A VEP diagnostic system for multi-dimensional decoupling modulation and pathway index quantitative evaluation

Abstract

The application discloses a VEP diagnosis system for multi-dimensional decoupling modulation and channel index quantitative evaluation, which comprises a stimulation generation module, a real-time correction unit, a signal acquisition module, an embedded processing module and an analysis and diagnosis module. The stimulation generation module is used for generating visual stimulation patterns which can be subjected to multi-dimensional parameter dynamic decoupling modulation of color contrast, brightness, spatial frequency and time frequency. The real-time correction unit is used for maintaining the brightness fluctuation of a non-target stimulation axis within a predetermined threshold. The signal acquisition module acquires multi-lead VEP signals. The embedded processing module is used for realizing phase locking and artifact separation. The analysis and diagnosis module receives the processed VEP signals, and extracts response characteristics of latency and standardized amplitude. Through multi-dimensional decoupling modulation of stimulation and streamlined data processing, the system realizes the isolation evaluation and functional quantification of the visual channel, and improves the objectivity and reliability of diagnosis.

Description

Technical Field

[0001] This application relates to the field of visual diagnostic technology, specifically to a VEP diagnostic system with multidimensional decoupling modulation and quantitative assessment of pathway index. Background Technology

[0002] In the field of medical diagnostics, the acquisition, processing, and evaluation of visual evoked potentials (VEPs) are of great significance for detecting visual pathway-related diseases. With the continuous advancement of medical research, the necessity of precise analysis and quantitative assessment of visual pathways has become increasingly recognized. Through VEP research, we can gain a deeper understanding of the functional status of the three major visual pathways: macrocytes (M cells), microcytes (P cells), and kinetes (K cells), providing crucial evidence for the diagnosis and treatment of visual system diseases. This helps doctors more accurately assess conditions, develop personalized treatment plans, improve treatment outcomes, and enhance patients' visual health.

[0003] Prior to the technical solution of this invention, various conventional methods existed for the research and application of visual evoked potentials (VEPs). In terms of stimulus generation, relatively simple visual stimulus patterns were typically used, but their parameter modulation was not flexible enough, making it difficult to achieve precise control over multiple dimensions of parameters. For example, parameters such as color contrast, brightness, spatial frequency, and temporal frequency could often only be adjusted in a single dimension, failing to achieve dynamic decoupling modulation of multiple parameters. During signal acquisition, electrode configurations might not be standardized, resulting in inaccurate and incomplete signal acquisition. Simultaneously, the noise suppression capability and input impedance of the acquisition equipment might be low, leading to a low signal-to-noise ratio at the acquisition end for weak VEP signals. In the signal processing stage, filtering methods might be relatively simple, unable to effectively remove various noises and artifacts. Signal analysis and diagnosis also primarily relied on qualitative methods, lacking precise quantitative evaluation indicators, making it difficult to accurately determine the functional state of the visual pathway.

[0004] The limitations of existing technologies lie in the difficulty of achieving physiological isolation and quantitative functional assessment of the three major visual pathways (M, P, and K) using conventional methods. The lack of flexibility in the parameter modulation of stimulus patterns prevents the precise induction of responses in specific visual pathways, making it difficult to obtain accurate characteristic information for each pathway. Insufficient signal acquisition and processing result in low-quality signals with significant noise and artifacts, affecting the accuracy of subsequent analysis and diagnosis. Furthermore, the lack of precise quantitative assessment indicators makes it difficult for physicians to objectively and accurately assess a patient's visual function, hindering disease diagnosis and treatment. Summary of the Invention

[0005] To address the technical problems in the prior art, this application provides a VEP diagnostic system with multidimensional decoupling modulation and quantitative assessment of pathway index.

[0006] The VEP diagnostic system provided in this application, which utilizes multidimensional decoupling modulation and quantitative assessment of pathway indices, employs the following technical solution:

[0007] A VEP diagnostic system based on multidimensional decoupling modulation and quantitative assessment of pathway indices includes:

[0008] The stimulus generation module is used to generate stimuli that can be adjusted for color contrast, brightness, and spatial frequency. and time frequency The module generates a visual stimulus pattern through multidimensional parameter dynamic decoupling modulation and outputs a stimulus synchronization signal; the module supports stepped contrast modulation to generate the stimulus sequence required for fitting an objective contrast sensitivity function.

[0009] A real-time correction unit is used to monitor and compensate for chromatic aberration in real time during the presentation of the visual stimulus pattern, so as to maintain the brightness fluctuation of the non-target stimulus axis within a predetermined threshold.

[0010] The signal acquisition module is used to acquire multi-lead VEP signals generated on the scalp of the test subject while viewing the visual stimulus pattern;

[0011] The embedded processing module, using an FPGA / ARM architecture, receives the stimulus synchronization signal and the multi-lead VEP signal to achieve phase locking and artifact separation.

[0012] The analysis and diagnosis module receives the VEP signal processed by the embedded processing module and extracts the latency based on the signal. The response characteristics of the normalized amplitude A were analyzed, and the visual pathway functional separation index was calculated. An objective contrast sensitivity function was fitted to perform quantitative diagnostic assessment.

