Pathology slide scanning imaging prescription map generation method, device, and storage medium

Abstract

The application discloses a pathological section scanning imaging prescription chart generation method, equipment and a storage medium, and relates to the technical field of image scanning. The pathological section scanning imaging prescription chart generation method comprises the following steps: at least two optical response channels are used to respectively illuminate pathological section samples, preview response images of the optical response channels are collected, the preview response images of the optical response channels are combined through registration, and a multi-channel preview response image is obtained; the multi-channel preview response image is divided into multiple image blocks, the optical characteristics of each image block are calculated, the optical characteristics are normalized, the normalized characteristics are combined, and a characteristic vector is obtained; the characteristic vector is input into a calibration matrix established in advance, and main scanning parameters are obtained; and the main scanning parameters are filled into a parameter layer aligned with a preview image space, and an imaging prescription chart is obtained. The application can achieve the technical effect of reducing parameter prediction deviation.

Description

Technical Field

[0001] This application relates to the field of image scanning technology, and in particular to a method, device and storage medium for generating prescription maps of pathological slide scanning imaging. Background Technology

[0002] Currently, digital pathology slide scanning technology is a key technology for digitizing traditional glass slides into high-resolution whole-slide images. In related technologies, before the formal high-magnification scanning of pathology slides, low-magnification pre-scans or pre-scan images are typically used to identify tissue regions and determine the scanning range. Based on this, focus maps are generated or parameters such as exposure time, gain, and white balance are estimated. These parameters are calculated and processed separately by different modules. The preview results are used for tissue localization or single-parameter estimation, leading to a disconnect between the preview and the main scan, and consequently, large deviations in parameter prediction. Summary of the Invention

[0003] The main objective of this application is to provide a method, device, and storage medium for generating prescription maps of pathological slide scanning imaging, aiming to solve the technical problem of large deviations in parameter prediction caused by the disconnect between the preview and the main scan.

[0004] To achieve the above objectives, this application provides a method for generating prescription maps of pathological slide scanning images. The method includes: At least two optical response channels are used to illuminate the pathological slide sample, and preview response images of each optical response channel are acquired. The preview response images of each optical response channel are then registered and combined to obtain a multi-channel preview response image. The multi-channel preview response image is divided into multiple image blocks, the optical features of each image block are calculated, the optical features are normalized, and the normalized features are combined to obtain a feature vector. The feature vectors are input into a pre-established calibration matrix to obtain the main scan parameters; The main scan parameters are filled into a parameter layer aligned with the multi-channel preview response image space to obtain an imaging prescription map.

[0005] In one embodiment, before acquiring preview response images of each optical response channel, the method further includes: Acquire dark field and white field images. The dark field image is the dark current image of each channel under no illumination, and the white field image is the response image of each channel acquired under uniform illumination without samples. After acquiring preview response images for each optical response channel, the process also includes: The preview response image is corrected based on the dark field image and the white field image to obtain the corrected preview response image.

[0006] In one embodiment, before inputting the feature vector into a pre-established calibration matrix to obtain the main scan parameters, the method further includes: Prepare multiple sets of pathological slide samples covering different staining types, tissue types, and staining intensities; For each group of pathological slide samples, obtain multi-channel preview response images and feature vectors; Perform a main scan on each group of pathological slide samples to obtain the target main scan parameters; The feature vectors are paired with the target main scan parameters to obtain the training sample set; The calibration matrix is ​​obtained by training a multivariate linear regression or neural network based on the training sample set.

[0007] In one embodiment, the master scan parameters are filled into a parameter layer aligned with the multi-channel preview response image space to obtain an imaging prescription map, including: The main scan parameters are filled into a parameter layer aligned with the multi-channel preview response image space to obtain the filled prescription map; The overlapping areas of the image blocks in the filled prescription map are weighted, fused, and smoothed to obtain the imaging prescription map.

[0008] In one embodiment, after filling the main scanning parameters into a parameter layer aligned with the multi-channel preview response image space to obtain the imaging prescription map, the method further includes: Based on the parameter similarity of the main scanning parameters at each location in the imaging prescription map, a clustering algorithm is used to divide the scanning area into multiple sub-regions; Plan the scanning path within each sub-region to obtain the path point sequence; The path point sequence is combined with the main scan parameters of the sub-region to generate a time sequence queue; Hardware control instructions are generated based on the timing queue and sent to the hardware module for control.

[0009] In one embodiment, the method for generating prescription maps from pathological slide scanning images further includes: After the formal scan begins, multiple preset sentinel areas are scanned, and the actual imaging results of each sentinel area are compared with the predicted results in the imaging prescription map to calculate the brightness deviation, saturation ratio deviation, sharpness deviation and focus deviation. When any deviation exceeds the corresponding preset parameter threshold, the prescription parameters of the subsequent scanned area are corrected according to the direction and magnitude of the deviation to obtain the corrected prescription parameters. Continue scanning the sentinel region based on the corrected prescription parameters, and continuously perform sentinel region scanning and prescription parameter correction; Each time a preset percentage of the area is scanned, a new sentry zone is set.

[0010] In one embodiment, when any deviation exceeds a corresponding threshold, the prescription parameters for the subsequent scanned area are corrected according to the deviation direction and magnitude to obtain corrected prescription parameters, including: When the number of sentinel regions whose deviations in the target parameters are consistent and whose deviations exceed the preset parameter threshold is greater than the first preset threshold, the prescription parameters of the subsequent regions to be scanned are globally corrected to obtain the corrected prescription parameters. When the number of sentinel regions whose deviation from the target parameter exceeds the preset parameter threshold is less than the second preset threshold, and the sentinel region appears in the target region, the prescription parameters of the target region are locally corrected to obtain the corrected prescription parameters. The target region is determined by the center coordinates of the deviated sentinel region and the Gaussian influence radius. When a systematic deviation in the same direction occurs, the calibration matrix is ​​updated, and the prescription parameters for the subsequent scanned areas are regenerated based on the updated calibration matrix. Among them, global correction, local correction and calibration matrix update can be selected and used in combination, and the first preset threshold is greater than the second preset threshold.

[0011] In one embodiment, before inputting the feature vector into a pre-established calibration matrix to obtain the main scan parameters, the method further includes: Obtain the staining type of the slide to be scanned, and select different calibration matrices according to the staining type.

[0012] In addition, to achieve the above objectives, this application also provides a pathological slide scanning imaging prescription map generation device, which includes: a stage, a beam-splitting oblique illumination module, a preview imaging module, an imaging-side beam-splitting module, a light intensity monitoring module, a processor, a main scanning controller, a memory, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the above-mentioned pathological slide scanning imaging prescription map generation method.

[0013] In addition, to achieve the above objectives, this application also provides a storage medium, which is a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the above-described method for generating prescription maps for pathological slide scanning imaging.

[0014] This application provides a method for generating prescription maps for pathological slide scanning imaging. The method involves illuminating the pathological slide sample using at least two optical response channels, acquiring preview response images for each channel, registering and combining these images to obtain a multi-channel preview response image. This multi-channel preview response image is then divided into multiple image blocks, and the optical features of each block are calculated. These optical features are normalized and combined to obtain a feature vector. The feature vector is input into a pre-established calibration matrix to obtain the main scanning parameters. Finally, the main scanning parameters are filled into a parameter layer aligned with the preview image space to obtain the imaging prescription map. By acquiring comprehensive optical information through multi-channel registration and combination, and establishing a mapping relationship between the preview and main scan based on image block-level features, the method can eliminate systematic biases caused by optical path differences and spatial misalignment, achieving collaborative parameter prediction and thus reducing parameter prediction bias. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating an embodiment of the method for generating prescription maps from pathological slide scanning imaging in this application. Figure 2 This is a schematic diagram of the imaging prescription map data structure provided in Embodiment 1 of the method for generating pathological slide scanning imaging prescription maps in this application; Figure 3 A simplified flowchart illustrating the method for generating prescription maps from pathological slide scanning imaging provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of the beam-splitting oblique illumination preview optical system provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the pathological slide scanning imaging prescription map generation device in the embodiments of this application.

[0018] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of this application and are not intended to limit this application.

[0020] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0021] Currently, digital pathology slide scanning technology is a key technology for digitizing traditional glass slides into high-resolution whole-slide images. In related technologies, before the formal high-magnification scanning of pathology slides, low-magnification pre-scans or pre-scan images are typically used to identify tissue regions and determine the scanning range. Based on this, focus maps are generated or parameters such as exposure time, gain, and white balance are estimated. These parameters are calculated and processed separately by different modules. The preview results are used for tissue localization or single-parameter estimation, leading to a disconnect between the preview and the main scan, and consequently, large deviations in parameter prediction.

[0022] The main solution of this application is as follows: Illuminate the pathological slide sample using at least two optical response channels, acquire preview response images for each optical response channel, register and combine these images to obtain a multi-channel preview response image; divide the multi-channel preview response image into multiple image blocks, calculate the optical features of each image block, normalize the optical features, and combine the normalized features to obtain a feature vector; input the feature vector into a pre-established calibration matrix to obtain the main scanning parameters; fill the main scanning parameters into a parameter layer aligned with the preview image space to obtain an imaging prescription map. By acquiring comprehensive optical information through multi-channel registration and combination, and establishing a mapping relationship between the preview and main scan based on image block-level features, systematic biases caused by optical path differences and spatial misalignment can be eliminated, achieving collaborative parameter prediction and thus reducing parameter prediction bias.

