Artifact removal method and artifact removal unit for stimulus-evoked signal

Abstract

The present invention belongs to the technical field of signal detection, and specifically relates to a method and unit for removing artifacts of a stimulus-induced signal. In the artifact removal process, automatic segmentation is used to detect multi-layer data segments, and segmented spline interpolation is applied to the fitting of stimulus recovery artifacts, which can ensure that fast high-amplitude data segments change monotonically without oscillation, and low-amplitude data segments will not be overfitted, retaining the original waveform shape, so that the restoration artifact fitting is more accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal detection, and in particular relates to a method and a unit for removing artifacts of a stimulus-induced signal. Background Art

[0002] As we all know, by detecting the key components of evoked potentials, the integrity of sensory conduction nerves can be detected. Electrical stimulation is often used to induce a tingling sensation. The resulting nerve impulse passes through a specific location at a specific time. Evoked potential signals can be collected at these locations. When the amplitude and latency (the time from stimulation to the propagation of excitation to the collection point) of the detected component in the signal waveform meet normal standards, it means that the sensory conduction pathway is normal. However, due to the conductivity of the human body, the electrical stimulation applied to the human body will affect the voltage at other sites. In particular, after the stimulation ends, the stimulation circuit discharges slowly, causing the voltage at the collection point to drift severely, which is called stimulus recovery artifact (slow recovery, a stimulus artifact that slowly recovers after the transient voltage of the stimulus causes signal saturation or a sudden increase). This stimulus recovery artifact may last for more than 8ms, affecting the detection of the N8 component (a negative wave with a latency of 8ms, used for the detection of peripheral damage).

[0003] In the related art, the stimulus-induced signal after stimulation is generally fitted with the entire data segment to remove artifacts. This unrestricted fitting ignores the signal characteristics of the stimulus recovery artifacts in the stimulus-induced signal, and is prone to cause oscillations, especially at the beginning and end of the fitting. The baseline fitting error is serious, resulting in inaccurate artifact removal, which affects the detection reliability of the evoked potential. Summary of the invention

[0004] The present invention provides a method and unit for removing artifacts of a stimulus-induced signal, which removes stimulus recovery artifacts by segmentally fitting baselines in a manner of identifying each data segment.

[0005] In order to solve the above technical problems, the present invention provides a method for removing artifacts of stimulation-induced signals, comprising: multi-layer detection of data segments, comprising: multi-layer detection of data segments, comprising: detecting data segments to be processed in the signal, detecting post-stimulation oscillation data segments, stimulation recovery artifact data segments, and unprocessed data segments in the data segments to be processed, and detecting high-amplitude data segments and low-amplitude data segments in the stimulation recovery artifact data segments; fitting baselines, comprising: performing monotonic piecewise cubic spline fitting on the high-amplitude data segments to obtain a second baseline, and\or performing non-monotonic piecewise cubic spline fitting on the low-amplitude data segments to obtain a third baseline; removing artifacts, comprising: not processing the post-stimulation oscillation data segments; removing artifacts in the stimulation recovery artifact data segments based on the second baseline and / or the third baseline to obtain artifact-removed data segments; combining data segments, i.e., baseline-aligning the artifact-removed data segments with the unprocessed data segments and splicing them with the post-stimulation oscillation data segments to obtain the electrical stimulation induced signals after artifact removal.

[0006] Furthermore, the data segments to be processed in the detection signal include: recording the stimulation-induced signals before and after the end of the stimulation and within the first set time as the data segments to be processed; and also include: preprocessing the data segments to be processed; the preprocessing includes: at least one of band-pass filtering, removing the first baseline, and deleting bad segment data; wherein the cutoff frequency of the band-pass filter is configured to be 2-1000Hz; the first baseline is configured to be the mean of the amplitude of the data segment to be processed within the second set time; the bad segment data is configured to be the data segment of the data segment to be processed that exceeds the set threshold within the third set time.

