An extracranial implantable brain-computer interface device and its control method
By using an extracranial implantable brain-computer interface device, employing minimally invasive puncture and wireless power supply technology, it is possible to acquire and decode dedicated signals from different brain regions. This solves the problems of large trauma, low signal-to-noise ratio, and power supply limitations of existing brain-computer interfaces, improves the safety and accuracy of brain-computer interaction, and facilitates clinical promotion.
Patent Information
- Application Number
- CN202610863507.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-15
- Publication Date
- 2026-11-10
AI Technical Summary
Existing brain-computer interface technologies suffer from problems such as high risks of craniotomy, significant trauma, low signal-to-noise ratio, poor anti-interference ability, and limitations in power supply and communication, making it difficult to meet the needs of clinical promotion and precise control.
Design an extracranial implantable brain-computer interface device that uses minimally invasive puncture to implant a multi-channel electrode array, combined with a wireless communication module and a machine learning model, to achieve dedicated signal acquisition and decoding for different brain regions. It adopts wireless power supply and laser welding fixation structure to avoid craniotomy and battery replacement.
It significantly reduces surgical risks, improves signal-to-noise ratio and decoding accuracy, simplifies operation and facilitates promotion, achieves high signal-to-noise ratio brain-computer interface control, and has clinical practicality and user-friendliness.
Smart Images

Figure CN123067698A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of brain-computer interface technology, specifically to an extracranial implantable brain-computer interface device and its control method. Background Technology
[0002] Currently, brain-computer interfaces (BCIs) are a technology that establishes a communication and control channel between the brain and external devices, using the brain's bioelectrical signals to directly control external devices or to regulate brain activity through external stimuli, thereby enhancing and improving brain function. BCIs are mainly implemented in three ways: invasive, non-invasive, and semi-invasive. However, existing technologies have the following problems: 1. Invasive brain-computer interface: It requires craniotomy to implant electrodes, which carries high surgical risks and is highly invasive. It is prone to complications such as infection and brain tissue damage, thus limiting its clinical application.
[0003] 2. Non-invasive brain-computer interface: The EEG signals collected by scalp electrodes have a low signal-to-noise ratio, poor anti-interference ability, and insufficient signal decoding accuracy, making it difficult to meet the needs of precise control.
[0004] 3. Existing semi-invasive methods: Mostly subdural / epidural implantation, still require drilling into the skull, resulting in a high degree of trauma, and lack of dedicated electrode layouts for different brain regions and multiple paradigms.
[0005] 4. Power and Communication Limitations: Some implantable devices rely on battery power, which has a limited lifespan; wired transmission is prone to infection, while wireless solutions suffer from low energy efficiency and susceptibility to signal interference. Therefore, designing a solution that is easy to use and promote in clinical practice while ensuring reliable and stable operation has become a pressing technical problem for those skilled in the art. Summary of the Invention
[0006] To address the aforementioned shortcomings, this invention discloses an extracranial implantable brain-computer interface device, which is simple to operate and use, and easy to promote.
[0007] The first aspect of this invention discloses an extracranial implantable brain-computer interface device, comprising: An implantable host is implanted in the subcutaneous region of the brain. The implantable host includes a host housing, internal circuit components, a power supply module, and electrode extension wires. The power supply module is used to supply power to the internal circuit components. The internal circuit components include a signal acquisition module, a power management module, and a wireless communication module. The wireless communication module is used to transmit the acquired EEG signals to an external receiver wirelessly. A multi-channel electrode array, comprising a first electrode and a second electrode; the first electrode is electrically connected to an internal circuit component via a first electrode extension wire; the second electrode is electrically connected to an internal circuit component via a second electrode extension wire. Both the first electrode and the second electrode are implanted subcutaneously in the head via minimally invasive puncture. The first electrode is used to collect electroencephalogram (EEG) signals from a first brain region; the second electrode is used to collect EEG signals from a second brain region.
[0008] As an optional implementation, in a first aspect of the present invention, the multi-channel electrode array includes a reference electrode, which is electrically connected to an internal circuit component via a third electrode extension wire, and is implanted into the subcutaneous region of the head via a minimally invasive puncture. The reference electrode is used to collect potential signals from the mastoid region and to provide a signal reference.
[0009] As an optional implementation, in the first aspect of the present invention, the ends of the first electrode, the second electrode and the reference electrode are all provided with fixing holes, and the fixing holes are used to fix the ends of the first electrode, the second electrode and the reference electrode to the subcutaneous part of the skull of the brain by suturing. The reference electrode is a tubular electrode; the number of the reference electrode is 1; the number of electrode contacts on the reference electrode is 1 to 3.
[0010] As an optional implementation, in a first aspect of the present invention, the first electrode is an elongated electrode, and the number of electrodes in the first electrode is two; the first electrode includes a first motion imagination region electrode and a second motion imagination region electrode. The first electrode is positioned within the area enclosed by points FC5, FC6, CP5, and CP6 in the anterior and posterior central regions of the brain. The electrode contacts of the first and second motor imagery area electrodes are selected from multiple collinear points within the enclosed area, with each electrode having 5 to 9 collinear points.
