Optical performance testing method of full-spectrum optical imaging endoscope

Abstract

This invention belongs to the field of optical testing and discloses a method for testing the optical performance of a full-spectrum optical imaging endoscope. The method acquires the original image and spectral response of the endoscope under multi-dimensional standard light sources, constructs an optical performance benchmark feature library including color reproduction error and signal-to-noise ratio, identifies the color separation critical points of the visible and near-infrared channels based on this feature library, quantifies the pseudo-color interference coefficient, and constructs a test spectral configuration priority matrix. It combines key band identifiers of the surgical scenario to screen and correct the test configuration, generating a simulation-predicted and optimized detection scheme. In automated testing, it monitors image sharpness and color gamut integrity in real time, and automatically backtracks and adjusts lens shading correction and white balance parameters when the indicators fail to meet the standards. This invention significantly improves the accuracy and efficiency of testing through quantitative simulation and closed-loop control, providing precise and reliable technical support for the research and development and quality control of high-performance endoscopes.

Description

Technical Field

[0001] This invention relates to the field of optical testing technology, and more specifically, to a method for testing the optical performance of a full-spectrum optical imaging endoscope. Background Technology

[0002] Existing performance testing methods for full-spectrum optical imaging endoscopes face multi-dimensional technical bottlenecks in comprehensive evaluation, particularly in the performance trade-offs and channel crosstalk issues encountered when processing heterogeneous imaging information during the fusion evaluation of multimodal optical data. With the integration of visible light and near-infrared dual-modal imaging functions into endoscopes, the optical performance indicators (such as color reproduction and signal-to-noise ratio), optimal operating conditions (such as color temperature and illuminance), and noise characteristics of the two modes differ significantly. Traditional independent static testing strategies cannot adapt to the variable lighting environment in clinical surgery.

[0003] In complex testing scenarios, when performance metrics of one mode (such as color reproduction) are prioritized for optimization, the system often ignores its potential impact on another mode (such as fluorescence detection sensitivity). To ensure the pass rate of basic tests, existing methods are forced to evaluate the performance of the visible light channel and the near-infrared channel in isolation, artificially ignoring the mutual influence between the two channels when they actually work together. This simplification strategy results in a serious loss of system-level performance characteristics, especially key performance defects such as optical crosstalk between the visible light and near-infrared channels and false color interference during multispectral fusion imaging, which cannot be effectively quantified and evaluated. Clinicians and R&D engineers often point out that although individual metrics may seem excellent, the color fidelity and lesion identification of images in fusion mode do not meet expectations. The test results appear accurate but lack effective guidance for comprehensive clinical applications.

[0004] In view of this, the present invention proposes an optical performance testing method for a full-spectrum optical imaging endoscope to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: A method for testing the optical performance of a full-spectrum optical imaging endoscope, comprising: Step S1: Obtain the raw image data of the endoscope under a standard light source and the spectral response curve of the sensor. The raw image data includes image data collected under different color temperatures and illuminance conditions in the visible light band and near-infrared light band. Extract the color reproduction error and signal-to-noise ratio features under different color temperatures and illuminance conditions, and construct an optical performance benchmark feature library based on the color reproduction error and signal-to-noise ratio features. Step S2: Identify the critical point of color separation between the visible light band and the near-infrared light band based on the optical performance benchmark feature library, calculate the pseudo-color interference coefficient during multispectral fusion imaging, and construct a test spectrum configuration priority matrix according to the color fidelity requirements of medical images. Step S3: Generate an initial detection scheme by combining key band identifiers of the surgical scene, obtain tissue recognition prediction indicators through image enhancement and fusion algorithm simulation, and identify test configuration nodes with optical crosstalk in the initial detection scheme; Step S4: Adjust the filter cutoff parameters or near-infrared color mapping weights for test configuration nodes with optical crosstalk to generate a corrected detection scheme; Step S5: During the automated testing process based on the modified detection scheme, the image sharpness index and wide color gamut coverage integrity of the acquired image data are monitored in real time. If the image sharpness index or wide color gamut coverage integrity is less than the corresponding standard threshold, the LSC lens shading correction parameters and AWB white balance parameters are automatically adjusted back, and the automated test is re-executed based on the adjusted LSC lens shading correction parameters and AWB white balance parameters to obtain the optical performance test results.

[0006] Furthermore, the process of extracting color reproduction error and signal-to-noise ratio features under different color temperatures and illuminance conditions includes: Near-infrared light and visible light standard light sources were injected into the incident section of the endoscope beam under different color temperatures and illuminance conditions. Optical molecular images generated after the near-infrared light was excited by the ICG targeted fluorescent group and reflected light images generated by the visible light standard light source mapped onto the standard color card were collected. Near-infrared channel signal-to-noise ratio features are extracted from the optical molecular image, and color reproduction error features are extracted from the reflected light image.

[0007] Furthermore, the process of constructing the optical performance benchmark feature library includes: A two-dimensional feature matrix with color temperature and illuminance as index dimensions is established, and the color reproduction error features and signal-to-noise ratio features extracted under different color temperature-illuminance combinations are classified and stored in the corresponding matrix nodes. Calculate the color difference statistics for the color reproduction error characteristics of each matrix node, and extract the mean signal-to-noise ratio and background noise power spectral density for the signal-to-noise ratio characteristics of each matrix node. The optimal working range for visible light and near-infrared light is selected based on the preset color difference threshold and the preset signal-to-noise ratio threshold, and the overlapping area of ​​the two is determined as the optimal working range for the whole spectrum. The environmental parameters, performance indicators, and sensor spectral response curves corresponding to the optimal working range of the full spectrum are associated and stored to form an optical performance benchmark feature library.

[0008] Furthermore, the process of identifying the color separation threshold includes: The beam splitter in the endoscope separates the light signal transmitted from the eyepiece into a visible light component of 400nm to 700nm and a near-infrared light component of 760nm to 1000nm. A cutoff filter is applied to the near-infrared light channel and its cutoff wavelength is stepped within the range of 835nm to 855nm. At each cutoff wavelength step point, a color image output by a high-definition color CCD and a black and white image output by a high-sensitivity black and white CCD are acquired respectively. Calculate the proportion of near-infrared leakage brightness in the color image and the proportion of visible light residue brightness in the black and white image at each step point; determine the cutoff wavelength corresponding to the product of the two proportions being less than the preset separation threshold as the color separation critical point.

[0009] Furthermore, the calculation of the pseudo-color interference coefficient includes: A 2D fused image is obtained by image registration and fusion of the color image and the black and white image, and the set of pixels of fluorescently marked regions is extracted from the 2D fused image. Calculate the chromaticity offset of each pixel in the pixel set of the fluorescently marked region, and determine the weighted average of the chromaticity offset as the pseudo-color interference coefficient.

[0010] Furthermore, the process of constructing the test spectrum configuration priority matrix includes: A three-dimensional configuration space is constructed using visible light illuminance level, near-infrared excitation light power level, and cutoff wavelength level of cutoff filter as dimensions; a medical image color fidelity score is calculated for each configuration node in the three-dimensional configuration space; Based on the color fidelity score of the medical image, the configuration nodes in the three-dimensional configuration space are arranged in descending order to generate the test spectrum configuration priority matrix, and the configuration nodes with a preset proportion before the arrangement are marked as priority test configurations.

[0011] Furthermore, the process of generating the initial detection scheme includes: Read the key band identifiers corresponding to the target surgical type from the pre-stored surgical scenario library. The key band identifiers include the peak wavelength of the reflectance spectrum and the valley wavelength of the absorption spectrum corresponding to the tissue types that need to be distinguished in the target surgical type. Each wavelength in the key band identifier is compared with the priority test configuration in the test spectrum configuration priority matrix to verify band coverage, and a subset of priority test configurations that have an effective response to each wavelength in the key band identifier is selected; the initial detection scheme is formed based on the subset of priority test configurations.

[0012] Furthermore, the process of obtaining organizational identification predictive metrics includes: The visible light image data and near-infrared fluorescence image data corresponding to each configured node in the initial detection scheme are input into the pre-constructed image registration simulation unit. The image registration simulation unit performs registration processing on the visible light image data and near-infrared fluorescence image data based on the stereo matching algorithm to generate simulated registration image pairs. The simulated registered image is input into the three-dimensional image fusion simulation module to generate a simulated 3D fused image. The grayscale contrast, edge sharpness, and fluorescence uniformity of the target lesion area relative to the normal tissue area in the simulated 3D fused image are extracted. The weighted comprehensive value of the three components is determined as the tissue recognition prediction index.

