Guided dual-scale coupling correction method for uneven illumination of fluorescence plate imaging

Abstract

The application discloses a fluorescent hole plate imaging light unevenness calibration guided double-scale coupling correction method, and belongs to the technical field of fluorescent imaging detection and image processing. The method comprises the following steps: collecting a standard fluorescent hole plate reference image, extracting the fluorescent intensity of each hole site, fitting and constructing a physical light basis model, and storing the hole site level calibration correction coefficient; for a to-be-corrected sample image, shielding the in-hole fluorescent region to estimate the light, and weighting and fusing the calibration model to obtain a mixed light estimation, thereby performing adaptive gamma correction to obtain a coarse correction image; extracting the fluorescent intensity of each hole site on the coarse correction image; and adaptively calculating a residual correction coefficient according to the effective hole site distribution to complete fine correction of the hole site level. The application deeply fuses the calibration prior and the adaptive algorithm, simultaneously solves the global light gradual change and the hole site local deviation problem through double-scale coupling correction, and significantly improves the uniformity and accuracy of the whole plate fluorescent quantification.

Description

Technical Field

[0001] This invention relates to the field of fluorescence imaging detection and image processing technology, and in particular to a calibration-guided dual-scale coupling correction method for non-uniform illumination in fluorescence well plate imaging. Background Technology

[0002] Fluorescent plate imaging systems (96-well plates, 384-well plates, 8-tube arrays, etc.) typically use LED point light sources as excitation sources. Due to the inherent characteristics of point light sources, the intensity of the excitation light illuminating the plate plane exhibits a distribution pattern of strong intensity at the center and weak intensity at the edges, leading to the following problems: excessively strong fluorescence signals at the central wells and weak signals at the edge wells, causing systematic bias in the overall plate quantitative analysis; decreased accuracy in target segmentation and quantitative fluorescence extraction in the low signal-to-noise ratio areas at the edge wells; and a lack of comparability of fluorescence signals between different wells, affecting the reliability of experimental conclusions.

[0003] The existing methods for correcting uneven illumination at the image processing level mainly include: (1) Flat-Field correction: dividing the image to be corrected by the reference image to achieve pixel-by-pixel correction, but it amplifies noise in dark areas severely and does not consider the discrete structural features of the aperture plate, thus failing to solve the local uneven deviation at the aperture position level.

[0004] (2) Retinex-type methods and two-dimensional gamma correction: Illumination components are estimated based on the illuminance-reflectance model, and then brightness adaptive adjustment is achieved through pixel-by-pixel gamma correction. Among them, the improved two-dimensional gamma function is used for adaptive brightness correction. After extracting illumination information through guided filtering, the two-dimensional gamma function is constructed to dynamically adjust the correction parameters. Some studies divide the image into different regions and establish regional adaptive target mean functions, and dynamically adjust the gamma parameters according to the illumination distribution characteristics of each region.

[0005] This type of method has achieved good results in general image enhancement scenarios, but it has the following drawbacks when applied to fluorescence plate imaging correction: When estimating illumination across the entire image, strong fluorescence signals inside the aperture are misidentified as illumination components, leading to an overestimation of illumination in the aperture region and a decrease in correction accuracy. The correction was not optimized for the discrete well position characteristics of the orifice plate, and only the visual effect was improved after correction, which could not meet the requirements for quantitative detection. It cannot resolve local non-uniform deviations at the orifice level (such as orifice wall obstruction and meniscus refraction of liquid).

[0006] Among existing patented technologies, some solutions involve reference division correction using a uniform fluorescent plate (such as US Patent US7666663 and Japanese Patent JP2005181244), which are pure calibration-driven open-loop corrections and lack the ability to adaptively adjust to the current image content; other solutions involve the hardware architecture of a multi-hole plate fluorescence imaging system (such as Chinese Patent CN107003245B), but mainly solve the imaging hardware architecture problem and do not involve the correction algorithm for illumination non-uniformity.

[0007] In summary, existing technologies do not organically combine standard calibration data with adaptive image processing algorithms, making it impossible to simultaneously achieve pixel-level continuous correction and aperture-level discrete fine correction. Furthermore, correction and segmentation are mutually restrictive, making it difficult to meet the high-precision requirements of fluorescence plate imaging systems for full-plate quantitative analysis. Summary of the Invention

[0008] To achieve the above-mentioned objectives and other advantages of the present invention, a first objective of the present invention is to provide a calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture plate imaging, comprising the following steps: Acquire reference images of standard fluorescent plates, extract the fluorescence intensity of each well, fit and construct a parameterized physical illumination model, and store the well-level calibration correction coefficients; For the image of the sample to be calibrated, the fluorescence signal region within the well is shielded, and illumination estimation is performed based on the background region between the wells; the physical illumination model is fused with the illumination estimation to obtain a hybrid illumination estimation; adaptive gamma correction is performed based on the hybrid illumination estimation to obtain a pixel-level coarsely calibrated image; Well segmentation is performed on the coarsely corrected image to extract the region of each well and calculate the average fluorescence intensity of each well. Based on the average fluorescence intensity of each well site and the well site-level calibration correction coefficient, the residual correction coefficient is adaptively calculated according to the effective well site distribution, and the well site regions in the coarse correction image are corrected to obtain the final corrected image with dual-scale coupling.