[0013] In some embodiments, the stimulation generation module can generate the following four specific stimulation patterns, which are determined by parameters. and The limitation of these pathways enables physiological isolation of the M, P, and K pathways, and is used to induce and obtain the specific VEP response characteristics required by the analysis and diagnostic module.

[0014] a) M-channel specific stimulation: a black and white sinusoidal grating, using time-frequency... Limited to Hz, spatial frequency Limited to cpd; the stimulus is used to obtain the latency characteristics of the M pathway. ;

[0015] b) P-channel specific stimulation: an isoluminous red-green sinusoidal grating, using time-frequency... Limited to Hz, spatial frequency Limited to cpd; The stimulation is used to obtain the latency and amplitude characteristics of the P pathway;

[0016] c) K-channel specific stimulation: An isoluminous blue-yellow sinusoidal grating, using time-frequency... Limited to Hz; the stimulation is used to obtain the latency and amplitude characteristics of the K pathway;

[0017] d) Luminance / color fusion channel stimulation: A color grating with luminance contrast is used at a high spatial frequency; the stimulation is used to obtain the amplitude characteristics of the fusion channel as a benchmark for the pathway separation index.

[0018] In some embodiments, the correction unit incorporates a spectroradiometer, and the correction algorithm employs iterative least squares to control the brightness fluctuation of the non-target stimulus axis to no higher than [a certain value] under different step contrast modulations. .

[0019] In some embodiments, the filtering section of the signal acquisition module employs a hybrid adaptive filtering strategy to process the acquired multi-lead VEP signal. The hybrid adaptive filtering strategy includes the following steps:

[0020] a) Using the multi-lead VEP signal as input, an integrated notch filter circuit or a digital notch filter is used to remove power frequency interference and its higher-order harmonics in real time, thereby obtaining a pre-filtered VEP signal.

[0021] b) Using the pre-filtered VEP signal as input, the dynamic wavelet transform algorithm is used for decomposition. The dynamic wavelet transform algorithm decomposes the signal into different frequency band scales by selecting specific wavelet basis functions to obtain the original wavelet coefficient set.

[0022] c) Low-frequency transient VEP noise suppression: Using the original wavelet coefficient set as input, the wavelet coefficients concentrated on the preset low-frequency scale are thresholded to eliminate baseline drift and breathing artifacts on the preset low-frequency scale, while preserving the preset low-frequency transient VEP components to the maximum extent, thus obtaining a pure wavelet coefficient set.

[0023] d) Effective bandwidth reconstruction: Using the pure wavelet coefficient set as input, the wavelet coefficients are reconstructed based solely on the preset effective physiological bandwidth using the inverse wavelet transform method to obtain the noise-removed VEP signal.

[0024] In some embodiments, the artifact separation in the embedded processing module uses an independent component analysis algorithm to process the multi-lead VEP signal, and the artifact separation step includes:

[0025] a) Using the multi-lead VEP signal as input, blind source separation is performed using the independent component analysis algorithm to obtain a set of mutually independent IC vectors;

[0026] b) Perform time-domain and frequency-domain feature analysis on the IC vector to identify and label artifact components that highly match the artifact features of electrooculography, electromyography, and electrocardiography;

[0027] c) Using the IC vector as input, during the inverse ICA transformation, the weights corresponding to the artifact components are set to zero or removed, and the signal is reconstructed using only the non-artifact components to obtain the VEP pure signal, which is the final output of the embedded processing module.

[0028] In some embodiments, the embedded processing module further includes the following functions:

[0029] a) Using the stimulus synchronization signal and the multi-lead VEP signal as inputs, an optical trigger and hardware interrupt mechanism are employed to time-stamp and synchronously superimpose the acquired VEP data, thereby limiting the time error between the stimulus and the VEP signal to within a certain range. ms, to obtain the synchronously superimposed VEP signal;

[0030] b) Using the multi-lead VEP signal as input, the low-frequency VEP signal is processed by real-time high-pass filtering combined with an adaptive threshold algorithm. By continuously monitoring and adjusting the zero-point baseline of the signal, the VEP signal after baseline zero-point correction is obtained.

[0031] In some embodiments, the embedded processing module further includes a spatial filtering unit, which processes the multi-lead VEP signal, and the spatial filtering step includes:

[0032] a) Based on the electrode position information of the multi-lead VEP signal and the preset head model, establish the transfer matrix between the cortical source and the scalp potential;

[0033] b) Using the multi-lead VEP signal and the transfer matrix as input, the conjugate gradient method is used to solve the least squares solution of the cortical source current density in order to estimate the source activity intensity of the primary visual cortex.

[0034] c) The source activity intensity of the primary visual cortex is mapped to an enhanced VEP signal to obtain an enhanced primary visual cortex-specific VEP response signal, which is used to extract the latency and amplitude characteristics of the M, P, and K pathways more accurately.