[0013] Furthermore, the process of identifying test configuration nodes with optical crosstalk in the initial detection scheme includes: For each configuration node in the initial detection scheme, the energy penetration rate of the near-infrared light channel to the visible light band and the energy penetration rate of the visible light channel to the near-infrared band are calculated respectively. Configuration nodes whose energy penetration rate of the near-infrared light channel to the visible light band exceeds a first preset penetration rate threshold or whose energy penetration rate of the visible light channel to the near-infrared band exceeds a second preset penetration rate threshold are marked as test configuration nodes with optical crosstalk.

[0014] Furthermore, the automatic retrospective adjustment of LSC lens shading correction parameters and AWB white balance parameters includes: During the automated testing process, high-definition color CCD output color images and high-sensitivity black-and-white CCD output black-and-white images are acquired respectively. The edge sharpness is extracted by performing Laplacian gradient operation on the color image, and the contrast of the fluorescence signal is extracted by performing grayscale variance analysis on the black and white image. The weighted average of the edge sharpness and the fluorescence signal contrast is used as the image sharpness index. The color image is mapped from the RGB color space to the Lab color space, and the ratio of the actual gamut volume covered in the Lab color space to the standard color gamut volume of the medical image is calculated as the wide color gamut coverage completeness. When the image sharpness index is less than the corresponding sharpness standard threshold, the edge brightness distribution curve of the currently acquired image is extracted, the relative brightness attenuation ratio between the image center region and the edge region is calculated, and the corresponding lens shadow correction compensation coefficient is matched in the preset LSC lens shadow correction parameter lookup table according to the relative brightness attenuation ratio, and the lens shadow correction compensation coefficient is updated to the lens shadow correction unit in the endoscope. When the wide color gamut coverage integrity is less than the corresponding coverage standard threshold, the mean value of each tone channel of the current color image is extracted, the color temperature deviation between the mean value of each tone channel and the standard white point reference value is calculated, the corresponding gain compensation value is matched in the preset AWB white balance parameter mapping table according to the color temperature deviation, and the gain compensation value is updated to the white balance adjustment unit until the image sharpness index and the wide color gamut coverage integrity both meet the corresponding standard threshold.

[0015] After adjusting the LSC lens shading correction parameters and AWB white balance parameters, the corresponding complete test process is re-executed based on the adjusted parameters. The image sharpness index and wide color gamut coverage integrity of the newly acquired image are monitored in real time. If any indicator is still less than the corresponding standard threshold, the parameters are iteratively adjusted. If the indicator still fails to meet the standard after three consecutive iterations, the current test configuration node is recorded as an unqualified node and marked as an anomaly. When all indicators meet the corresponding standard threshold, the optical performance test results, including the final LSC parameters, AWB parameters, image sharpness index, wide color gamut coverage integrity, 2D fused image quality score, and 3D stereoscopic image depth accuracy, are output.

[0016] The technical effects and advantages of the optical performance testing method for a full-spectrum optical imaging endoscope of the present invention are as follows: This invention scientifically integrates multimodal optical data to ensure that information such as color reproduction in visible light, signal-to-noise ratio in near-infrared light, optical crosstalk between channels, and pseudo-color interference in fused images can be accurately quantified and collaboratively evaluated. This significantly improves the accuracy and reliability of optical performance testing and provides precise and reliable technical support for the research and development and quality control of high-performance endoscopes.

[0017] Furthermore, in practical applications, even when faced with varying color temperature and illuminance environments in clinical settings, this method can still maintain a comprehensive assessment capability that meets both the requirements of color fidelity and near-infrared detection sensitivity. It eliminates the need for isolated static assessments or reliance on subjective judgment, making the final testing scheme more scientific and efficient.

[0018] In particular, the construction of the optical performance benchmark feature library and the automated closed-loop backtracking adjustment mechanism have broken through the limitations of traditional testing methods that cannot quantify the influence between channels and cannot dynamically compensate for fluctuations in the testing environment, effectively eliminating unreasonable phenomena such as distorted test results and disconnection from actual clinical manifestations. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a method for testing the optical performance of a full-spectrum optical imaging endoscope according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 Please see Figure 1 As shown, this embodiment provides a method for testing the optical performance of a full-spectrum optical imaging endoscope, including: Step S1: Obtain the raw image data of the endoscope under standard light source and the spectral response curve of the sensor. The raw image data includes image data collected under different color temperatures and illuminance conditions in the visible light band and near-infrared light band. Extract the color reproduction error and signal-to-noise ratio features under different color temperatures and illuminance conditions, and construct an optical performance benchmark feature library based on the color reproduction error and signal-to-noise ratio features.

[0022] Among them, raw image data refers to the set of raw light signal records collected by the endoscope through the image sensor device under standard light source illumination conditions without post-processing. It includes image data of visible light and near-infrared light corresponding bands collected under different color temperatures and illuminance conditions, which can truly reflect the imaging response characteristics of the endoscope in diverse lighting environments; the sensor spectral response curve refers to the quantum efficiency distribution curve of the image sensor for light signals of different wavelengths, which is a quantitative characterization of the sensor's ability to convert light energy of various wavelengths across the entire spectrum.

[0023] In embodiments of the present invention, the process of extracting color reproduction error and signal-to-noise ratio features under different color temperatures and illuminance conditions includes: Near-infrared and visible light standard light sources were injected into the incident section of the endoscope's beam guide under different color temperatures and illuminance conditions. Optical molecular images generated after the near-infrared light was excited by the ICG-targeted fluorescent group, and reflected light images generated by the visible light standard light source mapping onto the standard color chart were collected. Specifically, the process first configured standard test light sources covering multiple color temperature levels from low color temperature (2800K) to high color temperature (6500K) and multiple illuminance levels from low illuminance (500 lux) to high illuminance (10000 lux), forming a combined color temperature and illuminance test matrix. Under each color temperature and illuminance combination condition, two types of standard light sources were injected into the incident end of the endoscope's beam guide: the first type was near-infrared standard excitation light with a wavelength of 780±10nm, used to excite the targeted fluorescent group to produce a fluorescence response; the second type was a visible light standard light source covering 400nm to 700nm, used to illuminate the standard color chart to generate standard reflected light signals.

[0024] After injecting near-infrared standard excitation light, near-infrared fluorescence images (845±10nm) generated by ICG-targeted fluorescent groups are acquired, i.e., optical molecular images. These images are captured by a high-sensitivity black-and-white CCD in the near-infrared channel of the endoscope, recording the spatial distribution of fluorescence emission intensity. Simultaneously, a visible light standard light source is injected to illuminate a standard color chart, and reflected light images generated by each color block of the standard color chart are acquired. These reflected light images are captured by a high-definition color CCD in the visible light channel of the endoscope, recording the colorimetric response of the standard color chart under the current color temperature and illuminance conditions. The standard color chart uses a 24-color Macbeth color chart conforming to international standards, covering typical medical image reference color blocks including skin color, neutral grayscale, and saturated colors. Targeted fluorescent groups refer to biomarkers that can emit fluorescence in the 845±10nm band under near-infrared excitation light irradiation, used to simulate the fluorescence-guided lesion marking state in surgical scenarios.

[0025] Near-infrared channel signal-to-noise ratio (SNR) features are extracted from optical molecular images, and color reproduction error features are extracted from reflected light images. Specifically, the extraction process of near-infrared channel SNR features is based on the pixel grayscale distribution of fluorescence response images. First, fluorescently labeled regions and background regions are delineated in the fluorescence response images. The fluorescently labeled regions are selected from the connected regions with the highest fluorescence intensity, and the background regions are selected from the set of non-fluorescent pixels far from the fluorescently labeled regions. The mean of the grayscale values ​​of all pixels in the fluorescently labeled regions is calculated as the signal intensity, and the standard deviation of the grayscale values ​​of all pixels in the background regions is calculated as the noise intensity. The ratio of signal intensity to noise intensity is determined as the near-infrared channel SNR. The above calculation process is performed on all fluorescence response images acquired under different color temperature and illuminance combinations to form a near-infrared channel SNR feature sequence covering the complete test matrix.

[0026] The extraction process of color reproduction error features is based on the comparative analysis of reflected light images and standard chromaticity reference values. First, the color block regions of the reflected light images are located, and the pixel set corresponding to each color block in the standard color chart is extracted. The mean chromaticity coordinates in the Lab color space are calculated for the pixel set of each color block, which is taken as the measured chromaticity value of the color block under the current test conditions. The standard chromaticity value of the corresponding color block is read from the standard chromaticity database, and the color difference between the measured chromaticity value and the standard chromaticity value is calculated. The color difference is calculated using the CIEDE2000 color difference formula to accurately reflect the color deviation perceived by the human eye. The arithmetic mean of the color differences of the 24 color blocks is taken to obtain the average color reproduction error under the current test conditions. The above calculation process is performed on all reflected light images acquired under different color temperature and illuminance combinations to form a color reproduction error feature sequence covering the complete test matrix.

