Method and device for detecting wave I, wave III and wave V of auditory brainstem reaction, FPGA (Field Programmable Gate Array), equipment and medium

Abstract

The invention discloses an I-wave, III-wave and V-wave detection method and device for auditory brainstem reaction, an FPGA, equipment and a medium. The method comprises the following steps: acquiring an auditory brainstem response (ABR) signal, filtering the ABR signal to obtain a filtered signal, finding out a wave I, a wave III and a wave V from the filtered signal, checking whether wave intervals and latency periods of the wave I, the wave III and the wave V meet a preset reliability condition or not, and if so, determining that the reliability of the wave I, the wave III and the wave V meets the preset reliability condition. Taking the found wave I, the wave III and the wave V as final results, and outputting respective latency periods of the wave I, the wave III and the wave V; and if not, inputting the original signal into a pre-trained artificial neural network for processing, finding out a wave I, a wave III and a wave V from a processing result output by the artificial neural network, and outputting the incubation periods of the wave I, the wave III and the wave V found from the processing result as a final result. Therefore, automatic detection of the I wave, the III wave and the V wave is realized, the consumed time is short, the manual labor amount is reduced, and the working efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of signal processing, and particularly to a method, apparatus, FPGA, device and medium for detecting waves I, III and V of auditory brainstem response. Background Art

[0002] Auditory Brainstem Response (ABR) is an activation response of the auditory nerve and nerve fibers caused by sound stimulation, which is recorded by scalp electroencephalogram (EEG). Generally, the EEG signal collected within 10 ms after sound stimulation is regarded as the Auditory Brainstem Response (ABR) signal.

[0003] As an objective hearing assessment method, Auditory Brainstem Response (ABR) is widely used in situations where subjective results are difficult to obtain or unreliable, such as infant hearing screening, hearing assessment and injury assessment of patients with language disorders, etc. Auditory Brainstem Response (ABR) can record seven peaks, among which waves I, III and V have particularly high value in research and clinical applications due to their stability and amplitude.

[0004] In the related art, there is a lack of an automatic detection technical solution for waves I, III and V, and doctors need to manually annotate waves I, III and V, which is time-consuming and requires a large amount of manual labor.

[0005] The above statements are only used to provide background technical information related to the present application, and do not necessarily constitute prior art. Summary of the Invention

[0006] The purpose of the present application is to provide a method, apparatus, FPGA, device and medium for detecting waves I, III and V of auditory brainstem response. To provide a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary part is not a general review, nor is it intended to identify key / important constituent elements or delineate the protection scope of these embodiments. Its sole purpose is to present some concepts in a simple form as a prelude to the subsequent detailed description.

[0007] According to one aspect of the embodiments of the present application, there is provided a method for detecting waves I, III and V of auditory brainstem response, including:

[0008] Obtaining an Auditory Brainstem Response (ABR) signal;

[0009] Performing filtering processing on the ABR signal to obtain a filtered signal;

[0010] Finding waves I, III and V in the filtered signal, and checking whether the inter-wave intervals and latencies of waves I, III and V meet preset reliability conditions;

[0011] When the preset reliability condition is satisfied, the wave I, wave III, and wave V found in the filtered signal are used as the final results, and their respective latencies are output.

[0012] When the preset reliability condition is not satisfied, the ABR signal is input into a pre-trained artificial neural network for processing, and wave I, wave III, and wave V are found from the processing results output by the artificial neural network, and the latencies of wave I, wave III, and wave V found from the processing results are output as the final results.

[0013] In some embodiments of the present application, finding wave I, wave III, and wave V in the filtered signal includes:

[0014] Obtain wave I, wave III, and wave V in the filtered ABR signal;

[0015] Obtain the delays of wave I, wave III, and wave V respectively.

[0016] In some embodiments of the present application, obtaining wave I, wave III, and wave V in the filtered ABR signal includes:

[0017] For each of wave I, wave III, and wave V, determine all local maxima within the latency range of the wave in the signal passing through the band-pass filter;

[0018] For each of wave I, wave III, and wave V, select, from the delays of all local maxima of the wave through the band-pass filter, the point of the local maximum closest to the corresponding preset duration as the wave.

[0019] In some embodiments of the present application, obtaining the delays of wave I, wave III, and wave V respectively includes:

[0020] Obtain the delay of the wave based on the positions of wave I, wave III, and wave V in the found sequence and the sampling rate.

[0021] In some embodiments of the present application, the preset reliability condition includes: wave I, wave III, and wave V all exist, the latency of each of wave I, wave III, and wave V belongs to the corresponding preset interval, the inter-wave interval between wave I and wave III belongs to the corresponding preset interval, and the inter-wave interval between wave III and wave V belongs to the corresponding preset interval.

[0022] In some embodiments of the present application, the preset reliability conditions include: the presence of I wave, III wave, and V wave, the latency of the I wave being between 0.8 ms and 2 ms, the latency of the III wave being between 3.5 ms and 4.5 ms, the latency of the V wave being between 4.8 ms and 6 ms, the inter-wave interval between the I wave and the III wave being 2 ms, and the inter-wave interval between the III wave and the V wave being 2 ms.

[0023] In some embodiments of the present application, the artificial neural network includes a U-Net neural network.

[0024] In some embodiments of the present application, the U-Net neural network includes:

[0025] A first unit for expanding the data size to twice the original size through nearest neighbor interpolation;

[0026] A second unit for changing the number of channels through pointwise convolution, and using dilated convolution for all convolutions after the first two convolutional layers and the max pooling layer;

[0027] A third unit for passing the 128 points of the last layer through the Heaviside function to output 128 points with values of 0 or 1.