[0035] In some embodiments, the visual pathway functional separation index calculated by the analysis and diagnosis module includes at least the following three sets of quantitative indicators:

[0036] a) P / M channel relative latency delay index : ;

[0037] b) P / M channel relative amplitude suppression index : ;

[0038] c) K / P channel relative function ratio, wherein the channel relative function ratio includes and :

[0039] ;

[0040]

[0041] Where L is the VEP latency and A is the VEP normalized amplitude.

[0042] In some embodiments, the VEP-CSF objective fitting in the analysis and diagnosis module is performed on the M, P, and K channels, and the fitting step includes:

[0043] a) Using the VEP amplitude data acquired under the stepped contrast modulation as input, and standardizing the amplitude value, the VEP amplitude data includes the contrast level and the corresponding amplitude value;

[0044] b) Using the standardized amplitude data as input, a nonlinear regression model is selected to perform least squares fitting on the curve of VEP amplitude value changing with the logarithm of contrast, and the fitted curve equation is obtained.

[0045] c) Based on the fitted curve equation, extract at least the following three parameters as diagnostic feature vectors:

[0046] i) Sensitivity peak : The maximum amplitude when the curve saturates;

[0047] ii) Contrast threshold The contrast required for the curve to reach the preset minimum response;

[0048] iii) Contrast half-saturation The curve reached The required contrast at half.

[0049] In some embodiments, the analysis and diagnosis module employs an ensemble learning algorithm for final diagnosis, and the final diagnosis steps include:

[0050] a) Construct a multidimensional diagnostic feature vector using the visual pathway functional separation index and the VEP-CSF fitting parameters as input;

[0051] b) Model classification: Using the diagnostic feature vector as input, an ensemble learning algorithm is used to classify and analyze the visual function status of the test subjects;

[0052] c) Output diagnostic results: Based on the model classification results, output the following two types of diagnostic information:

[0053] i) Objective quantitative grading: color vision abnormalities or optic nerve lesions are classified into pre-defined levels;

[0054] ii) Etiological identification: Output the probability of the etiology.

[0055] In summary, this application includes at least one of the following beneficial technical effects:

[0056] 1. Through multidimensional parameter decoupling modulation and real-time correction, physiological isolation of the three major visual pathways M, P, and K was achieved, and VEP response characteristics specific to different pathways were obtained. The stimulus generation module dynamically decouples and modulates by limiting the time and spatial frequencies, increasing the activation level of the target pathway while reducing interference from non-target pathways.

[0057] 2. By employing a hybrid adaptive filtering, artifact separation, and time-locking approach, effective separation and suppression of various artifacts in the VEP signal are achieved, improving the signal-to-noise ratio and measurement stability of the VEP signal. This effect is achieved through multi-level embedded processing: First, the embedded processing module integrates notch filtering and adaptive filtering based on dynamic wavelet transform, performing multi-scale processing on the signal in both the frequency and time-frequency domains, particularly eliminating artifacts through thresholding. Low-frequency transient artifacts at Hz.

[0058] 3. An independent component analysis algorithm was employed to perform blind source separation on the multi-lead signals, identifying and removing artifacts such as electrooculography (EOG) and electromyography (EMG), and reconstructing the signal using only the non-artifact components to obtain a pure VEP signal. Simultaneously, phase locking was achieved through an optical trigger and hardware interrupt mechanism to control the time error between stimulus and signal acquisition, limiting it to within a certain range. The accuracy of VEP latency measurement was maintained within milliseconds (ms), providing a high-purity and stable VEP waveform for subsequent quantitative feature extraction.

[0059] 4. The analysis and diagnosis module first calculates the visual pathway functional separation index, directly quantifying the relative delay and inhibition levels between pathways. Next, it performs VEP-CSF objective fitting on the M, P, and K channels, extracting objective parameters such as sensitivity peak, contrast threshold, and contrast half-saturation. Finally, the system employs an ensemble learning algorithm, using the visual pathway functional separation index and VEP-CSF fitting parameters as multi-dimensional feature vectors input to the model. Through this data-driven fusion of objective indicators, this system achieves quantitative grading of color vision abnormalities and assists in the etiological differentiation of complex visual lesions, improving the objectivity of clinical diagnosis and its ability to support decision-making. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the structure of a VEP diagnostic system for multidimensional decoupling modulation and quantitative assessment of pathway index provided in one embodiment of this application;

[0061] Figure 2 yes Figure 1 Flowchart of the hybrid adaptive filtering strategy for the signal acquisition module;

[0062] Figure 3 yes Figure 1 The flowchart shows the process of constructing diagnostic feature vectors and classifying ensemble learning models in the analysis and diagnosis module. Detailed Implementation

[0063] Please refer to Figure 1 This invention discloses a diagnostic system for the acquisition, processing, and quantitative assessment of visual evoked potentials (VEPs). The system integrates streamlined functional modules, aiming to achieve physiological isolation and quantitative functional assessment of the three major visual pathways: M, P, and K. The system includes a stimulus generation module, a real-time correction unit, a signal acquisition module, an embedded processing module, and an analysis and diagnostic module.

