An apparatus and method for assisting in the analysis of immunohistochemically positive cells in lymphoma

Abstract

The application discloses an auxiliary lymphoma immunohistochemical positive cell positioning analysis device and method, and relates to the technical field of positive cell positioning.The application comprises an image acquisition and preprocessing module, a positive area preliminary screening module, a cell morphology analysis module, a three-dimensional coordinate positioning module and a report generation and visualization interaction module, is used for accessing a digital slice scanner or a digital pathology image management system, reading an immunohistochemical slice image in a WSI format, and being internally provided with a convolutional neural network model based on deep learning; the model is trained and optimized through a large number of annotated lymphoma immunohistochemical slice images, and can automatically identify suspicious areas with positive cell characteristics in the image.The double-layer analysis mechanism combining the deep learning model and the multi-dimensional screening of morphological parameters can significantly reduce the interference of non-specific staining and background impurities, and effectively reduce the missed detection rate and the false detection rate, compared with the traditional manual film reading or single threshold judgment method.

Description

Technical Field

[0001] This invention belongs to the field of positive cell localization, specifically, it relates to an auxiliary device and method for immunohistochemical localization analysis of positive cells in lymphoma. Background Technology

[0002] Pathology is the gold standard for lymphoma diagnosis, and immunohistochemistry is an important basis for lymphoma diagnosis. Without it, it is almost impossible to accurately classify lymphomas. However, there are many difficulties in the current immunohistochemical diagnosis of lymphoma, such as the inability of a single marker to distinguish related subtypes, the possibility that the same marker may be expressed in multiple types of lymphoma, and the possibility of false negatives due to small sample size. Rapid identification of immunohistochemically positive cells is fundamental, which can reduce problems such as time consumption and inaccurate localization, and improve the speed and accuracy of diagnosis.

[0003] However, in current clinical practice, the identification and counting of immunohistochemically positive cells still mainly rely on manual interpretation by pathologists under a microscope. This traditional method is not only labor-intensive and highly subjective, easily affected by factors such as physician experience, fatigue, and visual bias, leading to poor consistency of results; but also significantly increases the difficulty of manual identification for samples with low positive rates or atypical positive cells, easily resulting in missed detections or misjudgments. In addition, the manual counting process is time-consuming, especially when the sample size is large or when multiple markers need to be analyzed together, which makes it difficult to meet the clinical need for rapid diagnosis and is not conducive to subsequent in-depth quantitative analysis of the morphological characteristics and spatial distribution of positive cells, thus restricting further improvement in the accuracy of lymphoma diagnosis.

[0004] In view of this, the present invention is proposed. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide an auxiliary device and method for localizing lymphoma immunohistochemical positive cells, thereby solving the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows: An auxiliary device for localizing positive immunohistochemical cells in lymphoma includes: The image acquisition and preprocessing module is used to connect to a digital slide scanner or digital pathology image management system and read WSI format immunohistochemical slide images; The positive region screening module has a built-in deep learning-based convolutional neural network model. This model has been trained and optimized with a large number of labeled lymphoma immunohistochemical slice images and can automatically identify suspicious regions with positive cell characteristics in the image. The cell morphology analysis module performs edge detection and morphological parameter extraction on individual cells in the positive areas obtained from the initial screening. Combined with the morphological standards of lymphoma positive cells, it further distinguishes true positive cells from background impurities or non-specific stained cells. The three-dimensional coordinate positioning module accurately marks the coordinates of confirmed positive cells in the WSI image based on the physical coordinate information of the digital slices, and generates a positioning report containing cell location, quantity, and morphological parameters. The report generation and visualization interaction module can integrate positive cell location data, morphological parameters and statistical analysis results to automatically generate standardized test reports. It also provides an interactive visualization interface and allows manual addition, deletion or correction of classification results. The system will recalculate the analysis indicators based on the corrected results.

