEG depth electrodes.

The device achieves 256-channel ECoG signal acquisition with improved signal-to-noise ratios and lower impedance compared to prior generation designs, per the source. The NMPA has cleared it for clinical trial use via type-testing — a regulatory step that confirms the device meets product specifications for human use, but is distinct from a full NMPA market approval.

**Why 256 channels matters:** Channel count directly constrains the spatial resolution of cortical mapping and, downstream, the granularity of neural decoding. More channels mean finer-grained separation of neural populations, which translates to higher decoding accuracy for complex outputs like continuous speech or nuanced motor commands. The claim that this represents the highest channel density among devices in Chinese clinical trials is notable, though independently unverifiable from this source alone.

**The semi-invasive positioning is strategically deliberate.** Fully intracortical arrays — including those used by [Neuralink](https://bciintel.com/companies/neuralink) and [Blackrock Neurotech](https://bciintel.com/companies/blackrock-neurotech) — offer single-unit resolution and spike-level signals but require specialized neurosurgical infrastructure and carry elevated risks of glial scarring, infection, and electrode displacement over time. [Synchron](https://bciintel.com/companies/synchron)'s [endovascular](https://bciintel.com/glossary/endovascular) approach (Stentrode) sidesteps craniotomy entirely but trades spatial resolution for access. Brainsmart is threading between these poles: better signal fidelity than EEG, lower surgical barrier than intracortical, broader deployment potential than endovascular.

Brainsmart is also explicitly exploring hybrid integration — combining spike signals from invasive probes with ECoG data — which would allow the platform to scale toward intracortical-grade resolution without abandoning the semi-invasive access pathway.

---

## Animal and Early Human Data: Promising, but Pre-Clinical Scale

**Animal data:** Dogs implanted with the electrodes recovered within one day post-implantation and successfully distinguished auditory stimuli across frequencies. In a key experiment, researchers reconstructed the pitch contour of a Cantonese song from neural signals, achieving decoding accuracy of approximately 90%. Monkeys are currently being trained for hippocampal and motor decoding experiments.

**Human data:** Early trials conducted during awake brain tumor surgery captured language-related cortical signals. This is a well-established research paradigm — awake craniotomy provides a controlled window for cortical mapping — but it is important to be precise: this is investigator-initiated trial (IIT) data from intraoperative recordings, not a longitudinal implant study. The company reports having completed several IITs in Shenzhen and Guangzhou since March 2026.

**The honest calibration:** A ~90% pitch-contour reconstruction in an animal model is an impressive feasibility signal, but the leap from reconstructing tonal contours in trained animals to generating intelligible speech output in chronic human implants involves several unsolved problems — cross-session decoder stability, generalization to novel phonemes and speakers, artifact rejection in ambulatory settings, and long-term electrode-tissue interface integrity. Brainsmart's stated roadmap to generalize decoding to unfamiliar songs and enable cross-subject transfer acknowledges these gaps directly.

For patients with stroke, brain tumors, spinal cord injury, or [ALS](https://bciintel.com/glossary/als) who have lost vocal function, a working speech BCI would be clinically transformative — but the distance from intraoperative IIT data to a chronically implanted, regulatory-cleared communication device remains substantial.

---

## Regulatory and Manufacturing Infrastructure

Brainsmart's manufacturing is split between the Hetao Shenzhen-Hong Kong Science and Technology Innovation Cooperation Zone and Hong Kong Science and Technology Park (HKSTP), where electrode fabrication is carried out. The distributed structure is designed to capture Hong Kong's academic infrastructure and China's clinical and manufacturing scale.

The regulatory path is China-primary: NMPA type-testing has been cleared, GCP trials are targeted for 2027, and China's 15th National Five-Year Plan has designated BCI as one of six "future industries" — a policy tailwind that the company cites as potentially compressing the NMPA approval cycle to as little as one year. China has also introduced commercial pricing structures for BCI products, which signals that the reimbursement infrastructure is being built in parallel with the regulatory pathway.

The Series A proceeds are designated for R&D, GMP manufacturing facility build-out, and GCP trial execution. The company is structured to accept both international investment and RMB-denominated funds — a deliberate dual-currency strategy that broadens the LP pool while hedging against capital flow restrictions.

---

## Industry Implications: What Brainsmart's Trajectory Means for the Global BCI Market

**China as a parallel clinical development track.** The NMPA pathway, China's policy prioritization of BCI, and the country's large patient populations for stroke and ALS create conditions for faster clinical iteration than the FDA IDE → PMA pathway. If Brainsmart reaches GCP in 2027 on schedule, it could accumulate chronic implant safety and efficacy data at a pace that Western counterparts — constrained by slower IDE approvals and smaller trial sites — cannot match. This matters for the global technology race: Chinese companies have increasingly used domestic clinical data to accelerate international regulatory submissions.