[0023] It should be noted that the executing entity in this embodiment can be a pathological slide scanning imaging prescription map generation device, a computing service device with data processing, network communication, and program execution functions, or a pathological slide scanning imaging prescription map generation device capable of performing the above functions, etc. This embodiment does not specifically limit it in this way. The following uses a pathological slide scanning imaging prescription map generation device as the executing entity to describe this embodiment and the following embodiments.

[0024] Based on this, Embodiment 1 of this application proposes a method for generating prescription maps of pathological slide scanning imaging. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the method for generating prescription maps from pathological slide scanning imaging according to this application. The method includes steps S10 to S40: Step S10: Illuminate the pathological slide sample using at least two optical response channels, acquire preview response images of each optical response channel, and register and combine the preview response images of each optical response channel to obtain a multi-channel preview response image.

[0025] In this embodiment, an optical response channel refers to an optical acquisition path composed of a specific wavelength light source, polarization state, and detection method. A preview response image refers to a single-channel raw image acquired for a specific optical response channel, reflecting the optical characteristics of the preview area. A multi-channel preview response image refers to a structured multidimensional dataset containing response values ​​from multiple channels, formed by spatially registering and combining the preview response images of each optical response channel.

[0026] Specifically, before illuminating and acquiring the optical response channels, the illumination intensity, exposure time, camera gain, and polarization state of each channel are set. The illumination intensity is controlled by the LED drive current, ranging from 10 to 500 mA with a step of ≤1 mA. The exposure time ranges from 100 μs to 100 ms, preferably from 1 ms to 50 ms with a step of ≤10 μs. The camera gain ranges from 0 to 30 dB with a step of ≤0.1 dB. The polarization state is defined as follows: the polarizer is fixed, and the analyzer can be rotated from 0 to 90 degrees with an angular accuracy of ≤1 degree.

[0027] During the illumination process according to the preset channel sequence, 2%-10% of the light reflected by the beam splitter enters the photodiode, which records the actual illumination intensity of each wavelength in real time. The processor uses this light intensity data to normalize the original preview image to eliminate illumination intensity drift caused by light source aging or temperature fluctuations. Subsequently, the pathological slide sample generates absorption, transmission, reflection, and scattering responses to obliquely incident light. The response light enters the preview objective upwards, and the imaging side beam splitting module set behind the preview objective decomposes the image of the same field of view into multiple optical channels according to wavelength or polarization, acquiring multi-channel preview response images. The processor receives the preview images of each channel and the light intensity values ​​measured by the photodiode, performs normalization processing, and obtains multi-channel preview response data for subsequent generation of the main scanning imaging prescription map, providing high-quality multi-channel preview response images for subsequent feature extraction and prescription map generation.

[0028] Preferably, the optical response channels include a short-wave absorption channel, a mid-wave transmission channel, a long-wave transmission channel, a near-infrared transmission or scattering channel, a reflection / glare channel, and a polarization difference channel. The short-wave absorption channel has a wavelength λ1 = 430-480 nm and is mainly used to detect deeply stained areas and cell nuclei. The mid-wave transmission channel has λ2 = 500-560 nm and is mainly used for conventional HE-stained tissue imaging. The long-wave transmission channel has λ3 = 600-680 nm and is mainly used for lightly stained areas and thick tissue areas. The near-infrared transmission or scattering channel has λ4 = 700-900 nm and is mainly used for tissue thickness detection and scattering characteristic analysis. The reflection / glare channel, through an orthogonal configuration of the analyzer and polarizer, detects tissue surface reflection and coverslip glare. The polarization difference channel acquires images using both parallel and orthogonal configurations of the analyzer and polarizer, and calculates the polarization difference image I. diff =Iparallel -I perpendicular It is used to strip away surface reflected light and enhance tissue texture contrast, where I parallel The parallel polarization response intensity, i.e., the intensity of the response image of this optical channel acquired by the preview imaging module when the polarizer and analyzer are configured with their polarization directions parallel, is I. perpendicular The orthogonal polarization response intensity refers to the intensity of the response image of the optical channel acquired by the preview imaging module when the polarizer and polarizer are configured with their polarization directions orthogonal.

[0029] As an optional implementation, the pathological slide sample is illuminated by a short-wave absorption channel, a medium-wave transmission channel, a long-wave transmission channel, a near-infrared transmission or scattering channel, and a polarization difference channel, respectively. Preview response images of each optical response channel are acquired, and the preview response images of each optical response channel are registered and combined to obtain a multi-channel preview response image.

[0030] Specifically, a short-wave absorption channel (λ1=430-480nm), a mid-wave transmission channel (λ2=500-560nm), a long-wave transmission channel (λ3=600-680nm), a near-infrared transmission or scattering channel (λ4=700-900nm), and a polarization difference channel are employed. LEDs of the corresponding bands are illuminated sequentially according to a preset channel order, and their polarization states are adjusted. Preview response images for each channel are acquired. Spatial registration is then performed using feature-point-based image registration algorithms, such as SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features), or ORB (Oriented FAST and Rotated BRIEF), with any band channel as a reference frame. The registered channel images are then combined according to channel dimensions to obtain a multi-channel preview response image encompassing short-wave absorption, mid-wave transmission, long-wave transmission, near-infrared scattering, and polarization difference responses. It can comprehensively acquire the response information of tissue regions in multiple optical dimensions such as absorption, transmission, scattering and polarization difference, providing a rich data foundation for subsequent extraction of optical features and generation of high-precision master scan imaging prescription maps.

[0031] As another optional implementation, the pathological slide sample is illuminated by the transmission channel and the reflection / glare channel respectively, and the preview response images of each optical response channel are acquired. The preview response images of each optical response channel are then registered and combined to obtain a multi-channel preview response image.

[0032] Specifically, the λ2=500-560nm band was used as the main illumination band. Images of the reflection / glare channel were acquired using an orthogonal configuration of the analyzer and polarizer. Simultaneously, transmission channel images of this band were acquired via transmission detection, and thickness scattering channel images were acquired using the near-infrared λ4=700-900nm band. LEDs in each band were sequentially illuminated according to a preset channel order, and their polarization states were switched. Preview response images for each channel were acquired, and the transmission, reflection / glare, and near-infrared scattering channel images under the same preview coordinates were spatially registered using a feature-point-based image registration algorithm. This resulted in a multi-channel preview response image encompassing transmission response, surface reflection glare response, and thickness scattering response. With a low-hardware configuration using only two optical channels, the transmission optical response and surface reflection glare information of the tissue region were simultaneously acquired, providing the necessary data foundation for subsequent calculations of optical characteristics.

[0033] In addition, the controller first reads the current position of the stage and establishes the correspondence between the preview image coordinates and the physical coordinates of the stage; the stage can move along the entire preview path or along the preview path of the tissue candidate area; the preview acquisition area can cover the entire slide, and the standard slide preview range can be set from 20mm*40mm to 26mm*76mm. The preview acquisition area can also cover the tissue-bearing area in the slide. During local preview, a single preview field of view can cover from 1mm*1mm to 30mm*30mm; the spatial resolution of the preview acquisition can be set from 1μm / pixel to 100μm / pixel, preferably from 5μm / pixel to 30μm / pixel; to facilitate the subsequent conversion of the preview response into the main scanning parameters, the system records the physical coordinates of the stage, the coordinates of the preview image, the acquisition timestamp, the illumination intensity of each channel, the exposure time, the camera gain, and the ambient temperature in each preview field of view.

[0034] Step S20: Divide the multi-channel preview response image into multiple image blocks, calculate the optical features of each image block, normalize the optical features, and combine the normalized features to obtain a feature vector.

[0035] In this embodiment, optical features refer to a set of indicators calculated from the multi-channel preview response data of an image patch to quantify the optical properties of that region, including optical density (OD), channel absorption ratio (CAR), reflection risk index (RRI), dynamic range occupancy (DRO), local texture response, tint intensity index, and local sharpness index. Feature vectors refer to multi-dimensional vectors formed by normalizing and combining the various optical features of a single image patch.

[0036] As one implementation method, the multi-channel preview response image is divided into multiple image blocks, and the optical density, channel absorption ratio, reflection risk index, dynamic range occupancy, local texture response (LTR), staining depth index (SDI), and local sharpness index (LSI) of each image block are calculated. The above optical features are normalized and the normalized features are combined to obtain a feature vector.