[0009] Furthermore, the physical lighting model is constructed as follows: Using the center coordinates of each well location as the independent variable and the fluorescence intensity of each well location as the observed value, global surface fitting is performed using the least squares method. The fitting model adopts a two-dimensional Gaussian model or a third-order polynomial model.

[0010] Furthermore, the illumination estimation employs the multi-scale Retinex method; The fusion of the physical illumination model and the illumination estimate is a weighted fusion, and the fusion formula is as follows: in, This is the normalized value of the physical lighting fundamental model. λ is the estimated illumination value, and λ is the calibration weighting coefficient, 0≤λ≤1.

[0011] Furthermore, in the adaptive gamma correction step based on the mixed illumination estimation, the gamma coefficients are determined pixel-by-pixel according to the relative relationship between the mixed illumination estimation and the reference brightness value, where the reference brightness value is the 75th quantile of the mixed illumination estimation. The formula for calculating the gamma coefficients is: in, The 75th percentile of the estimated mixed illumination.

[0012] Furthermore, the step of adaptively calculating the residual correction coefficient based on the effective pore location distribution specifically includes: When the proportion of the number of effective wells to the total number of wells on the plate is greater than or equal to the preset threshold, the median fluorescence intensity of all effective wells is used as the target intensity to calculate the residual correction coefficient. When the proportion of the number of effective holes to the total number of holes on the board is less than the preset threshold, the hole-level calibration correction coefficient calibrated offline is used and the correction range is limited.

[0013] Furthermore, it also includes a background subtraction step: A preliminary mask for the aperture positions is generated based on the prior geometric parameters of the aperture plate. The blank background regions between the apertures are extracted. The median value of the pixels in each blank background region between the apertures is used as a node. The physical illumination model is used as a two-dimensional distribution constraint. An interpolation method is used to generate a pixel-level non-uniform background estimation map of the entire image. The sample image is subtracted from the background estimation map to obtain the fluorescence signal image after background subtraction.

[0014] Furthermore, the step of performing hole segmentation on the coarsely corrected image specifically includes: On the coarsely corrected image, the foreground region of the well site is extracted using Otsu threshold segmentation. Combined with the prior geometric constraints of the well plate rows and columns, the boundary of each well site is fitted with an ellipse or a circle to obtain the effective region of each well site. After excluding the edge region of the well site, the average fluorescence intensity within the effective region is calculated.

[0015] Furthermore, it also includes online time-drift dynamic compensation steps: No fewer than three reference well positions are preset in the well plate. The fluorescence intensity of the reference well positions is extracted each time the sample is imaged, and the time drift compensation factor is calculated by comparing it with the initial reference intensity calibrated offline. The physical illumination basic model and / or the well position level calibration correction coefficient are dynamically updated using the time drift compensation factor.

[0016] A second objective of the present invention is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0017] A third objective of the present invention is to provide a computer-readable storage medium having program instructions stored thereon, which, when executed, implement the method described above.

[0018] Compared with the prior art, the beneficial effects of the present invention are: Existing technologies only address illumination non-uniformity at a single scale (pixel level), failing to simultaneously resolve both global continuous gradients and localized discrepancies at the aperture location level. This invention constructs a closed-loop coupled framework of pixel-level coarse correction, precise segmentation, and aperture location-level fine correction: pixel-level improved two-dimensional gamma correction eliminates global illumination gradients characterized by strong light at the center and weak light around the edges; aperture location-level calibration correction compensates for localized deviations such as aperture wall occlusion, meniscus refraction, and edge effects. The two scales work in a positive cycle, with coarse correction improving segmentation accuracy and precise segmentation feeding back into fine correction. Compared to single-scale correction schemes, this significantly improves the uniformity of fluorescence intensity across all aperture locations and the accuracy of quantitative analysis. Experimental data shows that using this method significantly reduces the coefficient of variation (CV) of fluorescence intensity across all aperture locations, and significantly improves the comparability and accuracy of quantitative analysis for each aperture location.

[0019] Existing Retinex-based methods misinterpret strong fluorescence signals within wells as illumination components when estimating illumination across the entire image, leading to overestimation of illumination in these areas and decreased correction accuracy. This invention incorporates the physical calibration data of a standard fluorescence plate into the illumination estimation process. By using a well-position mask to shield fluorescence signal regions within the wells, only the blank background regions between the wells are used as the data source for illumination estimation, fundamentally eliminating the interference of fluorescence signals. Simultaneously, a weighted fusion of the offline-calibrated parameterized illumination model and multi-scale Retinex adaptive illumination estimation is performed. This ensures that the illumination component estimation incorporates both prior physical knowledge of the optical system and adaptive information from the current image, guaranteeing both physical accuracy and adaptability to different samples and imaging parameters.