[0007] Optionally, in the image acquisition and preprocessing module, the image is automatically denoised, contrast-enhanced, and color-normalized to eliminate image interference caused by differences in scanning equipment, uneven staining, or fluctuations in lighting conditions, ensuring the stability of image quality for subsequent analysis.

[0008] Optionally, in the initial screening module for positive areas, a confidence threshold can be set to screen out high-probability positive areas, thus narrowing down the scope of subsequent detailed analysis.

[0009] Optionally, in the cell morphology analysis module, morphological parameters include cell area, perimeter, nucleus-to-cytoplasm ratio, and staining intensity characteristics.

[0010] Optionally, in the 3D coordinate positioning module, the positioning report supports linkage with the digital pathology image management system, allowing users to jump to the corresponding slice location for viewing by clicking on the coordinates in the report.

[0011] Optionally, in the report generation and visualization interaction module, the report covers basic sample information, heat map of positive cell distribution, quantitative indicators of key morphological features, positive rate statistics and typical cell image examples. Users can view the analysis results under different conditions in real time by adjusting threshold parameters, region scaling, feature filtering and other operations, and can dynamically display the spatial distribution pattern of positive cells in the slide.

[0012] A method for localizing positive immunohistochemical cells in lymphoma, comprising an auxiliary immunohistochemical cell localization analysis device for lymphoma as described in any one of the above claims, including: Step 1: Read the WSI format lymphoma immunohistochemical slide images through a digital slide scanner or digital pathology image management system interface; Step 2: Based on a deep learning-based convolutional neural network model, automatically identify suspicious regions in the image that have positive cell features; Step 3: For the positive areas obtained from the initial screening, perform edge detection and morphological parameter extraction on individual cells, and combine with the morphological standards of lymphoma positive cells to further distinguish true positive cells from background impurities or non-specific stained cells. Step 4: Accurately mark the coordinates of the confirmed positive cells in the WSI image and generate a localization report containing cell location, number, and morphological parameters; Step 5: Integrate positive cell localization data, morphological parameters, and statistical analysis results to automatically generate a standardized test report. It also provides an interactive visualization interface and allows manual addition, deletion, or correction of classification results.

[0013] Optionally, in step one, the image needs to be automatically denoised, contrast-enhanced, and color-normalized to eliminate image interference caused by differences in scanning equipment, uneven staining, or fluctuations in lighting conditions, and to ensure the stability of image quality in subsequent analysis.

[0014] Optionally, in step three, the morphological parameter extraction specifically includes quantitative indicators such as cell area, perimeter, roundness, nucleocytoplasmic ratio, mean and standard deviation of staining intensity. The nucleocytoplasmic ratio is obtained by calculating the area ratio of the nucleus and cytoplasm after segmenting the cell nucleus and cytoplasm regions. The staining intensity is based on the lightness channel in the HSV color space, which is converted and statistically analyzed for grayscale values. A threshold range is set in combination with the typical dark brown staining characteristics of lymphoma-positive cells. Cells in the initial screening area are matched with multi-dimensional parameters, and cells with an area smaller than 50 μm are removed. 2 It can remove fragmented impurities, irregularly shaped cells with a roundness of less than 0.6, and weakly positive interference areas with staining intensity below the lower limit of the threshold, thereby improving the specificity of true positive cell recognition.

[0015] Optionally, in step five, the interactive visualization interface supports multi-level zooming and panning of WSI images, can intuitively display the spatial distribution density of positive cells through heatmap mode, present the morphological parameter clustering characteristics of individual cells through scatter plot mode, and supports filtering and highlighting by parameters such as staining intensity and nucleocytoplasmic ratio; the standardized test report includes basic patient information, sample number, test date, total number of positive cells, positive rate, morphological parameter statistical histogram, and typical positive cell example image, and the report format conforms to the requirements of the "Standard for Quantitative Analysis Report of Clinical Pathology"; the manual correction function supports adjusting the cell classification status by rectangular selection, polygonal outline, or point selection, and records the manual intervention operation log. The corrected results are synchronized to the location report and statistical data in real time to ensure data traceability and repeatability.