[0037] Specifically, the multi-channel preview response image is divided into image blocks of 64*64 to 512*512 pixels in size, with a block overlap rate of 0%-50%, and a square or rectangular shape. The image blocks can be adaptively divided according to tissue shape; preferably, the image block size is 128*128 to 256*256 pixels, with an overlap rate of 20%-30%. For each image block, optical density, channel absorptivity, reflection risk index, dynamic range occupancy, local texture response, staining depth index, and local sharpness index are calculated; where optical density OD(λ) = log10(I trans (λ) / I incident (λ)), where I trans (λ) represents the transmitted light intensity, I incident (λ) represents the incident light intensity, and optical density reflects the degree of absorption of light in this wavelength band by the tissue; Channel Absorption Ratio (CAR) = OD(λ1) / OD(λ3), where OD(λ1) is the short-wavelength optical density and OD(λ3) is the long-wavelength optical density. The short-wavelength to long-wavelength absorption ratio reflects the staining depth. CAR > 2 indicates a dark-stained area, and CAR < 0.5 indicates a light-stained area; Reflectance Risk Index (RRI) = I reflect / (I trans +I reflect ), where I reflect I represents the intensity of the reflection channel. trans For transmission channel intensity, RRI > 0.3 indicates a high-reflection-risk area; Dynamic Range Occupancy (DRO) = (I max I min ) / I fullscale , among which, I max and I min These represent the maximum and minimum strengths within the block, I. fullscaleFor full-scale cameras, a DRO > 0.8 indicates a high dynamic range region, requiring segmented exposure. Local texture response is calculated using a Gray-Level Co-occurrence Matrix (GLCM) or a Local Binary Pattern (LBP), with the local texture response LTR = Σij(i j) 2 *P(i,j), where P(i,j) is a GLCM matrix element. A larger LTR value indicates a more complex texture, requiring higher focus density. The Staining Depth Index (SDI) is calculated as: SDI = α*OD(λ1) + β*OD(λ2) + γ*OD(λ3), where α, β, and γ are weighting coefficients determined by the staining type. For example, for HE staining, α = 0.5, β = 0.3, and γ = 0.2. OD(λ1) represents the optical density at short wavelength λ1, OD(λ2) at medium wavelength λ2, and OD(λ3) at long wavelength λ3. The Local Sharpness Index (LSI) is calculated using the Laplacian operator or the Tenengrad function (Tenengrad evaluation function), LSI = Σxy(Gx 2 +Gy 2 In the LSI vector V, Gx and Gy represent the gradients in the x and y directions, respectively. A lower LSI value indicates a region is more likely to be out of focus. Each feature is normalized to a value between [0,1]. The normalized features are then combined into a feature vector V. feature =[OD norm CAR norm ,RRI norm DRO norm LTR norm SDI norm LSI norm ] T The feature vector is obtained. By converting the multi-channel preview response data into image block-level feature representations, a unified and standardized input basis is provided for subsequent conversion of the feature vectors into the master scanning parameters of each image block through the calibration matrix and the generation of high-quality imaging prescription maps.

[0038] Step S30: Input the feature vector into the pre-established calibration matrix to obtain the main scanning parameters.

[0039] In this embodiment, the calibration matrix refers to a mathematical model, such as a linear matrix or a neural network, that maps image patch feature vectors to main scan parameters. The main scan parameters refer to the set of parameters output by the calibration matrix for each image patch, used to drive the main scan hardware execution. The main scan parameters include the exposure time T. exp Camera gain G cam Illumination Intensity led White balance coefficient WB rWB g WB b Initial focus height Z focus Take the coke density D focus and scanning speed V scan The exposure time T exp The range is 0.1ms-100ms, preferably 1ms-50ms; camera gain G cam The range is 0-30 dB, preferably 0-20 dB; Illumination intensity I led The range is 10-500mA, preferably 50-300mA; white balance coefficient WB r WB g WB b The range is 0.5-2.0, and after normalization, it satisfies WB. r +WB g +WB b =3; Initial focus height Z focus The range is -50μm to +50μm, relative to the preview focal plane; the focal density D is taken as... focus The range is 1-20 points / mm², preferably 2-10 points / mm²; scanning speed V scan The range is 0.1-10 mm / s, preferably 0.5-5 mm / s.

[0040] As an optional implementation, the feature vector is input into a pre-established linear calibration matrix to obtain the main scan parameters.

[0041] Specifically, a linear calibration matrix P is used. main =M calib *V feature +b, where M calib b is a 9*7 calibration matrix, b is a 9*1 bias vector, and V feature P is a 7-dimensional feature vector. main The main scan parameter vector is 9-dimensional; the feature vector V of each image block is... feature Substituting into the linear formula, the corresponding 9-dimensional main scanning parameter vector is obtained. This pre-established linear calibration matrix is ​​obtained by training a multivariate linear regression algorithm based on the training sample set. It has the advantages of low computational cost and applicability to simple scenarios with relatively simple optical characteristics of the samples. It can quickly convert feature vectors into main scanning parameters for each image patch.

[0042] As another optional implementation, the feature vector is input into a pre-established nonlinear calibration matrix to obtain the main scanning parameters.

[0043] Specifically, a three-layer fully connected neural network is used, where the input layer of the three-layer fully connected neural network has 7 neurons, corresponding to a 7-dimensional feature vector V. featureThe hidden layers consist of 32 and 16 neurons, both using ReLU (Rectified Linear Unit) activation functions. The output layer has 9 neurons, corresponding to a 9-dimensional main scan parameter vector P. main The activation function is either Sigmoid or Linear; the loss function is mean squared error or weighted MSE; the optimizer is Adam (Adaptive Moment Estimation) with a learning rate of 0.001-0.01; the training epochs are 100-1000, with an early stopping patience of 20. This means that if the loss function on the validation set does not decrease further after 20 consecutive training epochs during the training of the neural network calibration matrix, training is terminated early to prevent overfitting. The feature vector V of each image patch is... feature The input is given to the neural network for forward propagation, and the 9-dimensional main scan parameter vector P is calculated layer by layer. main The pre-established neural network calibration matrix is ​​trained on the training sample set using the backpropagation algorithm. It has a strong nonlinear fitting capability and is suitable for complex scenarios across staining types and tissue types. It can accurately establish complex mapping relationships from preview response features to main scan parameters.

[0044] Furthermore, the main scanning parameters output by the calibration matrix must satisfy constraints on exposure time and gain, illumination intensity and exposure time, and scanning speed and focus density. The exposure time and gain constraint, specifically T... exp *10 (Gcam / 20) ≤T max *10 (Gmax / 20) To avoid overexposure, T exp Indicates exposure time, G cam T represents camera gain. max and G max These represent the maximum permissible exposure time and the maximum permissible gain, respectively; I represents the illumination intensity and exposure time constraints. led *T exp ≥E min To ensure a sufficient signal-to-noise ratio, where I led E represents lighting intensity. min This represents the minimum exposure threshold; the scanning speed and focus density are constrained by V. scan / D focus ≤T focus V scan D represents the scan speed. focusT represents the focal density. focus This indicates the maximum allowable time for a single focus attempt to ensure sufficient focusing time. For parameters that do not meet the above constraints, the Lagrange multiplier method or the projection gradient method are used for optimization and adjustment to ensure that the parameters fall within the feasible region, thus ensuring that the theoretical output of the calibration matrix can be transformed into executable master scan parameters that meet physical and hardware limitations.

[0045] Step S40: Fill the main scanning parameters into the parameter layer aligned with the preview response image space to obtain the imaging prescription map.

[0046] In this embodiment, the parameter layer aligned with the preview image space refers to a multi-layer floating-point parameter container with the same spatial resolution and coordinate system as the preview response image, used to fill the main scan parameters of each image block according to their spatial position. The imaging prescription map refers to a multi-layer parameter map that can directly drive the main scan controller to perform regional dynamic control. Please refer to... Figure 2 , Figure 2 This is a schematic diagram of the imaging prescription map data structure provided in Embodiment 1 of the pathological slide scanning imaging prescription map generation method of this application. The imaging prescription map includes nine parameters: exposure time, camera gain, illumination intensity, white balance coefficient R, white balance coefficient G, white balance coefficient B, initial focus height, focus density, and scanning speed.

[0047] As an alternative implementation, the main scanning parameters are filled into a parameter layer that is spatially aligned with the multi-channel preview response image to obtain an imaging prescription map, wherein the parameter layer adopts a spatial resolution that is exactly the same as that of the multi-channel preview response image.

[0048] Specifically, a 9-layer parameter layer is established, pixel-level aligned with the multi-channel preview response image, with each pixel position corresponding to a set of 9-dimensional master scan parameters. The vectors corresponding to the master scan parameters of each image block output from the calibration matrix are filled pixel-by-pixel into the spatial position of that image block in the 9-layer parameter layer. For overlapping areas of adjacent image blocks, fusion weights are calculated based on the distance from the current pixel to the center point of its respective image block, followed by weighted fusion and Gaussian smoothing to obtain an imaging prescription map with the same resolution as the original preview image. By setting the parameter layer to a spatial resolution completely consistent with the multi-channel preview response image, the details of the parameter spatial distribution are preserved to the maximum extent, providing the highest precision data support for subsequent region clustering, path planning, and regional dynamic control.

[0049] As an alternative implementation, the master scan parameters are filled into a parameter layer that is spatially aligned with the multichannel preview response image to obtain an imaging prescription map, wherein the parameter layer uses a lower spatial resolution scaled proportionally to the multichannel preview response image.

[0050] Specifically, a nine-layer parameter layer is established with a spatial resolution ranging from 1 / 10 to 1 of the preview response image, i.e., a scaling factor of 0.1-1. The master scan parameter vectors of each image patch output from the calibration matrix are filled according to the corresponding positions of that image patch in the downsampled parameter layer. Weighted fusion and Gaussian smoothing are also applied to the overlapping areas of adjacent image patches to obtain a down-resolution imaging prescription map. By setting the parameter layer to a lower spatial resolution that is scaled proportionally to the multi-channel preview response image, the data storage and computational overhead of the imaging prescription map are reduced while preserving the macroscopic spatial distribution characteristics of the parameters, thus improving the efficiency of prescription map generation and subsequent hardware timing queue generation.