[0007] Furthermore, detecting the post-stimulus oscillation data segment, the stimulation recovery artifact data segment, and the unprocessed data segment in the data segment to be processed includes: recording the data segment to be processed before the start time of the stimulation recovery artifact as the post-stimulus oscillation data segment; recording the data segment to be processed after the start time of the stimulation recovery artifact and within the fourth set time as the stimulation recovery artifact data segment; recording the data segment to be processed after the fourth set time as the unprocessed data segment; wherein the start time of the stimulation recovery artifact is configured to be the start time when the first-order derivative of the data segment to be processed is calculated and its first-order derivative is continuously less than the first set threshold within a first set length.

[0008] Furthermore, detecting high-amplitude data segments and low-amplitude data segments in the stimulus recovery artifact data segments includes: performing monotonic fitting on the data segments to be processed to obtain a fourth baseline; calculating the first-order derivative of the fourth baseline; when the sampling points within a second set length are continuously greater than a second set threshold, it is considered that the low-amplitude stage has been entered, and the first sampling point therein is recorded as the dividing point between the high-amplitude data segment and the low-amplitude data segment.

[0009] Furthermore, the monotonic piecewise cubic spline fitting of the high-amplitude data segment includes: segmenting the high-amplitude data segment again to obtain high-amplitude data sub-segments; using piecewise spline interpolation to fit the baseline of the high-amplitude data sub-segment as a stimulus recovery artifact, and limiting the fitting to a monotonic decrease during the fitting process.

[0010] Furthermore, the non-monotonic piecewise cubic spline fitting of the low-amplitude data segment includes: no longer segmenting the low-amplitude data segment, using piecewise spline interpolation to fit the baseline of the low-amplitude data segment as a stimulus recovery artifact, and not limiting the monotonicity of the fitting during the fitting process.

[0011] In a second aspect, the present invention provides an artifact removal unit for a stimulus-induced signal, comprising: a data storage module for storing the original stimulus-induced signal; a data segment distinguishing module for detecting a data segment to be processed, a post-stimulation oscillation data segment, a stimulus recovery artifact data segment, an unprocessed data segment, a high-amplitude data segment, and a low-amplitude data segment; a controller for running the artifact removal method to remove artifacts from the original stimulus-induced signal; and a data segment combining module for baseline-aligning the artifact-removed data segment with the unprocessed data segment and splicing them with the post-stimulation oscillation data segment.

[0012] In a third aspect, the present invention provides a stimulation evoked potential mapping machine, comprising: the artifact removal unit described above; a processor, which controls the controller to run the artifact removal method described above; a stimulation circuit, which releases electrical stimulation; an acquisition circuit, which acquires the original stimulation evoked signal of evoked electricity; and an output module, which outputs the electrical stimulation evoked signal after the artifacts are removed.

[0013] In a fourth aspect, the present invention provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0014] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method when executed by a processor.

[0015] In a sixth aspect, the present invention provides a computer program product, comprising a computer program, which implements the steps of the method when executed by a processor.

[0016] The beneficial effect of the present invention is that the artifact removal method and artifact removal unit of the stimulation induced signal of the present invention detect and distinguish three layers of data segments of the stimulation induced signal, that is, detect the data segment to be processed in the signal, detect the post-stimulation oscillation data segment, the stimulation recovery artifact data segment, and the unprocessed data segment in the data segment to be processed, and detect the high-amplitude data segment and the low-amplitude data segment in the stimulation recovery artifact data segment; then detect the corresponding baseline according to different data segments or adopt different artifact removal methods, and finally combine and splice the processed data segment with the unprocessed data segment to finally obtain the electrical stimulation induced signal after the artifact is removed. In the artifact removal process, automatic segmentation is used to detect multiple layers of data segments, and segmented spline interpolation is applied to the fitting of the stimulation recovery artifact, which can ensure that the fast high-amplitude data segment changes monotonically and does not produce oscillation, and the low-amplitude data segment will not be overfitted, retaining the original waveform shape, so that the restoration artifact fitting is more accurate.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 It is a principle block diagram of the present invention.

[0021] Figure 2 It is a schematic diagram for identifying the data segment to be processed according to the stimulus trigger.

[0022] Figure 3 It is a schematic diagram of bad segment data in the data segment to be processed.

[0023] Figure 4 is a schematic diagram of the first baseline in the data segment to be processed.

[0024] Figure 5 It is a schematic diagram of the data segment to be processed after preprocessing.