[0028] According to another aspect of the embodiments of the present application, there is provided a device for detecting I wave, III wave, and V wave of auditory brainstem response, including:

[0029] An ABR signal acquisition module for acquiring an auditory brainstem response ABR signal;

[0030] A filtering module for filtering the ABR signal to obtain a filtered signal;

[0031] A reliability check module for finding the I wave, III wave, and V wave in the filtered signal, checking whether the inter-wave intervals and latencies of the I wave, III wave, and V wave meet the preset reliability conditions, and when the preset reliability conditions are met, taking the I wave, III wave, and V wave found in the filtered signal as the final result and outputting their respective latencies;

[0032] A neural network module for, when the preset reliability conditions are not met, inputting the ABR signal into a pre-trained artificial neural network for processing, finding the I wave, III wave, and V wave from the processing result output by the artificial neural network, and outputting the latencies of the I wave, III wave, and V wave found from the processing result as the final result.

[0033] According to another aspect of the embodiments of the present application, there is provided an FPGA deployed with a filter, a reliability check module, and an artificial neural network;

[0034] The filter is used to filter the auditory brainstem response (ABR) signal to obtain a filtered signal;

[0035] The reliability check module is used to find waves I, III, and V in the filtered signal and check whether the inter-wave intervals and latencies of waves I, III, and V meet the preset reliability conditions;

[0036] When the preset reliability conditions are met, waves I, III, and V found in the filtered signal are used as the final results, and their respective latencies are output;

[0037] The artificial neural network module is used to, when the preset reliability conditions are not met, input the ABR signal into a pre-trained artificial neural network for processing, find waves I, III, and V from the processing results output by the artificial neural network, and output the latencies of waves I, III, and V found from the processing results as the final results.

[0038] According to another aspect of the embodiments of the present application, a detection system for waves I, III, and V of the auditory brainstem response is provided, including a signal acquisition chip and the FPGA according to any embodiment of the present application; the signal acquisition chip is connected to the FPGA; the signal acquisition chip is used to acquire the auditory brainstem response (ABR) signal.

[0039] According to another aspect of the embodiments of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the method for detecting waves I, III, and V of the auditory brainstem response according to any embodiment of the present application.

[0040] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program is executed by a processor to implement the method for detecting waves I, III, and V of the auditory brainstem response according to any embodiment of the present application.

[0041] The technical solution provided by one aspect of the embodiments of the present application may include the following beneficial effects:

[0042] The method for detecting the wave I, wave III, and wave V of the auditory brainstem response provided by the embodiment of the present application obtains the auditory brainstem response ABR signal, filters the ABR signal to obtain a filtered signal, finds the wave I, wave III, and wave V in the filtered signal, checks whether the inter-wave interval and latency of the wave I, wave III, and wave V meet the preset reliability conditions. When the preset reliability conditions are met, the wave I, wave III, and wave V found in the filtered signal are used as the final results, and their respective latencies are output. When the preset reliability conditions are not met, the ABR signal is input into a pre-trained artificial neural network for processing, and the wave I, wave III, and wave V are found from the processing results output by the artificial neural network, and the latencies of the wave I, wave III, and wave V found from the processing results are output as the final results, thereby realizing the automatic detection of the wave I, wave III, and wave V, with short time consumption, reduced manual labor, and improved work efficiency.

[0043] The above description is only an overview of the technical solution of the embodiment of the present application. In order to be able to understand the technical means of the embodiment of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of the embodiment of the present application more obvious and understandable, the following specifically gives the specific implementation manners of the present application. Brief Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0045] Figure 1 The flowchart of the method for detecting the wave I, wave III, and wave V of the auditory brainstem response according to an embodiment of the present application is shown.

[0046] Figure 2 The flowchart of filtering the ABR signal according to an embodiment of the present application is shown.

[0047] Figure 3 The structural block diagram of the device for detecting the wave I, wave III, and wave V of the auditory brainstem response according to an embodiment of the present application is shown.

[0048] Figure 4 The structural block diagram of the system for detecting the wave I, wave III, and wave V of the auditory brainstem response according to an embodiment of the present application is shown.

[0049] Figure 5 The structural schematic diagram of the U-Net network model according to an embodiment of the present application is shown.

[0050] As shown in Fig. 6(a), it is a bar chart of the accuracy comparison results between the U-Net and the BiLSTM neural network with an error of 0.1 millisecond and 0.2 millisecond.

[0051] Fig. 6(b) shows a bar chart of the comparison results of the number of parameters used by the U-Net and the number of parameters used by the BiLSTM.

[0052] Figure 7 The flowchart of the detection methods for the I wave, III wave, and V wave of the auditory brainstem response in an embodiment of the present application is shown.

[0053] Figure 8 The block diagram of the electronic device in an embodiment of the present application is shown.

[0054] Figure 9 The schematic diagram of the computer-readable storage medium in an embodiment of the present application is shown. Detailed implementation manners

[0055] In order to make the objectives, technical solutions, and advantages of the present application clearer, the present application will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present application.

[0056] Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present application belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted in an idealized or overly formal sense unless specifically defined as here.