[0064] The stimulus generation module is used to generate stimuli that can be adjusted for color contrast, brightness, and spatial frequency. ) and time frequency ( The module generates a multidimensional parameter dynamically decoupled and modulated visual stimulus pattern and outputs a stimulus synchronization signal; the module supports stepped contrast modulation to generate the stimulus sequence required for fitting the objective contrast sensitivity function (VEP-CSF).

[0065] The visual stimulus pattern generated by the stimulus generation module is monitored and compensated for chromatic aberration in real time during the presentation process by the real-time correction unit, so as to maintain the brightness fluctuation of non-target stimulus axes within a predetermined threshold.

[0066] The signal acquisition module is used to acquire multi-lead VEP signals generated on the scalp of the test subject while viewing the visual stimulus pattern;

[0067] The embedded processing module adopts an FPGA / ARM architecture and receives the stimulus synchronization signal and the multi-lead VEP signal to achieve phase locking and artifact separation.

[0068] The analysis and diagnosis module is used to receive the VEP signal after it has been processed by the embedded processing module, extract the response characteristics of latency L and normalized amplitude A based on the signal, calculate the visual pathway functional separation index, and perform objective contrast sensitivity function (VEP-CSF) fitting for quantitative diagnostic assessment.

[0069] The stimulus generation module is responsible for generating highly controllable visual stimulus patterns, such as sinusoidal gratings. The key technology of this module lies in its ability to control color contrast, brightness, and spatial frequency (…). ) and time frequency ( The parameters across these four dimensions are dynamically decoupled and modulated. Simultaneously, the module supports stepped contrast modulation to acquire VEP amplitude response data at specific contrast levels, thereby generating the stimulus sequence required for fitting the objective contrast sensitivity function (VEP-CSF). The module also synchronously outputs a stimulus synchronization signal for time locking in subsequent embedded modules.

[0070] The stimulus generation module decouples and modulates (color contrast, brightness, ...) parameters through multidimensional parameter decoupling. , This method, employing stepped contrast modulation, enables the flexible generation of specific stimuli for particular visual pathways and the stimulation sequences required for contrast sensitivity testing. This precise and controllable stimulus generation overcomes the inflexibility of stimulus parameter modulation in existing technologies, laying the foundation for subsequent physiological isolation and quantitative assessment of the M, P, and K pathways.

[0071] To achieve physiological isolation of the M, P, and K pathways, stimulation patterns are... and Parameters are precisely defined. For example, for M-channel specific stimulation, a black-and-white sinusoidal grating is used, with time and frequency precisely defined. Hz (high frequency) and spatial frequency cpd (low frequency). The VEP response evoked by this stimulus is used to obtain the latency characteristics of the M pathway. For P-channel specific stimulation, an isoluminance red-green sinusoidal grating was used, with time and frequency defined. Hz (low frequency) and spatial frequency cpd (high frequency). This stimulation is used to obtain the latency characteristics of the P pathway. ) and amplitude characteristics ( K-channel specific stimulation uses an isoluminance blue-yellow sinusoidal grating with limited time and frequency. Hz, used to obtain the latency characteristics of the K pathway ( ) and amplitude characteristics ( In addition, the system also incorporates a brightness / color fusion channel stimulus, employing a color grating with brightness contrast to define high spatial frequencies, used to acquire the amplitude characteristics of the fusion channel. ), as a reference for calculating the pathway separation index.

[0072] By precisely defining the spatial frequency ( ) and time frequency ( ) parameters, such as the M channel Hz and cpd, and P and K channels Hz, achieving physiological isolation of the M, P, and K pathways. This specific stimulation mode ensures the characteristic features of the acquired VEP signal ( , , , , It accurately reflects the functional status of each independent pathway, overcoming the shortcomings of conventional methods in isolating and quantitatively evaluating the three major pathways.

[0073] The real-time correction unit incorporates a built-in spectroradiometer, which continuously monitors the display's output chromaticity in real time. The correction algorithm employs an iterative least squares method, adjusting the display drive based on the monitoring results to maintain brightness fluctuations on non-target stimulus axes within a range not exceeding [a certain threshold]. Within the predetermined threshold, the accuracy of the equal brightness condition is improved.

[0074] The real-time correction unit incorporates a spectroradiometer and employs an iterative least squares algorithm to monitor and compensate for chromatic aberration in real time. This mechanism can control the brightness fluctuation of non-target stimulus axes under stepped contrast modulation to no higher than [the specified value]. Within a very small range, the accuracy and consistency of isoluminance stimuli (such as P and K channel stimuli) are greatly guaranteed, effectively avoiding the interference of chromaticity drift on VEP signal acquisition and ensuring the reliability of pathway separation.

[0075] The signal acquisition module is used to acquire multi-lead VEP signals generated on the subject's scalp while viewing visual stimulus patterns. To ensure signal quality and localization, the electrode configuration adopts a standardized 10-20 international system extension layout, including at least visual cortical locations such as Oz, POz, O1, and O2. The low-noise amplifier in the acquisition module is designed with extremely high common-mode rejection ratio (CMRR). dB) and input impedance ( G This allows the weak VEP signal to have high fidelity.