[0025] Figure 6It is a schematic diagram for distinguishing the post-stimulus oscillation data segment and the stimulus recovery artifact data segment according to the starting time of the stimulus recovery artifact.

[0026] Figure 7 It is a schematic diagram of distinguishing the post-stimulus oscillation data segment and the stimulus recovery artifact data segment after removing the baseline.

[0027] Figure 8 This is the effect diagram of not restricting the monotonic fitting baseline.

[0028] Fig. 9 This is a graph showing the effect of restricting monotonicity to fit the baseline.

[0029] Fig.10 It is a comparison diagram of whether the present invention is monotonically fitted and baseline removed.

[0030] Fig.11 It is a comparison diagram of whether the present invention is monotonically fitted to remove the baseline and background data distribution.

[0031] Fig.12 This is a schematic diagram of the results of monotonically fitting the entire data to remove the baseline when the slow recovery stage is a downward trend.

[0032] Fig.13 This is a schematic diagram of the results of monotonically fitting the entire data to remove the baseline when the slow recovery stage is an upward trend.

[0033] Fig.14 This is a schematic diagram of the results when the slow recovery stage is a downward trend and the slow recovery stage is fitted separately to remove the baseline.

[0034] Fig.15 This is a schematic diagram of the results when the slow recovery stage is an upward trend and the slow recovery stage is fitted separately to remove the baseline.

[0035] Fig.16 This is the effect of SEP simulation data without artifact removal.

[0036] Fig.17 This is the effect diagram of SEP simulation data after artifact removal.

[0037] Fig.18 This is the unprocessed overlay result of the real SEP data.

[0038] Fig.19 This is the overlay result diagram of the unprocessed data segment of the real SEP data using the monotonic fitting baseline of the entire data segment to remove artifacts.

[0039] Fig. 20 This is the superposition result of the real SEP data after segmentation processing and artifact removal. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Example 1.

[0042] See Figure 1-Figure 15 , this embodiment 1 provides a method for removing artifacts of a stimulation-induced signal, comprising: multi-layer detection of data segments, including: a first layer of detection, detecting the data segments to be processed in the signal; a second layer of detection, detecting the post-stimulation oscillation data segments, the stimulation recovery artifact data segments, and the unprocessed data segments in the data segments to be processed; a third layer of detection, detecting the high-amplitude data segments and the low-amplitude data segments in the stimulation recovery artifact data segments; fitting a baseline, including: performing monotonic piecewise cubic spline fitting on the high-amplitude data segments to obtain a second baseline, and\or performing non-monotonic piecewise cubic spline fitting on the low-amplitude data segments to obtain a third baseline; removing artifacts, including: not processing the post-stimulation oscillation data segments; removing the artifacts in the stimulation recovery artifact data segments based on the second baseline and / or the third baseline to obtain the artifact-removed data segments; combining data segments, i.e., aligning the baselines of the artifact-removed data segments with the unprocessed data segments and splicing them with the post-stimulation oscillation data segments to obtain the electrical stimulation-induced signal after artifact removal.

[0043] Optionally, the stimulation induced signal is configured as an evoked potential caused by electrical stimulation, such as somatosensory evoked potential (SEP), electromyography evoked potential and other physiological electrical signals. The stimulation recovery artifact (slow recovery) in the stimulation induced signal is a specific form of stimulation artifact, which can be understood as the slow recovery after the stimulation transient voltage causes signal saturation or a sudden increase. In Example 1, the stimulation recovery artifact of the corresponding data segments (high amplitude data segments, low amplitude data segments) can be removed respectively according to each baseline (such as the second baseline and the third baseline); or the baselines can be combined into a complete baseline to remove the stimulation recovery artifact of the stimulation recovery artifact data segment.

[0044] As an optional implementation of multi-layer detection of data segments.