[0011] As an optional implementation, in the first aspect of the present invention, the second electrode is an elongated electrode, the number of electrodes on the second electrode is 1, and the number of electrode contacts on the second electrode is 4 to 8; The electrode contacts of the second electrode are selected from 3 to 9 collinear points within the area enclosed by the occipital region of the brain, namely O1, O2, Oz, PO3, PO4, PO5, PO6, PO7, PO8, and POz.
[0012] As an optional implementation, in the first aspect of the present invention, the power supply module includes a coil antenna, which is used to receive external radio frequency energy and to power the entire internal circuit components through a power management module. As an optional implementation, in the first aspect of the present invention, the brain-computer interface device further includes an external unit, which includes an external power supply transmitting coil and an external data receiver. The external power supply transmitting coil is used to transmit radio frequency energy to a coil antenna, and the external data receiver is used to receive digitized EEG signals wirelessly transmitted from the internal circuit components and transmit them to a corresponding signal analysis module.
[0013] As an optional implementation, in the first aspect of the present invention, the implant host further includes a feedthrough structure; the feedthrough structure is fixed to the host housing by laser welding, and the feedthrough structure is fixed to the multi-channel electrode array and the coil antenna by laser integral welding; the multi-channel electrode array transmits the acquired EEG signals to the signal acquisition module through the feedthrough structure; the antenna coil transmits the acquired radio frequency energy to the internal circuit components through the feedthrough structure to achieve power supply.
[0014] As an optional implementation, in the first aspect of the present invention, the main unit housing is a titanium housing or a titanium alloy housing, and a silicone sleeve is also provided on the outside of the main unit housing.
[0015] A second aspect of this invention discloses a method for controlling an extracranial implantable brain-computer interface, comprising: Receives EEG signals collected from the parietal lobe area by the first electrode; The EEG signals are decoded using a machine learning model to identify motor imagery intentions, which include intentions to move the left and right hands and intentions to move the upper and lower limbs. The second electrode collects steady-state visual evoked potential signals induced by external visual stimuli from the occipital lobe region. The steady-state visual evoked potential signal is decoded using a steady-state visual evoked potential decoding algorithm to identify instructions associated with the frequency of visual stimuli. Based on the motion imagination intention obtained from decoding or the steady-state visual evoked potential command obtained from decoding, the corresponding control signal is output.
[0016] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: The extracranial implantable brain-computer interface device in this embodiment of the invention does not require craniotomy, only subcutaneous puncture and a small incision, which significantly reduces surgical risks and postoperative complications. Moreover, the extracranial electrodes are close to the cortex, and the signal-to-noise ratio is much higher than that of the scalp electrodes. The dual-paradigm dedicated layout enables zoned targeted signal acquisition, thereby supporting the improvement of decoding accuracy. Furthermore, the device in this embodiment of the invention is easy to implant and has a short learning curve, making it easy to promote in clinical practice in neurology and rehabilitation. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of the structure of the extracranial implantable brain-computer interface device disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of the brain-computer interface device implantation process disclosed in an embodiment of the present invention; Figure 3 This is a side view of the device disclosed in an embodiment of the present invention after implantation; Figure 4 This is a flowchart illustrating the control method for an extracranial implantable brain-computer interface disclosed in an embodiment of the present invention. The annotations in the attached figures are explained as follows: 101. Implant main unit; 102. Main unit fixing point; 103. Coil antenna; 104. Magnetic attraction point; 105. Silicone casing; 210. First motion visualization zone electrode; 211. First fixing hole; 212. Motion electrode contact; 213. First electrode extension wire; 220. Second motion visualization zone electrode; 221. Second fixing hole; 230. Visual evoked zone electrode; 231. Third fixing hole; 232. Visual electrode contact point; 233. Second electrode extension wire; 240. Reference electrode; 241. Reference electrode contact; 242. Third electrode extension wire. Detailed Implementation
[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] It should be noted that the terms "first," "second," "third," "fourth," etc., in the specification and claims of this invention are used to distinguish different objects, not to describe a specific order. The terms "comprising" and "having," and any variations thereof, in the embodiments of this invention are intended to cover non-exclusive inclusion. Exemplarily, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.
[0021] Example 1 like Figures 1 to 3 As shown, the first aspect of this invention discloses an extracranial implantable brain-computer interface device, comprising: The implant host 101 is implanted in the subcutaneous region of the brain. The implant host 101 includes a host shell, internal circuit components, a power supply module, and electrode extension wires. The power supply module is used to supply power to the internal circuit components. The internal circuit components include a signal acquisition module, a power management module, and a wireless communication module. The wireless communication module is used to transmit the acquired EEG signals to an external receiver wirelessly. A multi-channel electrode array, comprising a first electrode and a second electrode; the first electrode is electrically connected to an internal circuit component via a first electrode extension wire 213; the second electrode is electrically connected to an internal circuit component via a second electrode extension wire 233. Both the first electrode and the second electrode are implanted subcutaneously in the head via minimally invasive puncture. The first electrode is used to collect EEG signals from a first brain region; the second electrode is used to collect EEG signals from a second brain region. In one embodiment of the invention, the first electrode is used to collect EEG signals from the parietal lobe region related to motor imagery; the second electrode is used to collect steady-state visual evoked potential signals from the occipital lobe region.