[0020] Existing calibration methods do not consider the differentiated requirements of calibration strategies under different experimental scenarios. A single strategy is difficult to cover various application scenarios such as uniform concentration throughout the plate, different concentrations in each well, and some wells without samples. This invention automatically switches the calibration mode based on the effective well position ratio: in full-plate mode, it forces homogenization with the median value as the target, suitable for scenarios such as uniform concentration throughout the plate; in partial-fill mode, it uses a calibration coefficient and limits the calibration range, effectively preserving the true concentration differences between wells. The two modes adaptively switch without manual intervention, and are compatible with various experimental scenarios such as uniform concentration throughout the plate, different concentrations in each well, and some wells without samples, greatly improving the universality and practicality of the method.

[0021] This invention requires no modification to the optical hardware structure and can be directly upgraded or integrated into existing fluorescence plate imaging systems via software upgrades or algorithm integration. It supports various plate specifications, including 96-well plates, 384-well plates, and 8-tube arrays, and exhibits good compatibility with existing equipment. It boasts low deployment costs and broad application prospects.

[0022] This invention monitors changes in light intensity of the light source in real time by using preset reference aperture positions, dynamically updates the physical illumination model and correction coefficients, effectively compensates for system time-varying factors such as light source aging and temperature drift, eliminates the need for frequent full-board recalibration, and ensures the accuracy, stability and engineering reliability of the calibration method in long-term use.

[0023] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 Flowchart of a dual-scale coupling correction method for calibrating illumination non-uniformity in fluorescence plate imaging; Figure 2 This is a reference image of a standard fluorescent plate and a schematic diagram of automatic well position detection. Figure 3 A schematic diagram illustrating the construction of a calibration-guided hybrid illumination estimation model; Figure 4 This is a schematic diagram illustrating the process and effect of dual-scale coupling correction. Figure 5 A schematic diagram of an adaptive dual-mode aperture-level correction strategy; Figure 6 A bar chart comparing the uniformity of fluorescence intensity across all well sites using different correction schemes; Figure 7 A module architecture diagram of a dual-scale coupled correction algorithm for calibrating illumination non-uniformity in fluorescence plate imaging; Figure 8 This is a schematic diagram of a computer device. Figure 9 This is a schematic diagram of a computer-readable storage medium. Detailed Implementation

[0025] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0026] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0027] The drawing numbers in this application are only used to distinguish the steps in the scheme and are not used to limit the execution order of the steps. The specific execution order is as described in the specification.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0029] This invention proposes a calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging, such as... Figure 1 As shown, this method deeply integrates offline calibration and online correction stages, and solves the problem of uneven image illumination caused by point source excitation through dual-scale coupling of pixel-level global coarse correction and aperture-level fine correction. The specific scheme is as follows: Example 1 A calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence plate imaging includes the following steps: S100, Offline calibration stage: Acquire reference images of standard fluorescent plates, extract fluorescence intensity of each well, fit and construct a parameterized physical illumination model, and store well-level calibration correction coefficients; The offline calibration stage is fundamental to the entire calibration method. Its purpose is to establish a physical benchmark for the illumination distribution of the imaging system, providing global constraints and quantitative references for subsequent online calibration. This stage involves acquiring reference images using a standard fluorescence plate, constructing a parameterized physical illumination model, and storing the calibration correction coefficients for each well location.

[0030] Before acquiring reference images, a standard fluorescence plate is prepared. The standard fluorescence plate is a reference plate whose wells are filled with a standard fluorescent substance of the same concentration and high stability. The standard fluorescence plate is then placed in the target imaging system, and reference images are acquired under standard imaging parameters (including light source power, exposure time, gain, etc.). ,like Figure 2 As shown in (a) above. The standard imaging parameters should be consistent with those used in subsequent sample testing to ensure the validity of the calibration data.

[0031] Standard fluorescent materials should possess long-term photostability to reduce the impact of fluorescence decay on calibration accuracy during subsequent use. Standard fluorescent plates can be prepared in 96-well, 384-well, or 8-well tube configurations according to actual detection requirements.

[0032] After acquiring the reference image Next, the location of each well needs to be automatically detected from the image and the fluorescence intensity features extracted. The specific steps for automatic well location detection are as follows: First, background subtraction is performed on the reference image to enhance contrast. Since the fluorescence signal within the wells is strong and the inter-well areas are relatively dark in the reference image, background subtraction can highlight the difference between the well locations and the background. Morphological top-hat transformation or median filtering followed by differencing can be used for background subtraction. Since the fluorescence intensity at each well in the reference image is uniform and known, the background is relatively simple and can be effectively estimated using median filtering. In this embodiment, a large-size median filter is preferred to estimate the background. Subtracting the background image from the original image yields the contrast-enhanced reference image.