[0057] In the related art, there is a lack of an automatic detection technical solution for the I wave, III wave, and V wave. Doctors need to manually annotate the I wave, III wave, and V wave, which is time-consuming and requires a large amount of manual labor. In view of the problems existing in the related art, an embodiment of the present application provides a method for detecting the I wave, III wave, and V wave of the auditory brainstem response, which includes obtaining the auditory brainstem response ABR signal, performing filtering processing on the ABR signal to obtain a filtered signal, checking whether the filtered signal meets a preset reliability condition, and if the preset reliability condition is met, determining the filtered signal as the I wave, III wave, and V wave of the auditory brainstem response; if the preset reliability condition is not met, inputting the filtered signal into a pre-trained artificial neural network for processing, and using the processing result output by the artificial neural network as the I wave, III wave, and V wave of the auditory brainstem response, thereby realizing the automatic detection of the I wave, III wave, and V wave, which is time-saving, reduces the manual labor amount, and improves the work efficiency.

[0058] The following describes a method, device, FPGA, device, and medium for detecting the I wave, III wave, and V wave of the auditory brainstem response according to an embodiment of the present application with reference to the accompanying drawings.

[0059] Reference Figure 1 As shown, an embodiment of the present application provides a method for detecting the I wave, III wave, and V wave of the auditory brainstem response, which may include steps S10 to S50:

[0060] S10. Obtain the auditory brainstem response ABR signal.

[0061] The auditory brainstem response ABR signal can be obtained through a signal acquisition chip. Electrodes are connected to the signal acquisition chip, such as recording electrodes, reference electrodes, and ground electrodes, and the auditory brainstem response ABR signal can be obtained through the electrodes.

[0062] S20. Perform filtering processing on the ABR signal to obtain a filtered signal.

[0063] S30. Find the I wave, III wave, and V wave in the filtered signal, and check whether the inter-wave interval and latency of the I wave, III wave, and V wave meet the preset reliability condition.

[0064] Reference Figure 2 As shown, in some embodiments, finding the I wave, III wave, and V wave in the filtered signal may include steps S301 to S302:

[0065] S301. Obtain the I wave, III wave, and V wave in the filtered ABR signal.

[0066] Exemplarily, obtaining the wave I, wave III, and wave V in the filtered ABR signal includes: for each of the wave I, the wave III, and the wave V, determining all local maxima within the latency range of the wave in the signal passing through the band-pass filter;

[0067] For each of the wave I, the wave III, and the wave V, selecting, from the delays of all local maxima of the wave passing through the band-pass filter, the point of the local maximum closest to the corresponding preset duration as the wave.

[0068] S302. Obtaining the delays of the wave I, wave III, and wave V respectively.

[0069] Exemplarily, obtaining the delays of the wave I, wave III, and wave V respectively includes: obtaining the delay of the wave according to the positions of the wave I, wave III, and wave V in the found sequence and the sampling rate.

[0070] Exemplarily, the preset reliability conditions include: the wave I, wave III, and wave V all exist, the latency of each of the wave I, wave III, and wave V belongs to the corresponding preset interval, the inter-wave interval between the wave I and wave III belongs to the corresponding preset interval, and the inter-wave interval between the wave III and wave V belongs to the corresponding preset interval.

[0071] Specifically, the preset reliability conditions may include: the wave I, wave III, and wave V all exist, the latency of the wave I is between 0.8 ms and 2 ms, the latency of the wave III is between 3.5 ms and 4.5 ms, the latency of the wave V is between 4.8 ms and 6 ms, the inter-wave interval between the wave I and the wave III is 2 ms, and the inter-wave interval between the wave III and wave V is 2 ms.

[0072] Comparing the corresponding parameters of the filtered signal with each condition in the preset reliability conditions respectively to determine whether each condition is satisfied.

[0073] S40. When the preset reliability conditions are satisfied, taking the wave I, wave III, and wave V found in the filtered signal as the final results and outputting their respective latencies.

[0074] S50. When the preset reliability conditions are not satisfied, inputting the ABR signal into a pre-trained artificial neural network for processing, finding the wave I, wave III, and wave V from the processing results output by the artificial neural network, and outputting the latencies of the wave I, wave III, and wave V found from the processing results as the final results.

[0075] Specifically, among the six conditions that the I wave, the III wave, and the V wave all exist, the latency of the I wave is between 0.8 ms and 2 ms, the latency of the III wave is between 3.5 ms and 4.5 ms, the latency of the V wave is between 4.8 ms and 6 ms, the inter-wave interval between the I wave and the III wave is 2 ms, and the inter-wave interval between the III wave and the V wave is 2 ms, as long as one condition is not met, it is determined that the preset reliability condition is not satisfied.

[0076] In some embodiments, the artificial neural network includes a U-Net neural network.

[0077] Exemplarily, the U-Net neural network includes: a first unit for expanding the data size to twice the original size by nearest neighbor interpolation; a second unit for changing the number of channels by pointwise convolution, and all convolutions after the first two convolutional layers and the max-pooling layer use dilated convolutions; a third unit for passing the 128 points of the last layer through the Heaviside function to output 128 points with values of 0 or 1.

[0078] Exemplarily, the artificial neural network is further configured to identify continuous segments with a value of 1 within the delay range of the first wave to obtain multiple segments; the first wave is any one of the I wave, the III wave, and the V wave; find the midpoint index of each segment; compare the midpoint index of each segment with the index of the first preset duration, convert the midpoint index with the smallest difference to time, and use the waveform corresponding to the converted time as the final result.