[0027] In embodiments of the present invention, the process of constructing the optical performance benchmark feature library includes: A two-dimensional feature matrix with color temperature and illuminance as index dimensions is established. Color reproduction error features and signal-to-noise ratio features extracted under different color temperature-illuminance combinations are classified and stored in the corresponding matrix nodes. Specifically, the row dimension of the two-dimensional feature matrix corresponds to the color temperature discretization node, and the column dimension corresponds to the illuminance discretization node. Each node in the matrix stores a complete feature record, which includes multiple feature fields such as the mean color difference, the maximum color difference, the standard deviation of the color difference, the mean near-infrared signal-to-noise ratio, and the background noise power spectral density. During the matrix creation process, the matrix dimensions are first determined according to the preset color temperature sequence and illuminance sequence. Row indices are assigned to each node in the color temperature sequence, and column indices are assigned to each node in the illuminance sequence, establishing a bidirectional mapping relationship between parameter values ​​and matrix coordinates.

[0028] For each matrix node, color difference statistics are calculated based on the color reproduction error characteristics. For each matrix node, the mean signal-to-noise ratio (SNR) and background noise power spectral density are extracted. Specifically, during the color difference statistics calculation, the arithmetic mean, maximum color difference, and standard deviation of the color difference values ​​for the 24 color blocks stored in each color temperature-illuminance combination node are calculated. These three statistical values ​​together constitute the color difference statistics vector for that node. They are used to reflect the overall color reproduction level of the endoscope under the working conditions, reveal the degree of deviation of the most difficult-to-reproduce colors, and characterize the dispersion of color differences between different colors. The combination of these three can comprehensively describe the statistical characteristics of color reproduction error. For each matrix node, the mean SNR and background noise power spectral density are extracted. The mean SNR is obtained by time-domain averaging of the SNR values ​​of multiple frames of optical molecular images under the same color temperature-illuminance conditions. The background noise power spectral density is obtained by averaging and normalizing the spectrum of multiple noise regions. These two indicators together describe the signal quality of the near-infrared channel.

[0029] The optimal working range for visible light and near-infrared light is selected based on preset color difference thresholds and preset signal-to-noise ratio thresholds, and the overlapping area of ​​the two is determined as the optimal working range for the entire spectrum. The preset color difference threshold is set with reference to the color reproduction standard of medical imaging. Typically, color temperature-illuminance combination nodes with an arithmetic mean of less than 3.0 and a maximum color difference of less than 6.0 are defined as the optimal working range for visible light. Within this range, the color reproduction error of the endoscope is within the clinically acceptable range. The preset signal-to-noise ratio threshold is set with reference to the minimum requirements for lesion identification in near-infrared fluorescence imaging. Typically, color temperature-illuminance combination nodes with a signal-to-noise ratio mean of greater than 20dB and a background noise power spectral density lower than the preset noise upper limit are defined as the optimal working range for near-infrared imaging. The intersection of the set of color temperature-illuminance nodes corresponding to the optimal working range for visible light and the set of nodes corresponding to the optimal working range for near-infrared imaging is taken to obtain the full-spectrum synergistic optimal working range that simultaneously meets the dual requirements of visible light color fidelity and near-infrared detection sensitivity. If the intersection of the two working ranges is empty, the color difference threshold and the signal-to-noise ratio threshold are relaxed by one-tenth each and recalculated until a non-empty synergistic optimal working range is obtained.

[0030] The environmental parameters, performance indicators, and sensor spectral response curves corresponding to the full-spectrum optimal working range are associated and stored to form an optical performance benchmark feature library. Specifically, during the association and storage process, each color temperature-illuminance node in the full-spectrum optimal working range is used as a recording unit. The color temperature value, illuminance value, color difference statistical value vector, mean signal-to-noise ratio, background noise power spectral density, and the corresponding sensor spectral response curve sampling point sequence of that node are integrated into a complete benchmark record. The resulting optical performance benchmark feature library is a multi-indexed structured dataset that can comprehensively characterize the full-spectrum optical performance characteristics of endoscopes under different working conditions, providing accurate benchmark reference data for priority evaluation and parameter optimization of subsequent test scheme configurations.

[0031] Step S2: Identify the critical point of color separation between the visible light band and the near-infrared light band based on the optical performance benchmark feature library, calculate the pseudo-color interference coefficient during multispectral fusion imaging, and construct a test spectrum configuration priority matrix according to the color fidelity requirements of medical images.

[0032] In an embodiment of the present invention, the process of identifying the color separation critical point includes: The beam splitter in the endoscope separates the light signal transmitted from the eyepiece into a visible light component of 400nm to 700nm and a near-infrared light component of 760nm to 1000nm. Specifically, the beam splitter driving process first sends a beam splitter mode activation command to the beam splitter control interface of the endoscope, causing the beam splitter to enter the standard working state. Based on the principle of wavelength selective reflection, the beam splitter separates the mixed light signal transmitted from the eyepiece into two outputs according to the wavelength range: the first output is the visible light component from 400nm to 700nm, which is transmitted to the high-definition color CCD; the second output is the near-infrared light component from 760nm to 1000nm, which is transmitted to the high-sensitivity monochrome CCD. A wavelength-adjustable cutoff filter is installed in the near-infrared light channel. The cutoff wavelength adjustment range of the cutoff filter covers 845±10nm, and several cutoff wavelength step points are preset for adjustment accuracy. The function of the cutoff filter is to introduce a wavelength selective cutoff boundary in the near-infrared light channel, which is used to precisely control the passage and blocking state of different wavelength signals in the near-infrared channel, providing a continuously adjustable parameter basis for the subsequent identification of color separation critical points.

[0033] A cutoff filter is applied to the near-infrared light channel and its cutoff wavelength is stepped within the range of 835nm to 855nm. At each cutoff wavelength step point, a color image output by a high-definition color CCD and a black and white image output by a high-sensitivity black and white CCD are acquired respectively. Specifically, the step-scan acquisition process starts from the 835nm cutoff wavelength starting point, and adjusts the cutoff wavelength parameters of the cutoff filter sequentially at 1nm step intervals. At each cutoff wavelength step point, the target cutoff wavelength parameters are first written to the cutoff filter controller, and the cutoff filter is waited for to complete wavelength switching and enter a stable state. The stabilization waiting time is set to 500ms to eliminate the impact of transient response during the cutoff wavelength switching process on the quality of the acquired image. After the stabilization waiting is completed, the high-definition color CCD and the high-sensitivity monochrome CCD are simultaneously triggered to acquire images. The high-definition color CCD outputs a color image under the current cutoff wavelength conditions and records the chromaticity response state of the visible light channel. The high-sensitivity monochrome CCD outputs a monochrome image under the current cutoff wavelength conditions and records the intensity response state of the near-infrared light channel. The color image and monochrome image acquired at the same step point are combined into an image pair, and stored in the step acquisition result set with the current cutoff wavelength value as the key. The above process is repeated until the image acquisition at the 850nm cutoff wavelength termination point is completed, forming a complete step acquisition result set covering all cutoff wavelength step points.

[0034] The near-infrared leakage luminance percentage in the color image and the visible light residue luminance percentage in the black-and-white image are calculated at each step point. Specifically, the calculation of the near-infrared leakage luminance percentage is based on pixel luminance analysis of the color image. First, the color image undergoes color space conversion, changing the RGB color space to the CIE color space, and the luminance component is extracted. Within the luminance component of the color image, the pixel-by-pixel luminance difference between the corresponding color images is obtained by comparing the luminance differences with and without a near-infrared cutoff filter (i.e., subtracting the luminance value of the color image with the filter from the luminance value of the color image without the filter). This difference is used as the luminance contribution value at each pixel. The arithmetic mean of the luminance contribution values ​​at each pixel is then calculated to obtain the additional luminance contribution value caused by near-infrared light leakage at each step point. The ratio of the additional luminance contribution value to the average total luminance of the color image is determined as the near-infrared leakage luminance percentage. A higher near-infrared leakage luminance percentage indicates a more severe interference of near-infrared light with the visible light channel.

[0035] The calculation of the visible light residual brightness ratio is based on pixel grayscale analysis of black and white images. First, the near-infrared response region is delineated in the black and white image. By comparing the grayscale difference between the black and white image under the current cutoff wavelength condition and the black and white image when the near-infrared signal is completely cut off, the background brightness contribution value caused by visible light residual is calculated. The ratio of the background brightness contribution value to the average grayscale value of the black and white image is determined as the visible light residual brightness ratio. The higher the visible light residual brightness ratio, the more severe the interference of visible light on the near-infrared channel. The above two ratio calculations are performed on each cutoff wavelength step point in the step acquisition result set to form a near-infrared leakage brightness ratio sequence and a visible light residual brightness ratio sequence covering all step points. It should be noted that the calculation process of the background brightness contribution value is similar to the calculation method of the additional brightness contribution value. By subtracting the grayscale value of the black and white image when the near-infrared signal is completely cut off from the grayscale value of the black and white image under the current cutoff wavelength condition, a grayscale difference map caused by visible light residual is obtained. The arithmetic mean of the differences of all pixels in the grayscale difference map is used to obtain the background brightness contribution value.