[0022] The solution of this invention, by placing the implant main unit 101 in the subcutaneous region of the brain, rather than within the skull or brain tissue, avoids craniotomy and significantly reduces the risk of intracranial infection, hemorrhage, and nerve tissue damage. Simultaneously, the electrodes are implanted using a minimally invasive puncture method, further reducing surgical trauma and postoperative recovery time. In specific implementation, the implant main unit 101 is fixed in the corresponding position through the main unit fixation point 102. For example... Figure 2 As shown, this is a schematic diagram of device implantation, illustrating subcutaneous tunnel puncture, electrode implantation, and main unit implantation. The first motor imagery area electrode 210, the second motor imagery area electrode 220, the visual evoked area electrode 230, and the reference electrode 240 are all implanted electrodes after the subcutaneous tunnel is formed through puncture. The main unit 101 is then subcutaneously implanted after the skin is opened. After device implantation, as... Figure 3 As shown, it is a side view of the human body after implantation, which is used to show the layout of the implant host 101 and various electrodes in the head.
[0023] In one embodiment of the present invention, the first electrode is a motor imagery area electrode, and the second electrode is a visual evoked potential area electrode. Specifically, by setting up motor imagery area electrodes and visual evoked potential area electrodes 230, motor imagery EEG signals from the parietal lobe and steady-state visual evoked potential signals from the occipital lobe can be collected simultaneously, enabling parallel monitoring of different types of brain activity (active intention and passive evoked potential), which helps to improve the command recognition rate and control dimensions of the brain-computer interface. Through the dedicated electrode layout of the motor imagery area electrodes and visual evoked potential area electrodes, signal isolation and feature enhancement between paradigms are achieved, improving the subsequent decoding accuracy.
[0024] The implant integrates a wireless communication module, enabling wireless transmission of acquired EEG signals to an external receiver, avoiding the infection risks and inconvenience of wired interfaces. Furthermore, the entire device is subcutaneously implanted and invisible externally, offering excellent aesthetics and privacy protection. The motor imagery area electrodes and visual evoked area electrodes 230 are specifically designed for the parietal and occipital lobes, respectively, improving the signal-to-noise ratio and spatial specificity of signal acquisition. This benefits subsequent signal processing and decoding algorithm performance, leading to more precise brain-computer interface control.
[0025] More preferably, the multi-channel electrode array includes a reference electrode 240, which is electrically connected to internal circuit components via a third electrode extension wire 242. The reference electrode 240 is implanted subcutaneously into the head via minimally invasive puncture. The reference electrode 240 is used to acquire potential signals from the mastoid region and to provide a signal reference. Besides the mastoid region, the reference electrode can also be placed in the motor imagery area and the visual evoked area; however, it is more preferred to place it in the mastoid region.
[0026] By placing a dedicated reference electrode 240 in the relatively electrically quiet mastoid region, a stable, low-noise potential reference can be provided for the weak EEG signals (microvolts) acquired by the motor imagery electrode and the visual evoked electrode. This significantly improves the signal-to-noise ratio and acquisition accuracy.
[0027] The following reference and grounding scheme is applicable to both the motion imagination and steady-state visual evoked potential paradigms: Reference electrode 240: FCz (center of the forehead and top, midpoint of the line connecting Fz and Cz in the 10-20 system) - FCz is located between the motor cortex and the visual cortex, between the Fz and Cz electrodes. It is a common reference point for steady-state visual evoked potentials and is also close to the CPz electrode commonly used for motor imagery reference points, providing a stable reference benchmark.
[0028] Grounding electrode: Cz (center of the top of the head) - Cz is located at the apex of the sagittal midline, far away from the steady-state visual evoked potential signal source of the occipital lobe, and as a grounding point, it has the least interference to both paradigms; the FCz reference combined with Cz grounding can effectively suppress common-mode noise.
[0029] One reference line (at point 1-2 in the mastoid region of the temporal lobe) is implanted on the same side as the main unit, at a short distance. However, since left-hand and right-hand imagery tasks correspond to different brain regions with signal changes, the introduction of the reference line carries the risk of causing classification imbalance in left- and right-hand motor imagery classification, and its actual effect varies from person to person.
[0030] Therefore, the following two methods for using reference lines should be retained, and the appropriate method should be selected flexibly based on the results of clinical validation: Option A (Enable Reference Lines): Use reference line points for reference and grounding functions. This option is suitable for individuals whose verification results show that reference lines do not affect the balance of left and right hand motor imagination classification. Option B (without reference line): Reference and grounding rely entirely on electrode points on the motion-evoked potential (PEP) data acquisition line. Whether the reference line is connected to the signal loop will be ultimately determined during the individualized evaluation phase (through motion-evoked potential classification and equalization testing).