[0045] In the first level of detection, since the amplitude of the stimulus recovery artifact will gradually decrease to a certain extent over time and be at the same order of magnitude as the original data, not all stimulus-induced signals at all times after stimulation need to have the artifacts removed. At the same time, in order to obtain complete stimulus-induced signals, the intercepted data segments to be processed include post-stimulation signals containing stimulus recovery artifacts and pre-stimulation signals without stimulus recovery artifacts, and the original data is saved for later use during data processing. The signals before and after the stimulus and within the first set time are recorded as data segments to be processed. The specific detection process includes: the end time of the stimulus can be identified by stimulus marking (or stimulus trigger) and the data segment to be processed is determined in combination with the duration of the stimulus recovery artifact. Generally, based on empirical values, the first set time is set to 0.3s, i.e., 0.1s before stimulus and 0.2s after stimulus, so as to completely cover the time period containing stimulus recovery artifacts. See Figure 2 and Figure 3 , for example, the stimulus trigger in the acquisition signal (i.e. Figure 2 The middle a line is located at 0.1s, and the image is captured 0.1s before the stimulus trigger (i.e. Figure 3 0-0.1s) and 0.2s after the stimulus trigger (i.e. Figure 3 The signal of 0.1-0.3s) in the middle and the data of 0-0.3s (corresponding to the first set time) are taken as the data segment to be processed.

[0046] Further, the second layer detection is performed after the data segment to be processed is preprocessed. The preprocessing includes: at least one of bandpass filtering, removing the first baseline, and deleting bad segment data. Optionally, the cutoff frequency of the bandpass filtering is configured to be 2-1000Hz, and a Butterworth second-order filter is used to remove high-frequency fluctuations in the data segment to be processed. Optionally, the first baseline is configured as the data segment to be processed within the second set time (such as Figure 4 The mean value of the amplitude of the data segment between the origin and the e line in the middle), for example, the mean value of the data in 0-0.05s (corresponding to the second set time). Optionally, the bad segment data is configured as a data segment that exceeds a set threshold within a third set time. For example, a data segment with any data point >30μV or <-30μV within 0.11-0.3s (corresponding to the third set time) is automatically marked as a bad segment data (such as Figure 3 It can be seen that the third set time of the bad segment data starts from the end of the stimulus recovery artifact. Figure 5 ,Through preprocessing, the first baseline and high-frequency fluctuations can be removed, and the influence of high-energy noise on the superposition average can be avoided, such as Figure 3 There are high-amplitude glitches in the waveform (causing the waveform line in the figure to be very thick), and Figure 5After preprocessing, these glitches are suppressed (the lines are thin and clear). However, preprocessing cannot eliminate stimulus restoration artifacts, so it is still necessary to remove artifacts from the stimulus restoration artifact data segment.

[0047] In the second layer of detection, since the stimulus recovery artifact is not completely monotonic, high-amplitude oscillations may occur at the end of the stimulation due to circuit instability and other conditions (i.e., the post-stimulation oscillation data segment). This part of the data is also seriously disturbed like the stimulation stage. Whether the high-amplitude oscillation is included in the comparison results after baseline removal is as follows: Figure 7 As shown, obviously, if the oscillation is introduced into the subsequent analysis, it may affect the fitting of the recovery artifact and cause inaccurate baseline removal. Therefore, it is necessary to detect the post-stimulation oscillation data segment and the stimulation recovery artifact data segment in the data segment to be processed. The post-stimulation oscillation data segment is not processed directly, and no baseline fitting is performed. For the stimulation recovery artifact data segment, the third layer of detection needs to be continued to distinguish between high-amplitude data segments and low-amplitude data segments. The specific detection process includes: recording the data segment to be processed before the start time of the stimulation recovery artifact as the post-stimulation oscillation data segment; recording the data segment to be processed after the start time of the stimulation recovery artifact and within the fourth set time as the stimulation recovery artifact data segment; recording the data segment to be processed after the fourth set time as the unprocessed data segment; wherein the start time of the stimulation recovery artifact is configured to be the start time of taking the first-order derivative of the data segment to be processed and its first-order derivative is continuously less than the set threshold within the set length. Optionally, the set length is the number of sampling points in the data segment, and the set threshold is the set amplitude of the data point, for example, 5 sampling points, 11μV. As Figure 2 As shown, line a represents the end of stimulation or the stimulation trigger, line b represents the start of the stimulation recovery artifact, and line c represents the end of the stimulation recovery artifact (according to the actual duration of the stimulation recovery artifact, the recovery artifact stimulation generally lasts for about 10ms after stimulation, that is, 0.1-0.11s corresponds to the fourth setting time). The part between line a and line b can be considered as the post-stimulation oscillation data segment, the part between line b and line c can be considered as the stimulation recovery artifact data segment, and the part after line c (to 0.3s on the horizontal axis) can be considered as the unprocessed data segment. Or see Figure 6 , line f represents the end time of stimulation, line g represents the start time of stimulation recovery artifact, then the part between line f and line g can be considered as the post-stimulation oscillation data segment; the part after line g and within the fourth set time can be considered as the stimulation recovery artifact data segment.