[0036] The cutoff wavelength corresponding to the point where the product of the two proportions is less than a preset separation threshold is determined as the color separation critical point. The critical point determination process first iterates through all cutoff wavelength step points, extracting the corresponding near-infrared leakage brightness proportion and visible light residual brightness proportion for each step point, calculating their product to obtain the bidirectional crosstalk comprehensive index at that step point. The bidirectional crosstalk comprehensive index, by coupling the channel leakage effects in two directions, accurately reflects the overall separation performance between the two channels under a specific cutoff wavelength condition. The bidirectional crosstalk comprehensive index at each step point is compared with the preset separation threshold, which is pre-set by those skilled in the art based on medical image color fidelity standards. In the step scan sequence... For the first time, the cutoff wavelength value corresponding to the cutoff wavelength step point where the bidirectional crosstalk comprehensive index is lower than the preset separation threshold is found, which is determined as the color separation critical point between the visible light band and the near-infrared band. The color separation critical point is a key benchmark parameter for subsequent construction of the test spectrum configuration priority matrix and adjustment of filter cutoff parameters, and is recorded in the optical performance benchmark feature library. This step can accurately locate the boundary wavelength of effective isolation of the dual-channel spectrum in a fine step scanning method, providing an accurate numerical basis for subsequent filter cutoff parameter configuration, and effectively ensuring the mutual independence of dual-channel image acquisition in the multispectral imaging system.

[0037] In an embodiment of the present invention, the process of obtaining the pseudo-color interference coefficient includes: A 2D fused image is obtained by registering and fusing a color image and a black-and-white image. The set of pixels representing fluorescently labeled regions is then extracted from the 2D fused image. Specifically, the generation process first involves image registration of the color and black-and-white images acquired at the color separation threshold. Image registration employs a homography transformation method based on feature point matching, extracting SIFT feature points from both images and performing cross-image feature matching. The homography transformation matrix aligns the black-and-white image to the color image coordinate system. A homography transformation is then applied to the black-and-white image to generate a registered black-and-white image that is strictly aligned with the color image in pixel coordinates. After image registration, the registered black-and-white image is superimposed as a near-infrared channel intensity map onto the luminance channel of the color image. Weighted mixing is then performed according to a preset initial fusion weight to generate a 2D fused image containing visible light color information and near-infrared fluorescence intensity information.

[0038] The pixel set extraction process for fluorescently labeled regions is based on the spatial distribution characteristics of near-infrared fluorescence intensity in the 2D fused image. First, the near-infrared fluorescence intensity value corresponding to each pixel is extracted from the registered black and white image. Using the near-infrared fluorescence intensity threshold as the segmentation boundary, all pixels with near-infrared fluorescence intensity values ​​higher than the threshold are marked as candidate pixels for fluorescently labeled regions. Connectivity analysis is performed on the candidate pixel set to retain connected regions with areas exceeding the minimum connected region area threshold and filter out isolated noise pixels. The coordinate set of all pixels within the filtered connected regions is determined as the pixel set of the fluorescently labeled region. The pixel set refers to the set of coordinate positions and corresponding chromaticity values ​​of all pixels within the fluorescently labeled region, which is the basic data unit for calculating the pseudo-color interference coefficient.

[0039] The chromaticity shift of each pixel within the pixel set of the fluorescently marked region is calculated, and the weighted average of the chromaticity shifts is determined as the pseudo-color interference coefficient. Specifically, the calculation process for the chromaticity shift first performs a color space conversion on the 2D fused image, converting the RGB color space to the CIE color space, and obtains the chromaticity coordinates of each pixel. For each pixel in the pixel set corresponding to the fluorescently marked region, its measured chromaticity coordinates in the 2D fused image are extracted, and the reference chromaticity coordinates of the corresponding anatomical position under pure visible light imaging conditions are read from the standard reference database. The Euclidean distance between the measured chromaticity coordinates and the reference chromaticity coordinates is calculated as the chromaticity shift at that pixel. The magnitude of the chromaticity shift directly reflects the degree of interference of near-infrared fluorescence signal fusion on the accuracy of visible light color at that position.

[0040] The weight determination process in the weighted mean calculation is based on the ratio of near-infrared fluorescence intensity to visible light brightness at each pixel. First, the near-infrared fluorescence intensity value of the pixel is read from the registered black and white image, and the visible light brightness value of the pixel is read from the brightness channel of the color image. The ratio of near-infrared fluorescence intensity to visible light brightness is calculated as the fusion weight of the pixel. The chromaticity offset of all pixels in the set of fluorescently marked pixels is weighted and summed according to the corresponding fusion weight, and then divided by the sum of all fusion weights to obtain the weighted mean chromaticity offset, which is determined as the pseudo-color interference coefficient under the current spectral configuration. It should be noted that when the pseudo-color interference coefficient is greater than the preset interference threshold, the corresponding spectral configuration is marked as a high-interference configuration and recorded in the optical performance benchmark feature library for subsequent construction of the test spectral configuration priority matrix as a basis for exclusion or weight reduction.

[0041] In an embodiment of the present invention, the process of constructing the test spectrum configuration priority matrix includes: A three-dimensional configuration space is constructed using visible light illuminance level, near-infrared excitation light power level, and cutoff filter cutoff wavelength level as dimensions. Specifically, the construction process of the three-dimensional configuration space first determines the discrete level division scheme of the three dimensions: visible light illuminance level is divided into three levels: low illuminance (500 lux to 2000 lux), medium illuminance (2000 lux to 6000 lux), and high illuminance (6000 lux to 10000 lux); near-infrared excitation light power level is divided into three levels: low power (0 to 5mW), medium power (5mW to 15mW), and high power (15mW to 30mW); cutoff filter cutoff wavelength level is based on the color separation critical point, and is divided into three levels near the critical point: below the critical point, at the critical point, and above the critical point. The level combinations of the three dimensions form corresponding configuration nodes, and each configuration node corresponds to a specific spectral test configuration scheme. Using the configuration node number as an index, a data structure for the three-dimensional configuration space is established to store the parameter combination of each configuration node and the medical image color fidelity score obtained in subsequent calculations.

[0042] For each configuration node in the three-dimensional configuration space, a medical image color fidelity score is calculated. The medical image color fidelity score is obtained by weighted summation of three components: tissue color reproduction accuracy, fluorescence lesion recognition contrast, and false color interference suppression. The tissue color reproduction accuracy component is calculated based on color reproduction error feature records in the optical performance benchmark feature library that match the visible light illuminance level of the current configuration node. The corresponding average color reproduction error value is extracted and normalized and mapped to a preset full-score benchmark value for color reproduction error. The fluorescence lesion recognition contrast component is calculated based on near-infrared channel signal-to-noise ratio feature records in the optical performance benchmark feature library that match the near-infrared excitation power level of the current configuration node. The corresponding near-infrared... The average signal-to-noise ratio (SNR) of the channel is normalized and mapped to a preset full-score SNR benchmark. The calculation of the pseudo-color interference suppression component is based on the cutoff wavelength level of the cutoff filter of the current configuration node and the pseudo-color interference coefficient record. If the current configuration node has been marked as a high-interference configuration, the pseudo-color interference suppression component is assigned a penalty value; otherwise, it is reverse-normalized based on the magnitude of the pseudo-color interference coefficient. The three components are weighted and summed according to preset weight coefficients. The weight coefficients are preset by professionals in the field based on medical image quality evaluation standards. The sum of the weight coefficients of the three components is 1, which yields the medical image color fidelity score of the current configuration node and stores it in the three-dimensional configuration space data structure.

[0043] Based on the color fidelity score of medical images, the configuration nodes in the 3D configuration space are sorted in descending order to generate a test spectral configuration priority matrix. Configuration nodes at the top of the sorting are marked as priority test configurations. Specifically, the descending sorting process traverses all 27 configuration nodes in the 3D configuration space, extracts the color fidelity score of each node, and sorts all configuration nodes in descending order of score, forming a one-dimensional ordered configuration node sequence. This ordered configuration node sequence is defined as the test spectral configuration priority matrix, with the first position corresponding to the configuration node with the highest score, representing the spectral configuration scheme with the best overall imaging quality under the current optical performance benchmark feature library conditions. According to a preset proportion parameter (usually the top 30% to 50%), the configuration nodes at the top of the sorting are marked as priority test configurations. The priority test configuration subset is the basic configuration pool for generating the initial detection scheme by combining the key band identifiers of the surgical scene. The marking results are stored in the identifier field of the test spectral configuration priority matrix for subsequent steps.