[0031] More preferably, the ends of the motor imagery area electrode, the visual evoked area electrode 230, and the reference electrode 240 are all provided with fixing holes. The fixing holes are used to fix the ends of the motor imagery area electrode, the visual evoked area electrode 230, and the reference electrode 240 to the subcutaneous position of the brain by suturing. Specifically, the first motor imagery area electrode 210 is provided with a first fixing hole 211, and the electrode is provided with an array of motor electrode contacts 212; the second motor imagery area electrode 220 is provided with a second fixing hole 221, and the electrode is also provided with an array of motor electrode contacts 212; the visual evoked area electrode 230 is provided with a third fixing hole 231, and the visual evoked area electrode 230 is provided with a visual electrode contact 232; the reference electrode 240 is provided with a reference electrode contact 241. The reference electrode 240 is a long strip electrode; the number of reference electrodes 240 is 1; the number of electrode contacts on the reference electrode 240 is 1 to 3.
[0032] Subcutaneous tissue has a certain degree of fluidity (e.g., scalp movement, external friction), and relying solely on puncture implantation may result in electrode micromovement or dislodgement over time. Fixation holes allow the use of medical sutures to securely suture the electrode tip to the subcutaneous fascia or periosteum, achieving mechanical anchoring, effectively preventing electrode displacement, and ensuring that the multi-channel electrode array remains aligned with the target brain region (parietal lobe, occipital lobe, temporal lobe) after long-term implantation, thus ensuring the stability and repeatability of signal recording.
[0033] More preferably, the first electrode is a long strip electrode, and the number of electrodes in the first electrode can be 1 to 3, with the most preferred number being 2; the first electrode includes a first motion imagery area electrode 210 and a second motion imagery area electrode 220; The first electrode is positioned within the area enclosed by points FC5, FC6, CP5, and CP6 in the precentral and postcentral regions of the brain. The electrode contacts of the first and second motor imagery electrodes are selected from multiple collinear points within the enclosed area, with each electrode having 5 to 9 collinear points. Specifically, the first motor imagery electrode 210 is positioned in the brain region in the sequence of FC5, FC3, FC1, FC2, FC4, and FC6, and its direction of travel is from the left temporofrontal region approach towards the medial frontal central region and across the midline to the right frontal region. The second motor imagery area electrode 220 is located in the brain region in sequence as C5, C3, C1, C2, C4, and C6, and its direction of travel is from the left central area approach along the central sulcus inward and across the midline to the right central area; wherein, the first motor imagery area electrode 210 and the second motor imagery area electrode 220 are both implanted through a subcutaneous tunnel across the midline after a single puncture on the left side.
[0034] Specifically, two long strip electrodes (each with up to 18 contacts) can extend linearly along the main functional zones of the cerebral cortex (premotor frontal lobe and sensorimotor central sulcus), enabling simultaneous, high-density sampling of ipsilateral and contralateral limb motor imagery-related cortical areas. Compared to single-point or a few point electrodes, this design significantly improves the ability to capture motor imagery EEG patterns.
[0035] The design logic of the above-mentioned electric shock covers as many effective brain regions as possible (providing sufficient data for the decoding algorithm), and does not cause the host size to be too large due to an excessive number of shocks.
[0036] Specifically, the solution of this invention uses a minimally invasive puncture point (left side) to implant two high-density contact point arrays extending along the long axis of the premotor zone, auxiliary motor zone, sensorimotor cortex, and central sulcus, respectively, to achieve high spatial resolution, low trauma, and long-term stable synchronous acquisition of the bilateral hemisphere motor imagery related cortex (FC area + C area).
[0037] More preferably, the visual evoked zone electrode 230 is an elongated electrode, and the number of electrodes on the visual evoked zone electrode 230 is 1; the number of electrode contacts on the visual evoked zone electrode 230 is 4 to 8. The electrode contacts of the second electrode are selected from 3 to 9 collinear points within the area enclosed by O1, O2, Oz, PO3, PO4, PO5, PO6, PO7, PO8, and POz in the occipital region of the brain. Specifically, the visual evoked area electrode 230 is implanted into the primary visual cortex and extrastriate cortex of the occipital lobe.
[0038] The specific locations here can be combined with the standard locations of the occipital lobe and apical occipital region in the 10-20 system or its extended 10-10 system.
[0039] The solution of this invention involves carefully implanting a long strip electrode (4-8 contacts) along the topological axis of the occipital visual cortex (V1, V2 / V3, V4 / extrastrandal region) and selecting 3-4 physiologically optimal collinear points (such as Oz+O1+PO7). This allows for the simultaneous capture of both robust standard steady-state visual evoked potentials and highly specific feature-modulated steady-state visual evoked potentials, significantly improving the instruction set capacity, recognition speed, and noise resistance of the visual evoked potential interface while maintaining low invasiveness of the implantation.
[0040] More preferably, the power supply module includes a coil antenna 103, which is used to receive external radio frequency energy and power the entire internal circuit components through the power management module. The signal acquisition module is used to filter, amplify, and perform analog-to-digital conversion on the acquired EEG signals.