[0016] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art. Of course, any product implementing the present invention does not necessarily need to achieve all of the following advantages at the same time: 1. Improve the accuracy of positive cell identification. Through a two-layer analysis mechanism that combines deep learning models with multi-dimensional screening of morphological parameters, compared with traditional manual slide reading or single threshold judgment methods, it significantly reduces the interference of non-specific staining and background impurities, and effectively reduces the rate of missed detection and false detection. 2. To achieve automated analysis of the entire slice, for ultra-large field-of-view images in WSI format, the system narrows down the analysis range through a positive area screening module, avoiding indiscriminate traversal of the entire slice; 3. Precise positioning and backtracking in three-dimensional coordinates: Based on the positioning function of the physical coordinates of digital slides, it supports linkage with the digital pathology system to realize bidirectional jump between "report and image", which makes it easy for pathologists to directly backtrack to the original slide position for verification, solving the problem of unclear specimen position records and difficulty in reproduction in traditional slide reading; 4. It combines the flexibility of automation and human interaction. On the basis of automated analysis, it provides an interactive visual interface and manual correction function, which supports pathologists to intervene and adjust the model classification results. After correction, the system automatically updates the statistical data, which retains the efficiency of artificial intelligence and incorporates expert experience to ensure the rigor of diagnosis, in line with the collaborative model of "machine assistance + human decision-making" in clinical pathology. 5. Standardized reporting and data traceability: The generated test reports strictly follow industry standards, including quantitative indicators, distribution heatmaps, and typical cell examples. They also record manual correction logs and model iteration data, enabling full traceability of the analysis process and providing standardized data support for multi-center research data sharing, efficacy evaluation, and prognostic analysis.

[0017] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0018] The accompanying drawings described below are merely some embodiments. Those skilled in the art can obtain other drawings based on these drawings without any creative effort. In the drawings: Figure 1 Schematic diagram of an immunohistochemistry-positive cell localization analysis device; Figure 2 This is a schematic diagram of the immunohistochemistry-positive cell localization analysis method.

[0019] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation

[0020] The invention will now be described in further detail with reference to the accompanying drawings.

[0021] Please see Figure 1As shown, this embodiment provides an auxiliary device for localizing lymphoma immunohistochemically positive cells, including: The image acquisition and preprocessing module is used to connect to a digital slide scanner or digital pathology image management system, read immunohistochemical slide images in WSI format, and perform automatic noise reduction, contrast enhancement and color standardization on the images to eliminate image interference caused by differences in scanning equipment, uneven staining or fluctuations in lighting conditions, and ensure the image quality stability of subsequent analysis. The positive region screening module has a built-in deep learning-based convolutional neural network model. This model has been trained and optimized with a large number of labeled lymphoma immunohistochemical slide images. It can automatically identify suspicious regions with positive cell characteristics in the image and filter out high-probability positive regions by setting a confidence threshold, thus narrowing down the scope of subsequent fine analysis. The cell morphology analysis module performs edge detection and morphological parameter extraction on the positive areas obtained from the initial screening. Combined with the morphological standards of lymphoma positive cells, it further distinguishes true positive cells from background impurities or non-specific stained cells. The morphological parameters include cell area, perimeter, nucleus-cytoplasm ratio, and staining intensity characteristics. The three-dimensional coordinate positioning module accurately marks the coordinates of confirmed positive cells in the WSI image based on the physical coordinate information of the digital slide, and generates a positioning report containing cell location, quantity and morphological parameters. The positioning report supports linkage with the digital pathology image management system, so that clicking on the coordinates in the report can jump to the corresponding slide location for viewing. The report generation and visualization interaction module can integrate positive cell location data, morphological parameters, and statistical analysis results to automatically generate standardized test reports. It also provides an interactive visualization interface, allowing users to manually add, delete, or correct classification results. The system will recalculate the analysis indicators based on the corrected results. The report covers basic sample information, a heat map of positive cell distribution, quantitative indicators of key morphological features, positive rate statistics, and typical cell image examples. Users can view the analysis results under different conditions in real time by adjusting threshold parameters, region zooming, feature filtering, etc. It can dynamically display the spatial distribution pattern of positive cells in the slide.