[0033] Secondly, Otsu thresholding combined with morphological operations is used to extract the aperture regions. The Otsu method can automatically calculate the optimal segmentation threshold based on the image grayscale histogram, binarizing the image into foreground (aperture regions) and background (inter-aperture regions). Since there may be noise or stray light interference in the perforated plate image, morphological opening operations are used after binarization to remove small noise areas, and then closing operations are used to fill the small holes in the aperture regions to obtain complete connected regions of the apertures.

[0034] Then, the detected well locations are assigned to a regular grid using a row and column sorting method. Since the well locations in the standard well plate are arranged in a regular row and column grid, the centroid coordinates of the connected regions of each well location are first clustered and sorted according to the vertical coordinate to determine the row number; then, they are clustered and sorted according to the horizontal coordinate to determine the column number. This method maps well locations detected in any order to their corresponding row and column grid positions (i,j), achieving an accurate correspondence between well locations and grid positions, as shown in the figure. Figure 2 As shown in (b) of the diagram.

[0035] Finally, the average fluorescence intensity within the effective region of each well site was calculated. After determining the regions of each aperture, to eliminate areas where the aperture edges are obscured by the aperture wall or interfered with by edge effects, morphological erosion can be performed on each aperture region to extract the effective area of ​​the aperture's center. The average grayscale value of all pixels within the effective area is then calculated as the reference fluorescence intensity for that aperture. The extraction results are presented in the form of a three-dimensional histogram, such as... Figure 2 As shown in (c), it can be clearly seen that the fluorescence intensity of the central pore is significantly higher than that of the edge pore.

[0036] After obtaining the reference fluorescence intensity for all well sites. After obtaining the corresponding center coordinates of the apertures, a continuous parametric illumination model is constructed using a surface fitting method. This model characterizes the global illumination distribution of an imaging system under standard imaging parameters, caused by the characteristics of a point source.

[0037] The specific fitting method is as follows: using the center coordinates of each well as the corresponding reference fluorescence intensity as the independent variable. For the observed values, global surface fitting is performed using the least squares method.

[0038] In this embodiment, a two-dimensional Gaussian model is preferably used for fitting, as this model can well describe the Gaussian distribution of light intensity under point light source illumination. Alternatively, a third-order polynomial model can be used for surface fitting to handle more complex illumination distribution patterns. In practical applications, a suitable fitting model can be selected based on the illumination distribution characteristics of the actual imaging system.

[0039] After completing the model fitting, store the following calibration baseline data for subsequent online calibration: physical illumination basic model parameters, including the model type used and the corresponding fitting parameters; reference intensity values ​​for each aperture location. ; and hole position calibration correction coefficient This coefficient is calculated based on the reference intensity at each hole location in the reference image and the uniform intensity of the target. Specifically, the median value of the reference intensity at all hole locations in the reference image is calculated. ,but This coefficient characterizes the degree of deviation of each hole location from the target uniform strength. The larger the value, the greater the reinforcement required for that hole location; the smaller the value, the greater the suppression required. In subsequent partial filling modes, it is used to limit the hole location-level correction amplitude.

[0040] S200, Pixel-level coarse correction steps in the online correction stage: In the image of the sample to be corrected, the fluorescence signal area inside the hole is shielded, and the illumination is estimated based on the background area between the holes; the physical illumination model is fused with the illumination estimate to obtain a mixed illumination estimate; adaptive gamma correction is performed based on the mixed illumination estimate to obtain a pixel-level coarse correction image; The online correction stage performs dual-scale coupled correction on the sample image to be corrected. Figure 7 This paper demonstrates the core modules and data flow relationships of the correction method of this invention. This stage mainly includes functional modules such as background estimation, pixel-level coarse correction, pore location segmentation, pore location fine correction, and time drift compensation. These modules are connected in series via data flow to form a complete correction chain: After background subtraction and pixel-level coarse correction, the sample image is input into the pore location segmentation module. The segmentation result and calibration parameters jointly drive the pore location fine correction module, ultimately outputting the corrected image. The three steps of coarse correction, segmentation, and fine correction form a positively coupled closed loop: coarse correction improves image uniformity, thereby increasing segmentation accuracy, while precise segmentation provides accurate pore location data for fine correction, and the two promote each other.

[0041] For sample images with strong background fluorescence or floor noise in the inter-well region, background subtraction can be performed first to reduce the interference of the background on illumination estimation and fluorescence quantification. The specific procedure for background subtraction is as follows: First, based on the prior geometric parameters of the perforated plate and the physical lighting fundamental model. The center coordinates are used to generate a preliminary mask for the borehole positions. Specifically, based on the known orifice plate specifications (96-hole plate, 384-hole plate, or 8-tube connector), geometric parameters such as the row and column spacing and borehole diameter are obtained, combined with... The provided center coordinates of the perforated plate are used to generate a preliminary position mask for each perforation, marking the area where each perforation is located in the image.

[0042] Secondly, based on the preliminary mask of the hole position, the blank background area between the holes is extracted. The pixels in the mask area are excluded, and the remaining pixels are the pixels of the blank background area between the holes.