[0076] The signal acquisition module employs a high-performance, low-noise amplifier and a standardized multi-lead electrode configuration, which effectively improves the signal-to-noise ratio and accuracy of weak VEP signals, and solves the problem of potentially low noise suppression capability and input impedance in existing acquisition devices.

[0077] The embedded processing module adopts an FPGA / ARM architecture, receives stimulus synchronization signals and multi-lead VEP signals, and performs multi-stage preprocessing functions. This module has the following auxiliary functions: a high-precision phase-locking function uses stimulus synchronization signals and multi-lead VEP signals as inputs, employs optical triggers and hardware interrupt mechanisms to time-stamp and synchronously superimpose the acquired VEP data, thus limiting the time error between the stimulus and VEP signals to within a certain range. The VEP signal is obtained by ms and synchronous superposition. The baseline drift correction function takes the multi-lead VEP signal as input, adopts real-time high-pass filtering combined with adaptive threshold algorithm, and obtains the VEP signal after baseline zero-point correction by continuously monitoring and adjusting the zero-point baseline of the signal.

[0078] The embedded processing module employs an FPGA / ARM architecture and an optical trigger / hardware interrupt mechanism to achieve high-precision phase locking, limiting the time error between the stimulus and the VEP signal to within a certain range. Within milliseconds, the accuracy of VEP signal superposition was ensured. Simultaneously, baseline zero-point correction was performed using real-time high-pass filtering combined with an adaptive threshold algorithm, eliminating the impact of low-frequency drift on VEP latency and amplitude extraction, thus guaranteeing the reliability of subsequent feature analysis.

[0079] Please refer to Figure 2 The filtering section of the signal acquisition module employs a hybrid adaptive filtering strategy to process the multi-lead VEP signal. This strategy first uses the multi-lead VEP signal as input, integrating a notch filter circuit or digital notch filter to remove power frequency interference and its higher-order harmonics in real time, obtaining a pre-filtered VEP signal. Secondly, using this pre-filtered VEP signal as input, a Dynamic Wavelet Transform (DWT) algorithm is used for decomposition. This algorithm selects specific wavelet basis functions to decompose the signal into different frequency band scales, obtaining the original wavelet coefficient set. Subsequently, low-frequency transient VEP noise suppression is performed, using the original wavelet coefficient set as input, targeting signals concentrated at a preset low-frequency scale (corresponding to...). The wavelet coefficients at a frequency of 0.1 Hz are thresholded to eliminate baseline drift and breathing artifacts at the preset low-frequency scale, while preserving the preset low-frequency transient VEP components to the maximum extent, resulting in a pure wavelet coefficient set. Finally, effective bandwidth reconstruction is performed. Using the pure wavelet coefficient set as input, the wavelet coefficients are reconstructed using the inverse wavelet transform method, based only on the wavelet coefficients within the preset effective physiological bandwidth (limited to a scale of 0.1 Hz to 100 Hz), to obtain a noise-removed VEP signal.

[0080] A hybrid adaptive filtering strategy, by integrating notch filtering and dynamic wavelet transform algorithms, effectively removes various types of noise, including power frequency interference, baseline drift, and breathing artifacts. In particular, it improves the removal of low-frequency scale noise (…). Thresholding of wavelet coefficients (Hz) was performed to suppress low-frequency noise and preserve low-frequency VEP components. By performing inverse wavelet transform reconstruction within an effective physiological bandwidth of 0.1Hz to 100Hz, a VEP signal with a high signal-to-noise ratio was obtained, which made up for the shortcomings of existing filtering methods that are singular and unable to effectively remove various noises and artifacts.

[0081] The artifact separation in the embedded processing module employs Independent Component Analysis (ICA) to process the multi-lead VEP signal. This step first uses the multi-lead VEP signal as input and performs blind source separation using the ICA algorithm, obtaining a set of independent IC vectors. Next, time-domain and frequency-domain feature analysis is performed on the IC vectors to identify and label artifact components that highly match the artifact characteristics of electrooculography (EOG), electromyography (EMG), and electrocardiography (ECG). Finally, during signal reconstruction, using the IC vectors as input, the weights corresponding to the artifact components are set to zero or removed during inverse ICA transformation, and signal reconstruction is performed using only non-artifact components to obtain a clean VEP signal, which serves as the final output of the embedded processing module.

[0082] Artifact separation employs Independent Component Analysis (ICA) algorithm, which decomposes multi-lead VEP signals into independent components through blind source separation, thereby identifying and removing physiological artifacts such as electrooculography (EOG), electromyography (EMG), and electrocardiography. This method can preserve the effective information of the VEP signal to the greatest extent, obtaining a high-purity VEP signal, providing reliable input for subsequent quantitative analysis.

