NeuroXess's fully implanted, fully wireless subdural BCI with a chest-pocket battery, in a registration trial at Huashan Hospital since July 2026. Most figures are company disclosures from Chinese-language sources.
Independent, source-linked catalog sheet. Not a manufacturer-issued datasheet, regulatory decision or instructions for clinical use. Human evidence does not establish approval. Source-specific restrictions, conflicts and missing specifications are retained below.
NeuroXess Triple-F fully implanted subdural BCI
The Triple-F system, called 三全 in Chinese for fully implanted, fully wireless and fully functional, is NeuroXess’s successor to the wired research array on the 256-channel sheet. It started a registration trial in July 2026. The English-language press and the English company pages used for the earlier sheet do not describe it. Sources here are Chinese-language: hospital and state reporting, the company’s Chinese news pages and ClinicalTrials.gov. Accuracy figures are company disclosures.
Identity
Field
Value and source scope
Device
NeuroXess Triple-F BCI, 三全 (fully implanted, fully wireless, fully functional) system; filed under the generic name 植入式脑机接口手部运动功能代偿系统 (implantable BCI hand motor function compensation system). Components: flexible subdural electrode, skull-side unit, and a battery and processor unit in a chest pocket [1, 2, 3]
Manufacturer
NeuroXess Technology (Shanghai) Co., Ltd. (脑虎科技), sponsor on ClinicalTrials.gov, with Huashan Hospital as collaborator [1]
Interface class
Subdural cortical-surface flexible electrode (ECoG), wireless, with a rechargeable battery implanted under the chest skin [2, 3]
Origin
NeuroXess with Huashan Hospital, Fudan University; the first case was led by Mao Ying and Chen Liang’s team [3, 4]
First demonstrated
October 30, 2025: first patient implanted at Huashan Hospital, a 28-year-old man paralysed for 8 years; first power-on 5 days after surgery; 5.2 bits per second on a 32 by 32 Webgrid test after 17 days of training, per the hospital director [4]. Reported at a Shanghai conference on December 13, 2025 [3, 5]
First human implant
October 30, 2025 at Huashan Hospital, per the hospital director’s December 5, 2025 talk [4]. A December 13, 2025 report called it a recent first implant [3]
Species studied
Human
Regulatory status
Registration trial under way, not approved. Company announcement: GCP registration trial started July 7, 2026 at Huashan Hospital, principal investigator Mao Ying, with 15 tertiary hospitals planned; entered NMPA CMDE innovative-device special review, announced July 2026 [2, 6]. Registry: NCT07720882, recruiting, start July 7, 2026, estimated 32 participants [1]. NMPA published its BCI product classification guidance on June 30, 2026, treating implanted BCI as Class III, per the company [2]
Function
Records subdural cortical surface signals from the motor area and decodes hand movement intent; used for cursor control, wheelchair control and an exoskeleton glove [2, 3]
Target tissue
Cortical surface under the dura over the motor area; the company says it does not enter brain tissue [2]
Geometry and architecture
Field
Value and source scope
Interface type
Flexible electrode on the cortical surface under the dura, no brain penetration [2, 6]
Array layout
Electrode count
64 channels, with 100 percent of channels conducting at 6 months in the first patient and at 4 months in the second, per a Xinmin Evening News report of company figures [6]. The company’s own news page does not state the channel count [2]
Pitch
Electrode lengths
Shank width and thickness
Tip and exposed site geometry
Contact coating
Insulation
Insertion method
DBS-style surgery with no dedicated surgical robot, per the company [2]; the battery unit is placed in a subcutaneous chest pocket [3, 4]
Anchoring and fixation
Electrode and channel physics
Field
Value and source scope
Exposed site area
Electrode material
Impedance (with measurement frequency)
Noise floor or SNR
Recording modality
Subdural ECoG [2, 4]
Sampling rate
Stimulation capability
Charge injection limit
Reference and ground
Tissue interface and bioresponse
Field
Value and source scope
Target tissue
Insertion trauma and BBB disruption
Vascular disruption risk
Micromotion sensitivity
Gliosis and encapsulation
Neuron loss near sites
Foreign-body response mitigation
Typical failure modes
System architecture
Field
Value and source scope
Onboard electronics
Battery, processor and heat-producing units are in a separate chest-pocket module; the company also describes its own operating system, XessOS [2, 4]
Data path
Fully wireless: wireless data and wireless power, no external cable or connector through the skin [2, 3]
Telemetry bandwidth
Sampling rate
Power
Rechargeable internal battery in the chest; short wireless charging supports daily use, per the company [2, 3]
Thermal management
Heat-producing units are kept away from the brain by placing them in the chest, per the company [2, 3]
Packaging and hermeticity
MRI compatibility
Surgical complexity
First patient moved to a general ward 1 day after surgery [3, 4]. DBS-style procedure per the company [2]
Output connectors
None through the skin; no percutaneous connector [2, 3]
Performance envelope
Field
Value and source scope
Acute yield
Chronic yield
Company-reported decoding accuracy 96.1 percent at over 6 months in the first patient (28 years old, C3 spinal cord injury) and 91.2 percent at 4 months in the second (26 years old, C4, Nanchang) [2, 6]
Stability over time
Longevity
Longest reported follow-up is over 6 months in the first patient as of July 2026 [6]
Revision and explant experience
Adverse events
The company says safety was sufficiently verified; no adverse-event data published [2, 6]
Notable demonstrations
5.2 bits per second Webgrid result at Huashan after 17 days of training, matched by a cursor decoding figure of 5.2 in the company’s July 2026 account [2, 4]. Company claim of latency under 50 ms [2]. In June 2026 the Shanghai and Nanchang patients played chess against each other over 800 km, one by controlling a virtual board, the other an exoskeleton glove (company claim) [2]
Clinical and preclinical evidence
Field
Value and source scope
Human subjects
At least two patients with this system per the company (Shanghai 28-year-old, Nanchang 26-year-old) [2, 6]. The registry estimates 32 participants [1]. The company also says it has completed 54 human implants across its BCI products as of December 2025 [3] and 60 clinical trials as of July 2026 [2]; these count all NeuroXess products, not this system alone
Preclinical cohort
Follow-up duration
Registry: follow-up at 1, 2, 3 and 6 months after implantation; primary completion January 30, 2027 [1]
Indications
Upper-limb functional replacement in tetraplegia caused by spinal cord injury [1]
Registry primary outcome: proportion of subjects whose BCI-assisted ARAT score at 3 months, minus baseline unassisted ARAT, reaches the minimal clinically important difference. Secondary outcomes include ARAT with and without BCI at 2, 3 and 6 months [1]
Key limitations
Accuracy and channel figures are company disclosures, not peer reviewed. Very few patients so far. Electrode dimensions, impedance and sampling rate are unpublished. The 64-channel figure comes from a press report; neither the registry nor the company news page states it
Engineering tradeoffs
Field
Value and source scope
Strengths
No transcutaneous connector or external unit, heat source kept away from the brain, subdural placement that avoids brain tissue, and a conventional DBS-style operation, per the company [2]
Limitations
Surface potentials have lower spatial resolution than penetrating arrays; a chest-pocket battery needs periodic wireless charging and a second implant site; evidence is early [1, 2]
Scaling constraints
Version boundary
This sheet covers the fully implanted system with a chest-pocket battery, first implanted October 30, 2025. The wired 256-channel research array from 2024 and 2025 is a different configuration. The Chinese company page and the registry do not give the same level of hardware detail; where they differ, both are noted.