[0043] Then, using the median pixel value of the blank background area between each hole as a node, and based on the physical lighting model... As a two-dimensional distribution constraint, an interpolation method is used to generate a pixel-level non-uniform background estimation map of the entire image. Specifically, the median value of the pixels in the blank background region between each hole is extracted as the discrete node value. Radial basis function interpolation or thin-plate spline interpolation methods are used, and... The distribution pattern is used as a regularization constraint for interpolation to generate a smooth full-image background estimation map. This constrained interpolation method can ensure that the background estimation map is faithful to the actual pixel values ​​in the inter-aperture region and follows the smooth transition trend of the illumination model in the intra-aperture region, avoiding distortion caused by over-interpolation.

[0044] Finally, the background estimation image is subtracted from the sample image to obtain the fluorescence signal image after background subtraction. : After background subtraction, the image mainly retains the fluorescence signal in the well location area and a small amount of residual noise, providing a cleaner data source for subsequent illumination estimation and fluorescence quantification. For sample images with weak background fluorescence and no background subtraction required, the original sample image can also be used directly as input for subsequent steps.

[0045] The pixel-level global coarse correction step is one of the core components of the entire correction method. Its purpose is to address the uneven global illumination gradient, where the excitation source is strong at the center and weak around the edges, through a calibration-guided improved two-dimensional gamma correction algorithm. This step eliminates interference from fluorescence signals by eliminating fluorescence signal interference within the shielding aperture and weightedly fuses the offline calibrated physical illumination model with the adaptive Retinex illumination estimation, thus enabling the illumination component estimation to possess both physical prior constraints and adaptive characteristics.

[0046] Illumination component estimation is performed on the sample image. The purpose of illumination estimation is to obtain the brightness distribution in the image caused by the illumination system so that it can be compensated for later by gamma correction.

[0047] To prevent strong fluorescence signals within the apertures from being misidentified as illumination components, an initial masking technique is used to shield the fluorescence signal region within the apertures during illumination estimation, retaining only the background region between the apertures as the data source for illumination estimation. Then, a multi-scale Gaussian filter is applied to perform an initial illumination estimate of the region between the apertures, which is then combined with a physical illumination model. Constrained interpolation is performed to complete the illumination components within the aperture region, thus obtaining the illumination estimate. The result is as follows Figure 3 As shown in (a) of the diagram. Multi-scale Gaussian filtering can estimate illumination components at different scales, taking into account both the macroscopic trend and local details of the illumination distribution. From Figure 3 As can be seen in (a), the estimated illumination value of the well site area is significantly higher due to the interference of fluorescence signal inside the well.

[0048] Physical lighting fundamental model Normalized and multi-scale Retinex illumination estimation Weighted fusion is performed to obtain the mixed illumination estimate. The result is as follows Figure 3 As shown in (c): in, The normalized value of the physical lighting fundamental model is... Mapped to the [0,1] interval. λ is the multi-scale Retinex illumination estimate, which is also normalized to the [0,1] interval. λ is the calibration weight coefficient, with a value range of 0≤λ≤1. Figure 3 (b) shows the calibration parametric lighting model. The results reflect the ideal light distribution pattern.

[0049] The calibration weighting coefficient λ is preset based on the stability of the imaging system. If the hardware parameters of the imaging system, such as the light source and optical components, are relatively stable, the value of λ can be appropriately increased to enhance the constraint of physical priors; if the imaging system exhibits significant time-varying characteristics or large differences between different batches, the value of λ can be appropriately decreased to enhance the response capability of adaptive illumination estimation. In this embodiment, for a stable laboratory imaging system, λ can be set to 0.5~0.7; for systems that require frequent movement or have significant light source attenuation, λ can be set to 0.3~0.5.

[0050] Hybrid illumination estimation combines the global accuracy of the physical calibration model with the scene flexibility of the adaptive algorithm. The physical calibration model is based on the fitting of measured data from standard fluorescence plates, possessing clear physical meaning and global accuracy; the adaptive Retinex illumination estimation is calculated based on the actual pixel values ​​of the current image, reflecting subtle changes under current imaging conditions. The illumination component estimate obtained by fusing the two is unaffected by fluorescence signals within the wells and is adaptable to different samples and imaging parameters. Figure 3 The profile comparison diagram shown in (d) further verifies the fusion effect. The profile of the mixed illumination estimation not only retains the smooth trend of the calibration model, but also reflects the adaptive information of the image in local details.

[0051] Based on mixed illumination estimation Calculate the pixel-by-pixel adaptive gamma coefficients and perform pixel-by-pixel gamma correction.

[0052] First, the 75th percentile of the mixed illumination estimate is used as the reference brightness value. The 75th quantile was chosen instead of the maximum or mean because it better represents the illumination level of brighter areas in the image, avoids interference from isolated bright spots or noise that might occur with the maximum value, and prevents over-correction due to overestimation of illumination levels. Through comparative verification with multiple sets of experimental data, the 75th quantile achieved the optimal balance between correction effect and algorithm stability; therefore, it is the preferred option in this embodiment.