[0079] Reference Figure 3 As shown, another embodiment of the present application provides a detection device for the I wave, III wave, and V wave of auditory brainstem response, including:

[0080] An ABR signal acquisition module for acquiring an auditory brainstem response ABR signal;

[0081] A filtering module for filtering the ABR signal to obtain a filtered signal;

[0082] A reliability check module for finding the I wave, III wave, and V wave in the filtered signal, checking whether the inter-wave intervals and latencies of the I wave, III wave, and V wave meet the preset reliability conditions, and in the case of meeting the preset reliability conditions, using the I wave, III wave, and V wave found in the filtered signal as the final result and outputting their respective latencies;

[0083] A neural network module, which is used to process the ABR signal by inputting it into a pre-trained artificial neural network when a preset reliability condition is not met, and find the wave I, wave III, and wave V from the processing results output by the artificial neural network, and output the latencies of the wave I, wave III, and wave V found from the processing results as the final result.

[0084] Exemplarily, the neural network module is further used to identify continuous segments with a value of 1 within the delay range of the first wave to obtain multiple segments; the first wave is any one of wave I, wave III, and wave V;

[0085] Find the midpoint index of each segment;

[0086] Compare the midpoint index of each segment with the index of the first preset duration, convert the midpoint index with the smallest difference into time, and use the waveform corresponding to the converted time as the final result.

[0087] Exemplarily, the reliability check module includes:

[0088] A first filtering unit, which is used to obtain wave I, wave III, and wave V in the filtered ABR signal;

[0089] A delay acquisition unit, which is used to acquire the delays of wave I, wave III, and wave V respectively.

[0090] Exemplarily, acquiring the delays of wave I, wave III, and wave V includes:

[0091] For each of wave I, wave III, and wave V, determine all local maxima within the latency range of the wave in the signal passing through the band-pass filter;

[0092] For each of wave I, wave III, and wave V, select, through the band-pass filter, the point of the local maximum closest to the corresponding preset duration from the delays of all local maxima of the wave as the wave, and obtain the delay of the wave according to the positions of wave I, wave III, and wave V in the found sequence and the sampling rate.

[0093] Exemplarily, the first filtering unit is further specifically used for:

[0094] For each of wave I, wave III, and wave V, determine all local maxima within the latency range of the wave in the signal passing through the band-pass filter;

[0095] The delay acquisition unit is further specifically used for: for each of wave I, wave III, and wave V, select, through the band-pass filter, the delay of the local maximum closest to the corresponding preset duration from the delays of all local maxima of the wave as the delay of the wave.

[0096] Exemplarily, the preset reliability conditions include: the presence of waves I, III, and V, the latency of each of waves I, III, and V belonging to a corresponding preset interval, the inter-wave interval between waves I and III belonging to a corresponding preset interval, and the inter-wave interval between waves III and V belonging to a corresponding preset interval.

[0097] Exemplarily, the preset reliability conditions include: the presence of waves I, III, and V, the latency of wave I being between 0.8 ms and 2 ms, the latency of wave III being between 3.5 ms and 4.5 ms, the latency of wave V being between 4.8 ms and 6 ms, the inter-wave interval between wave I and wave III being 2 ms, and the inter-wave interval between wave III and wave V being 2 ms.

[0098] Exemplarily, the artificial neural network includes a U-Net neural network.

[0099] Exemplarily, the U-Net neural network includes:

[0100] A first unit for expanding the data size to twice the original size by nearest neighbor interpolation;

[0101] A second unit for changing the number of channels by pointwise convolution, and using dilated convolution for all convolutions after the first two convolutional layers and the max-pooling layer;

[0102] A third unit for passing the 128 points of the last layer through the Heaviside function to output 128 points with values of 0 or 1.

[0103] The device for detecting waves I, III, and V of the auditory brainstem response provided by the embodiments of the present application can acquire the auditory brainstem response ABR signal, perform filtering processing on the ABR signal to obtain a filtered signal, find waves I, III, and V in the filtered signal, check whether the inter-wave intervals and latencies of waves I, III, and V meet the preset reliability conditions. When the preset reliability conditions are met, waves I, III, and V found in the filtered signal are used as the final results, and their respective latencies are output. When the preset reliability conditions are not met, the ABR signal is input into a pre-trained artificial neural network for processing, and waves I, III, and V are found from the processing results output by the artificial neural network, and the latencies of waves I, III, and V found from the processing results are output as the final results, thereby realizing the automatic detection of waves I, III, and V, with short time consumption, reduced manual labor, and improved work efficiency.

[0104] The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments. Their similarities or resemblances can be referred to each other. For the sake of brevity, they will not be elaborated herein.

[0105] Another embodiment of the present application provides an FPGA (Field Programmable Gate Array) deployed with a filter, a reliability check module, and an artificial neural network. The filter is used to filter the auditory brainstem response (ABR) signal to obtain a filtered signal. The reliability check module is used to find waves I, III, and V in the filtered signal and check whether the inter-wave intervals and latencies of waves I, III, and V meet preset reliability conditions. When the preset reliability conditions are met, waves I, III, and V found in the filtered signal are used as the final results, and their respective latencies are output. The artificial neural network module is used to, when the preset reliability conditions are not met, input the ABR signal into a pre-trained artificial neural network for processing, find waves I, III, and V from the processing results output by the artificial neural network, and output the latencies of waves I, III, and V found from the processing results as the final results.

[0106] The FPGA provided by the embodiment of the present application can realize the automatic detection of waves I, III, and V, which takes a short time, reduces the manual labor amount, and improves the work efficiency.