[0044] In this embodiment, the pseudo-color interference coefficient is determined by calculating the weighted average of the chromaticity offset of each pixel within the fluorescently marked region, and a test spectral configuration priority matrix is ​​generated based on the medical image color fidelity score of each node in the three-dimensional configuration space. This scheme can accurately evaluate the comprehensive impact of each spectral configuration on the color fidelity of medical images in a quantitative manner. By guiding the reasonable sorting of test configurations through the priority matrix, it ensures that the spectral configuration with the best performance is verified first, effectively improving the utilization efficiency of test resources and the overall quality of the detection scheme.

[0045] Step S3: Generate an initial detection scheme by combining key band identifiers of the surgical scene, obtain tissue recognition prediction indicators through image enhancement and fusion algorithm simulation, and identify test configuration nodes with optical crosstalk in the initial detection scheme.

[0046] In an embodiment of the present invention, the process of generating an initial detection scheme includes: The system retrieves key band identifiers corresponding to the target surgical type from a pre-stored surgical scene library. Specifically, the surgical scene library is a pre-established structured database containing spectral feature information for various common surgical types. Each surgical type corresponds to one surgical scene record, which includes the surgical type name, a list of tissue types requiring special attention, and key band identifiers for each tissue type. The key band identifiers consist of two types of wavelength parameters: the first is the peak wavelength of the reflectance spectrum, representing the characteristic wavelength with the highest reflectance intensity under visible or near-infrared irradiation for that tissue type; the second is the valley wavelength of the absorption spectrum, representing the characteristic wavelength with the highest absorption rate for a specific wavelength of light for that tissue type. The retrieval process first receives the input parameter for the target surgical type, then performs a precise query in the surgical scene library using the surgical type name as the key, extracting key band identifiers for all tissue types in the corresponding surgical scene record to form a set of key band identifiers for the current surgical type. This set of key band identifiers covers all spectral wavelengths requiring special attention in the target surgical scene and serves as the benchmark for band coverage verification during the initial detection scheme generation process.

[0047] Each wavelength in the key band identifier is compared with the priority test configuration in the test spectrum configuration priority matrix to verify the band coverage, and a subset of priority test configurations that have an effective response to each wavelength in the key band identifier is selected. Specifically, the band coverage verification process first extracts all priority test configuration nodes from the test spectral configuration priority matrix to form a priority test configuration candidate pool. For each priority test configuration node in the candidate pool, each wavelength parameter in the current surgical type key band identifier set is traversed. For each wavelength parameter, it is determined whether the spectral response range of the configuration node covers the wavelength. The judgment criteria are as follows: if the wavelength to be verified falls within the effective response band (400nm to 700nm) corresponding to the visible light illuminance level of the configuration node, then it is checked whether the quantum efficiency of the sensor spectral response curve at that wavelength exceeds the preset effective response threshold; if the wavelength to be verified falls within the effective response band (760nm to 1000nm) corresponding to the near-infrared excitation light power level, then it is checked whether the near-infrared channel signal-to-noise ratio exceeds the preset minimum usable signal-to-noise ratio. The above verification is performed on all wavelength parameters in the key band identifier set. Only when a certain priority test configuration node has an effective response to all wavelengths in the set is it included in the configuration subset that has passed the band coverage verification, thus obtaining the priority test configuration subset.

[0048] An initial detection scheme is formed based on a priority subset of test configurations. Specifically, the initial detection scheme is arranged by first reordering the configuration subsets that have passed the band coverage verification according to the descending order of the test spectrum configuration priority matrix, ensuring that the configuration nodes with higher medical image color fidelity scores are placed first. The sorted configuration subsets are then arranged sequentially into a test sequence for the initial detection scheme. Each test sequence entry records the visible light illuminance level parameters, near-infrared excitation light power level parameters, cutoff filter cutoff wavelength level parameters, and medical image color fidelity scores for the corresponding configuration node. The initial detection scheme is organized in the form of a test sequence, indicating that the configuration nodes placed earlier should be prioritized for testing during automated testing. If optical crosstalk or other problems exist in the preceding configuration nodes, the system will switch to subsequent configuration nodes in turn. The generated results of the initial detection scheme are stored in the detection scheme data structure as input data for the subsequent generation of detection schemes after optical crosstalk identification and correction.

[0049] In an embodiment of the present invention, the process of obtaining the tissue identification prediction index includes: The visible light image data and near-infrared fluorescence image data corresponding to each configured node in the initial detection scheme are input into the pre-constructed image registration simulation unit. The input preparation process of the image registration simulation unit first traverses the test sequence of the initial detection scheme. For each configuration node, typical visible light image data and near-infrared fluorescence image data corresponding to the visible light illuminance level and near-infrared excitation power level of that node are extracted from the optical performance benchmark feature library. The typical visible light image data includes the color response image of the standard tissue model under the current visible light illuminance level, and the typical near-infrared fluorescence image data includes the optical molecular image of the targeted fluorescent group under the current near-infrared excitation power level. The visible light image data and near-infrared fluorescence image data are combined into an image input pair and fed into the image registration simulation unit for processing. The image registration simulation unit performs registration processing on the input image pair based on the stereo matching algorithm. The stereo matching algorithm extracts corresponding feature points in the two images and calculates the optimal transformation parameters to align the coordinate system of the near-infrared fluorescence image to the coordinate system of the visible light image, generating a geometrically precisely aligned simulated registered image pair. The visible light image and the registered near-infrared fluorescence image in the simulated registered image pair correspond strictly in pixel coordinates to ensure the spatial consistency of information in each band in subsequent fusion operations.

[0050] It should be noted that the optimal transformation parameters refer to the set of transformation model parameters obtained by using a stereo matching algorithm during image registration, which minimizes the total reprojection error between all inlier matching pairs in the original image (black and white image) and their corresponding matching points in the target image (color image) after geometric transformation. In this invention, the optimal transformation parameters are specifically represented by a homography transformation matrix, which fully describes the geometric transformation relationship required to map the black and white image coordinate system to the color image coordinate system.

[0051] The simulated registered image pairs are input into a pre-constructed 3D image fusion simulation unit to generate a simulated 3D fused image. Specifically, after receiving the simulated registered image pairs, the 3D image fusion simulation unit first determines the fusion contribution weight of the near-infrared channel based on the cutoff wavelength level parameter of the cutoff filter of the currently configured node. The closer the cutoff wavelength is to the color separation critical point, the higher the fusion contribution weight of the near-infrared channel. Using the RGB channel data of the visible light image as the color basis and the grayscale data of the registered near-infrared fluorescence image as the depth enhancement layer, weighted mixing is performed according to the fusion contribution weight. At the same time, edge enhancement processing based on image gradient is introduced to highlight the complementary features of visible light and near-infrared information in space. The mixing result is projected as a texture mapping in the simulated 3D scene, and combined with the existing distortion model in the endoscope optical system to generate a simulated 3D fused image with spatial depth. The simulated 3D fused image retains the color accuracy of visible light while fusing the lesion marking information of near-infrared fluorescence, which is used for the quantitative extraction of subsequent tissue identification prediction indicators.

[0052] Three components—grayscale contrast, edge sharpness, and fluorescence uniformity—of the target lesion region relative to the normal tissue region are extracted from the simulated 3D fused image. The weighted average value of these three components is determined as the tissue identification prediction index. Specifically, the extraction process of the grayscale contrast component first delineates the target lesion region in the simulated 3D fused image based on near-infrared fluorescence intensity threshold segmentation. A normal tissue reference region is formed by extending the boundary of the target lesion region outward by a preset pixel distance. The mean grayscale value of all pixels in the target lesion region is calculated separately from the mean grayscale value of all pixels in the normal tissue reference region. The ratio of the difference between the two to the sum of the two is calculated to obtain the normalized grayscale contrast component. The higher the grayscale contrast component, the more significant the brightness difference between the target lesion region and the normal tissue region in the fused image, and the stronger the visibility of the lesion.

[0053] The process of extracting edge sharpness components involves applying the Sobel gradient operator to the simulated 3D fused image and calculating the gradient magnitude at each pixel. The gradient magnitude distribution is extracted at the boundary between the target lesion region and the normal tissue reference region, and the mean of the gradient magnitude within the boundary region is calculated as the edge sharpness component. The higher the edge sharpness component, the clearer and sharper the boundary of the target lesion region in the fused image, which is helpful for assisting surgeons in accurately judging the lesion boundary.