**The semi-invasive segment is heating up.** [Precision Neuroscience](https://bciintel.com/companies/precision-neuroscience) is pursuing a thin-film cortical surface array in the U.S. with a similarly minimally invasive implantation philosophy. Brainsmart's 256-channel ECoG positions it as a direct analog in the Chinese market, with the added wrinkle of the hybrid ECoG-plus-spike integration roadmap. For investors tracking the global competitive set, the semi-invasive tier is likely to be the highest-volume clinical segment over the next decade — lower surgical barrier means addressable patient populations that fully intracortical systems cannot reach at scale.

**Neural foundation models as a differentiator.** The inclusion of a researcher with BCI neural foundation model expertise on the founding team is a signal worth tracking. Cross-subject, cross-session neural decoding generalization is one of the field's hardest open problems. A foundation model approach — borrowing from large-scale transfer learning in AI — could reduce the calibration burden on clinical deployment. If this delivers even partial generalization, it addresses a real commercialization bottleneck that has hamstrung earlier ECoG-based speech BCIs. Motor BCI applications that eventually interface with robotic prosthetics or assistive exoskeletons may also benefit from such generalizable decoders — for broader context on that intersection, see [humanoidintel.ai](https://humanoidintel.ai).

**Risk factors for investors:** The company is 15 months old. NMPA type-testing cleared is a meaningful milestone, but it is the first gate in a multi-gate process. The founding team's cohesion as an operating unit is unproven at scale. GMP facility build-out and GCP trial execution simultaneously is operationally demanding. And the ~90% animal decoding accuracy, while compelling, needs replication in chronic human implants across diverse patient populations before it informs a commercial valuation.

---

## Key Takeaways

- Brainsmart is seeking approximately **RMB 200 million** in Series A funding following a **USD 10 million** angel round (Q1 2026) led by Bayland Capital and Lanchi Ventures.
- Its **256-channel flexible ECoG array** has cleared NMPA type-testing for clinical trial use — claimed to be the highest channel density in Chinese clinical trials.
- Animal experiments achieved approximately **90% accuracy** reconstructing Cantonese pitch contours from neural signals; early human data comes from intraoperative awake-surgery recordings, not chronic implants.
- GCP clinical trials are targeted for **2027**; NMPA type-testing of the fully implanted product is targeted by **year-end 2026**.
- China's 15th Five-Year Plan designates BCI as a priority "future industry," with potential NMPA approval cycles as short as one year per the source — a regulatory tailwind with no direct FDA equivalent.
- The semi-invasive ECoG tier — below intracortical in surgical complexity, above EEG in signal fidelity — is becoming the most competitively active segment globally.
- A neural foundation model approach to cross-subject decoding generalization is explicitly part of Brainsmart's algorithmic strategy and could address a core commercialization barrier for the field.

---

*Medical disclaimer: All findings cited are from investigator-initiated trials and animal feasibility studies. No Brainsmart device has received market authorization from the NMPA or any other regulatory body. This article does not constitute medical advice. Results from small feasibility studies and intraoperative recordings should not be extrapolated to chronic implant performance in larger, controlled trials.*

---

## Frequently Asked Questions

**What is Brainsmart's BCI technology and how does it work?**
Brainsmart makes a flexible, thin electrode array that sits on the brain's surface (ECoG) and is implanted through small burr holes rather than a large craniotomy. Its 256-channel design captures high-resolution neural signals from the cortex, which are then decoded by algorithms to reconstruct intended speech or motor commands. The device is semi-invasive — more signal-rich than scalp EEG but less surgically demanding than intracortical arrays.

**How does Brainsmart compare to Neuralink and Synchron?**
Neuralink uses fully intracortical Utah-style probes for single-neuron spike resolution; Synchron's Stentrode is endovascular, requiring no open brain surgery but with lower spatial resolution. Brainsmart's flexible ECoG sits between these poles — better spatial resolution than endovascular, lower surgical risk than intracortical. It is also developing hybrid ECoG-plus-spike integration to partially close the resolution gap with intracortical systems.

**What regulatory milestones has Brainsmart achieved?**
As of mid-2026, Brainsmart's electrode has cleared NMPA type-testing for clinical trial use in China and completed several investigator-initiated trials (IITs) in Shenzhen and Guangzhou. The company is targeting NMPA type-testing submission for its fully implanted product by end-2026 and GCP clinical trial initiation in 2027. No market authorization has been granted.

**What is the significance of China's BCI policy designation?**
China's 15th Five-Year Plan lists BCI as one of six priority "future industries," and Chinese authorities have already introduced commercial pricing for BCI products. The source reports the NMPA approval cycle could potentially be compressed to approximately one year under this policy framework — significantly faster than the typical FDA PMA pathway, which could give Chinese BCI developers a clinical data advantage in the global market.

**What patient populations is Brainsmart targeting?**
The company's stated focus is restoring speech and motor function in patients with stroke, brain tumors, spinal cord injury, or ALS. Early human data was collected during awake brain tumor surgery. ALS and spinal cord injury patients with preserved cortical function but lost motor output represent the core addressable population for fully implanted, chronic BCI systems.