A multi-channel signal-to-noise ratio enhancement and time base correction analysis method and system for steady-state auditory evoked potentials

Abstract

The application discloses a multi-channel signal-to-noise ratio enhancement and time base correction analysis method and system for steady-state auditory evoked potential. The method extracts a hardware DRDY timestamp from an electroencephalogram frame and unwraps the hardware timestamp into a hardware time sequence, and estimates a measured sampling rate preferentially; narrow-band sub-bin refinement is performed on a target modulation frequency, and an integer period epoch is constructed according to the refined frequency; then, the multi-channel epoch is subjected to synchronous artifact rejection, detrending and variance reciprocal weighted average, and combined with a robust frequency domain SNR, PLV, Hotelling T2 and sequential stop output significance conclusion. The scheme can reduce the influence of sampling drift and artifacts on ASSR detection.

Description

Technical Field

[0001] This invention relates to multi-channel signal enhancement, hardware time base correction, frequency refinement, and statistical decision-making for steady-state auditory evoked potentials (ASSRs), and particularly to a multi-channel signal-to-noise ratio enhancement and time base correction analysis method and system for steady-state auditory evoked potentials. Background Technology

[0002] The ASSR signal is weak, and there is drift in the sampling clock and sound card clock. Traditional single-channel or single-frequency point detection is prone to target frequency point offset, artifact contamination, and duplicate testing errors in long-window analysis. This case addresses how to use hardware DRDY timestamps, frequency refinement, multi-channel collaborative elimination, and Hotelling T² sequential statistics to steadily improve SNR and decision reliability.

[0003] Existing publicly available solutions have covered techniques such as impedance testing, ASSR statistical testing, or auditory evoked potential fitting, but they usually lack the combination of engineering closed loop, protocol constraints, and data traceability emphasized in this case. Summary of the Invention

[0004] The technical problem this invention aims to solve is that the ASSR signal is weak, and there is drift in the sampling clock and sound card clock. Traditional single-channel or single-frequency point detection is prone to target frequency point offset, artifact contamination, and duplicate testing errors in long-window analysis. This invention addresses how to use hardware DRDY timestamps, frequency refinement, multi-channel collaborative elimination, and Hotelling T² sequential statistics to stably improve SNR and decision reliability.

[0005] To address the above problems, the present invention provides the following technical solution:

[0006] The 31B EEG frame carries a DRDY 1μs lower 16-bit timestamp, which is dewound by the host computer into a perf_ts_hw second value and is used primarily for offline estimation of the measured sampling rate.

[0007] The true ASSR frequency is refined using zero-padded FFT and three-point parabolic interpolation within the target ±0.5Hz range.

[0008] Construct integer-period epochs based on refined frequencies, synchronously eliminate artifacts across channels, and perform linear detrending and variance-weighted averages on each epoch.

[0009] By combining multitaper PSD, robust noise estimation, PLV, F-test, Hotelling T², and Bonferroni sequential stopping, the significance of ASSR and SNR are output.

[0010] The present invention has at least the following beneficial effects:

[0011] Reduce target bin drift caused by sound card and ADS1299 clock drift.

[0012] Transform multi-channel coherent information into joint statistical advantages to reduce misjudgment of random peaks in single channels.

[0013] When a significant response is achieved, the process can be stopped early, or a conservative termination conclusion can be given if there is no response.

[0014] This invention pertains to the technical solutions for EEG signal acquisition, processing, quality control, and host computer regulation, and is not intended to directly provide disease diagnosis conclusions or treatment plans; any descriptions involving auditory response or hearing threshold measurement are used to illustrate signal processing, experimental control, or auxiliary evaluation scenarios. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the system composition of the present invention.

[0016] Figure 2 This is a schematic diagram of the method flow of the present invention.

[0017] Figure 3 This is a schematic diagram of the module relationships in the ASSR host computer project of the present invention. Detailed Implementation

[0018] In one embodiment, the system runs on a Python / PyQt host computer and connects to an 8-channel EEG acquisition hardware based on the ADS1299. The host computer includes modules for serial port acquisition, signal processing, real-time display, quality control, recording and export, and closed-loop control. The slave computer is responsible for sending EEG data, impedance data, or calibration status according to a pre-defined frame format, while the host computer is responsible for parsing, processing, and outputting the data based on the frame type, time information, and current operating mode.

[0019] The core steps of this embodiment include:

[0020] It receives EEG frames containing hardware sampling timestamps and multi-channel EEG data, and unwinds the low-order hardware timestamps into a monotonically increasing hardware time series.

[0021] The measured sampling rate is estimated based on the priority of hardware time series, PC performance timing series, and nominal time series.

[0022] The filtered EEG signal is frequency refined within the neighborhood of the preset target modulation frequency, and an epoch containing an integer number of cycles is constructed according to the refined frequency.

[0023] Synchronous artifact removal, detrending, inverse variance weighted averaging, and frequency domain analysis are performed on the epochs of multiple channels to obtain the target frequency amplitude and noise estimation.

[0024] Hotelling T² test is performed on the complex components of the target frequency points of multiple channels, and a sequential stop decision is made.

[0025] In the project implementation, the relevant logic has been implemented in the following files:

[0026] core / serial_worker.py: Parses perf_ts_u16 and 8-channel data from 31B EEG frames.

[0027] main.py: Unwraps the hardware timestamp and writes it to the perf_ts_hw column.