[0051] This embodiment provides a method for generating prescription maps for pathological slide scanning imaging. First, the pathological slide sample is illuminated using at least two optical response channels, and preview response images of each optical response channel are acquired. These preview response images are then registered and combined to obtain a multi-channel preview response image. The multi-channel preview response image is divided into multiple image blocks, and the optical features of each image block are calculated. These optical features are normalized, and the normalized features are combined to obtain a feature vector. The feature vector is input into a pre-established calibration matrix to obtain the main scanning parameters. The main scanning parameters are then filled into a parameter layer aligned with the preview image space to obtain the imaging prescription map. By acquiring comprehensive optical information through multi-channel registration and combination, and establishing a mapping relationship between the preview and main scan based on image block-level features, systematic deviations caused by optical path differences and spatial misalignment can be eliminated, achieving parameter collaborative prediction and thus reducing parameter prediction bias.

[0052] Based on Embodiment 1, in Embodiment 2 of this application, the content that is the same as or similar to that in Embodiment 1 can be referred to the above description, and will not be repeated hereafter. In addition, before acquiring the preview response images of each optical response channel, the following steps are also included: Step S101: Acquire dark field image and white field image. The dark field image is the dark current image of each channel under no illumination, and the white field image is the response image of each channel acquired under uniform illumination under no sample state.

[0053] As one implementation method, all lighting sources are turned off, and the detection cameras in each channel acquire dark current images under no-light conditions, which are used as dark field images I. dark Then, with no pathological slide samples placed on the stage, the beam-splitting oblique illumination module is controlled to sequentially illuminate each wavelength of light source according to a preset channel sequence, uniformly illuminating the sample-free area. The response images are then captured by the detector cameras in each channel, serving as the white field image I. whiteBy obtaining a one-to-one corresponding dark field image and white field image for each optical response channel, reference data is provided for subsequent dark field subtraction, white field correction, and illumination intensity normalization processing of the preview response image.

[0054] After acquiring preview response images for each optical response channel, the process also includes: Step S102: Correct the preview response image based on the dark field image and the white field image to obtain the corrected preview response image.

[0055] As one implementation method, dark field subtraction and white field correction are performed on the preview response images acquired by each optical response channel. Dark field subtraction involves subtracting the corresponding dark field image of that channel pixel by pixel to eliminate dark current noise from the image sensor. White field correction involves dividing the image after dark field subtraction by the difference between the white field image and the dark field image to eliminate camera pixel response unevenness and illumination spatial uniformity errors. Simultaneously, the actual illumination intensity recorded in real time by the light intensity monitoring module is used, and the image is normalized by combining it with a reference illumination intensity to eliminate intensity drift caused by light source aging or fluctuations, resulting in the corrected preview response image I for each channel. corrected =(I raw -I dark ) / (I white -I dark )*I monitorref / I monitor , where I raw For the preview response image captured, I monitorref For reference lighting intensity, I monitor This represents the actual illumination intensity. By obtaining the corrected preview response image, an accurate and reliable data foundation is provided for subsequent multi-channel image registration, optical feature extraction, and main scan parameter prediction.

[0056] In this embodiment, dark current noise is eliminated by dark field subtraction, and pixel response non-uniformity and illumination spatial non-uniformity are eliminated by white field correction. Light intensity normalization is combined to eliminate light source fluctuations, thereby obtaining a preview response image that can accurately reflect the optical characteristics of the sample, providing a high-quality data foundation for subsequent multi-channel registration and optical feature extraction.

[0057] Based on any of the above embodiments of this application, Embodiment 3 of this application proposes a method for generating prescription maps for pathological slide scanning imaging, which can be referred to the above description and will not be repeated hereafter. Based on this, before inputting the feature vector into a pre-established calibration matrix to obtain the main scanning parameters, the method further includes: Step S301: Prepare multiple sets of pathological slide samples covering different staining types, different tissue types, and different staining depths.

[0058] As one implementation method, N sets of standard pathological slide samples are selected, where N≥50, preferably N≥100. The pathological slide samples cover various staining types, including HE staining (Hematoxylin and Eosin Stain), IHC staining (Immunohistochemistry Stain), and TCT (ThinPrep Cytology Test), encompassing various tissue types such as densely packed nuclei, abundant cytoplasm, fibrous tissue, and uneven thickness, and including samples with different staining intensities, such as deep, light, and medium staining. By constructing a representative sample library, a data foundation is provided for subsequent acquisition of multi-channel preview response images, extraction of feature vectors, and determination of optimal master scan parameters.

[0059] Step S302: Obtain multi-channel preview response images and feature vectors for each group of pathological slide samples.

[0060] In one implementation, each group of pathological slide samples is placed on a stage, and the beam-splitting oblique illumination module is controlled to illuminate the slide at an oblique incident angle of 5-60 degrees relative to the slide surface, preferably 10-30 degrees, according to a preset channel sequence. The preview imaging module acquires preview response images for each optical response channel, and corrects and normalizes the light intensity based on dark-field and white-field images. The corrected preview response images for each channel are then spatially registered and combined to obtain a multi-channel preview response image for the sample. This multi-channel preview response image is then divided into multiple image blocks, and optical features are calculated for each image block. After normalization of these optical features, they are combined to obtain the corresponding feature vector for that image block. By obtaining multi-channel preview response images and feature vectors corresponding to each image block for each group of pathological slide samples, input data is provided for subsequent pairing with target master scan parameters to train the calibration matrix.

[0061] Step S303: Perform a main scan on each group of pathological slide samples to obtain the target main scan parameters.

[0062] In this embodiment, the target master scan parameters refer to the set of master scan parameters obtained after performing a master scan on each group of pathological slide samples, which can enable the image block to obtain the best imaging effect. These parameters include exposure time, camera gain, illumination intensity, white balance coefficient, initial focus height, focus density, and scan speed.

[0063] In one implementation, each group of pathological slide samples is placed in the main scanning optical path. Based on the optimal imaging effect annotated by human experts, pathological imaging experts evaluate the quality of the main scanning image and manually adjust the parameters until the optimal whole-slice image is obtained, then record the corresponding parameters to obtain the target main scanning parameters. Alternatively, an automatic optimization algorithm is used. By defining an imaging quality evaluation function and using grid search, Bayesian optimization, or a genetic algorithm to automatically search for the main scanning parameters that optimize the evaluation function, the optimal main scanning parameters are determined as the target main scanning parameters for the corresponding image block. By obtaining the target main scanning parameters corresponding to the feature vector of each image block of each group of pathological slide samples, supervised labels are provided for subsequent pairing of feature vectors with target main scanning parameters and training of the calibration matrix.

[0064] Step S304: Pair the feature vector with the target main scanning parameters to obtain the training sample set.

[0065] Step S305: Based on the training sample set, obtain the calibration matrix by training multiple linear regression or a neural network.

[0066] As one implementation method, for each group of pathological slide samples, the feature vectors of each image block in the sample are spatially mapped one-to-one with the target master scan parameters of the corresponding image block to form paired samples. The paired samples from all calibration sample groups are then aggregated to obtain a training sample set. Based on the training sample set, in simple scenarios, a multiple linear regression algorithm is used to solve for the calibration matrix M. calib And the bias vector b, such that P main =M calib *V feature +b minimizes the mapping error. In complex scenarios, a three-layer fully connected neural network is used for forward and backward propagation training with 7-dimensional feature vectors as input and 9-dimensional master scan parameters as output, thereby obtaining the calibration matrix. By training with multiple linear regression or a three-layer fully connected neural network to obtain the calibration matrix, the mapping relationship from preview response features to master scan parameters is accurately established, providing a reliable basis for converting the feature vectors of each image patch into master scan parameters and generating high-quality imaging prescription maps during formal scanning.

[0067] In this embodiment, by establishing a high-precision mapping relationship between preview response features and main scanning parameters, a reliable mathematical model foundation is provided for accurately converting the feature vectors of each image block into main scanning parameters and generating high-quality imaging prescription maps during formal scanning.

[0068] Based on any of the above embodiments of this application, Embodiment 4 of this application proposes a method for generating a pathological slide scanning imaging prescription map, which can be referred to the above description and will not be repeated hereafter. Based on this, the main scanning parameters are filled into a parameter layer aligned with the multi-channel preview response image space to obtain the imaging prescription map, including: Step S41: Fill the main scan parameters into the parameter layer aligned with the multi-channel preview response image space to obtain the filled prescription map.

[0069] Step S42: The overlapping areas of the image blocks in the filled prescription map are weighted, fused, and smoothed to obtain the imaging prescription map.

[0070] As one implementation method, a parameter layer with the same spatial resolution as the preview response image or scaled proportionally is established. The main scan parameters corresponding to each image block obtained through the calibration matrix are converted into main scan parameter vectors, and written layer by layer into the corresponding positions in the parameter layer according to the spatial position of the image block in the preview response image to obtain the filled prescription map.

[0071] For overlapping regions between adjacent image patches, a weighted fusion method based on the distance from the pixel to the center point of the image patch is used for parameter fusion, where the weight w = 1 - (d / d) max ) 2 d is the distance from the current pixel to the center of the block. max The image block width is half the side length of the image block. The fused parameter layer is then Gaussian smoothed, with a smoothing coefficient σ ranging from 0.5 to 2.0 pixels, preferably 1.0 pixel, resulting in an imaging prescription map spatially aligned with the preview response image. By performing distance-weighted fusion and Gaussian smoothing on the overlapping areas of the image blocks, parameter jumps at the image block boundaries are eliminated, making the imaging prescription map spatially continuous and smooth, avoiding difficulties in main scanning hardware execution or image quality degradation due to abrupt parameter changes.