[0083] The embedded processing module also includes a spatial filtering unit. This spatial filtering step first establishes a transfer matrix between cortical sources and scalp potentials based on the electrode location information of the multi-lead VEP signal and a preset head model. Then, using the multi-lead VEP signal and the transfer matrix as input, the conjugate gradient method (CGM) algorithm is employed to solve for the least-squares solution of the cortical source current density to estimate the source activity intensity of the primary visual cortex (V1 region). Finally, the source activity intensity of the V1 region is mapped to an enhanced VEP signal, resulting in an enhanced primary visual cortex-specific VEP response signal. This signal is used to more accurately extract the latency and amplitude characteristics of the M, P, and K pathways.

[0084] The spatial filtering unit estimates and maps the source activity intensity of the primary visual cortex (V1 area) by establishing the transfer matrix between cortical sources and scalp potentials and employing the conjugate gradient method, thereby obtaining an enhanced VEP signal. This method effectively improves the spatial specificity and signal-to-noise ratio of the signal, making the extraction of latency and amplitude features of the M, P, and K pathways more accurate.

[0085] The analysis and diagnosis module receives the VEP signal processed by the embedded processing module, extracts the response characteristics of latency L and normalized amplitude A based on the signal, and performs a final quantitative evaluation.

[0086] The module calculates the visual pathway functional separation index, which quantifies the relative functional status among the three major pathways (M, P, and K) and is one of the core quantitative indicators. This index includes at least three sets of quantitative indicators:

[0087] P / M channel relative latency delay index :

[0088]

[0089] This index is used to assess the conduction velocity delay of the P pathway relative to the M pathway.

[0090] P / M channel relative amplitude suppression index :

[0091]

[0092] This index is used to assess the degree of inhibition of P pathway function relative to fusion channel function.

[0093] K / P channel relative function ratio ( and ):

[0094] and

[0095] This index is used to compare the relative performance of the K and P pathways in terms of latency and amplitude.

[0096] The analysis and diagnosis module calculates the visual pathway functional separation index, including the P / M channel relative latency delay index (P / M channel). ), P / M channel relative amplitude suppression index ( ) and the relative function ratio of the K / P channel ( and This technology enables precise quantitative assessment of the functional status of the three major pathways: M, P, and K. These indices provide objective and quantifiable indicators, compensating for the lack of precise quantitative assessment indicators in existing technologies and providing crucial evidence for the diagnosis of visual system diseases.

[0097] The objective fitting of VEP-CSF in the analysis and diagnosis module is performed on the M, P, and K channels, and the process is as follows: First, data preparation is performed, using VEP amplitude data acquired under stepped contrast modulation as input, and the amplitude values ​​are standardized. The VEP amplitude data includes contrast levels and corresponding amplitude values. Then, nonlinear fitting is performed, using the standardized amplitude data as input, and a nonlinear regression model (e.g., the Naka-Rushton equation) is selected to perform least-squares fitting on the curve of VEP amplitude value changing with the logarithm of contrast, obtaining the fitted curve equation. Finally, based on the fitted curve equation, at least the following three parameters are extracted as diagnostic feature vectors: peak sensitivity ( ), that is, the maximum amplitude when the curve saturates; contrast threshold ( ), that is, the contrast ratio required for the curve to reach the preset minimum response; and the contrast half-saturation ( ), that is, the curve reaches The required contrast at half.

[0098] The VEP-CSF objective fitting achieves a quantitative assessment of the contrast sensitivity function (CSF) of each pathway (M, P, K) by performing nonlinear regression least squares fitting on standardized VEP amplitude data acquired under stepped contrast modulation. This is achieved by extracting... , and These three key parameters, serving as diagnostic feature vectors, can accurately describe the functional characteristics of each pathway, providing detailed quantitative evidence for distinguishing different pathological types of visual function impairment.

[0099] Please refer to Figure 3 The analysis and diagnosis module employs an ensemble learning algorithm for final diagnosis. This diagnostic step first constructs feature vectors based on the visual pathway functional separation index (VDI). ) and the VEP-CSF fitting parameters ( Using the diagnostic feature vector as input, a multidimensional diagnostic feature vector is constructed. Next, model classification is performed, using the diagnostic feature vector as input and employing an ensemble learning algorithm (such as gradient boosting trees) to classify and analyze the visual function status of the test subjects. Finally, the diagnostic results are output. Based on the model classification results, two types of diagnostic information are output: objective quantitative grading, classifying color vision abnormalities or optic nerve lesions into preset levels; and etiological discrimination, outputting the probability of the etiology, such as distinguishing between congenital red-green color vision abnormalities and acquired blue-yellow color vision abnormalities, providing accurate decision support for clinical practice.

[0100] The final diagnosis employs an ensemble learning algorithm, constructing a multidimensional diagnostic feature vector by combining the visual pathway functional separation index and VEP-CSF fitting parameters. This ensemble learning classification method can comprehensively analyze multidimensional quantitative indicators, outputting objective quantitative grading and etiological discrimination, effectively improving the accuracy and objectivity of the diagnosis. This, in turn, assists doctors in more accurately assessing the condition and developing personalized treatment plans.