[0048] The third level of detection, when the amplitude of the stimulus restoration artifact gradually decreases to a certain extent, the change trend further slows down (that is, the slope becomes smaller), and the stimulus restoration artifact will also be in the same order of magnitude as the original data, that is, the baseline is jointly determined by the stimulus restoration artifact (decline) and the original data (either rise or fall). If the entire segment of data uses monotonic fitting, it is impossible to fit the rise caused by the original data, and only the descending baseline can be removed, and the ascending baseline cannot be removed; therefore, it is necessary to detect the high-amplitude data segment and the low-amplitude data segment in the stimulus restoration artifact data segment to fit the baselines respectively. The specific detection process includes: performing monotonic fitting on the data segment to be processed to obtain the fourth baseline; finding the first-order derivative of the fourth baseline; when the sampling points within the second set length are continuously greater than the second set threshold, it is considered to enter the low-amplitude stage, and the first sampling point therein is recorded as the dividing point between the high-amplitude data segment and the low-amplitude data segment. For example, when the first-order derivative values ​​of 10 sampling points are continuously greater than -0.1, it is considered to have entered the low amplitude stage, and the first sampling point among the 10 consecutive sampling points greater than -0.1 is recorded as the dividing point between the high amplitude and the low amplitude. Figure 6 , the h line represents the moment when the stimulus recovery artifact and the original data are at the same order of magnitude, then the portion between the g line and the h line can be considered as a high-amplitude data segment; the portion after the h line and within the fourth set time can be considered as a low-amplitude data segment. Figure 2 It can also be clearly seen that the i line is the dividing line between high amplitude and low amplitude, and the signal fluctuations on both sides of the i line are obviously different.

[0049] As an optional implementation of baseline fitting.

[0050] As mentioned above, in this case, the stimulus-induced signal is distinguished by three-layer detection, and finally four data segments are obtained, namely, the post-stimulation oscillation data segment, the high-amplitude data segment, the low-amplitude data segment, and the unprocessed data segment. Since the unprocessed data segment does not contain the stimulus recovery artifact, no baseline fitting is performed. The post-stimulation oscillation data segment is also severely disturbed like the stimulation stage. If the oscillation is introduced into the subsequent analysis, it may affect the fitting of the recovery artifact and needs to be directly excluded, and no baseline fitting is performed. Therefore, baseline fitting is performed only for the high-amplitude data segment and the low-amplitude data segment in the stimulus recovery artifact data segment. Figure 8 Shows the effect of not restricting monotonicity to fit the baseline, Fig. 9 shows the effect of restricted monotonicity fitting the baseline, Fig.10 Shows whether the monotonic fit is to baseline comparison, Fig.11 It shows whether the monotonic fit is to the baseline and background data distribution. Figure 8-Figure 11 By comparison, we can draw the following conclusions.

[0051] (1) Segmented spline fitting divides the entire data into small segments and fits them separately. The data changes in the small segments are more stable and the fitting is more accurate.

[0052] (2) Piecewise spline fitting based on the Fritsch-Carlson method can ensure the monotonicity of the fitting curve.

[0053] (3) When non-monotonic fitting is used in the entire data segment, all low-frequency components are eliminated, which may also remove important components; compared with non-monotonic fitting, the amplitude distribution after monotonic fitting to remove the baseline is closer to the amplitude distribution of the background data before stimulation.