[0072] In this embodiment, by filling the main scanning parameters of each image block into a parameter layer aligned with the preview response image space, a precise mapping from discrete image block parameters to a continuous spatial parameter map is achieved.

[0073] Based on any of the above embodiments of this application, Embodiment 5 of this application proposes a method for generating a pathological slide scanning imaging prescription map, which can be referred to the above description and will not be repeated hereafter. Based on this, after filling the main scanning parameters into a parameter layer aligned with the multi-channel preview response image space to obtain the imaging prescription map, the method further includes: Step S401: Based on the parameter similarity of the main scanning parameters at each location in the imaging prescription map, a clustering algorithm is used to divide the scanning area into multiple sub-regions.

[0074] In one implementation, the main scan controller reads the main scan parameter vector corresponding to each spatial location in the imaging prescription map. It calculates the similarity between the main scan parameter vectors of any two spatial locations using Euclidean distance or weighted Euclidean distance; a smaller distance indicates higher parameter similarity. Then, a K-means clustering algorithm or a region growing algorithm is used to divide the entire scanning area into several sub-regions according to parameter similarity. The number of clusters K ranges from 3 to 10, preferably from 5 to 7. By dividing the area into multiple sub-regions, the main scan parameter vectors at locations within the same sub-region are similar to each other, while the main scan parameter vectors between different sub-regions show significant differences. Dividing the scanning area into multiple sub-regions lays the foundation for subsequent targeted scan control within each sub-region.

[0075] Step S402: Plan the scanning path within each sub-region to obtain the path point sequence.

[0076] In this embodiment, the path point sequence refers to the ordered set of control points formed by the coordinates of the stage target, which are generated by discretizing the selected scanning path and arranged in the scanning order.

[0077] As one implementation method, for each sub-region, a suitable scanning path type is selected based on the sub-region's geometry and prescription map parameter variation characteristics. For rectangular sub-regions, a serpentine path is used to reduce idle travel and improve scanning efficiency; for circular or irregular sub-regions, a spiral path is used to scan from the center outwards to cover the entire region; for sub-regions with drastic parameter changes in the prescription map, an adaptive path is used to dynamically adjust the scanning direction according to parameter changes, prioritizing scanning high-priority regions. After path planning, each path is discretized into a continuous sequence of path points, each path point containing the stage target coordinates (X, Y, Z). By planning scanning paths matching the sub-region's geometry and parameter characteristics within each sub-region and generating path point sequences, an ordered spatial trajectory is provided for each sub-region, ensuring efficient coverage during the scanning process.

[0078] Step S403: Combine the path point sequence with the main scan parameters of the sub-region to generate a time sequence queue.

[0079] In this embodiment, the timing queue refers to the ordered control sequence generated by the main scanning controller after binding the path point sequence of each sub-region with the main scanning parameters of that sub-region in the scanning order. Each timing point includes the stage target coordinates (X,Y,Z), exposure time, camera gain, illumination intensity, white balance coefficient, focus height, focus density, scanning speed, and hardware trigger signal timing, which are used to synchronously drive the camera, light source, stage, and focusing actuator to achieve regional dynamic control.

[0080] As one implementation method, for each sub-region's path point sequence, the corresponding master scanning parameters of that sub-region in the imaging prescription map are bound to each path point in the path point sequence, forming an ordered control sequence, i.e., a timing queue, containing the stage target coordinates (X, Y, Z), various master scanning parameters, and trigger signal timings. By combining the path point sequence with the master scanning parameters of the sub-region to generate the timing queue, the spatial parameter distribution is transformed into a hardware control sequence executed sequentially in time, enabling each hardware module to work collaboratively under a unified trigger timing.

[0081] Step S404: Generate hardware control instructions based on the timing queue and send the hardware control instructions to the hardware module for control.

[0082] In one implementation, the main scanning controller parses each timing point in the timing queue, converting the parameters contained in each timing point into instruction signals executable by each hardware module. Exposure control sets the exposure time via the camera SDK (Software Development Kit), with an accuracy ≤1μs; light source control sets the current for each wavelength band via the LED driver circuit, with a response time ≤10μs and current accuracy ≤0.1mA; camera gain control sets analog or digital gain via the camera SDK, with an accuracy ≤0.1dB; stage motion control sets the target position and speed via a motion controller, with a positioning accuracy ≤0.1μm and a speed range of 0.1-100mm / s; focusing control adjusts the objective lens height via piezoelectric ceramics or a stepper motor, with an adjustment range of ±100μm and an accuracy ≤0.1μm; scanning speed control dynamically adjusts the stage speed according to the prescription map, reducing the speed in areas of drastic parameter changes and increasing the speed in uniform areas. By transforming the imaging prescription map into precise coordinated control of physical hardware such as the camera, light source, stage, and focusing mechanism, it enables regional dynamic exposure, dynamic gain, dynamic focusing, and dynamic scanning path control.

[0083] Furthermore, to ensure the synchronous execution of all hardware modules, a hardware triggering mechanism is employed. This mechanism utilizes the camera's external trigger input, with the motion controller or FPGA (Field Programmable Gate Array) generating trigger pulses. The triggering mode selects line triggering or frame triggering based on the scanning strategy. The triggering sequence is: exposure start, illumination on, exposure end, illumination off, stage movement, and next position trigger. During scanning, the system monitors image brightness in real time, detecting overexposure or underexposure by statistically analyzing the mean and histogram of each frame; monitors image sharpness, detecting defocus by calculating the gradient or frequency domain energy of each frame; monitors stage position, monitoring the deviation between the actual and target positions via encoder feedback; and monitors illumination stability, using a light intensity monitoring module to monitor illumination intensity in real time. If an anomaly is detected, minor anomalies are recorded and scanning continues; moderate anomalies pause scanning, automatically adjust parameters, and then resume; severe anomalies stop scanning and trigger an alarm to alert the operator. This hardware triggering mechanism achieves precise synchronous execution of the hardware modules and tiered execution based on the severity of the anomaly, ensuring the collaborative accuracy of all hardware components and the stability and reliability of the system during the main scanning process.

[0084] In this embodiment, by converting the multi-layer parameters in the imaging prescription map into timing control instructions executable by each hardware module, regional dynamic scanning control of the entire pathological slide is realized, thereby solving the problem of poor adaptability of the unified strategy to complex samples and improving the quality of scanned images and scanning efficiency.

[0085] Based on any of the above embodiments of this application, Embodiment Six of this application proposes a method for generating prescription maps of pathological slide scanning imaging, which can be referred to the above description and will not be repeated hereafter. In addition, the method for generating prescription maps of pathological slide scanning imaging further includes: Step S50: After the formal scan begins, multiple preset sentinel areas are scanned, and the actual imaging results of each sentinel area are compared with the predicted results in the imaging prescription map to calculate the brightness deviation, saturation ratio deviation, sharpness deviation, and focus deviation.

[0086] In this embodiment, the sentinel region refers to several verification regions that are pre-selected based on the imaging prescription map before the formal scan begins and are scanned preferentially after the formal scan starts. These regions are used to compare the actual imaging results with the predicted results in the imaging prescription map, providing a basis for deviation correction.

[0087] As one implementation method, after the formal scan is initiated, the main scan controller prioritizes scanning multiple sentinel regions pre-defined based on the imaging prescription map. Three to ten sentinel regions are set, evenly distributed within the scanning area, and regions with drastic parameter changes or low confidence in the prescription map are preferentially selected. Preferably, the number of sentinel regions is 5 to 7. Each sentinel region covers 1 to 5 scanning fields of view, preferably 2 to 3 fields of view. For each sentinel region, the actual scanning imaging result is acquired, and the actual imaging result is compared with the prescription map prediction result to calculate brightness deviation, saturation ratio deviation, sharpness deviation, and focus deviation. Wherein, brightness deviation ΔL = |L actual -L predicted | / L predicted , where L actual L represents the average brightness of the actual image. predicted Predict brightness for prescription maps, brightness deviation threshold ΔL th =10%-30%, preferably, the brightness deviation threshold is 15%-20%; saturation ratio deviation ΔS=|S actual -S predicted |, where S actual S represents the actual saturation pixel ratio. predicted To predict the saturation ratio, the saturation ratio threshold ΔS th =1%-5%, preferably, the saturation ratio threshold is 2%-3%; Sharpness deviation ΔC=|C actual -C predicted | / C predicted , where C actual For actual sharpness metrics, C predicted To predict sharpness, a sharpness deviation threshold ΔC is set. th =10%-30%, preferably, the sharpness deviation threshold is 15%-20%; focus deviation ΔZ=|Z actual -Z predicted |, where Z actual Z is the actual optimal focal position. predicted Predict the focal position for the prescription map, with a focal deviation threshold ΔZ. th =0.5-2.0μm, preferably, the focal deviation threshold is 1.0μm. By obtaining the quantitative deviation index of each sentinel region in multiple dimensions, a closed-loop correction decision basis is provided for subsequent judgment on whether the deviation exceeds the threshold, and thus for the correction of prescription parameters of the subsequent scanned region.

[0088] Step S60: When any deviation exceeds the corresponding preset parameter threshold, the prescription parameters of the subsequent scanned area are corrected according to the deviation direction and deviation magnitude to obtain the corrected prescription parameters.