[0054] The extraction process of the fluorescence uniformity component involves calculating the spatial variation coefficient of near-infrared fluorescence intensity within the target lesion area. The spatial variation coefficient is calculated as the ratio of the standard deviation of fluorescence intensity to the mean fluorescence intensity. The spatial variation coefficient is then reverse-normalized; a smaller coefficient indicates a higher fluorescence uniformity component. A higher fluorescence uniformity component indicates a more uniform distribution of fluorescent markers within the target lesion area, and stronger reliability of the lesion markers. The grayscale contrast component, edge sharpness component, and fluorescence uniformity component are weighted and summed according to preset weight coefficients. These weight coefficients are pre-set by experts in the field based on the clinical needs of the surgical scenario. The sum of the weight coefficients of the three components is 1, yielding the tissue identification prediction index for the current configuration node. The tissue identification prediction index is recorded in the corresponding configuration node entry of the initial detection scheme, providing a quantitative evaluation basis for subsequent optical crosstalk identification.

[0055] In embodiments of the present invention, the process of identifying test configuration nodes with optical crosstalk includes: For each configuration node in the initial detection scheme, the energy penetration rate of the near-infrared light channel to the visible light band and the energy penetration rate of the visible light channel to the near-infrared light band are calculated separately. Specifically, the calculation process of the energy penetration rate of the near-infrared light channel to the visible light band first extracts the color image data corresponding to the current configuration node from the optical performance benchmark feature library; in the color image data, by comparing the brightness difference between two sets of color images acquired in the near-infrared excitation light on state and the near-infrared excitation light off state, the additional brightness distribution caused by the leakage of the near-infrared light channel is extracted; the ratio of the average value of the additional brightness distribution of the whole image to the average value of the total brightness of the color image is calculated to obtain the energy penetration rate of the near-infrared light channel to the visible light band; this energy penetration rate reflects the degree to which near-infrared light bypasses the cutoff filter to enter the visible light channel and affects the color imaging quality.

[0056] The process of calculating the energy penetration rate of the visible light channel to the near-infrared band first extracts the black-and-white image data corresponding to the current configuration node from the optical performance benchmark feature library. In the black-and-white image data, the background gray-scale distribution caused by visible light signal penetration is extracted by comparing the gray-scale difference between two sets of black-and-white images acquired when the visible light standard light source is on and when the visible light standard light source is off. The ratio of the mean of the background gray-scale distribution of the entire image to the mean of the total gray-scale of the black-and-white image is calculated to obtain the energy penetration rate of the visible light channel to the near-infrared band. This energy penetration rate reflects the degree to which the visible light signal penetrates the cutoff boundary of the near-infrared channel and enters the near-infrared imaging area, forming background interference. The above two energy penetration rate calculations are performed sequentially on all configuration nodes in the initial detection scheme test sequence, and the calculation results are stored in the penetration rate field of the corresponding configuration node entry.

[0057] Configuration nodes whose near-infrared light channel's energy penetration rate to the visible light band exceeds a first preset penetration rate threshold, or whose visible light channel's energy penetration rate to the near-infrared band exceeds a second preset penetration rate threshold, are marked as test configuration nodes with optical crosstalk. Specifically, the optical crosstalk marking process traverses the penetration rate fields of all configuration nodes in the initial detection scheme's test sequence, performing dual-path penetration rate threshold judgments on each configuration node. The first preset penetration rate threshold is the upper limit of tolerance for near-infrared light channel energy penetration to the visible light band, pre-set by those skilled in the art based on medical image color fidelity standards; the second preset penetration rate threshold is the upper limit of tolerance for visible light channel energy penetration to the near-infrared band, pre-set by those skilled in the art based on near-infrared fluorescence imaging sensitivity requirements. For a given configuration node, if its near-infrared light channel's energy penetration rate to the visible light band exceeds a second preset penetration rate threshold, then it is considered a test configuration node with optical crosstalk. If the energy penetration rate of a wavelength band exceeds the first preset penetration rate threshold, the configuration node has a crosstalk problem of near-infrared interference with visible light; if the energy penetration rate of its visible light channel to the near-infrared wavelength band exceeds the second preset penetration rate threshold, the configuration node has a crosstalk problem of visible light interference with near-infrared. Configuration nodes that meet any of the above conditions are marked as test configuration nodes with optical crosstalk in the initial detection scheme, and the crosstalk type identifier (near-infrared crosstalk to visible light, visible light crosstalk to near-infrared, or bidirectional crosstalk) is recorded for subsequent targeted adjustment of filter cutoff parameters or near-infrared color mapping weights.

[0058] Step S4: Adjust the filter cutoff parameters or near-infrared color mapping weights for test configuration nodes with optical crosstalk to generate a corrected detection scheme.

[0059] Among them, the filter cutoff parameter refers to key parameters such as the wavelength position and cutoff slope of the cutoff filter to achieve effective cutoff within the spectral range. By adjusting the cutoff parameter, the filter's ability to block light signals in a specific band can be changed, thereby reducing the degree of energy penetration between the two channels. The near-infrared color mapping weight refers to the weight coefficient assigned when the near-infrared band image signal is mapped to the visible light color image space during the multispectral image fusion process. By adjusting this weight, the color rendering intensity of the near-infrared fluorescence signal in the fused image can be controlled to reduce interference to the visible light color channel. The corrected detection scheme refers to the test configuration sequence that meets the spectral isolation requirements after adjusting the parameters of the test configuration nodes with optical crosstalk. It serves as the optimization basis for subsequent automated testing.

[0060] In one embodiment, for a test configuration node marked as having optical crosstalk, the type of crosstalk source is first determined. If the energy penetration rate of the near-infrared light channel into the visible light band exceeds a first preset penetration rate threshold, the cutoff parameters of the cutoff filter corresponding to the configuration node are adjusted to tighten the cutoff wavelength or increase the cutoff slope in the short-wave direction to enhance the suppression capability of near-infrared light penetration into the visible light band. If the energy penetration rate of the visible light channel into the near-infrared band exceeds a second preset penetration rate threshold, the near-infrared color mapping weight corresponding to the configuration node is reduced to weaken the residual influence of visible light signals in the near-infrared channel. After the parameter adjustment is completed, the energy penetration rate of the adjusted configuration node is re-evaluated. After confirming that both types of penetration rates are lower than the corresponding preset thresholds, the adjusted configuration node is updated to the corresponding position in the initial detection scheme to generate a corrected detection scheme.

[0061] Step S5: During the automated testing process based on the modified detection scheme, the image sharpness index and wide color gamut coverage integrity of the acquired image data are monitored in real time. If the image sharpness index or wide color gamut coverage integrity is less than the corresponding standard threshold, the LSC lens shading correction parameters and AWB white balance parameters are automatically adjusted back, and the automated test is re-executed based on the adjusted LSC lens shading correction parameters and AWB white balance parameters to obtain the optical performance test results.

[0062] Among them, the image sharpness index is a composite index that comprehensively quantifies the sharpness of images acquired by the endoscope and the contrast quality of fluorescence signals. It is calculated by weighted average of the edge sharpness of the color image and the fluorescence signal contrast of the black and white image, and can reflect the overall imaging sharpness level of the endoscope under the current parameter configuration; the wide color gamut coverage integrity refers to the ratio of the actual color gamut volume covered by the endoscope color image in the Lab color space to the standard color gamut volume of medical images, and is used to measure the endoscope's ability to reproduce and cover the standard color gamut of medical images; the LSC lens shading correction parameter refers to the set of correction coefficients used to compensate for the image peripheral brightness attenuation effect caused by the lens optical system. By adjusting this parameter... It can eliminate the relative brightness attenuation difference between the central and edge areas of an image, improving the uniformity of image brightness; AWB white balance parameters refer to a set of gain compensation coefficients used to eliminate the influence of different color temperature lighting environments on the overall tonal shift of the image. By adjusting the gain compensation value of each tone channel, the endoscope can accurately reproduce the true color of the target object under different lighting conditions; the optical performance test results refer to the comprehensive performance evaluation report output after completing all automated testing processes, which includes the final LSC lens shadow correction parameters, AWB white balance parameters, image sharpness index, wide color gamut coverage integrity, 2D fusion image quality score, and 3D stereoscopic image depth accuracy.

[0063] In an embodiment of the present invention, the process of automatically retrospectively adjusting the LSC lens shading correction parameters and AWB white balance parameters includes: Color images from a high-definition color CCD and black-and-white images from a high-sensitivity black-and-white CCD were acquired separately. Laplacian gradient calculation was performed on the color images to extract edge sharpness, and grayscale variance analysis was performed on the black-and-white images to extract fluorescence signal contrast. The weighted average of edge sharpness and fluorescence signal contrast was used as the image sharpness index.