[0041] The coil antenna 103 is fixed to the corresponding position via the magnetic attachment point 104. In practical implementation, the coil antenna 103 receives the radio frequency radiation energy from an external device (such as a head-mounted transmitting coil) and converts it into electrical energy. This eliminates the need for a built-in battery (or only a miniature backup battery) in the implant, completely avoiding the problem of needing a second surgery to replace the battery after it is depleted.
[0042] By eliminating the battery (or using only a very small energy storage capacitor), the main unit casing can be significantly reduced in size, making it more suitable for implantation in the limited space under the scalp, reducing the feeling of a foreign body and postoperative discomfort.
[0043] The solution of this invention achieves wireless power supply through coil antenna 103, rectifier, and LDO, completely eliminating the limitation of battery life and realizing true permanent implantation; at the same time, the complete analog front end composed of filtering, amplification and ADC modulates the microvolt-level EEG signal into a pure, digital, high-quality data stream inside the implant.
[0044] More preferably, the brain-computer interface device further includes an external unit, which includes an external power supply transmitting coil and an external data receiver. The external power supply transmitting coil is used to transmit radio frequency energy to the coil antenna 103, and the external data receiver is used to receive digitized EEG signals wirelessly transmitted from the internal circuit components and transmit them to the corresponding signal analysis module.
[0045] Transcutaneous wireless power transmission is achieved through an externally powered transmitting coil, eliminating battery dependence and enabling lifelong use, miniaturization, and low maintenance. Real-time, low-latency, and high-performance external data reception and analysis are achieved through an external data receiver, forming a complete closed-loop system of power supply, acquisition, transmission, decoding, and feedback. The synergy of these two components makes this extracranial implantable brain-computer interface highly clinically practical and user-friendly—patients only need to wear the external unit for immediate use, without any wired connections, battery replacements, or secondary surgery.
[0046] More preferably, the implant host 101 further includes a feedthrough structure; the feedthrough structure is fixed to the host housing by laser welding, and the feedthrough structure is fixed to the multi-channel electrode array and the coil antenna 103 by laser integrated welding; the multi-channel electrode array transmits the acquired EEG signals to the signal acquisition module through the feedthrough structure; the antenna coil transmits the acquired radio frequency energy to the internal circuit components through the feedthrough structure to achieve power supply.
[0047] Specifically, the feedthrough structure is a sealed electrical connector that runs through the main housing, allowing the electrode extension wire and coil antenna 103 to pass from the outside (subcutaneous tissue environment) into the housing and connect to the circuit board.
[0048] Laser welding is used to fix it to the shell, which can achieve medical implant-grade airtightness and ensure that the internal circuit components are not corroded in the human body for a long time.
[0049] Laser welding fuses the metal frame of the feedthrough structure with the main unit housing (usually titanium or medical-grade stainless steel) at the atomic level, forming a permanent connection that is far superior to adhesive bonding or mechanical pressing. The welded joint has high strength and can withstand the manipulation forces during implantation surgery, daily head movements (such as massage, washing hair, and minor impacts), and the stress generated by long-term tissue healing without loosening or falling off.
[0050] Laser welding eliminates the need for solder or adhesives, avoiding the potential for chronic inflammation or rejection reactions caused by additional chemicals such as epoxy resins and cyanoacrylates.
[0051] In practical implementation, the connection between the feedthrough and the electrode / antenna is also achieved using laser welding, rather than traditional crimping, brazing, or conductive adhesive bonding. Its unique advantages are as follows: The electrode wire ends are directly fused to the feedthrough pins to form a metal-to-metal connection. The connection point has no chemically active interface, exhibits extremely strong corrosion resistance, and has a lifespan comparable to the base metal itself.
[0052] The traditional failure chain consists of the electrode wire, crimp sleeve / solder, feedthrough pin, feedthrough internal conductor, and internal housing circuitry. Each connection is a potential failure point. Laser welding directly fuses the electrode wire and feedthrough pin, eliminating intermediate connectors and transition structures, significantly reducing assembly complexity and failure probability.
[0053] Laser welding is suitable for automated production lines. Welding parameters (power, pulse width, focus) are precisely controllable, ensuring consistent connection quality for each product and meeting the stringent quality control requirements of implantable medical devices.
[0054] Using laser welding technology, the feedthrough structure is hermetically fixed to the main unit housing and permanently integrated with the multi-channel electrode array and coil antenna 103, forming a sealed electrical interface for the implant that is all-metal, glue-free, and pressure-contact-free. This design completely solves two major pain points that have long existed in traditional implantable electronic devices: fluid leakage and corrosion, and connection point micro-motion failure, thereby ensuring the integrity of microvolt-level EEG signals, the stability of radio frequency energy transmission, and the stability of the entire implant in the body.
[0055] More preferably, the main unit housing is a titanium housing or a titanium alloy housing, and a silicone sleeve 105 is also provided on the outside of the main unit housing.
[0056] Specifically, the titanium shell measures 30mm × 20mm × 5mm, is made of titanium alloy, has laser-welded feedthroughs, and integrates acquisition, communication, and power circuits on the internal PCB board.