[0022] To improve the accuracy of positive cell identification, a two-layer analysis mechanism combining deep learning models and multi-dimensional screening of morphological parameters is adopted. Compared with traditional manual slide reading or single threshold judgment methods, it significantly reduces the interference of non-specific staining and background impurities, and effectively reduces the rate of missed detection and false detection. To achieve automated analysis of the entire slice, for ultra-large field-of-view images in WSI format, the system narrows the analysis range through a positive area screening module, avoiding indiscriminate traversal of the entire slice; Precise three-dimensional coordinate positioning and backtracking, based on the positioning function of physical coordinates of digital slides, supports linkage with digital pathology system to realize bidirectional jump between "report and image", which makes it easy for pathologists to directly backtrack the original slide position for verification, and solves the problem of unclear specimen position record and difficulty in reproduction in traditional slide reading; It combines the flexibility of automation and human interaction. On the basis of automated analysis, it provides an interactive visual interface and manual correction function, which supports pathologists to intervene and adjust the model classification results. After correction, the system automatically updates the statistical data, which retains the efficiency of artificial intelligence and incorporates expert experience to ensure the rigor of diagnosis, in line with the collaborative model of "machine assistance + human decision-making" in clinical pathology. Standardized reporting and data traceability: The generated test reports strictly adhere to industry standards, including quantitative indicators, distribution heatmaps, and typical cell examples. They also record manual correction logs and model iteration data, ensuring full traceability of the analysis process and providing standardized data support for multi-center research data sharing, efficacy assessment, and prognostic analysis. Please see Figure 2 As shown, an auxiliary method for localizing positive lymphoma immunohistochemistry cells includes an auxiliary lymphoma immunohistochemistry positive cell localization analysis device having any of the above embodiments, comprising: Step 1: Read the WSI format lymphoma immunohistochemical slide images through a digital slide scanner or digital pathology image management system interface. Automatic noise reduction, contrast enhancement, and color standardization of the images are required to eliminate image interference caused by differences in scanning equipment, uneven staining, or fluctuations in lighting conditions, and to ensure the image quality stability of subsequent analysis. Step 2: Based on a deep learning-based convolutional neural network model, automatically identify suspicious regions in the image that have positive cell features; Step 3: For the positive areas obtained from the initial screening, perform edge detection and morphological parameter extraction on individual cells, and combine with the morphological standards of lymphoma positive cells to further distinguish true positive cells from background impurities or non-specific stained cells. Morphological parameter extraction specifically includes quantitative indicators such as cell area, perimeter, roundness, nucleocytoplasmic ratio, mean and standard deviation of staining intensity. The nucleocytoplasmic ratio is obtained by calculating the area ratio of the nucleus and cytoplasm after segmenting the cell nucleus. Staining intensity is based on the lightness channel in the HSV color space, which is converted and statistically analyzed for grayscale values. A threshold range is set in combination with the typical dark brown staining characteristics of lymphoma-positive cells. Cells in the initial screening area are matched with multi-dimensional parameters, and cells with an area smaller than 50 μm are removed. 2 It can remove fragments and impurities, irregularly shaped cells with a roundness of less than 0.6, and weakly positive interference areas with staining intensity below the lower limit of the threshold, thereby improving the specificity of true positive cell recognition. Step 4: Accurately mark the coordinates of the confirmed positive cells in the WSI image and generate a localization report containing cell location, number, and morphological parameters; Step 5: Integrate positive cell localization data, morphological parameters, and statistical analysis results to automatically generate a standardized test report. It also provides an interactive visualization interface and allows manual addition, deletion, or correction of classification results. The interactive visualization interface supports multi-level zooming and panning of WSI images. It can intuitively display the spatial distribution density of positive cells through heatmap mode and present the morphological parameter clustering characteristics of individual cells through scatter plot mode. It also supports filtering and highlighting based on parameters such as staining intensity and nucleocytoplasmic ratio. The standardized test report includes basic patient information, sample number, test date, total number of positive cells, positive rate, morphological parameter statistical histogram, and typical positive cell example image. The report format complies with the requirements of the "Standard for Quantitative Analysis Reporting in Clinical Pathology". The manual correction function allows adjustment of cell classification status through rectangular selection, polygonal outline, or point selection, and records the manual intervention operation log. The corrected results are synchronized to the location report and statistical data in real time to ensure data traceability and repeatability.