[0101] The technical effects of the technical solution provided in this application include:

[0102] (1) Through multidimensional parameter decoupling modulation and real-time correction, the physiological isolation of the three major visual pathways M, P and K was achieved, and the VEP response characteristics specific to different pathways were obtained. The stimulation generation module dynamically decouples and modulates by limiting the time frequency and spatial frequency, increasing the activation degree of the target pathway and reducing the interference of non-target pathways.

[0103] (2) By combining adaptive filtering, artifact separation, and time locking, various artifacts in the VEP signal are effectively separated and suppressed, improving the signal-to-noise ratio and measurement stability of the VEP signal. This effect is achieved through multi-level embedded processing: First, the embedded processing module integrates notch filtering and adaptive filtering based on dynamic wavelet transform, and performs sub-scaling processing on the signal in the frequency domain and time-frequency domain, especially eliminating artifacts through thresholding. Low-frequency transient artifacts at Hz.

[0104] (3) Independent component analysis (ICA) was used to perform blind source separation on the multi-lead signals, identify and remove artifacts such as electrooculography (EOG) and electromyography (EMG), and reconstruct the signal using only the non-artifact components to obtain a pure VEP signal. Simultaneously, phase locking was achieved through an optical trigger and a hardware interrupt mechanism to control the time error between stimulus and signal acquisition, limiting it to within a certain range. The accuracy of VEP latency measurement was maintained within milliseconds (ms), providing a high-purity and stable VEP waveform for subsequent quantitative feature extraction.

[0105] (4) The analysis and diagnosis module first calculates the visual pathway functional separation index, directly quantifying the relative delay and inhibition levels between pathways. Then, it performs VEP-CSF objective fitting on the M, P, and K channels to extract objective parameters such as sensitivity peak, contrast threshold, and contrast half-saturation. Finally, the system uses an ensemble learning algorithm, inputting the visual pathway functional separation index and VEP-CSF fitting parameters as multidimensional feature vectors into the model. Through this data-driven fusion of objective indicators, the system achieves quantitative grading of color vision abnormalities and assists in the etiological discrimination of complex visual lesions, improving the objectivity of clinical diagnosis and the ability to support decision-making.

[0106] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Any other corresponding changes and modifications made based on the technical concept of this application should be included within the scope of protection of this application.

Claims

1. A VEP diagnostic system based on multidimensional decoupling modulation and quantitative assessment of pathway indices, characterized in that, include: The stimulus generation module is used to generate stimuli that can be adjusted for color contrast, brightness, and spatial frequency. and time frequency The system generates visual stimulus patterns through multidimensional parameter dynamic decoupling modulation and outputs stimulus synchronization signals. The stimulus generation module supports stepped contrast modulation to generate stimulus sequences required for fitting an objective contrast sensitivity function. The stimulus generation module can generate four specific stimulus patterns, which are determined by parameters... and The limitations of these pathways enable physiological isolation of the M, P, and K pathways, and are used to induce and obtain the specific VEP response characteristics required for the analysis and diagnostic modules. a) M-channel specific stimulation: a black and white sinusoidal grating, using time-frequency... Limited to Hz, spatial frequency Limited to cpd; The M-channel specific stimulation is used to obtain the latency characteristics of the M pathway. ; b) P-channel specific stimulation: an isoluminous red-green sinusoidal grating, using time-frequency... Limited to Hz, spatial frequency Limited to cpd; The P-channel specific stimulation is used to obtain the latency and amplitude characteristics of the P pathway; c) K-channel specific stimulation: An isoluminous blue-yellow sinusoidal grating, using time-frequency... Limited to Hz; The K-channel specific stimulation is used to obtain the latency and amplitude characteristics of the K pathway; d) Luminance / color fusion channel stimulation: A color grating with luminance contrast is used with a high spatial frequency; the luminance / color fusion channel stimulation is used to obtain the amplitude characteristics of the fusion channel as a benchmark for the pathway separation index; A real-time correction unit is used to monitor and compensate for chromatic aberration in real time during the presentation of the visual stimulus pattern, so as to maintain the brightness fluctuation of the non-target stimulus axis within a predetermined threshold. The signal acquisition module is used to acquire multi-lead VEP signals generated on the scalp of the test subject while viewing the visual stimulus pattern; The embedded processing module, using an FPGA / ARM architecture, receives the stimulus synchronization signal and the multi-lead VEP signal to achieve phase locking and artifact separation. The analysis and diagnosis module receives the VEP signal processed by the embedded processing module and extracts the latency based on the signal. Based on the response characteristics of the normalized amplitude A, the visual pathway functional dissociation index is calculated, which includes at least the following three sets of quantitative indicators: a) P / M channel relative latency delay index : ; b) P / M channel relative amplitude suppression index : ; c) K / P channel relative function ratio, wherein the K / P channel relative function ratio includes and : ； ， Where L is the VEP latency and A is the VEP normalized amplitude; The analysis and diagnosis module performs objective contrast sensitivity function fitting on the M, P, and K channels and extracts VEP-CSF fitting parameters. It then uses an ensemble learning algorithm to input the visual pathway function separation index and VEP-CSF fitting parameters as multidimensional feature vectors into the model.