[0107] In the related art, the clinical ABR devices used to obtain waves I, III, and V are large in volume, heavy in weight, and high in power consumption, making them not suitable for application scenarios such as ICU or remote or outdoor hearing screening scenarios. Compared with the related art, the FPGA provided by the embodiment of the present application is small in volume, light in weight, and low in power consumption, and is suitable for application scenarios such as ICU or remote or outdoor hearing screening scenarios.

[0108] Another embodiment of the present application provides a detection system for waves I, III, and V of auditory brainstem response, including a signal acquisition chip and the FPGA described above. The signal acquisition chip is connected to the FPGA. The signal acquisition chip is used to acquire the auditory brainstem response (ABR) signal. The detection system for waves I, III, and V of auditory brainstem response provided by the embodiment of the present application is small in volume, light in weight, and low in power consumption, and is suitable for application scenarios such as ICU or remote or outdoor hearing screening scenarios.

[0109] The detection system for waves I, III, and V of auditory brainstem response provided by the embodiment of the present application can realize the automatic detection of waves I, III, and V, which takes a short time, reduces the manual labor amount, and improves the work efficiency.

[0110] The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments. Their similarities or resemblances can be referred to each other. For the sake of brevity, they will not be elaborated herein.

[0111] Another embodiment of the present application proposes an automatic identification device for wave I, wave III, and wave V of ABR based on FPGA. This device integrates an ABR automatic identification algorithm, which inputs the delays of wave I, wave III, and wave V and the ABR waveform into the FPGA for processing to identify wave I, wave III, and wave V. This algorithm consists of two parts: a filtering algorithm and a U-Net neural network algorithm, and it achieves an average accuracy of 91.96% on a self-built dataset.

[0112] Compared with the related technology that only uses the U-Net neural network algorithm, this method reduces the energy consumption by 19.64% on the self-built dataset and has great practical value in scenarios that require long-term or repeated automatic detection tasks of wave I, wave III, and wave V of ABR.

[0113] In the related technology, in order to promote the portability and automation of the auditory brainstem response (ABR) detection system, a wireless EEG transmission system is used to collect data and send the collected data to the host for processing, but none of these systems integrate an automatic detection algorithm. The automatic identification device for wave I, wave III, and wave V of ABR based on FPGA is a high-precision and low-power on-chip signal processing system.

[0114] Since the voltage of the auditory brainstem response (ABR) is extremely low (lower than 1 μV), the usual method is to perform multiple stimulations and average the EEG signals to reduce noise. This method assumes that the signal components in the EEG (except for the ABR signal) are Gaussian noise. However, in real EEG signals, the noise is not always Gaussian distributed. Therefore, even after multiple averaging, the characteristics of wave I, wave III, and wave V are still not clear, showing as small peaks or inflection points, which will affect the identification of ABR waves. The ABR wave identification methods in the related technology rely on identifying peaks and inflection points to determine wave I, wave III, and wave V. However, these methods require complex algorithms, such as the slope range on both sides of the inflection point, the peak amplitude range, and the criteria for selecting between multiple peaks or inflection points, which makes them impractical in clinical applications. In order to simplify the algorithm and improve the robustness and generality, some other technical solutions in the related technology use filter reconstruction to reduce noise, which performs multi-level wavelet decomposition and reconstruction at a specific wavelet scale. However, this method has two problems. First, the filtered ABR wavelet may be missing. Second, the peak points of wave I, wave III, and wave V in the filtered ABR may deviate significantly from the original values, resulting in errors.

[0115] To improve the recognition accuracy, some technical solutions in the related art use a BiLSTM network to recognize the I wave, III wave, and V wave of ABR, achieving a clinical available accuracy of 92.91%. However, this method requires a large number of network parameters, making it unfriendly to the hardware of portable devices.

[0116] In view of the problems existing in the related art, the embodiment of the present application proposes a detection system for the I wave, III wave, and V wave of ABR that combines filtering processing and a neural network implemented on an FPGA. The detection system includes a U-Net network model and an FIR filter. The detection system can receive the input of the ABR signal and then output the delays of the I wave, III wave, and V wave. The proposed algorithm can be implemented on a Xilinx Ultra96, which is a lightweight FPGA chip with a weight of only 123.3 grams.

[0117] Exemplarily, as Figure 4 shown, the I wave, III wave, and V wave detection system includes a signal acquisition chip and an FPGA; a recording electrode, a reference electrode, and a ground electrode are connected to the signal acquisition chip, and the signal acquisition chip is connected to the FPGA; the signal acquisition chip obtains the auditory brainstem response ABR signal through the electrodes and inputs the data of the ABR signal to the FPGA through the SPI interface. The data is preprocessed in the PS (Processing System) part and then processed in the PL (Programmable Logic) part, the data is filtered, and the result is checked by a reliability check module. If the check is passed, the result is output; otherwise, the original data is input into the U-Net neural network, and the I wave, III wave, and V wave are selected from the output of the U-Net neural network. The final result will be transmitted back from the PL side to the PS side and wirelessly sent to the host.

[0118] In one example, the training data used comes from 65 patients using commercial devices and the proposed device. The provided ABR data is downsampled to 15625Hz, and the first 128 points (about 8ms) are regarded as the final ABR data. A total of 204 pieces of data are obtained by presenting click sounds of 90 or 80dB to the subjects. This data set includes common ABR clinical waveforms, such as waveforms containing fused double waves, fuzzy characteristic waveforms, and waveforms lacking characteristic waves. To ensure the consistency between the data collected from different devices, min-max normalization is applied to normalize the data to 0-15 bit unsigned integers.