[0089] As one implementation method, the brightness deviation, saturation ratio deviation, sharpness deviation, and focus deviation of each sentinel region are compared with their corresponding preset parameter thresholds. When any deviation exceeds the corresponding threshold, a correction strategy is selected based on the directional consistency and spatial distribution characteristics of the deviations in each sentinel region. If the deviations of multiple sentinel regions are in the same direction, a global offset correction is performed on the prescription parameters of the subsequent scanned areas. The average deviation of each sentinel region is multiplied by a correction coefficient and then superimposed onto the original prescription parameters. If the deviation only occurs in the target area, a local correction is performed with the deviating sentinel region as the center and different correction weights are assigned according to the distance from the center. The closer the distance, the greater the correction magnitude, and the farther the distance, the smaller the correction magnitude. If the deviation occurs systematically, the calibration matrix is ​​updated. By correcting the prescription parameters, the predicted values ​​of the main scanning parameters of the subsequent scanned areas are matched with the actual optical characteristics, thereby reducing the rescan rate and improving scanning efficiency.

[0090] Step S70: Continue scanning the sentinel region based on the corrected prescription parameters, and continue scanning the sentinel region and correcting the prescription parameters.

[0091] Each time a preset percentage of the area is scanned, a new sentry zone is set.

[0092] In one implementation, during the formal scanning process, the main scanning controller continues to perform main scanning on subsequent areas to be scanned using the corrected prescription parameters. As the scanning progresses, a new sentinel area is set after each preset percentage area is completed, such as 10%-20% of the scan. This new sentinel area is scanned first, and its actual imaging result is compared with the predicted result in the current corrected imaging prescription map to calculate brightness deviation, saturation ratio deviation, sharpness deviation, and focus deviation. If any deviation still exceeds the corresponding threshold, the prescription parameters of the subsequent areas to be scanned are corrected again or an alarm is triggered based on the direction and magnitude of the deviation. If the deviation is reduced to below the threshold after correction, scanning continues according to the corrected prescription map. This forms a continuous closed-loop correction process of scanning sentinel areas, calculating deviations, correcting prescription parameters, continuing scanning, and setting new sentinel areas, so that the imaging prescription map continuously approaches the actual optical characteristics throughout the scanning process, thereby reducing the rescan rate and improving scanning efficiency.

[0093] Furthermore, for scanned areas, if subsequent sentinel region closed-loop correction reveals severe parameter deviations—that is, when the deviation exceeds a severe threshold, such as a brightness deviation > 50% or a focus deviation > 5μm—the image quality of that scanned area is deemed unacceptable. A local rescanning strategy is employed, rescanning only the deviated areas and utilizing the effective information from the already scanned image of those areas to specifically adjust key parameters such as exposure time or focus position, avoiding the repeated execution of the complete multi-parameter scanning process. This approach minimizes rescanning time and data redundancy while ensuring the quality of the entire slice image, further reducing scanning overhead and improving overall scanning efficiency.

[0094] In this embodiment, after the formal scan is started, a preset sentinel area is scanned first to obtain the measured deviation. The prescription parameters of the subsequent areas to be scanned are corrected, and a new sentinel area is set after each preset proportion area is scanned to continuously perform closed-loop correction. This makes the imaging prescription map continuously approach the actual optical characteristics throughout the scanning process, thereby reducing the rescan rate and improving scanning efficiency and image quality.

[0095] Based on any of the above embodiments of this application, Embodiment Seven of this application proposes a method for generating prescription maps for pathological slide scanning imaging, which can be referred to the above description and will not be repeated hereafter. On this basis, when any deviation exceeds the corresponding threshold, the prescription parameters of the subsequent area to be scanned are corrected according to the direction and magnitude of the deviation, resulting in corrected prescription parameters, including: Step S61: When the number of sentinel regions whose deviation directions on the target parameters are consistent and whose deviation magnitude exceeds the preset parameter threshold is greater than the first preset threshold, the prescription parameters of the subsequent scanned regions are globally corrected to obtain the corrected prescription parameters.

[0096] In this embodiment, the consistent deviation direction refers to the same parameter. The signs of the deviations of the multiple sentinel areas involved in the judgment are exactly the same. That is, compared with the predicted value, the actual value is either higher or lower than the actual value. There is no situation where the positive and negative directions cancel each other out.

[0097] As one implementation method, for the brightness deviation, saturation ratio deviation, sharpness deviation, and focus deviation of each sentinel region, the number of sentinel regions with consistent deviation directions and deviation magnitudes exceeding the preset threshold corresponding to the target parameter is counted. When this number exceeds a first preset threshold, for example, accounting for 50% or more of the total number of sentinel regions, a global parameter prediction deviation is determined to exist. Preferably, when there are 5-7 sentinel regions, at least 3 sentinel regions with consistent deviation directions exceeding the threshold are considered as multiple sentinel regions with consistent deviation directions. The average deviation ΔP of all sentinel regions with consistent deviation directions and exceeding the preset parameter threshold is calculated, and this average deviation is multiplied by a global correction coefficient α and superimposed on the original prescription parameters of the subsequent scanned area, wherein the global correction coefficient ranges from 0.1 to 1.0; preferably, the global correction coefficient is 0.5; that is, P corrected =P original The formula +α*ΔP, where ΔP is the average deviation and α is the correction coefficient ranging from 0.1 to 1.0, is used to globally offset the prescription parameters of the subsequent scanned areas, resulting in the corrected prescription parameters. This global correction eliminates systematic biases, allowing the predicted parameter values ​​of the subsequent scanned areas to quickly approximate the actual optical characteristics, reducing the rescan rate and improving scanning efficiency.

[0098] Step S62: When the number of sentinel regions whose deviation from the target parameter exceeds the preset parameter threshold is less than the second preset threshold, and the sentinel region appears in the target region, the prescription parameters of the target region are locally corrected to obtain the corrected prescription parameters. The target region is determined by the center coordinates of the deviated sentinel region and the Gaussian influence radius.

[0099] In this embodiment, local correction is triggered only when the deviation occurs in the target region, meaning the number of sentinel regions exceeding the threshold is less than most of the thresholds required for global correction, for example, only 1-2 sentinel regions. The majority of the remaining sentinel regions are consistent with the prescription map prediction, indicating that the deviation is spatially localized rather than a universal phenomenon across the entire area. The target region refers to the spatial range jointly defined by the center coordinates of the deviating sentinel region and the Gaussian influence radius. That is, the geometric center (x0, y0) of the deviating sentinel region is used as the correction center. A Gaussian distribution is used to calculate the spatial weight function, so that the correction amount decreases exponentially with the distance from the center (x0, y0), and is significantly effective only within a range of approximately 2-3σ centered at (x0, y0), thereby strictly constraining the correction to the target region and its neighborhood.

[0100] As one implementation method, the number of sentinel regions whose deviation from the target parameters exceeds a preset parameter threshold is counted. If this number is less than a second preset threshold, such as 1-2 regions, a local anomaly is determined. The target region is defined with the center coordinates (x0, y0) of the deviating sentinel region as the origin and the Gaussian influence radius σ as the spatial range. The prescription parameters of subsequent scanned regions within this target region are locally corrected using the correction formula P. corrected (x,y)=P original (x,y)+β*ΔP(x,y)*w(x,y), where β is a local correction coefficient, ranging from 0.1 to 1.0; w(x,y) is the spatial weight function, calculated using a Gaussian distribution, w(x,y)=exp( ((x x0) 2 +(y y0) 2 ) / (2σ 2 In this equation, (x0, y0) represents the center of the sentinel region, σ is the radius of influence (range 1-10 mm), ΔP(x, y) represents the local deviation within the target region, and (x, y) are the spatial coordinates on the prescription map. By performing smooth attenuation correction on the target region, it is possible to repair local parameter deviations while avoiding interference with parameters in the surrounding normal region.

[0101] Step S63: When a systematic deviation in the same direction occurs, the calibration matrix is ​​updated, and the prescription parameters for the subsequent scanned area are regenerated based on the updated calibration matrix.

[0102] Among them, global correction, local correction and calibration matrix update can be selected and used in combination, and the first preset threshold is greater than the second preset threshold.

[0103] In this embodiment, the systematic occurrence of deviation refers to deviations that are not isolated or sporadic, but rather occur repeatedly and consistently during the scanning process. This manifests as any of the following: deviations in the same direction are consistently reproduced across multiple batches of sentinel areas; even after a global correction has been performed, deviations in the same direction reappear in new scanned areas, indicating that a single offset compensation cannot eliminate the error; the proportion of sentinel areas affected by the deviation is significantly higher, such as approaching or exceeding all sentinel areas, and the direction is highly consistent. These situations indicate that the source of the error is not a characteristic of an individual area, but rather a systematic deviation in the calibration matrix itself, such as insufficient compensation for differences between the preview optical path and the main scanning optical path, batch drift of the light source or sensor, or overall shift in the staining batch.

[0104] As one implementation method, during the formal scanning process, when a systematic prediction bias is determined to exist, the calibration matrix is ​​updated online, that is, by using the measured bias ΔP of the most recent round of sentinel regions and the feature vector V of the corresponding image patch.feature Calculate the parameters M of the new calibration matrix. new =M old +γ*(ΔP*V feature T ), where γ is the learning rate, ranging from 0.001 to 0.1. This update can be performed online, gradually improving prediction accuracy. Then, the updated calibration matrix replaces the original calibration matrix. Based on the updated calibration matrix, the mapping of feature vectors to master scanning parameters is re-executed for each image block in the subsequent scanning region, and the imaging prescription map is regenerated. This fundamentally eliminates systematic bias, allowing the predicted parameter values ​​of subsequent scanning regions to re-match with the actual optical characteristics, further improving the accuracy of closed-loop correction and scanning efficiency.