2. The VEP diagnostic system with multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The real-time correction unit incorporates a spectroradiometer, and the correction algorithm employs iterative least squares to control the brightness fluctuation of the non-target stimulus axis to no higher than [a certain value] under different step contrast modulations. .

3. The VEP diagnostic system based on multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The filtering section of the signal acquisition module employs a hybrid adaptive filtering strategy to process the acquired multi-lead VEP signal. The hybrid adaptive filtering strategy includes the following steps: a) Using the multi-lead VEP signal as input, an integrated notch filter circuit or a digital notch filter is used to remove power frequency interference and its higher-order harmonics in real time, thereby obtaining a pre-filtered VEP signal. b) Using the pre-filtered VEP signal as input, the dynamic wavelet transform algorithm is used for decomposition. The dynamic wavelet transform algorithm decomposes the signal into different frequency band scales by selecting specific wavelet basis functions to obtain the original wavelet coefficient set. c) Low-frequency transient VEP noise suppression: Using the original wavelet coefficient set as input, the wavelet coefficients concentrated on the preset low-frequency scale are thresholded to eliminate baseline drift and breathing artifacts on the preset low-frequency scale, while preserving the preset low-frequency transient VEP components to the maximum extent, thus obtaining a pure wavelet coefficient set. d) Effective bandwidth reconstruction: Using the pure wavelet coefficient set as input, the wavelet coefficients are reconstructed based solely on the preset effective physiological bandwidth using the inverse wavelet transform method to obtain the noise-removed VEP signal.

4. The VEP diagnostic system based on multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The artifact separation in the embedded processing module uses an independent component analysis algorithm to process the multi-lead VEP signal. The artifact separation steps include: a) Using the multi-lead VEP signal as input, blind source separation is performed using the independent component analysis algorithm to obtain a set of mutually independent IC vectors; b) Perform time-domain and frequency-domain feature analysis on the IC vector to identify and label artifact components that highly match the artifact features of electrooculography, electromyography, and electrocardiography; c) Using the IC vector as input, during the inverse ICA transformation, the weights corresponding to the artifact components are set to zero or removed, and the signal is reconstructed using only the non-artifact components to obtain the VEP pure signal, which is the final output of the embedded processing module.

5. The VEP diagnostic system based on multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The embedded processing module also has the following functions: a) Using the stimulus synchronization signal and the multi-lead VEP signal as inputs, an optical trigger and hardware interrupt mechanism are employed to time-stamp and synchronously superimpose the acquired VEP data, thereby limiting the time error between the stimulus and the VEP signal to within a certain range. ms, to obtain the synchronously superimposed VEP signal; b) Using the multi-lead VEP signal as input, the low-frequency VEP signal is processed by real-time high-pass filtering combined with an adaptive threshold algorithm. By continuously monitoring and adjusting the zero-point baseline of the signal, the VEP signal after baseline zero-point correction is obtained.

6. The VEP diagnostic system based on multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The embedded processing module further includes a spatial filtering unit, which processes the multi-lead VEP signal. The spatial filtering steps include: a) Based on the electrode position information of the multi-lead VEP signal and the preset head model, establish the transfer matrix between the cortical source and the scalp potential; b) Using the multi-lead VEP signal and the transfer matrix as input, the conjugate gradient method is used to solve the least squares solution of the cortical source current density in order to estimate the source activity intensity of the primary visual cortex. c) The source activity intensity of the primary visual cortex is mapped to an enhanced VEP signal to obtain an enhanced primary visual cortex-specific VEP response signal, which is used to extract the latency and amplitude characteristics of the M, P, and K pathways more accurately.

7. The VEP diagnostic system for multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The fitting steps include: a) The VEP amplitude data acquired under the stepped contrast modulation is used as input, and the amplitude value is standardized. The VEP amplitude data includes the contrast level and the corresponding amplitude value. b) Using standardized amplitude data as input, a nonlinear regression model is selected to perform least squares fitting on the curve of VEP amplitude value changing with the logarithm of contrast, and the fitted curve equation is obtained. c) Based on the fitted curve equation, extract at least the following three parameters as diagnostic feature vectors: i) Sensitivity peak : The maximum amplitude when the curve saturates; ii) Contrast threshold The contrast required for the curve to reach the preset minimum response; iii) Contrast half-saturation The curve reached The required contrast at half.

8. The VEP diagnostic system for multidimensional decoupling modulation and quantitative assessment of pathway index according to claim 1, characterized in that, The analysis and diagnosis module uses an ensemble learning algorithm for final diagnosis, and the final diagnosis steps include: a) Construct a multidimensional diagnostic feature vector using the visual pathway functional separation index and the VEP-CSF fitting parameters as input; b) Model classification: Using the multidimensional diagnostic feature vector as input, an ensemble learning algorithm is used to classify and analyze the visual function status of the test subjects; c) Output diagnostic results: Based on the model classification results, output the following two types of diagnostic information: i) Objective quantitative grading: color vision abnormalities or optic nerve lesions are classified into pre-defined levels; ii) Etiological identification: Output the probability of the etiology.

Images