[0057] Electrode parameters: The electrode wire is made of platinum-iridium alloy with a diameter of 0.1mm and the contacts are gold-plated; each motion visualization electrode has 8 contacts, the visual electrode has 6 contacts, and the reference electrode has 2 contacts.
[0058] Wireless power supply: The external coil transmits radio frequency energy, which is received by the internal coil and outputs a stable voltage of 3.3V-5V through a rectifier bridge and LDO, with a power supply efficiency of ≥60%.
[0059] The implantation procedure can be performed following these steps: First, the patient is placed in a supine position and given local anesthesia. A 2cm transverse incision is made in the temporal bone region behind the ear. Second, use a puncture needle to create subcutaneous tunnels from the incision toward the parietal lobe, occipital lobe, and mastoid region, and push the electrode to the target position; Third, the implant host 101 is placed into the subcutaneous cavity under the incision, the electrode and the coil are connected, and the incision and the electrode end fixation point are sutured. Fourth, sutures are removed 7 days after surgery, and external equipment is used to test signal acquisition and wireless communication functions.
[0060] The extracranial implantable brain-computer interface device in this embodiment of the invention requires no craniotomy, only subcutaneous puncture and a small incision, significantly reducing surgical risks and postoperative complications. Furthermore, the extracranial electrodes are close to the cortex, resulting in a much higher signal-to-noise ratio than scalp electrodes, and the dual-paradigm dedicated layout improves decoding accuracy. In addition, the device in this embodiment is easy to implant, has a short learning curve, and is convenient for clinical application in neurology and rehabilitation departments.
[0061] Example 2 Please see Figure 4 , Figure 4 This is a flowchart illustrating the control method for an extracranial implantable brain-computer interface disclosed in an embodiment of the present invention. The execution entity of the method described in this embodiment is an execution entity composed of software and / or hardware. This execution entity can receive relevant information wirelessly and send certain instructions. It may also have certain processing and storage functions. This execution entity can control multiple devices, such as remote physical servers or cloud servers and related software, or local hosts or servers and related software that perform related operations on devices located in a certain location. In some scenarios, it can also control multiple storage devices, which may be placed in the same location as the device or in different locations. Figure 4 As shown, the control method based on an extracranial implantable brain-computer interface includes the following steps: S101: Receives the electroencephalogram (EEG) signals collected by the first electrode from the parietal lobe region; S102: Use a machine learning model to decode the EEG signal to identify motor imagery intentions, which include left and right hand movement intentions and upper and lower limb movement intentions; S103: Receives steady-state visual evoked potential signals induced by external visual stimuli from the occipital lobe region by the second electrode; S104: Decode the steady-state visual evoked potential signal using a steady-state visual evoked potential decoding algorithm to identify instructions associated with the visual stimulus frequency; S105: Based on the motion imagination intention obtained from decoding or the steady-state visual evoked potential command obtained from decoding, output the corresponding control signal.
[0062] Specifically, basic data preprocessing can be performed during EEG signal processing using the following methods. In the subsequent embodiments of this invention, SSVEP refers to steady-state visual evoked potentials, and MI specifically refers to motor imagery.
[0063] Bandpass filtering: MI channel 1–45Hz, SSVEP channel 1–60Hz (covering the fundamental frequency and harmonics of the stimulus). Power frequency notch filtering: 50Hz notch filtering to eliminate power grid interference; Artifact removal: Electroocular (EOG) regression removal or ICA decomposition; Electromyographic (EMG) wavelet thresholding for noise reduction; Signal segmentation (Epoching): MI was performed 0.5–4.5 s after labeling (4 s window); SSVEP was performed 0.5–2.5 s after stimulation (2 s window), both including baseline correction; Baseline correction: The mean value 0.5 seconds before stimulation / cue was used as the baseline for removal.
[0064] In this embodiment of the invention, the machine learning model and the MI decoding algorithm are used. The MI decoding algorithm can be divided into two main categories: traditional signal processing methods and deep learning methods. Traditional signal processing algorithms include common space mode, regularized CSP, filter bank common space mode, power spectral density and Riemannian geometry methods.
[0065] Specifically, the common spatial pattern jointly diagonalizes the covariance matrices of the two types of MI signals, extracts the spatial components with the largest variance differences, and feeds the features into LDA / SVM for classification. Regularized CSP introduces a regularization term on the basis of CSP to alleviate the small sample covariance estimation bias. The filter bank common spatial pattern extracts CSP features from multiple frequency bands (μ, β, etc.) and then fuses them before feeding them into the classifier. Power spectral density calculates the power of each channel in the μ / β frequency band, extracts the ERD / ERS index, and feeds it into LDA / SVM. Riemannian geometric methods use the covariance matrix as features to calculate the distance on the Riemannian manifold for classification (such as MDM).