[0105] In this embodiment, by using a hierarchical and combined approach of three correction strategies—global correction, local correction, and calibration matrix update—adaptive correction strategies are adopted for deviations of different natures and ranges. This ensures correction accuracy while avoiding over-correction or under-correction, continuously improving the matching degree between the imaging prescription map and the actual optical characteristics, effectively reducing the rescan rate and improving scanning efficiency and image quality.

[0106] Based on any of the above embodiments of this application, Embodiment 8 of this application proposes a method for generating prescription maps for pathological slide scanning imaging, which can be referred to the above description and will not be repeated hereafter. Based on this, before inputting the feature vector into a pre-established calibration matrix to obtain the main scanning parameters, the method further includes: Step S306: Obtain the staining type of the slide to be scanned, and select different calibration matrices according to the staining type.

[0107] As one implementation method, before or during the preview of the slide to be scanned, the staining type of the slide is obtained. Based on the obtained staining type, a calibration matrix corresponding to the staining type and its corresponding feature weight coefficients are selected from a variety of pre-trained calibration matrices. The staining depth index is calculated using the corresponding feature weight coefficients α, β, and γ. For example, for HE staining, α=0.5, β=0.3, and γ=0.2. Dedicated calibration matrices are established for HE staining, IHC staining, and TCT staining. Each calibration matrix is ​​trained based on standard pathological slide calibration samples of the corresponding staining type. When the feature vectors of the image blocks are subsequently input into the calibration matrix to obtain the main scanning parameters, the system uses the calibration matrix corresponding to the staining type for mapping, ensuring that the generated imaging prescription map accurately matches the optical characteristics of the current staining.

[0108] In this embodiment, by obtaining the staining type of the slide to be scanned and selecting the calibration matrix and feature weight coefficients corresponding to the staining type, the mapping relationship between the preview features and the main scanning parameters can be adapted to the optical characteristics of the current staining, solving the problem of poor adaptability of the unified strategy to complex samples and improving the parameter prediction accuracy of the imaging prescription map.

[0109] For example, to help understand the technical concept or principle of the pathological slide scanning imaging prescription map generation method after combining this embodiment with the above embodiments, please refer to Figure 3 , Figure 3 A simplified flowchart illustrating the method for generating prescription maps from pathological slide scanning imaging provided in this application embodiment is shown below: First, at least two optical response channels are used to illuminate the slide at oblique incident angles of 5-60 degrees relative to the slide surface, and preview response images of each channel are acquired. Based on dark-field and white-field image correction and registration, a multi-channel preview response image is obtained. The multi-channel preview response image is divided into multiple image blocks, and the optical density, channel absorptivity, reflection risk index, dynamic range occupancy, local texture response, staining depth index, and local sharpness index of each image block are calculated. These are then normalized and combined to obtain a feature vector. The feature vector is input into a pre-established calibration matrix to obtain the master scanning parameters for each image block. The master scanning parameters are filled into a parameter layer spatially aligned with the multi-channel preview response image. After weighted fusion of overlapping areas and Gaussian smoothing, an imaging prescription map is obtained. Based on the similarity of parameters in the imaging prescription map, a method is adopted... Clustering algorithms divide the scanning sub-regions, plan scanning paths to generate path point sequences, and combine these sequences with the main scanning parameters of the sub-regions to generate a time-series queue. Hardware control commands are generated based on the time-series queue and sent to hardware such as the camera, light source, stage, and focusing mechanism to execute regional dynamic scanning. Multiple preset sentinel regions are scanned, and the actual imaging results are compared with the prescription map prediction results to calculate brightness deviation, saturation ratio deviation, sharpness deviation, and focus deviation. If the deviation exceeds the corresponding threshold, the prescription parameters of the subsequent scanned regions are corrected according to the direction and magnitude of the deviation, and the calibration matrix is ​​updated. Scanning continues after correction. After scanning a preset proportion of regions, a new sentinel region is set, and closed-loop correction is continuously performed. If the deviation does not exceed the threshold, the scanning is considered complete; if not, scanning continues; if complete, data is output and the process ends. By acquiring the multi-dimensional optical response of the tissue region, image block-level feature extraction and calibration matrix mapping generate a multi-layer parametric prescription map aligned with the preview image space. Based on the closed-loop correction of the sentinel regions, the prescription parameters are continuously optimized, thereby establishing a precise mapping relationship between the preview response and the main scanning parameters. This enables regional collaborative control, effectively reducing parameter prediction deviation and improving scanning efficiency and image quality.

[0110] For example, for HE-stained pathological section scanning, a standard HE-stained gastric tissue section was selected as the sample, with a section thickness of 4 μm and a coverslip thickness of 0.17 mm. A four-band oblique illumination preview was used, with bands λ1=450nm, λ2=530nm, λ3=630nm, and λ4=780nm, a preview resolution of 10 μm / pixel, and a preview time of 30 seconds. The multi-channel response image obtained from the preview was divided into 256*256 pixel image blocks, and 7-dimensional optical feature vectors were extracted. A pre-trained neural network calibration matrix was used to map the feature vectors of each image block to 9-dimensional master scanning parameters and generate a 9-layer parameter prescription map with spatial resolution consistent with the preview. The first hidden layer contains 32 neurons, and the second hidden layer contains 16 neurons. A 40x objective lens and a NA (Numerical) sensor were used. Aperture (numerical aperture) = 0.75, with a main scan speed of 1 mm / s and a focus density of 5 points / mm². Five sentinel regions were set, and closed-loop correction was performed using a global correction factor α = 0.5 and a local correction factor β = 0.3. Experimental results show that, compared with a uniform exposure strategy, the dynamic exposure strategy reduces overexposed areas by 85%, underexposed areas by 78%, and improves the overall image quality score by 23%.

[0111] For IHC-stained pathological section scanning, Ki-67 immunohistochemically stained breast tissue sections were selected as samples. During the preview stage, a near-infrared channel with λ4=780nm was added to detect the DAB (3,3′-Diaminobenzidine) staining depth. A dedicated calibration matrix and weighting coefficients optimized for IHC staining were used, with a focus on optimizing illumination intensity and exposure time parameters to avoid overexposure of the DAB-stained areas. After generating the corresponding imaging prescription map, the main scan was performed. Experimental results showed that after using the dedicated IHC calibration matrix and optimization strategy, the positive cell recognition rate increased from 89% to 96%, and the false negative rate decreased by 60%.

[0112] This application provides a system for generating prescription maps from scanned images of pathological slides. Please refer to [link / reference]. Figure 4 , Figure 4This is a schematic diagram of the structure of the beam-splitting oblique illumination preview optical system provided in the embodiments of this application. The multi-band light source employs a multi-color LED array, or white LEDs and filter wheels, laser diode arrays, liquid crystal tunable filters, etc., to provide illumination light in at least two different bands. Preferably, the following four bands are used: λ1 = 430-480nm, corresponding to the deeply stained region and cell nucleus; λ2 = 500-560nm, corresponding to conventional HE staining; λ3 = 600-680nm, corresponding to the lightly stained region and thick tissue; λ4 = 700-900nm, corresponding to near-infrared scattering / thickness detection. The LED driving current range for each band is 10-500mA, preferably 50-300mA; the driving current adjustment resolution is ≥8 bits, preferably ≥12 bits. The beam combiner uses a dichroic mirror group to combine light of different wavelengths into a single coaxial beam, ensuring that each wavelength band illuminates the same area of ​​the slide. The dichroic mirror has a cutoff wavelength accuracy of ±5nm, a transmittance of ≥90%, and a reflectance of ≥95%. CCD1 and CCD2 are two independent imaging channels separated by the dichroic mirror. The CCD is a charge-coupled device (CCD). CCD1 is the main imaging channel, located on the main axis of the optical path, receiving light reflected by the dichroic mirror group; CCD2 is the polarization imaging channel, located on the side axis of the optical path, receiving light transmitted through the dichroic mirror group. A collimating lens converts the divergent light after beam combining into parallel light, with a collimated beam divergence angle ≤2 degrees. The light homogenizing component employs an integrating rod or a diffuser. The integrating rod has a length of 50-150mm and a cross-section of 3mm*3mm to 10mm*10mm. The diffuser has a haze of ≥80%, ensuring that the illumination spot intensity has a spatial uniformity of ≥90% within a Φ10mm range, eliminating LED hotspots and edge dark areas. The adjustable aperture ranges from Φ1mm to Φ20mm, with an adjustment accuracy of ≤0.1mm. This limits the size of the illumination area, preventing light from illuminating the edges of the slide, the label area, or the main scanning optical path, and ensuring the sensor operates in the linear region. The polarizer has an extinction ratio of ≥100:1, generating linearly polarized light for subsequent polarization differential imaging. The beam splitter is placed at a 45-degree angle, with a splitting ratio of transmission:reflection = 90%-98%:2%-10%, preferably 95%:5%. This allows a small portion of the light to be reflected to the photodiode, while the remaining light is transmitted to the obliquely incident mirror. The reflector refracts the horizontal transmitted light downwards to illuminate the preview area at an angle of 5-60 degrees relative to the surface of the glass slide. Preferably, the incident angle is 10-30 degrees. Single-sided or double-sided optical paths are used to ensure that multiple wavelengths illuminate the same coordinate area. The reflector surface is coated with a broadband antireflection film with a reflectivity of ≥95% in the 430-900nm wavelength band.