[0023] This method, through the deep integration of multimodal image processing and intelligent analysis technologies, automates the entire process from whole-slice scanning to quantitative analysis of immunohistochemically positive lymphoma cells. In the image preprocessing stage, a combination of adaptive median filtering and the Retinex contrast enhancement algorithm effectively suppresses scanning noise and improves the detail recognition of weakly positive areas. Color standardization is based on the CIELAB color space to calibrate the hematoxylin-eosin staining channel, ensuring that the staining depth deviation between different batches of slices is controlled within a clinically acceptable range of ΔE < 3. The deep learning model employs an improved U-Net architecture, introducing an attention mechanism module in the encoder to enhance the feature extraction weights for positive cell nuclei. Furthermore, by employing a transfer learning strategy and pre-training with 100,000 annotated pathological images from the TCGA database, the system achieved a positive region recognition recall of 94.7% and a precision of 89.3% on the test set. The cell segmentation process innovatively combines the watershed algorithm with edge constraints from cell membrane marker staining channels, resolving the adhesion and separation problem of overlapping cells. The edge detection error of a single cell is controlled within ±2 pixels. The morphological parameter analysis module incorporates a dynamic threshold adjustment mechanism, automatically matching parameter weights based on the morphological differences of different lymphoma subtypes (such as diffuse large B-cell lymphoma and follicular lymphoma). For example, for the small lymphocyte characteristics of mantle cell lymphoma, the area screening threshold is lowered to 30 μm. 2The system also increases the nucleocytoplasmic ratio weighting coefficient to 0.35 and has a time series analysis function, which can perform longitudinal comparison of slide data from different treatment stages of the same patient. By analyzing the change curve of the number of positive cells and the fluctuation of the morphological parameter entropy value, it can help assess the treatment response and prognostic risk. In the clinical validation experiment, compared with the manual counting results of three senior pathologists, the method achieved an intragroup correlation coefficient (ICC) of 0.92, and the analysis time was shortened from an average of 45 minutes for traditional manual slide reading to 8 minutes. Moreover, it still maintained a detection sensitivity of 87.6% in samples with a low positive rate (<5%), which significantly reduced the risk of missed diagnosis.

[0024] This invention is not limited to the embodiments described above. Anyone should understand that structural changes made under the guidance of this invention, and any technical solutions that are the same as or similar to this invention, fall within the protection scope of this invention. Technical aspects, shapes, and structures not described in detail in this invention are all publicly known technologies.

Claims

1. A device for assisting in the localization analysis of immunohistochemically positive cells in lymphoma, characterized in that, include: The image acquisition and preprocessing module is used to connect to a digital slide scanner or digital pathology image management system and read WSI format immunohistochemical slide images; The positive region screening module has a built-in deep learning-based convolutional neural network model. This model has been trained and optimized with a large number of labeled lymphoma immunohistochemical slice images and can automatically identify suspicious regions with positive cell characteristics in the image. The cell morphology analysis module performs edge detection and morphological parameter extraction on individual cells in the positive areas obtained from the initial screening. Combined with the morphological standards of lymphoma positive cells, it further distinguishes true positive cells from background impurities or non-specific stained cells. The three-dimensional coordinate positioning module accurately marks the coordinates of confirmed positive cells in the WSI image based on the physical coordinate information of the digital slices, and generates a positioning report containing cell location, quantity, and morphological parameters. The report generation and visualization interaction module can integrate positive cell location data, morphological parameters and statistical analysis results to automatically generate standardized test reports. It also provides an interactive visualization interface and allows manual addition, deletion or correction of classification results. The system will recalculate the analysis indicators based on the corrected results.