[0053] The formula for calculating the pixel-by-pixel gamma coefficient is: The formula is designed based on the principle that when the estimated illumination value at a certain pixel location is... Approximately reference brightness value When the gamma coefficient is close to 0, the correction magnitude is moderate. When the estimated illumination value is lower than the reference value (i.e., the location is darker), the exponent term is positive (0 < γ < 1), and gamma correction will increase the brightness of the pixel. When the estimated illumination value is higher than the reference value (i.e., the location is brighter), the exponent term is negative (γ > 1), and gamma correction will suppress the brightness of the pixel. In this way, an adaptive correction effect of brightening dark areas and suppressing bright areas is achieved.

[0054] Then, the fluorescence signal image after background subtraction... Perform pixel-by-pixel gamma correction to obtain a pixel-level coarsely corrected image. .

[0055] For cases where background subtraction was not performed, the original sample image can be directly corrected. After gamma correction, the global brightness gradient caused by uneven light source distribution in the image is significantly compensated, and the overall uniformity of the image is significantly improved.

[0056] S300, Well site segmentation step in the online calibration stage: Perform well site segmentation on the coarse calibration image, extract each well site region and calculate the average fluorescence intensity of each well site; The precise segmentation of aperture positions in the coarse correction image The coarse-corrected image is segmented to provide accurate fluorescence intensity data for each well, enabling subsequent fine-tuning at the well level. Since the coarse-corrected image has already undergone illumination non-uniformity compensation, the contrast and uniformity of each well region are significantly improved, creating favorable conditions for precise segmentation.

[0057] First, in the coarsely corrected image The Otsu thresholding method is used to extract the foreground region of the hole. The Otsu method automatically calculates the optimal segmentation threshold based on the image's grayscale histogram, binarizing the image into foreground and background regions for the hole. After coarse correction, the contrast of each region in the image is enhanced, and Otsu thresholding can more accurately separate the hole region from the background.

[0058] Secondly, elliptical or circular fitting is performed on the boundary of each orifice using the prior geometric constraints of the orifice plate's rows and columns. Since the orifices in the standard orifice plate are arranged in a regular row and column grid with consistent diameters, this prior knowledge can be used to constrain and optimize the segmentation results. Specifically, connected component analysis is first performed on the binarized foreground region to obtain the contour point set of each candidate orifice region; then, based on the row and column prior, each contour is assigned to its corresponding grid position; finally, elliptical or circular fitting is performed on each contour to determine the precise location of the orifice boundary.

[0059] Then, the effective area mask for each hole is obtained. In order to eliminate the area of ​​the hole edge affected by local factors such as hole wall obstruction and meniscus refraction, morphological etching operation is performed on the fitted boundary of each hole, and only the central area of ​​the hole is retained as the effective area.

[0060] Finally, the average fluorescence intensity of all pixels within the effective area of ​​each aperture position is calculated. The fluorescence intensity value at that well location is used for subsequent fine calibration. Selecting an effective region can effectively reduce the impact of well edge inhomogeneity on the accuracy of quantitative fluorometry.

[0061] S400. Fine calibration steps at the pore level during the online calibration phase: Based on the average fluorescence intensity of each pore and the pore level calibration coefficient, the residual calibration coefficient is adaptively calculated according to the effective pore distribution, and each pore region in the coarse calibration image is calibrated to obtain the final calibrated image with dual scale coupling.

[0062] The aperture-level fine correction step is another core component of the entire correction method. While pixel-level coarse correction addresses the global illumination gradient problem, residual deviations may still exist between apertures due to local factors such as aperture wall occlusion, meniscus refraction, and edge effects. This step uses an adaptive dual-mode strategy to calculate the residual correction coefficients for each aperture location, performing multiplicative correction on each aperture region in the coarse-corrected image to achieve fine homogenization at the aperture level. Figure 4 (a) in the diagram illustrates the complete process of dual-scale coupling correction.

[0063] Before calculating the correction coefficient, the effective well sites in the sample image are first detected, i.e., the well sites where fluorescence signals are present. The method for determining the effective well sites is as follows: A well is considered valid when the average fluorescence intensity of the well exceeds 5% of the maximum fluorescence intensity of the entire plate. This threshold effectively distinguishes between valid wells containing fluorescent samples and empty wells (those without sample or with failed sample loading), preventing empty wells from being included in the correction calculation and affecting the accuracy of the statistical results.

[0064] The well-position level fine calibration employs an adaptive dual-mode strategy, automatically selecting the calibration mode based on the distribution of effective well positions. The design logic behind this strategy is that the fluorescence intensity distribution patterns at well positions differ under different experimental scenarios, necessitating differentiated calibration strategies.