[0064] Among them, Laplacian gradient operation refers to the image processing operation method that applies the Laplacian operator to a color image to perform second-order differential calculation, and extracts pixel regions with abrupt changes in brightness to quantify the edge sharpness of the image; edge sharpness refers to the statistical quantification result of the edge response value in the color image after Laplacian gradient operation. The larger the value, the sharper the image edge and the higher the lens imaging sharpness; gray-level variance analysis refers to the statistical calculation of the variance of the gray values ​​of the fluorescent marked area and the surrounding background area in a black and white image, and the analysis method that quantifies the contrast intensity of the fluorescence signal relative to the background noise by the degree of dispersion of the gray-level distribution; fluorescence signal contrast refers to the ratio of the difference between the mean gray value of the fluorescent marked area and the mean gray value of the background area to the sum of the two, which can reflect the degree of distinction between lesions and normal tissues in near-infrared fluorescence imaging.

[0065] In one embodiment, during automated testing, a high-definition color CCD and a high-sensitivity monochrome CCD are simultaneously triggered to complete image acquisition. A Laplacian operator is applied to the acquired color image to perform gradient calculations, and the root mean square value of the gradient response map is calculated as the edge sharpness. For the acquired monochrome image, the grayscale mean values ​​of the fluorescently marked region and the background region are extracted respectively, and the fluorescence signal contrast is calculated. Based on preset edge sharpness weighting coefficients and fluorescence signal contrast weighting coefficients, a weighted average of edge sharpness and fluorescence signal contrast is performed to obtain the image clarity index of the currently acquired image. The image clarity index is compared with the corresponding clarity standard threshold to determine whether to trigger the LSC lens shadow correction parameter backtracking adjustment process.

[0066] The color image is mapped from the RGB color space to the Lab color space, and the ratio of the actual gamut volume covered in the Lab color space to the standard color gamut volume of the medical image is calculated as the wide color gamut coverage completeness.

[0067] Among them, gamut volume refers to the three-dimensional convex hull volume formed by all pixel color coordinates of a color image in the Lab color space, which can quantify the breadth of the color coverage of the image; medical imaging standard gamut volume refers to the minimum gamut coverage volume that an endoscope image should achieve in the Lab color space according to the industry standard for color fidelity of medical images, and is a reference benchmark for judging the integrity of wide gamut coverage.

[0068] In one embodiment, the RGB color values ​​of all pixels in a color image acquired by a high-definition color CCD are mapped to the Lab color space using a standard color space transformation matrix to obtain the corresponding three-dimensional color coordinates of each pixel in the Lab color space. The three-dimensional convex hull volume of the set of all pixel three-dimensional coordinate points is calculated to obtain the actual color gamut volume covered by the current color image in the Lab color space. The actual covered color gamut volume is divided by the pre-stored standard color gamut volume of medical images to obtain the wide color gamut coverage completeness. The wide color gamut coverage completeness is compared with the corresponding coverage standard threshold to determine whether to trigger the AWB white balance parameter retrospective adjustment process.

[0069] When the image sharpness index is less than the corresponding sharpness standard threshold, the edge brightness distribution curve of the currently acquired image is extracted, the relative brightness attenuation ratio between the image center region and the edge region is calculated, and the corresponding lens shadow correction compensation coefficient is matched in the preset LSC lens shadow correction parameter lookup table according to the relative brightness attenuation ratio, and the lens shadow correction compensation coefficient is updated to the lens shadow correction unit in the endoscope.

[0070] Among them, the edge brightness distribution curve refers to the distribution curve of the average brightness value at each sampling position along the direction from the center area to the edge area of ​​the color image, which can intuitively reflect the brightness gradient law caused by the lens optical system; the relative brightness attenuation ratio refers to the inverse of the ratio of the average brightness value of the image edge area to the average brightness value of the center area minus one, which is used to quantify the degree of attenuation of the brightness of the image periphery due to the lens shadow effect; the LSC lens shadow correction parameter lookup table refers to a pre-established mapping data table indexed by the relative brightness attenuation ratio and with the corresponding lens shadow correction compensation coefficient as the query result, which can quickly locate the correction parameter that matches the current shadow attenuation level; the lens shadow correction unit refers to the functional unit in the endoscope that is specifically responsible for compensating the brightness uniformity of the acquired image, and eliminates the brightness attenuation of the image periphery by applying the lens shadow correction compensation coefficient.

[0071] In one embodiment, when the image sharpness index is less than the corresponding sharpness standard threshold, a luminance sampling sequence along the radial direction of the image is extracted from the currently acquired color image to generate an edge luminance distribution curve. The image is divided into a central region and an edge region, and the average luminance value of all pixels in each region is calculated. The relative luminance attenuation ratio is calculated based on the average luminance of the edge region and the average luminance of the central region. Using the relative luminance attenuation ratio as an index, the corresponding lens shading correction compensation coefficient is looked up in a preset LSC lens shading correction parameter lookup table. The found lens shading correction compensation coefficient is written into the parameter register of the lens shading correction unit in the endoscope, completing the real-time update of the LSC lens shading correction parameters.

[0072] When the wide color gamut coverage integrity is less than the corresponding coverage standard threshold, the mean value of each tone channel of the current color image is extracted, the color temperature deviation between the mean value of each tone channel and the standard white point reference value is calculated, the corresponding gain compensation value is matched in the preset AWB white balance parameter mapping table according to the color temperature deviation, and the gain compensation value is updated to the white balance adjustment unit until the image sharpness index and the wide color gamut coverage integrity both meet the corresponding standard threshold.

[0073] Among them, the hue channel mean refers to the arithmetic mean of all pixel values ​​of each hue channel (red channel, green channel, blue channel) in the RGB color space of the color image, which can reflect the overall hue shift direction and degree of the current image; the standard white point reference value refers to the ideal mean of each hue channel collected on a standard white reference target, representing the offset-free color reference response that the endoscope should produce under standard light source; the color temperature deviation refers to the quantitative value that describes the overall color temperature shift degree and direction of the current image, which is converted from the difference between each hue channel mean and the standard white point reference value, and is the basis for determining the direction and magnitude of white balance compensation; the AWB white balance parameter mapping table refers to a pre-established mapping data table that uses color temperature deviation as the index and corresponding hue channel gain compensation value as the query result, which can quickly locate the gain compensation parameter that matches the current color temperature deviation; the white balance adjustment unit refers to the functional unit in the endoscope image processing module that is specifically responsible for applying gain compensation to each hue channel to eliminate the influence of color temperature shift.

[0074] In one embodiment, when the wide color gamut coverage integrity is less than the corresponding coverage standard threshold, the pixel mean values ​​of the red, green, and blue tonal channels of the current color image are calculated separately, and the difference between each channel and the pre-stored standard white point reference value is calculated. The differences of each channel are then combined and converted into a color temperature deviation. Using the color temperature deviation as an index, the gain compensation value corresponding to each red, green, and blue channel is looked up in the preset AWB white balance parameter mapping table. The found gain compensation value is written to the gain register of the white balance adjustment unit, completing the real-time update of the AWB white balance parameters. After updating the LSC lens shading correction parameters and AWB white balance parameters, image acquisition is retried. The image sharpness index and wide color gamut coverage integrity of the newly acquired image are calculated. It is determined whether both indicators meet the corresponding standard thresholds. If either indicator still does not meet the standard threshold, the above backtracking adjustment process is iteratively executed. When an indicator still fails to meet the standard after three consecutive iterations, the current test configuration node is recorded as an unqualified node and marked as an anomaly. When both indicators meet the corresponding standard thresholds, the optical performance test results, including the final LSC lens shading correction parameters, AWB white balance parameters, image sharpness index, wide color gamut coverage integrity, 2D fused image quality score, and 3D stereoscopic image depth accuracy, are output.

[0075] In this embodiment, by real-time monitoring of image sharpness index and wide color gamut coverage integrity, automatic retrospective adjustment of LSC lens shading correction parameters and AWB white balance parameters is triggered when the indicators fail to meet the standards. A defective node marking mechanism is also set up for three failed iterations. This scheme can perceive the dynamic changes in image quality in real time during automated testing, quickly eliminate the influence of lens shading and color temperature shift on test results through a closed-loop parameter adjustment mechanism, ensure the accuracy and reliability of optical performance test results, and guarantee the complete traceability of the testing process through the defective node marking mechanism.