[0054] Therefore, in the baseline fitting process, this case adopts a monotonic-nonmonotonic segmented fitting method: the removal algorithm is designed according to the change law of the artifact. The baseline of the high-amplitude part is determined by the artifact. Using shorter data for monotonic fitting can remove the artifact more thoroughly. The baseline of the low-amplitude part is affected by the artifact and the original signal. Using longer data for non-monotonic fitting can only remove the approximate baseline and better preserve the original waveform. The entire data segment is divided into small segments through segmented spline fitting, and the fitting is performed separately; the data changes in the small segment are more stable and the fitting is more accurate. The specific fitting process is as follows.

[0055] Optionally, the monotone segmented cubic spline fitting of the high amplitude data segment includes: segmenting the high amplitude data segment again to obtain high amplitude data sub-segments; using segmented spline interpolation to fit the baseline of the high amplitude data sub-segment as the stimulus recovery artifact, and limiting the fitting to a monotone decrease during the fitting process. For example, the first sampling point of each high amplitude data sub-segment (epoch) is recorded as 0, and the last sampling point is recorded as 1, where [0, 0.05, 0.3, 0.5, 0.2, 0.9, 1] are the endpoints of all high amplitude data sub-segments. Segmented spline interpolation is a widely used fitting technique, and monotone fitting can limit the parameters in the fitting process according to the Fritsch-Carlson method to achieve a monotone fitting result. Applying segmented spline interpolation to the removal of high amplitude stimulus artifacts is more suitable for baseline fitting of high-speed changing stimulus recovery artifacts. Compared with non-monotonic fitting, the amplitude distribution after monotone fitting to remove the baseline is closer to the amplitude distribution of background data before stimulation.

[0056] Optionally, the non-monotonic piecewise cubic spline fitting of the low-amplitude data segment includes: the low-amplitude data segment is no longer segmented, and the baseline of the low-amplitude data segment is fitted using piecewise spline interpolation as a stimulus recovery artifact, and the monotonicity of the fitting is not restricted during the fitting process. For example, the first sampling point of the low-amplitude data segment is recorded as 0, the last sampling point is recorded as 1, [0, 1] are the endpoints of all low-amplitude data segments, and the baseline is fitted using a universal piecewise spline interpolation as a stimulus recovery artifact. Since the time span involved in the fitting is large, the baseline obtained by fitting will not affect the components with a shorter duration. When the low-amplitude data segment is fitted separately, both the descending baseline and the rising baseline can be removed. At the same time, using a longer data segment for non-monotonic cubic spline fitting can also remove the recovery artifact together with the SEP baseline itself.

[0057] See Fig.12 and Fig.13 , when monotone fitting is used for the entire data segment, all low-frequency components are eliminated, but important components may also be removed. However, the baseline fitting of this embodiment is based on the three-layer detection and differentiation results of the stimulus-induced signal, and baseline fitting is only performed on the high-amplitude data segment and the low-amplitude data segment. In addition, the baseline fitting method is designed according to the change law of the artifact, and a monotone piecewise cubic spline fitting is used for the high-amplitude data segment, and a non-monotone piecewise cubic spline fitting is used for the low-amplitude data segment (see Fig.14 and Fig.15 ), the segmented spline interpolation is applied to the baseline fitting of the stimulus recovery artifact, which makes the baseline fitting more accurate. Moreover, the segmented fitting can also obtain accurate fitting results when the slope of the original data changes rapidly. It is not only suitable for the baseline fitting of the high-speed change stage (corresponding to the high-amplitude data segment), but also ensures that the slow recovery stage (corresponding to the low-amplitude data segment) will not be overfitted, and the original waveform is retained.

[0058] In summary, in order to improve accuracy, this case proposes to use different fitting methods in stages according to the waveform of the stimulus recovery artifact. First, the segmented results, such as Figure 5, the signal data of 0-0.3s (stimulus trigger is at 0.1s, 0-0.1s before stimulation, 0.1-0.3s after stimulation) is intercepted as the data to be processed and preprocessed. Secondly, a fitting method with limited monotonicity is proposed to fit the artifacts in the recovery stage, and the physical properties of this type of artifacts (monotonic change) are fully utilized to suppress the oscillation of the fitting error. However, when the monotonic decline is strictly restricted for fitting on the entire data segment, it is easy to fit the downward trend as the baseline in the slow recovery stage, while ignoring the upward trend, which eventually leads to errors in the superimposed average result. Finally, in order to further improve the accuracy, the recovery artifacts are divided into a fast recovery stage (high amplitude data segment) and a slow recovery stage (low amplitude data segment), and different fitting methods are used respectively. The fast recovery stage limits the monotonicity of the fitting, while the slow recovery stage does not limit the monotonicity. This scheme further improves the accuracy of the fitting and avoids the above errors.