[0119] The doctor manually marked the I wave, III wave, and V wave in the originally collected ABR waveform. Based on these tags, two points before and after each marked point were also marked. The maximum deviation between the later marked points and the originally marked points was 0.128 ms, which is clinically acceptable. Then, the doctor reviewed these points and deleted those that deviated significantly from the original characteristics. If the detection result of the automatic algorithm fell within the marked range, it was considered correct.

[0120] The energies of the ABR I wave, III wave, and V wave are usually concentrated in the range of 100 - 2 kHz. To avoid excessive high - frequency signal attenuation, which may cause a significant shift in the ABR peak point, a 40 - order linear FIR band - pass filter with a frequency range of 100 - 3000 Hz was used. After filtering, the delay error caused by group delay was eliminated through delay compensation.

[0121] The ABR data first passed through a band - pass filter. The band - pass filter identified all local maxima within the latency ranges of the I wave, III wave, and V wave. If there were multiple maxima within these delay ranges, the maximum points closest to 1.5 ms, 3.5 ms, and 5.5 ms were respectively selected as the final delays. Then, a reliability check was performed on the results.

[0122] The six conditions for meeting the reliability requirements are as follows:

[0123] 1) The I wave, III wave, and V wave must all be present.

[0124] 2) The latency of the I wave should be between 0.8 ms and 2 ms.

[0125] 3) The latency of the III wave should be between 3.5 ms and 4.5 ms.

[0126] 4) The latency of the V wave should be between 4.8 ms and 6 ms.

[0127] 5) The inter - wave interval between the I wave and the III wave should be approximately 2 ms.

[0128] 6) The inter - wave interval between the III wave and the V wave should be approximately 2 ms.

[0129] These six conditions represent the typical characteristics of the ABR waveform of healthy individuals without obvious hearing impairments. If the results do not meet these conditions, it indicates that the ABR waveform is complex and will trigger the neural network for subsequent processing. In the dataset, the proportion of ABR data passing the reliability check was 23.53%.

[0130] The detection accuracy of the characteristic waves of the data passing the reliability check is shown in Table I.

[0131] The filtering processing result passing the reliability check has high reliability, making it suitable for the automatic detection of ABR characteristic waves.

[0132] To reduce the hardware power consumption, the neural network algorithm in this example is different from the algorithms adopted in the related technologies in two main aspects.

[0133] First, the data sampling rate used in this example is 15625Hz, which is lower than the sampling rate used in the related technologies. The reduction of the sampling rate makes the neural network model in this embodiment smaller, thus reducing the overall power consumption. Second, this example adopts the U-Net neural network structure. Due to its encoding and decoding architecture, it can effectively extract global features and local details. This structure is particularly suitable for small sample data sets and shows excellent performance in image segmentation and sequence segmentation tasks. In addition, dilated convolutions are incorporated into the design to expand the receptive field, enabling the neural network to capture dependencies at different time points.

[0134] The training data of the neural network is a 1×128 sequence, which can be expressed as

[0135] ABR ori (128) = {a1, a2, …, a 128} (1),

[0136] where a i represents a 15-bit unsigned number obtained after min-max normalization. The label is also a 1×128 sequence, which can be expressed as

[0137] ABR target (128) = {b1, b2, …, b 128} (2),

[0138] where, corresponding to ABR ori , b i is 0 or 1. For ABR target , the marked points are set to 1, while other marked points are set to 0.

[0139] The U-Net network structure used in this embodiment is as Figure 5 shown. The upsampling process adopted in this embodiment includes two steps.

[0140] The first step is nearest neighbor interpolation to expand the data size to twice the original size, and then the second step is pointwise convolution to change the number of channels. All convolutions after the two convolutional layers and the max pooling layer use dilated convolutions. The 128 points in the last layer pass through the Heaviside function to output 128 points with values of 0 or 1.

[0141] Taking waveform III as an example, the algorithm for extracting waveforms I, III, and V from the neural network output is as follows: 1) Identify continuous segments with a value of 1 within the delay range of waveform III.

[0142] Find the midpoint index of each segment; if the segment has an even number of points, use the point with the smaller index among the two middle points.

[0143] Compare the midpoint index of each segment with the index at 3.5 ms, select the midpoint index with the smallest difference, and convert it to time as the final result.

[0144] The method for identifying waveforms I and V is similar. If there are multiple candidate points within the corresponding waiting time, the comparison metric points are the points at 1.5 ms and 5.5 ms respectively.

[0145] The neural network algorithm was cross-validated five times on the dataset, and the detection accuracies of waveforms I, III, and V are shown in Table I. The average accuracies of waveforms I, III, and V are approximately 91.56%, and the detection accuracies of the three waveforms basically reach the clinically available level.

[0146] To verify the effectiveness of the method in this embodiment and its hardware implementation advantages, the average accuracy of feature wave detection was compared with the best accuracy achieved by the BiLSTM neural network in previous work and the number of parameters in the two networks. As shown in Fig. 6(a), it is a bar chart of the accuracy comparison results between the U-Net and the BiLSTM neural network under 0.1 ms and 0.2 ms errors, and Fig. 6(b) shows a bar chart of the comparison results of the number of parameters used by the U-Net and the number of parameters used by the BiLSTM. It can be seen that at an error of 0.128 ms, the accuracy of the method proposed in this embodiment is 6.1% higher than the accuracy of the BiLSTM at an error of 0.1 ms and only 1.35% lower than the BiLSTM at an error of 0.2 ms.