[0076] It should be noted that the parameter thresholds used in this application (such as the preset effective response threshold, the preset minimum usable signal-to-noise ratio, and the clear standard threshold) were all calibrated in advance by those skilled in the art based on historical work experience and relevant simulation experiments.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0078] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0079] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for testing the optical performance of a full-spectrum optical imaging endoscope, characterized in that, include: Step S1: Obtain the raw image data of the endoscope under a standard light source and the spectral response curve of the sensor. The raw image data includes image data collected under different color temperatures and illuminance conditions in the visible light band and near-infrared light band. Color reproduction error and signal-to-noise ratio features under different color temperatures and illuminance conditions are extracted, and an optical performance benchmark feature library is constructed based on the color reproduction error and signal-to-noise ratio features. Step S2: Identify the critical point of color separation between the visible light band and the near-infrared light band based on the optical performance benchmark feature library, calculate the pseudo-color interference coefficient during multispectral fusion imaging, and construct a test spectrum configuration priority matrix according to the color fidelity requirements of medical images. Step S3: Generate an initial detection scheme by combining key band identifiers of the surgical scene, obtain tissue recognition prediction indicators through image enhancement and fusion algorithm simulation, and identify test configuration nodes with optical crosstalk in the initial detection scheme; Step S4: Adjust the filter cutoff parameters or near-infrared color mapping weights for test configuration nodes with optical crosstalk to generate a corrected detection scheme; Step S5: During the automated testing process based on the modified detection scheme, the image sharpness index and wide color gamut coverage integrity of the acquired image data are monitored in real time. If the image sharpness index or wide color gamut coverage integrity is less than the corresponding standard threshold, the LSC lens shading correction parameters and AWB white balance parameters are automatically adjusted back, and the automated test is re-executed based on the adjusted LSC lens shading correction parameters and AWB white balance parameters to obtain the optical performance test results.

2. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The process of extracting color reproduction error and signal-to-noise ratio features under different color temperatures and illuminance conditions includes: Near-infrared light and visible light standard light sources were injected into the incident section of the endoscope beam under different color temperatures and illuminance conditions. Optical molecular images generated after the near-infrared light was excited by the ICG targeted fluorescent group and reflected light images generated by the visible light standard light source mapped onto the standard color card were collected. Near-infrared channel signal-to-noise ratio features are extracted from the optical molecular image, and color reproduction error features are extracted from the reflected light image.

3. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 2, characterized in that, The process of constructing the optical performance benchmark feature library includes: A two-dimensional feature matrix with color temperature and illuminance as index dimensions is established, and the color reproduction error features and signal-to-noise ratio features extracted under different color temperature-illuminance combinations are classified and stored in the corresponding matrix nodes. Calculate the color difference statistics for the color reproduction error characteristics of each matrix node, and extract the mean signal-to-noise ratio and background noise power spectral density for the signal-to-noise ratio characteristics of each matrix node. The optimal working range for visible light and near-infrared light is selected based on the preset color difference threshold and the preset signal-to-noise ratio threshold, and the overlapping area of ​​the two is determined as the optimal working range for the whole spectrum. The environmental parameters, performance indicators, and sensor spectral response curves corresponding to the optimal working range of the full spectrum are associated and stored to form an optical performance benchmark feature library.

4. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The process of identifying the color separation critical point includes: The beam splitter in the endoscope separates the light signal transmitted from the eyepiece into a visible light component of 400nm to 700nm and a near-infrared light component of 760nm to 1000nm. A cutoff filter is applied to the near-infrared light channel and its cutoff wavelength is stepped within the range of 835nm to 855nm. At each cutoff wavelength step point, a color image output by a high-definition color CCD and a black and white image output by a high-sensitivity black and white CCD are acquired respectively. Calculate the proportion of near-infrared leakage brightness in the color image and the proportion of visible light residue brightness in the black and white image at each step point; determine the cutoff wavelength corresponding to the product of the two proportions being less than the preset separation threshold as the color separation critical point.

5. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 4, characterized in that, The calculation of the pseudo-color interference coefficient includes: A 2D fused image is obtained by image registration and fusion of the color image and the black and white image, and the set of pixels of fluorescently marked regions is extracted from the 2D fused image. Calculate the chromaticity offset of each pixel in the pixel set of the fluorescently marked region, and determine the weighted average of the chromaticity offset as the pseudo-color interference coefficient.

6. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The process of constructing the test spectrum configuration priority matrix includes: A three-dimensional configuration space is constructed using visible light illuminance level, near-infrared excitation light power level, and cutoff wavelength level of cutoff filter as dimensions; a medical image color fidelity score is calculated for each configuration node in the three-dimensional configuration space; Based on the color fidelity score of the medical image, the configuration nodes in the three-dimensional configuration space are arranged in descending order to generate the test spectrum configuration priority matrix, and the configuration nodes with a preset proportion before the arrangement are marked as priority test configurations.

7. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The process of generating an initial detection scheme includes: Read the key band identifiers corresponding to the target surgical type from the pre-stored surgical scenario library. The key band identifiers include the peak wavelength of the reflectance spectrum and the valley wavelength of the absorption spectrum corresponding to the tissue types that need to be distinguished in the target surgical type. Each wavelength in the key band identifier is compared with the priority test configuration in the test spectrum configuration priority matrix to verify band coverage, and a subset of priority test configurations that have an effective response to each wavelength in the key band identifier is selected; the initial detection scheme is formed based on the subset of priority test configurations.

8. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The process of obtaining organizational identification predictive metrics includes: The visible light image data and near-infrared fluorescence image data corresponding to each configured node in the initial detection scheme are input into the pre-constructed image registration simulation unit. The image registration simulation unit performs registration processing on the visible light image data and near-infrared fluorescence image data based on the stereo matching algorithm to generate simulated registration image pairs. The simulated registered image is input into the three-dimensional image fusion simulation module to generate a simulated 3D fused image. The grayscale contrast, edge sharpness, and fluorescence uniformity of the target lesion area relative to the normal tissue area in the simulated 3D fused image are extracted. The weighted comprehensive value of the three components is determined as the tissue recognition prediction index.

9. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The process of identifying test configuration nodes with optical crosstalk in the initial detection scheme includes: For each configuration node in the initial detection scheme, the energy penetration rate of the near-infrared light channel to the visible light band and the energy penetration rate of the visible light channel to the near-infrared band are calculated respectively. Configuration nodes whose energy penetration rate of the near-infrared light channel to the visible light band exceeds a first preset penetration rate threshold or whose energy penetration rate of the visible light channel to the near-infrared band exceeds a second preset penetration rate threshold are marked as test configuration nodes with optical crosstalk.

10. The method for testing the optical performance of a full-spectrum optical imaging endoscope according to claim 1, characterized in that, The automatic retrospective adjustment of LSC lens shading correction parameters and AWB white balance parameters includes: During the automated testing process, high-definition color CCD output color images and high-sensitivity black-and-white CCD output black-and-white images are acquired respectively. The edge sharpness is extracted by performing Laplacian gradient operation on the color image, and the contrast of the fluorescence signal is extracted by performing grayscale variance analysis on the black and white image. The weighted average of the edge sharpness and the fluorescence signal contrast is used as the image sharpness index. The color image is mapped from the RGB color space to the Lab color space, and the ratio of the actual gamut volume covered in the Lab color space to the standard color gamut volume of the medical image is calculated as the wide color gamut coverage completeness. When the image sharpness index is less than the corresponding sharpness standard threshold, the edge brightness distribution curve of the currently acquired image is extracted, the relative brightness attenuation ratio between the image center region and the edge region is calculated, and the corresponding lens shadow correction compensation coefficient is matched in the preset LSC lens shadow correction parameter lookup table according to the relative brightness attenuation ratio, and the lens shadow correction compensation coefficient is updated to the lens shadow correction unit in the endoscope. When the wide color gamut coverage integrity is less than the corresponding coverage standard threshold, the mean value of each tone channel of the current color image is extracted, the color temperature deviation between the mean value of each tone channel and the standard white point reference value is calculated, the corresponding gain compensation value is matched in the preset AWB white balance parameter mapping table according to the color temperature deviation, and the gain compensation value is updated to the white balance adjustment unit until the image sharpness index and the wide color gamut coverage integrity both meet the corresponding standard threshold. After adjusting the LSC lens shading correction parameters and AWB white balance parameters, the corresponding complete test process is re-executed based on the adjusted parameters. The image sharpness index and wide color gamut coverage integrity of the newly acquired image are monitored in real time. If any indicator is still less than the corresponding standard threshold, the parameters are iteratively adjusted. If the indicator still fails to meet the standard after three consecutive iterations, the current test configuration node is recorded as an unqualified node and marked as an anomaly. When all indicators meet the corresponding standard threshold, the optical performance test results, including the final LSC parameters, AWB parameters, image sharpness index, wide color gamut coverage integrity, 2D fused image quality score, and 3D stereoscopic image depth accuracy, are output.

Images