[0059] Example 2.

[0060] On the basis of Example 1, this Example 2 provides an artifact removal unit for a stimulation induced signal, comprising: a data storage module for storing the original stimulation induced signal; a data segment distinguishing module for detecting a data segment to be processed, an unprocessed data segment, a post-stimulation oscillation data segment, a stimulation recovery artifact data segment, a high-amplitude data segment, and a low-amplitude data segment; a controller for running the artifact removal method to remove artifacts from the original stimulation induced signal; a data segment combining module for baseline-aligning the artifact-removed data segment with the unprocessed data segment and splicing it with the post-stimulation oscillation data segment, and finally superimposing the results to form an electrical stimulation induced signal after the artifacts are removed.

[0061] Example 3.

[0062] Based on Example 1, this Example 3 provides a stimulation evoked potential mapping machine, including: the artifact removal unit; a processor, which controls the controller to run the artifact removal method; a stimulation circuit, which releases electrical stimulation; an acquisition circuit, which acquires the original stimulation evoked signal of evoked electricity; and an output module, which outputs the electrical stimulation evoked signal after the artifacts are removed.

[0063] Example 4.

[0064] Based on the first embodiment, the fourth embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0065] Example 5.

[0066] Based on Example 1, this Example 5 provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented.

[0067] Example 6.

[0068] Based on Embodiment 1, this Embodiment 6 provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the method are implemented.

[0069] Test Example 1: Determine the N8 component from the simulation data.

[0070] Basic signal: SEP simulation data is used to add a recovery artifact lasting 0.011s; the starting point of the recovery artifact fluctuates randomly within 20 sampling points before and after 0.101s, and the negative half-wave of the sine wave is added at 0.109ms as the N8 component, with an amplitude of 1.5μV and a duration of 2ms.

[0071] Test condition 1: see Fig.16 ,The superposition results without artifact removal show that the negative wave is located at the restoration artifact, and it is difficult to determine whether the N8 component exists.

[0072] Test condition 2: see Fig.17 After preprocessing, data segment multilayer detection and segment fitting were used to remove artifacts. After removing the stimulus restoration artifact, the artifact was suppressed to 2ms, and the N8 component was located at 0.1093s, indicating that the artifact removal method in this case can retain the original SEP component while effectively suppressing the stimulus restoration artifact after removing the stimulus restoration artifact.

[0073] Experimental Example 2: Artifact removal effect of real data.

[0074] Basic signal: Using real somatosensory evoked potential data.

[0075] Test condition 3: see Fig.18 , the superposition result of the unprocessed data.

[0076] Test condition 4: see Fig.19 After preprocessing, the high-speed change phase and the slow recovery phase are not distinguished, and the final superposition result of the whole data segment using monotone fitting is used. When monotone fitting is used for the whole data segment, an increase occurs at the end of the fitting (at 0.108s). This is because the downward trend is artificially removed, but the upward trend is not removed.

[0077] Test condition 5: see Fig. 20After preprocessing, the high-speed change phase and the slow recovery phase are distinguished, and the slow recovery phase is fitted separately to obtain the final superposition result. The stimulus recovery artifact is shortened from the original 7-8ms to 2ms, and the waveform after removing the baseline shows that no other obvious noise is introduced.