[0147] In terms of the number of parameters, the method in this embodiment uses approximately 837 times fewer parameters than a 3-layer BiLSTM with an input length of 321 and a hidden layer size of 512, which greatly facilitates hardware implementation and reduces energy consumption.

[0148] In a specific example, the overall method flow is as Figure 7 shown.

[0149] First, use a filtering algorithm to process the preprocessed ABR data to obtain the delay of the feature wave, and then perform a reliability check on the delay of the feature wave.

[0150] If the delay of the feature wave passes the reliability check, it is accepted as the final result.

[0151] If the delay of the characteristic wave fails the reliability check, the ABR data is input into the U-Net neural network for processing, and the processing result output by the U-Net neural network is used as the final result.

[0152] The entire algorithm was subjected to five-fold cross-validation, and the average accuracy was 91.96%. Compared with the technical solutions that only use neural networks in the related technologies, the accuracy has been greatly improved. The detection accuracies of the I wave, III wave, and V wave are shown in Table I respectively. It can be seen that after applying the reliability check filtering process, the detection accuracies of the I wave, III wave, and V wave remain at a high level, which is suitable for clinical research.

[0153] Table I Detection Accuracy of Characteristic Waveforms

[0154]

[0155] In a specific example, this embodiment implements quantization-aware training on the neural network, and the final weights are quantized to 8-bit signed integers. At this stage, the accuracy remains almost unchanged, and the total parameter size is approximately 11.95 KB, so that the hardware overhead can be reduced while maintaining a high accuracy. 40 filter parameters are expanded to 1024 times their original values, rounded, quantized to 8-bit signed integers, and stored in the FPGA.

[0156] The dynamic energy consumption of one calculation of the neural network is 0.6 mJ, while the filtering algorithm requires approximately 0.0245 mJ for each run. Applying the overall algorithm to the self-built dataset of this embodiment, 23.53% of the dataset passed the reliability check of the filtering algorithm, and 76.47% of the dataset requires neural network processing.

[0157] Compared with the technical solutions that only use neural networks in the related technologies, the energy consumption of performing one identification of the I wave, III wave, and V wave on the entire dataset using the combined algorithm of the filtering algorithm and the neural network is reduced to 80.36%. It can be seen that combining the filtering process and the neural network algorithm provides a greater degree of optimization in terms of energy consumption, especially in scenarios where multiple ABR detections are required. The method of this embodiment is more obvious in saving energy consumption.

[0158] This embodiment proposes a U-Net detector based on EEG for automatically identifying characteristic waves in ABR. Compared with the neural network models in the related technologies, the proposed method significantly reduces the number of parameters while maintaining almost the same accuracy level, greatly reducing the hardware overhead and energy consumption.

[0159] In addition, this embodiment adopts a combined algorithm that uses filtering and neural network methods for ABR feature waveform detection. The average accuracy rate of this method on the self-built dataset is 91.96%. By using filtering processing before triggering U-Net on the self-built dataset, the power consumption is reduced by 19.64%.

[0160] The method of this embodiment has high practical value in clinical and scientific research scenarios where it is necessary to repeat the identification tasks of the I wave, III wave, and V wave of ABR by reducing the overall energy consumption.

[0161] The descriptions of the above embodiments tend to emphasize the differences between the embodiments. Their similarities can be referred to each other. For the sake of brevity, they will not be elaborated herein.

[0162] Another embodiment of this application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the computer program to implement the method of any of the above embodiments.

[0163] Refer to Figure 8 As shown, the electronic device 10 may include: a processor 100, a memory 101, a bus 102, and a communication interface 103. The processor 100, the communication interface 103, and the memory 101 are connected through the bus 102; a computer program that can run on the processor 100 is stored in the memory 101. When the processor 100 runs the computer program, it executes the method provided by any of the foregoing embodiments of this application.

[0164] Among them, the memory 101 may include a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 103 (which can be wired or wireless), a communication connection is realized between this device network element and at least one other network element, and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0165] The bus 102 may be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Among them, the memory 101 is used to store programs. After receiving an execution instruction, the processor 100 executes the program. The method disclosed in any of the foregoing embodiments of this application can be applied to the processor 100 or implemented by the processor 100.

[0166] The processor 100 may be an integrated circuit chip with the ability to process signals. In the implementation process, each step of the above method may be completed by the integrated logic circuit of the hardware in the processor 100 or the instructions in the form of software. The above-mentioned processor 100 may be a general-purpose processor, which may include a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application may be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 101, and the processor 100 reads the information in the memory 101 and combines its hardware to complete the steps of the above method.

[0167] The electronic device provided by the embodiments of the present application and the method provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the method adopted, run or implemented by it.

[0168] Another embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. The computer program is executed by a processor to implement the method of any of the above embodiments. Refer to Figure 9 As shown, the computer-readable storage medium shown is an optical disc 20, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will execute the method provided by any of the foregoing embodiments.

[0169] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here one by one.

[0170] The computer-readable storage medium provided by the above embodiments of the present application and the method provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run, or implemented by the application programs stored therein.

[0171] It should be noted that:

[0172] The term "module" is not intended to be limited to a specific physical form. Depending on the specific application, a module can be implemented as hardware, firmware, software, and / or a combination thereof. In addition, different modules can share common components or even be implemented by the same components. There may or may not be a clear boundary between different modules.