[0113] This application provides a pathological slide scanning imaging prescription map generation device, which includes: a stage, a beam-splitting oblique illumination module, a preview imaging module, an imaging-side beam-splitting module, a light intensity monitoring module, a processor, a main scanning controller, a memory, and a computer program stored in the memory and executable on the processor; wherein, the memory stores instructions executable by at least one processor, and the instructions are executed by at least one processor to enable at least one processor to execute the pathological slide scanning imaging prescription map generation method in the above embodiment 1.

[0114] The beam-splitting oblique illumination module includes at least two illumination sources of different wavelengths, a beam combiner, a collimator, a homogenizer, a polarization modulation component, and an oblique incidence light guide component. Illumination light of different wavelengths is combined, collimated, and homogenized before illuminating the slide preview area at an oblique incidence angle of 5-60 degrees relative to the slide surface. The preview imaging module collects the response light generated by the tissue section in response to the oblique incidence light. The imaging-side beam-splitting module decomposes the image of the same preview area into multiple optical channels. The light intensity monitoring module records the actual illumination intensity of each wavelength in real time. The processor receives the preview response images of multiple optical channels, normalizes the images of each channel according to the actual illumination intensity recorded by the light intensity monitoring module, divides the preview area into multiple image blocks and extracts optical features, and converts the multi-channel preview response of each image block into main scanning parameters through a calibration matrix to generate a multi-layer parameter prescription map. The main scanning controller generates a hardware timing queue based on the imaging prescription map and performs regional dynamic exposure, dynamic gain, dynamic focusing, and dynamic scanning path control.

[0115] The following is for reference. Figure 5 It shows a schematic diagram of a pathological slide scanning imaging prescription map generation device suitable for implementing the embodiments of this application. Figure 5 The pathological slide scanning imaging prescription map generation device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0116] like Figure 5As shown, the pathological slide scanning imaging prescription map generation device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The random access memory 1004 also stores various programs and data required for the operation of the pathological slide scanning imaging prescription map generation device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the pathology slide scanning imaging prescription map generation device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a pathology slide scanning imaging prescription map generation device with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0117] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0118] The pathological slide scanning imaging prescription map generation device provided in this application, employing the pathological slide scanning imaging prescription map generation method in the above embodiments, can solve the technical problem of large parameter prediction deviation caused by the disconnect between preview and main scan. Compared with the prior art, the beneficial effects of the pathological slide scanning imaging prescription map generation device provided in this application are the same as those of the pathological slide scanning imaging prescription map generation device provided in the above embodiments, and other technical features in this pathological slide scanning imaging prescription map generation device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0119] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0120] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0121] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the pathological slide scanning imaging prescription map generation method in the above embodiments.

[0122] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), or any suitable combination thereof.

[0123] The aforementioned computer-readable storage medium may be included in the pathological slide scanning imaging prescription map generation device; or it may exist independently and not be assembled into the pathological slide scanning imaging prescription map generation device.

[0124] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the pathological slide scanning imaging prescription map generation device, the device causes the following actions: illuminating the pathological slide sample using at least two optical response channels; acquiring preview response images of each optical response channel; registering and combining the preview response images of each optical response channel to obtain a multi-channel preview response image; dividing the multi-channel preview response image into multiple image blocks; calculating the optical features of each image block; normalizing the optical features; and combining the normalized features to obtain a feature vector; inputting the feature vector into a pre-established calibration matrix to obtain the main scanning parameters; and filling the main scanning parameters into a parameter layer aligned with the preview image space to obtain an imaging prescription map.

[0125] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0128] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described method for generating prescription maps of pathological slide scanning images. This solves the technical problem of large parameter prediction deviations caused by the disconnect between the preview and the main scan. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the pathological slide scanning imaging prescription map generation method provided in the above embodiments, and will not be elaborated upon here.

[0129] This application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the pathological slide scanning imaging prescription map generation method described above.

[0130] The computer program product provided in this application can solve the technical problem of large parameter prediction deviation caused by the disconnect between preview and main scan. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the pathological slide scanning imaging prescription map generation method provided in the above embodiments, and will not be repeated here.

[0131] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent scope of this application.

Claims

1. A method for generating prescription maps from pathological slide scanning imaging, characterized in that, The method for generating prescription maps from pathological section scanning imaging includes: At least two optical response channels are used to illuminate the pathological slide sample to obtain a dark field image and a white field image. The dark field image is the dark current image of each optical response channel under no illumination, and the white field image is the response image of each optical response channel acquired under uniform illumination in the absence of a sample. Preview response images of each optical response channel are acquired, and the preview response images are corrected based on the dark field image and the white field image to obtain corrected preview response images. The corrected preview response images of each optical response channel are registered and combined to obtain multi-channel preview response images. The multi-channel preview response image is divided into multiple image blocks, the optical features of each image block are calculated, the optical features are normalized, and the normalized features are combined to obtain a feature vector. The feature vector is input into a pre-established calibration matrix to obtain the main scan parameters; The main scanning parameters are filled into a parameter layer aligned with the multi-channel preview response image space to obtain an imaging prescription map.

2. The method for generating prescription maps for pathological slide scanning imaging as described in claim 1, characterized in that, Before inputting the feature vector into a pre-established calibration matrix to obtain the main scanning parameters, the process further includes: Prepare multiple sets of pathological slide samples covering different staining types, different tissue types, and different staining intensities; For each group of pathological slide samples, obtain the multi-channel preview response image and the feature vector; Perform a main scan on each group of pathological slide samples to obtain the target main scan parameters; The feature vectors are paired with the target main scanning parameters to obtain a training sample set; Based on the training sample set, the calibration matrix is ​​obtained by training using multiple linear regression or a neural network.

3. The method for generating prescription maps for pathological slide scanning imaging as described in claim 1, characterized in that, The step of filling the main scanning parameters into a parameter layer aligned with the multi-channel preview response image space to obtain an imaging prescription map includes: The main scanning parameters are filled into a parameter layer aligned with the space of the multi-channel preview response image to obtain the filled prescription image; The overlapping areas of the image blocks in the filled prescription map are weighted, fused, and smoothed to obtain the imaging prescription map.

4. The method for generating prescription maps for pathological slide scanning imaging as described in claim 1, characterized in that, After filling the main scanning parameters into a parameter layer aligned with the multi-channel preview response image space to obtain the imaging prescription map, the method further includes: Based on the parameter similarity of the main scanning parameters at each position in the imaging prescription map, a clustering algorithm is used to divide the scanning area into multiple sub-regions; Plan a scanning path within each sub-region to obtain a sequence of path points; The path point sequence is combined with the main scan parameters of the sub-region to generate a time-series queue; Hardware control instructions are generated based on the timing queue, and then sent to the hardware module for control.

5. The method for generating prescription maps for pathological slide scanning imaging as described in claim 1, characterized in that, The method for generating prescription maps from pathological section scanning imaging also includes: After the formal scan begins, multiple preset sentinel areas are scanned, and the actual imaging results of each sentinel area are compared with the predicted results in the imaging prescription map to calculate the brightness deviation, saturation ratio deviation, sharpness deviation and focus deviation. When any deviation exceeds the corresponding preset parameter threshold, the prescription parameters of the subsequent scanned area are corrected according to the direction and magnitude of the deviation to obtain the corrected prescription parameters. The sentinel region is scanned again based on the corrected prescription parameters, and the scanning of the sentinel region and the correction of the prescription parameters are continued. In this process, after scanning a preset proportion of the area, a new sentinel area is set.

6. The method for generating prescription maps for pathological slide scanning imaging as described in claim 5, characterized in that, When any deviation exceeds the corresponding threshold, the prescription parameters of the subsequent scanned area are corrected according to the direction and magnitude of the deviation to obtain the corrected prescription parameters, including: When the number of sentinel regions whose deviations on the target parameters are consistent in direction and whose deviations exceed the preset parameter threshold is greater than the first preset threshold, the prescription parameters of the subsequent scanned regions are globally corrected to obtain the corrected prescription parameters. When the number of sentinel regions whose deviation from the target parameter exceeds a preset parameter threshold is less than a second preset threshold, and the sentinel region appears in the target region, the prescription parameter of the target region is locally corrected to obtain the corrected prescription parameter. The target region is determined by the center coordinates of the deviated sentinel region and the Gaussian influence radius. When a systematic deviation in the same direction occurs, the calibration matrix is ​​updated, and the prescription parameters for the subsequent scanned region are regenerated based on the updated calibration matrix. The global correction, the local correction, and the calibration matrix update can be selected and used in combination, and the first preset threshold is greater than the second preset threshold.

7. The method for generating prescription maps for pathological slide scanning imaging as described in claim 1, characterized in that, Before inputting the feature vector into a pre-established calibration matrix to obtain the main scanning parameters, the process further includes: Obtain the staining type of the slide to be scanned, and select different calibration matrices according to the staining type.

8. A device for generating prescription maps from scanned images of pathological sections, characterized in that, The pathological slide scanning imaging prescription map generation device includes: a stage, a beam-splitting oblique illumination module, a preview imaging module, an imaging-side beam-splitting module, a light intensity monitoring module, a processor, a main scanning controller, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the pathological slide scanning imaging prescription map generation method as described in any one of claims 1 to 7.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method for generating a pathological slide scanning imaging prescription map as described in any one of claims 1 to 7.

Images