[0065] Full-plate mode: When the proportion of effective wells to the total number of wells on the plate is greater than or equal to a preset threshold (80% in this embodiment), it indicates that most wells on the plate contain effective fluorescent samples. In this scenario, it is reasonable to assume that the fluorescence intensity of the samples in each well is theoretically consistent, such as the same concentration across the entire plate, or that the concentration distribution of each well has a certain global statistical regularity. Using the median fluorescence intensity of all effective wells as the target uniform intensity is reasonable. Calculate the residual correction coefficients: This coefficient uniformly corrects the fluorescence intensity of each well to the median level, achieving forced homogenization of all wells across the plate. Compared to the mean, the median is more robust, less susceptible to extreme values, and better reflects the central trend of the dataset. Figure 5 Image (a) shows a full-plate sample image of 96 wells (all wells contain samples of the same concentration). Figure 5 (b) in the image shows the image where the intensity of all holes tends to be consistent after correction.

[0066] Partial Fill Mode: When the proportion of effective wells to the total number of wells in the plate is less than a preset threshold (e.g., 80%), it indicates that only some wells contain effective fluorescent samples, such as experimental designs using only some wells or failure to add samples to some wells. In this scenario, there may be inherent concentration differences between the effective wells. Forcibly using the median value as a target for homogenization will obscure the true concentration differences, leading to erroneous quantitative results. In this case, an offline-calibrated well-level calibration correction factor should be used. And limit the range of correction amplitude: This amplitude limiting operation is based on the following considerations: local deviations between apertures caused by factors such as aperture wall obstruction and meniscus refraction of the liquid surface typically do not exceed ±50%. Limiting the correction amplitude within this range can effectively compensate for local non-uniformity without excessively altering the original concentration differences between apertures. The correction amplitude range can be adjusted according to the aperture-level deviation characteristics of the actual imaging system; the range of 0.5 to 2.0 times given in this embodiment is a preferred range. Figure 5 Image (c) shows the sample image of 3 rows of 8 tubes (only 24 wells have samples). Figure 5 (d) in the image shows the original concentration difference of the effective pore sites after correction.

[0067] By combining the pixel-level gamma correction results with the aperture-level residual correction coefficients, multiplicative correction is performed only in the effective aperture region: ,∀ ∈ Hole Position Region The pixel-level coarse correction and the aperture-position-level fine correction are coupled through a process of coarse correction, precise segmentation, and fine correction: coarse correction provides high-quality images for segmentation, precise segmentation provides aperture intensity data for fine correction, and fine correction further eliminates residual aperture-position biases. Final corrected image. It not only eliminates the gradual change in global illumination but also compensates for local deviations at the well level, significantly improving the uniformity and quantitative accuracy of fluorescence intensity across the entire plate. Figure 4 (b) in the image is the original image. Figure 4 (c) in the figure represents the pixel-level correction result only. Figure 4 (d) in the figure represents the final result after dual-scale coupling correction according to the present invention. Figure 6 The bar chart shows the comparison of fluorescence intensity uniformity across all well sites using different correction schemes. The CV value of the scheme in this invention is significantly lower than that of existing schemes.

[0068] During use, the light source power of the imaging system may drift slowly due to factors such as aging and temperature changes, causing a deviation between the offline calibrated illumination baseline model and the current actual illumination distribution. To maintain the long-term stability of calibration accuracy, reference aperture positions can be preset in the aperture plate to achieve dynamic compensation for time drift.

[0069] Specifically, at least three reference wells are pre-set in the well plate, and these wells are filled with highly stable standard fluorescent material. Setting multiple reference wells provides redundant observation and reduces measurement deviations caused by accidental factors at a single well. The distribution of the reference wells in the well plate should cover multiple areas from the center to the edge as much as possible to simultaneously monitor illumination changes at different locations.

[0070] During each sample imaging, the fluorescence intensity of the reference well sites is extracted simultaneously. The extracted fluorescence intensity of the reference well sites is compared with the initial reference intensity recorded during offline calibration to calculate the time drift compensation factor α. The median ratio of the intensities of multiple reference well sites is used as the time drift compensation factor.

[0071] Using the time-drift compensation factor α to apply the physical illumination fundamental model The calibration correction coefficients for hole position level are dynamically updated.

[0072] By using time-drift dynamic compensation, the accuracy of the calibration model can be maintained without frequent full-plate recalibration, thus improving the engineering practicality and long-term stability of the method.

[0073] The technical solution of this invention is applicable to various specifications of fluorescence well plate imaging systems, such as 96-well plates, 384-well plates, and 8-tube arrays. For different specifications of well plates, only the geometric parameters such as the number of rows and columns and the aperture in the well position detection algorithm need to be adjusted, while the core calibration process remains unchanged.

[0074] For different application scenarios, such as the same concentration across the entire plate, different concentrations in each well, or no sample in some wells, the adaptive dual-mode strategy can automatically match the appropriate calibration mode without manual intervention and has broad scenario compatibility.

[0075] Example 2 A computer device 500, such as Figure 8 As shown, the system includes a memory 510, a processor 520, and a computer program 530 stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging. For a detailed description of the method, please refer to the corresponding description in the above method embodiments; it will not be repeated here.