[0173] The algorithms and displays provided herein are not inherently related to any particular computer, virtual device, or other equipment. Various general-purpose devices can also be used in conjunction with the examples based herein. Based on the above description, the structure required to construct such devices is obvious. In addition, the present application is not directed to any particular programming language. It should be understood that the content of the present application described herein can be implemented using various programming languages, and the description of a specific language above is for disclosing the best implementation mode of the present application.

[0174] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown sequentially in the direction of the arrows, these steps are not necessarily executed sequentially in the direction of the arrows. Unless otherwise clearly stated in this document, the execution of these steps is not strictly limited in order and can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time but can be executed at different times, and their execution order is not necessarily sequential but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0175] The above embodiments only represent the implementation modes of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for detecting wave I, wave III, and wave V of auditory brainstem response, characterized in that, Including: Obtaining an auditory brainstem response (ABR) signal; Performing filtering processing on the ABR signal to obtain a filtered signal; Finding waves I, III, and V in the filtered signal, and checking whether the inter-wave intervals and latencies of waves I, III, and V meet preset reliability conditions; When the preset reliability conditions are met, taking waves I, III, and V found in the filtered signal as the final results, and outputting their respective latencies; When the preset reliability conditions are not met, inputting the ABR signal into a pre-trained artificial neural network for processing, finding waves I, III, and V from the processing results output by the artificial neural network, and outputting the latencies of waves I, III, and V found from the processing results as the final results.

2. The method according to claim 1, wherein The finding waves I, III, and V in the filtered signal includes: Obtaining waves I, III, and V in the filtered ABR signal; Respectively obtaining the delays of waves I, III, and V.

3. The method according to claim 2, wherein The obtaining waves I, III, and V in the filtered ABR signal includes: For each of waves I, III, and V, determining all local maxima within the latency range of the wave in the signal passing through the band-pass filter; For each of waves I, III, and V, selecting, from the delays of all local maxima of the wave through the band-pass filter, the point of the local maximum closest to the corresponding preset duration as the wave.

4. The method according to claim 2, wherein The respectively obtaining the delays of waves I, III, and V includes: Obtaining the delay of the wave according to the positions of waves I, III, and V in the found sequence and the sampling rate.

5. The method according to any one of claims 1-4, characterized in that, The preset reliability conditions include: waves I, III, and V all exist, the latency of each of waves I, III, and V belongs to a corresponding preset interval, the inter-wave interval between waves I and III belongs to a corresponding preset interval, and the inter-wave interval between waves III and V belongs to a corresponding preset interval.

6. The method according to claim 5, wherein The preset reliability conditions include: waves I, III, and V all exist, the latency of wave I is between 0.8 ms and 2 ms, the latency of wave III is between 3.5 ms and 4.5 ms, the latency of wave V is between 4.8 ms and 6 ms, the inter-wave interval between wave I and wave III is 2 ms, and the inter-wave interval between wave III and wave V is 2 ms.

7. The method according to any one of claims 1-4, characterized in that The artificial neural network includes a U-Net neural network.

8. The method according to claim 7, characterized in that The U-Net neural network includes: A first unit for expanding the data size to twice the original size through nearest neighbor interpolation; A second unit for changing the number of channels through pointwise convolution, and using dilated convolution for all convolutions after the first two convolutional layers and the max-pooling layer; A third unit for passing the 128 points of the last layer through the Heaviside function to output 128 points with values of 0 or 1.

9. An apparatus for detecting wave I, wave III, and wave V of auditory brainstem response, characterized in that, Including: An ABR signal acquisition module for obtaining an auditory brainstem response (ABR) signal; A filtering module for performing filtering processing on the ABR signal to obtain a filtered signal; A reliability check module, which is used to find wave I, wave III, and wave V in the filtered signal, check whether the inter-wave intervals and latencies of wave I, wave III, and wave V meet the preset reliability conditions, and when the preset reliability conditions are met, take wave I, wave III, and wave V found in the filtered signal as the final results and output their respective latencies; A neural network module, which is used when the preset reliability conditions are not met, input the ABR signal into a pre-trained artificial neural network for processing, find wave I, wave III, and wave V from the processing results output by the artificial neural network, and output the latencies of wave I, wave III, and wave V found from the processing results as the final results.

10. An FPGA, characterized in that, It is deployed with a filter, a reliability check module, and an artificial neural network; The filter is used to filter the auditory brainstem response (ABR) signal to obtain a filtered signal; The reliability check module is used to find wave I, wave III, and wave V in the filtered signal and check whether the inter-wave intervals and latencies of wave I, wave III, and wave V meet the preset reliability conditions; When the preset reliability conditions are met, take wave I, wave III, and wave V found in the filtered signal as the final results and output their respective latencies; When the preset reliability conditions are not met, the artificial neural network inputs the ABR signal into a pre-trained artificial neural network for processing, finds wave I, wave III, and wave V from the processing results output by the artificial neural network, and outputs the latencies of wave I, wave III, and wave V found from the processing results as the final results.

11. An auditory brainstem response I wave, III wave, and V wave detection system, characterized in that, It includes a signal acquisition chip and the FPGA as described in claim 10; the signal acquisition chip is connected to the FPGA; the signal acquisition chip is used to acquire the auditory brainstem response (ABR) signal.

12. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the method for detecting wave I, wave III, and wave V of the auditory brainstem response as described in any one of claims 1-8.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by the processor to implement the method for detecting wave I, wave III, and wave V of the auditory brainstem response as described in any one of claims 1-8.

Images