[0028] core / analysis_workers.py: Offline analysis prioritizes perf_ts_hw, frequency refinement, epoch, hotelling T², and sequential stopping.

[0029] core / analysis_utils.py, core / freq_tracking.py, core / sequential_test.py: Algorithm utility functions.

[0030] tests / test_freq_tracking.py, tests / test_sequential_test.py, tests / test_signal_processing.py: Regression tests of key algorithms.

[0031] As an optional implementation, the host computer can run on a desktop computer, portable computing terminal, or edge computing device; the EEG simulation front end can be replaced with equivalent hardware with multi-channel EEG acquisition and framed output capabilities; the threshold, window length, number of channels, sampling rate, file format, and statistical test parameters can be adjusted according to the experimental purpose, as long as the data flow, control flow, and quality constraint relationship described in this invention are retained.

[0032] The above embodiments are only used to illustrate the technical solutions of the present invention. For those skilled in the art, equivalent substitutions can be made for the sampling rate, number of channels, threshold, statistical tests, export format, or hardware front-end model without departing from the concept of the present invention.

Claims

1. A multi-channel signal-to-noise ratio enhancement and time-base correction analysis method for steady-state auditory evoked potentials, characterized in that, include: a. Acquire multi-channel EEG data streams, and estimate the software proxy impedance based on the power frequency amplitude and RMS of the nearest window for each channel under normal acquisition conditions. b. The host computer receives the multi-channel EEG acquisition data stream and calculates the software proxy impedance of each channel based on the amplitude characteristics of the EEG signal within at least one time window under normal acquisition conditions; b. In response to the hardware impedance measurement trigger, send the first control command to the lower-level computer to make the EEG simulation front end enter the hardware lead-off impedance measurement state, and set the impedance mode prompt in the serial port parser of the upper-level computer; c. When an ambiguous field that can simultaneously correspond to EEG data frames and impedance data frames is detected in the serial port receive buffer, the parsing priority of the EEG data frames and impedance data frames is determined according to the impedance mode prompt. d. When a. receives EEG data frames containing hardware sampling timestamps and multi-channel EEG data; b. Unwind the low-order count value of the hardware sampling timestamp into a monotonically increasing hardware time series; c. Estimate the measured sampling rate based on the priority of hardware time series, host computer performance timing series, and nominal time series; d. The EEG signal is refined to the target frequency within a preset neighborhood of the target modulation frequency to obtain the corrected steady-state response frequency; e. Construct an EEG epoch containing an integer number of response cycles based on the measured sampling rate and the corrected steady-state response frequency; f. Perform synchronization artifact removal, detrending, and weighted averaging on EEG epochs of multiple channels to obtain complex responses of target frequencies in multiple channels; g. Calculate the signal-to-noise ratio and phase-locking index based on the complex response of the target frequency of the multi-channel, and perform a multi-channel joint significance test; h. Based on the cumulative test results, execute the sequential stop decision and output the steady-state auditory evoked potential response conclusion.

2. The method according to claim 1, characterized in that, The hardware sampling timestamp is the low-order count value of the DRDY timestamp sent by the lower-level machine along with the EEG data frame, and the unwinding includes accumulating one count cycle when the count value is detected to be rewinding.

3. The method according to claim 1, characterized in that, The measured sampling rate is estimated first based on the difference between adjacent samples in the hardware time series; when the hardware time series is missing or invalid, it is estimated by using the host computer performance timing series or the nominal time series in turn.

4. The method according to claim 1, characterized in that, The target frequency refinement includes performing a spectrum search within the positive and negative preset bandwidth range of the target modulation frequency, and obtaining a corrected frequency with sub-frequency accuracy through zero-fill fast Fourier transform and peak interpolation.

5. The method according to claim 1, characterized in that, The length of the EEG epoch is determined based on the modified steady-state response frequency, so that each epoch contains an integer number of response cycles to reduce spectral leakage and phase shift across epochs.

6. The method according to claim 1, characterized in that, The synchronous artifact removal includes synchronously removing any channel from all channels when the amplitude range of any channel in the same epoch exceeds an adaptive threshold.

7. The method according to claim 1, characterized in that, The weighted average involves determining the weight for each epoch based on the inverse of its variance, and then limiting the weights to form the channel average response.

8. The method according to claim 1, characterized in that, The multi-channel joint significance test involves forming a feature vector from the real and imaginary parts of multiple channels at the target frequency and performing a Hotelling T² test.

9. The method according to claim 1, characterized in that, The sequential stopping decision includes updating the significance test results after each accumulation of a predetermined number of epochs and performing Bonferroni correction on the repeated test threshold.

10. A multi-channel signal-to-noise ratio enhancement and time-base correction analysis system for steady-state auditory evoked potentials, characterized in that, include: a. Frame receiving module, used to receive EEG data frames containing hardware sampling timestamps and multi-channel EEG data; b. Time base correction module, used to perform hardware timestamp unwinding and measured sampling rate estimation; c. Frequency refinement module, used to obtain the corrected steady-state response frequency in the neighborhood of the target modulation frequency; d. Multi-channel enhancement module, used to construct epochs, remove artifacts, and perform weighted averaging; e. Statistical decision module, used to output signal-to-noise ratio, phase-locked index, multi-channel joint significance, and sequential stopping conclusion.

11. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 9.

Images