[0076] Example 3 A computer-readable storage medium, such as Figure 9 As shown, a computer program is stored thereon. When executed by a processor, the computer program implements the steps of a calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here.

[0077] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0078] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

[0079] The apparatus, computer device, and non-volatile computer storage medium and method provided in the embodiments of this specification are corresponding. Therefore, the apparatus, computer device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, computer device, and non-volatile computer storage medium will not be repeated here.

[0080] Besides implementing the controller using purely computer-readable program code, the same functionality can be achieved by logically programming the method steps, making the controller function as logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices within it that implement various functions can also be considered structures within that hardware component. Alternatively, the devices used to implement various functions can be viewed as both software units implementing the method and structures within a hardware component.

[0081] The systems, apparatuses, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above apparatuses are described separately as various units based on their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0082] Those skilled in the art will understand that the embodiments of this specification can be provided as methods or computer program products. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0087] This specification may be described in the general context of computer-executable instructions, such as program units, that are executed by a computer. Generally, program units include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification may also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program units may reside in local and remote computer storage media, including storage devices.

[0088] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0089] The above description is merely an embodiment of this specification and is not intended to limit the scope of one or more embodiments of this specification. Various modifications and variations can be made to one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of one or more embodiments of this specification.

Claims

1. A calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging, characterized in that, Includes the following steps: Acquire reference images of standard fluorescent plates, extract the fluorescence intensity of each well, fit and construct a parameterized physical illumination model, and store the well-level calibration correction coefficients; For the image of the sample to be calibrated, the fluorescence signal region within the well is shielded, and illumination estimation is performed based on the background region between the wells; the physical illumination model is fused with the illumination estimation to obtain a hybrid illumination estimation; adaptive gamma correction is performed based on the hybrid illumination estimation to obtain a pixel-level coarsely calibrated image; Well segmentation is performed on the coarsely corrected image to extract the region of each well and calculate the average fluorescence intensity of each well. Based on the average fluorescence intensity of each well site and the well site-level calibration correction coefficient, the residual correction coefficient is adaptively calculated according to the effective well site distribution, and the well site regions in the coarse correction image are corrected to obtain the final corrected image with dual-scale coupling.

2. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, The physical lighting model is constructed as follows: Using the center coordinates of each well location as the independent variable and the fluorescence intensity of each well location as the observed value, global surface fitting is performed using the least squares method. The fitting model adopts a two-dimensional Gaussian model or a third-order polynomial model.

3. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, The illumination estimation employs the multi-scale Retinex method; The fusion of the physical illumination model and the illumination estimate is a weighted fusion, and the fusion formula is as follows: in, This is the normalized value of the physical lighting fundamental model. λ is the estimated illumination value, and λ is the calibration weighting coefficient, 0≤λ≤1.

4. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, In the adaptive gamma correction step based on the mixed illumination estimation, the gamma coefficient is determined pixel-by-pixel according to the relative relationship between the mixed illumination estimation and the reference brightness value, where the reference brightness value is the 75th percentile of the mixed illumination estimation. The formula for calculating the gamma coefficient is: in, The 75th percentile of the estimated mixed illumination.

5. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, The step of adaptively calculating the residual correction coefficient based on the effective pore location distribution specifically includes: When the proportion of the number of effective wells to the total number of wells on the plate is greater than or equal to the preset threshold, the median fluorescence intensity of all effective wells is used as the target intensity to calculate the residual correction coefficient. When the proportion of the number of effective holes to the total number of holes on the board is less than the preset threshold, the hole-level calibration correction coefficient calibrated offline is used and the correction range is limited.

6. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, It also includes a background subtraction step: A preliminary mask for the aperture positions is generated based on the prior geometric parameters of the aperture plate. The blank background regions between the apertures are extracted. The median value of the pixels in each blank background region between the apertures is used as a node. The physical illumination model is used as a two-dimensional distribution constraint. An interpolation method is used to generate a pixel-level non-uniform background estimation map of the entire image. The sample image is subtracted from the background estimation map to obtain the fluorescence signal image after background subtraction.

7. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, The step of performing hole segmentation on the coarsely corrected image specifically includes: On the coarsely corrected image, the foreground region of the well site is extracted using Otsu threshold segmentation. Combined with the prior geometric constraints of the well plate rows and columns, the boundary of each well site is fitted with an ellipse or a circle to obtain the effective region of each well site. After excluding the edge region of the well site, the average fluorescence intensity within the effective region is calculated.

8. The calibration-guided dual-scale coupling correction method for illumination non-uniformity in fluorescence aperture imaging as described in claim 1, characterized in that, It also includes online time-drift dynamic compensation steps: No fewer than three reference well positions are preset in the well plate. The fluorescence intensity of the reference well positions is extracted each time the sample is imaged, and the time drift compensation factor is calculated by comparing it with the initial reference intensity calibrated offline. The physical illumination basic model and / or the well position level calibration correction coefficient are dynamically updated using the time drift compensation factor.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, It stores program instructions that, when executed, implement the method as described in any one of claims 1 to 8.

Images