[0078] In the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0079] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0080] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0081] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for removing artifacts of stimulus-induced signals, characterized in that: include: Multi-layer detection of data segments, including: detecting the data segments to be processed in the signal, detecting the post-stimulus oscillation data segments, the stimulation recovery artifact data segments, and the unprocessed data segments in the data segments to be processed, and detecting the high-amplitude data segments and the low-amplitude data segments in the stimulation recovery artifact data segments; Fitting the baseline includes: performing monotone piecewise cubic spline fitting on the high-amplitude data segment to obtain a second baseline, and performing non-monotone piecewise cubic spline fitting on the low-amplitude data segment to obtain a third baseline; Removing artifacts, including: not processing the post-stimulation oscillation data segment; removing artifacts in the stimulation recovery artifact data segment based on the second baseline and the third baseline to obtain the artifact-removed data segment; The combined data segments are to align the baselines of the artifact-removed data segments with the unprocessed data segments and to splice them with the post-stimulation oscillation data segments to obtain the electrical stimulation induced signals after artifact removal.

2. The artifact removal method according to claim 1, characterized in that: The data segments to be processed in the detection signal include: recording the stimulation induced signals before and after the end time of the stimulation and within the first set time as the data segments to be processed; The artifact removal method further comprises: preprocessing the data segment to be processed; The preprocessing includes: at least one of bandpass filtering, removing the first baseline, and deleting bad segment data; wherein The cutoff frequency of the bandpass filter is configured to be 2-1000 Hz; The first baseline is configured as the mean value of the amplitude of the data segment to be processed within a second set time; The bad segment data is configured as a data segment whose to-be-processed data segment exceeds a set threshold within a third set time.

3. The artifact removal method according to claim 1 or 2, characterized in that: Detecting the post-stimulus oscillation data segment, the stimulus recovery artifact data segment, and the unprocessed data segment in the data segment to be processed includes: The data segment to be processed before the start time of the stimulus recovery artifact is recorded as the post-stimulus oscillation data segment; Recording the data segment to be processed after the start time of the stimulus recovery artifact and within the fourth set time as the stimulus recovery artifact data segment; The data segment to be processed after the fourth set time is recorded as the unprocessed data segment; wherein The starting time of the stimulus recovery artifact is configured as the starting time when the first-order derivative of the data segment to be processed is calculated and the first-order derivative is continuously less than a first set threshold within a first set length.

4. The artifact removal method according to claim 1, characterized in that: The high-amplitude data segment and the low-amplitude data segment in the detection stimulus restoration artifact data segment include: Perform monotonic fitting on the data segment to be processed to obtain a fourth baseline; Take the first derivative with respect to the fourth baseline; When the sampling points within the second set length are continuously greater than the second set threshold, it is considered to have entered the low amplitude stage, and the first sampling point therein is recorded as the boundary point between the high amplitude data segment and the low amplitude data segment.

5. The artifact removal method according to claim 1 or 4, characterized in that: The monotone piecewise cubic spline fitting of the high amplitude data segment comprises: Segmenting the high-amplitude data segment again to obtain high-amplitude data sub-segments; The baseline of the high-amplitude data subsegment was fitted using piecewise spline interpolation as a stimulus reinstatement artifact, constraining the fit to be monotonically decreasing during the fitting process.

6. The artifact removal method according to claim 1 or 4, characterized in that: The non-monotone piecewise cubic spline fitting of the low amplitude data segment comprises: The low-amplitude data segments are no longer segmented, and the baseline of the low-amplitude data segments is fitted using piecewise spline interpolation as the stimulus recovery artifact. The monotonicity of the fitting is not restricted during the fitting process.

7. A unit for removing artifacts of stimulus-induced signals, characterized in that: include: A data storage module, used for storing original stimulation induced signals; A data segment distinguishing module, used for detecting a data segment to be processed, a post-stimulation oscillation data segment, a stimulation recovery artifact data segment, an unprocessed data segment, a high-amplitude data segment, and a low-amplitude data segment; A controller, configured to execute the artifact removal method according to claim 1 to remove artifacts from the original stimulus induced signal; The data segment combination module baseline-aligns the artifact-removed data segment with the unprocessed data segment and splices it with the post-stimulus oscillation data segment.

8. A stimulation evoked potential imaging machine, characterized in that: include: The artifact removal unit according to claim 7; A processor, controlling the controller to execute the artifact removal method according to claim 1; Stimulate the circuit to release electrical stimulation; A collection circuit collects the original stimulation-induced signal of evoked electricity; The output module outputs the electrical stimulation evoked signal after artifact removal.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to claim 1.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

Images