[0066] Deep learning algorithms include EEGNet, LMDA-Net deep learning methods, and ShallowConvNet / DeepConvNet, EEG-Conformer / ATCNet, and graph neural networks; EEGNet is a lightweight CNN that uses depthwise separable convolutions (temporal convolution + spatial convolution) to decode the raw EEG end-to-end. LMDA-Net is a lightweight multidimensional attention network that combines depthwise / pointwise convolutions with attention mechanisms to extract time-frequency-space features. ShallowConvNet / DeepConvNet are shallow / deep CNNs under the Braindecode framework, respectively simulating the filter bank + log-variance feature process. EEG-Conformer / ATCNet fuses CNN local feature extraction with Transformer global temporal modeling to capture long-range dependencies and graph neural networks to model the electrode topology as a graph structure, using GCN / GAT and other methods to capture inter-channel relationships.
[0067] When implementing the algorithm, the choice can be made based on the actual situation. For small datasets (early implantation stage, <100 trials / class): prioritize RCSP / FBCSP + LDA / SVM, which converges quickly and is less prone to overfitting. For large datasets (>200 trials / class): introduce deep learning methods such as EEGNet and LMDA-Net to improve the classification ceiling. For high real-time / power consumption requirements (implantation-side inference): prioritize lightweight models such as EEGNet or traditional methods such as CSP+LDA. For high stability requirements: introduce Riemannian geometric methods or transfer learning strategies.
[0068] In this embodiment of the invention, the SSVEP decoding algorithm in step S104 is Fast Fourier Transform (FFT), Power Spectral Density (PSD), Canonical Correlation Analysis (CCA), Filter Bank Canonical Correlation Analysis (FBCCA), and Multistimulus Canonical Correlation Analysis (MsetCCA).
[0069] Specifically, Fast Fourier Transform (FFT) is performed on the EEG epoch to compare the power peaks at each stimulus frequency and its harmonics, and the frequency corresponding to the maximum value is taken as the target; Power Spectral Density (PSD) is estimated using the Welch method or multi-window spectrum estimation to improve the quality of the spectrum estimation before peak detection; Canonical Correlation Analysis (CCA) calculates the maximum canonical correlation coefficient between the multi-channel EEG and the sinusoidal reference signal (including harmonics) at each frequency, and the frequency corresponding to the maximum ρ is taken; Filter Bank Canonical Correlation Analysis (FBCCA) divides the EEG into multiple frequency band sub-bands, and each sub-band is CCAed separately and then weighted and fused, with the weights based on the harmonic SNR; Multi-Stimulus Canonical Correlation Analysis (MsetCCA) uses the superposition and averaging of multiple trials to construct an individualized reference signal to replace the universal sinusoidal reference.
[0070] In addition to the aforementioned decoding algorithms, it also includes Task-Related Component Analysis (TRCA), Ensemble Task-Related Component Analysis (e-TRCA), SSVEPNet or lightweight convolutional neural networks designed for steady-state visual evoked potentials, and Transformer-based models. Task-Related Component Analysis (TRCA) extracts the spatial components with the highest repetition across trials as templates. Template matching is used for identification, and integrated task-related component analysis (e-TRCA) is used to jointly optimize TRCA filters for multiple frequency categories to reduce cross-category interference. SSVEPNet or a lightweight convolutional neural network designed for steady-state visual evoked potentials is used for a lightweight CNN designed for SSVEP to extract joint features in the frequency and spatial domains for classification. A Transformer-based model uses an attention mechanism to capture global correlations between multiple channels and frequencies for SSVEP identification.
[0071] When implementing the SSVEP algorithm, the selection criteria are as follows: For algorithms with no training data available: FFT, CCA, and FBCCA are plug-and-play and suitable for rapid deployment in the initial stages of implantation; For algorithms with a small amount of calibration data (5–15 trials per class): MsetCCA and TRCA significantly outperform unsupervised methods; For algorithms requiring high recognition speed (window < 1 s): TRCA / e-TRCA combined with a short time window is currently the optimal choice; For algorithms with dense stimulus frequencies (frequency spacing < 2 Hz): FBCCA or TRCA is superior to FFT / CCA (due to frequency resolution limitations).
[0072] Specifically, the two paradigms are divided as follows:
[0073] The control method of this invention uses a finite state machine (FSM) to manage dual paradigm switching, and its basic structure is three states: Idle state: Continuously monitors the biparametric signal baseline, with no command output, waiting for trigger conditions. MI state: Enters after detecting the MI intention signal. SSVEP state: Enters after the user gazes at the stimulus interface.
[0074] State switching strategy (specific thresholds are calibrated individually based on the experiment): Idle state, MI / SSVEP: Based on ERD amplitude threshold triggering (MI) or external trigger signal (SSVEP) MI / SSVEP, Idle State: Decoding confidence is below the rejection threshold, or there is no valid response after timeout. Rejection mechanism: Each paradigm has a confidence threshold. If the confidence level is below the threshold, the output is "Reject" instead of forced classification, to prevent false triggering.
[0075] In practical implementation, motor imagery involves acquiring signals in the mu band (8–13Hz) and decoding the intentions of left and right hand movements using a CNN model, achieving an accuracy of ≥90%. SSVEP involves acquiring visual stimulus-evoked signals in the 4–60Hz band and decoding them using canonical correlation analysis (CCA), with a command recognition delay of ≤1s and an accuracy of ≥95%.
