NMPA released the industry standard “Test method for sensing and feedback performance of the interactive neurostimulator for medical equipment using brain-computer interface technology” on September 28, 2025 which takes effect on October 1, 2025. The draft version was published in February 2025 for feedback.
For China’s first brain-computer interface medical device terminology standard, click HERE
This standard references to the FDA guidance of “Technical Considerations for Medical Devices with Physiologic Closed-Loop Control Technology”. According to NMPA Center for Medical Device Standardization Administration, “closed-loop implantable neuroregulation products represent cutting-edge technologies worldwide. For regulatory authorities, standardized testing methods can improve regulatory efficiency, reduce the complexity and costs caused by inconsistent standards, and promote healthy and orderly market competition.
The standard provides a rigorous and multi-faceted framework for qualifying the next generation of intelligent neurostimulators. It moves beyond testing simple output parameters to validating a complex, adaptive system that senses, decides, and acts. By standardizing test methods for perception hardware, stimulation output, closed-loop response time, and the core intelligence algorithms, the document aims to foster the development of safe, effective, and reliable BCI-based closed-loop therapies.
Scope and Applicability
The standard applies to implantable neurostimulators that are both:
- BCI-based: They interface directly with the brain or nervous system.
- Closed-loop: They possess the capability to sense neural signals and use an automated algorithm to adjust stimulation parameters accordingly.
While focused on closed-loop systems, the standard notes that its sensing performance tests can also be referenced for evaluating the sensing functions of open-loop implantable neurostimulators, if they have them.
Core Terminology
The document establishes a crucial foundation by defining key terms. Important definitions include:
- Closed-loop neurostimulator: A device that monitors brain signals and has an automatic algorithm to identify patterns and maintain control.
- BCI Medical Equipment (BCI-ME): An active medical device that connects to the brain or nerves for bidirectional interaction or one-way stimulation, used for treating neurological diseases or restoring function.
- Algorithm Performance: Encompasses metrics like accuracy, robustness, real-time capability, and computational efficiency.
- Feedback: Information about the results of a previous action used for subsequent improvement.
- Various neuro-signals are defined, such as Electroencephalogram (EEG), Local Field Potential (LFP), and Spikes (Action Potentials).
Perception Performance Testing
This section details how to test the device’s ability to accurately sense or read neural signals. The tests simulate various conditions to ensure the device’s sensing hardware performs reliably.
- Amplitude-Frequency Response: Verifies that the device records signals accurately across a range of frequencies (e.g., 1Hz-60Hz for EEG, 500Hz-30kHz for spikes) without undesired amplification or attenuation.
- Noise Immunity (Anti-Noise Capability): Tests the device’s ability to detect a small, relevant neural signal in the presence of underlying physiological noise, ensuring it can distinguish true signals from background interference.
- Polarization Voltage Tolerance: Assesses whether the device’s signal recording remains accurate when exposed to DC polarization voltages (±300 mV), which can occur at the electrode-tissue interface.
- Input Impedance: Measures the impedance of the device’s input circuits. High input impedance is crucial to prevent the device from “loading” the signal source and distorting the recorded neural data.
- Common-Mode Rejection Ratio (CMRR): Evaluates the device’s ability to reject electrical noise that is common to all recording electrodes (like ambient line noise), amplifying only the differential neural signal of interest.
- Impedance Sampling Test: A method to verify that the device can accurately measure the impedance of its own electrode contacts, which is vital for ensuring proper electrode connection and functionality.
Feedback Performance Testing
This section shifts focus from sensing to stimulation, defining standardized methods for measuring the characteristics of the electrical pulses delivered by the device.
- Stimulation Amplitude: The strength of each pulse, measured in volts (V) or milliamps (mA).
- Stimulation Pulse Width: The duration of each pulse, measured in microseconds (µs).
- Stimulation Frequency: The rate at which pulses are delivered, measured in Hertz (Hz).
- Stimulation Time: The total duration of a stimulation train, including any ramp-up or ramp-down periods.
These parameters are fundamental, as they directly influence the therapy’s effect and safety. The tests require the device to be connected to a simulated load resistor, and measurements are taken using an oscilloscope.
Closed-Loop Control and Response Testing
This is the core of the standard, testing the integrated “close-the-loop” functionality. It assesses the system’s ability to dynamically respond to changing neural states.
- Signal Source: Test signals should mimic real physiological conditions, including both “calm” periods (no stimulation needed) and “onset” periods (where stimulation is required). These can be synthetic or provided by the manufacturer with justification.
- System Performance: A critical test measures the response delay—the time between the device’s detection of a target physiological signal (e.g., an abnormal spike indicating a seizure) and the subsequent adjustment of its stimulation output. This ensures the system reacts quickly enough to be clinically useful.
Algorithm Testing Requirements
Given that the intelligence of the closed-loop system resides in its software, the standard dedicates significant attention to testing the signal processing and control algorithms.
- Testing Platform and Datasets: Algorithms must be tested on well-defined, diverse, and independent datasets (which can be from open-source databases like BCI Competition or PhysioNet) in both typical and extreme operating conditions.
- Key Performance Metrics: The standard lists several algorithm-specific metrics that must be evaluated:
- Accuracy, Repeatability, and Reproducibility: The algorithm must perform consistently and correctly across multiple tests and datasets.
- Robustness and Real-time Anomaly Detection: The algorithm must remain stable and functional in the presence of noise, hardware variations, or potential software attacks.
- Computational Efficiency: The processing time for a typical case must be fast enough to meet the real-time demands of closed-loop control.
- Signal Quality Metrics: Specific calculations like Energy Density and Sharpness Ratio of the acquired signals are to be tested for accuracy and speed.
- Error Statistics and Data Control: All errors during testing must be recorded, categorized, and statistically analyzed to identify systemic biases. Furthermore, stringent data quality control and patient privacy protection are mandated.