[0076] This invention employs implanted electrodes to collect mu rhythms (motor imagery) in the parietal lobe and SSVEP (visual evoked responses) in the occipital lobe. Machine learning models and SSVEP decoding algorithms are then used to recognize multiple types of motor intentions (left and right hands, upper and lower limbs) and multi-frequency commands. Finally, single-modal or fused control signals are flexibly output based on the user's context. This method organically combines visually independent and cognitively natural motor imagery with training-free and highly information-rich SSVEP, enabling the brain-computer interface to be used for continuous control and rehabilitation training in visually unsupported conditions, as well as for high-speed discrete selection and communication with visual support. This significantly improves the system's versatility, robustness, and user experience. It can be applied to devices in specific scenarios such as rehabilitation robots and smart home control.
[0077] The above provides a detailed description of the extracranial implantable brain-computer interface device and method disclosed in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. An extracranial implantable brain-computer interface device, characterized in that, include: An implantable host is implanted in the subcutaneous region of the brain. The implantable host includes a host shell, internal circuit components, a power supply module, and electrode extension wires. The power supply module is used to supply power to the internal circuit components. The internal circuit components include a signal acquisition module, a power management module, and a wireless communication module. The wireless communication module is used to transmit the acquired EEG signals to an external receiver wirelessly. A multi-channel electrode array, comprising a first electrode and a second electrode; the first electrode is electrically connected to an internal circuit component via a first electrode extension wire; the second electrode is electrically connected to an internal circuit component via a second electrode extension wire. Both the first electrode and the second electrode are implanted subcutaneously in the head via minimally invasive puncture. The first electrode is used to collect electroencephalogram (EEG) signals from a first brain region, and the second electrode is used to collect EEG signals from a second brain region.
2. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The multi-channel electrode array includes a reference electrode, which is electrically connected to an internal circuit component via a third electrode extension wire. The reference electrode is implanted subcutaneously in the head via a minimally invasive puncture. The reference electrode is used to acquire potential signals in the mastoid region and to provide a signal reference.
3. The extracranial implantable brain-computer interface device as described in claim 2, characterized in that, The first electrode, the second electrode, and the reference electrode are all provided with fixing holes at their ends. The fixing holes are used to fix the ends of the first electrode, the second electrode, and the reference electrode to the subcutaneous part of the skull of the brain by suturing. The reference electrode is a tubular electrode; the number of the reference electrodes is 1. The number of electrode contacts on the reference electrode is 1 to 3.
4. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The first electrode is a long strip electrode, and the number of electrodes in the first electrode is 2; the first electrode includes a first motion imagery area electrode and a second motion imagery area electrode; The first electrode is positioned within the area enclosed by points FC5, FC6, CP5, and CP6 in the anterior and posterior central regions of the brain. The electrode contacts of the first and second motor imagery area electrodes are selected from multiple collinear points within the enclosed area, with each electrode having 5 to 9 collinear points.
5. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The second electrode is a long strip electrode, and the number of electrodes in the second electrode is 1; The electrode contacts of the second electrode are selected from 3 to 9 collinear points within the area enclosed by the occipital region of the brain, namely O1, O2, Oz, PO3, PO4, PO5, PO6, PO7, PO8, and POz.
6. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The power supply module includes a coil antenna, which is used to receive external radio frequency energy and then use the received external radio frequency energy to power the entire internal circuit components through the power management module.
7. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The brain-computer interface device also includes an external unit, which includes an external power supply transmitting coil and an external data receiver. The external power supply transmitting coil is used to transmit radio frequency energy to the coil antenna, and the external data receiver is used to receive digitized EEG signals wirelessly transmitted from the internal circuit components and transmit them to the corresponding signal analysis module.
8. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The implant host also includes a feedthrough structure; the feedthrough structure is fixed to the host housing by laser welding, and the feedthrough structure is fixed to the multi-channel electrode array and the coil antenna by laser integrated welding; the multi-channel electrode array transmits the acquired EEG signals to the signal acquisition module through the feedthrough structure; the antenna coil transmits the acquired radio frequency energy to the internal circuit components through the feedthrough structure to achieve power supply.
9. The extracranial implantable brain-computer interface device as described in claim 1, characterized in that, The main unit housing is a titanium housing or a titanium alloy housing, and a silicone sleeve is also provided on the outside of the main unit housing.
10. A method for controlling an extracranial implantable brain-computer interface, characterized in that, include: Receives EEG signals collected from the parietal lobe area by the first electrode; The EEG signals are decoded using a machine learning model to identify motor imagery intentions, which include intentions to move the left and right hands and intentions to move the upper and lower limbs. The second electrode collects steady-state visual evoked potential signals induced by external visual stimuli from the occipital lobe region. The steady-state visual evoked potential signal is decoded using a steady-state visual evoked potential decoding algorithm to identify instructions associated with the frequency of visual stimuli. Based on the motion imagination intention obtained from decoding or the steady-state visual evoked potential command obtained from decoding, the corresponding control signal is output.