2. The auxiliary lymphoma immunohistochemical positive cell localization analysis device according to claim 1, characterized in that, In the image acquisition and preprocessing module, images are automatically denoised, contrast-enhanced, and color-normalized to eliminate image interference caused by differences in scanning equipment, uneven staining, or fluctuations in lighting conditions, ensuring the stability of image quality for subsequent analysis.

3. The auxiliary lymphoma immunohistochemical positive cell localization analysis device according to claim 1, characterized in that, In the initial screening module for positive areas, high-probability positive areas are selected by setting a confidence threshold, thus narrowing down the scope of subsequent detailed analysis.

4. The auxiliary lymphoma immunohistochemical positive cell localization analysis device according to claim 1, characterized in that, In the cell morphology analysis module, morphological parameters include cell area, perimeter, nucleus-to-cytoplasm ratio, and staining intensity characteristics.

5. The auxiliary lymphoma immunohistochemical positive cell localization analysis device according to claim 1, characterized in that, In the 3D coordinate positioning module, the positioning report supports linkage with the digital pathology image management system, allowing users to jump to the corresponding slice location by clicking on the coordinates in the report.

6. The auxiliary lymphoma immunohistochemical positive cell localization analysis device according to claim 1, characterized in that, In the report generation and visualization interaction module, the report covers basic sample information, heat map of positive cell distribution, quantitative indicators of key morphological features, positive rate statistics and typical cell image examples. It allows users to view the analysis results under different conditions in real time by adjusting threshold parameters, region scaling, feature filtering and other operations, and can dynamically display the spatial distribution pattern of positive cells in the slide.

7. A method for localizing positive immunohistochemical cells in lymphoma, applicable to any of the auxiliary immunohistochemical cell localization devices for lymphoma as described in claims 1-6, comprising: Step 1: Read the WSI format lymphoma immunohistochemical slide images through a digital slide scanner or digital pathology image management system interface; Step 2: Based on a deep learning-based convolutional neural network model, automatically identify suspicious regions in the image that have positive cell features; Step 3: For the positive areas obtained from the initial screening, perform edge detection and morphological parameter extraction on individual cells, and combine with the morphological standards of lymphoma positive cells to further distinguish true positive cells from background impurities or non-specific stained cells. Step 4: Accurately mark the coordinates of the confirmed positive cells in the WSI image and generate a localization report containing cell location, number, and morphological parameters; Step 5: Integrate positive cell localization data, morphological parameters, and statistical analysis results to automatically generate a standardized test report. It also provides an interactive visualization interface and allows manual addition, deletion, or correction of classification results.

8. The method for auxiliary immunohistochemical localization analysis of lymphoma positive cells according to claim 7, characterized in that, In step one, the image needs to be automatically denoised, contrast-enhanced, and color-normalized to eliminate image interference caused by differences in scanning equipment, uneven coloring, or fluctuations in lighting conditions, and to ensure the stability of image quality for subsequent analysis.

9. The method for auxiliary immunohistochemical localization analysis of lymphoma positive cells according to claim 7, characterized in that, In step three, morphological parameter extraction specifically includes quantitative indicators such as cell area, perimeter, roundness, nucleocytic ratio, mean and standard deviation of staining intensity. The nucleocytic ratio is obtained by calculating the area ratio of the nucleus and cytoplasm after segmenting the cell nucleus and cytoplasm regions. Staining intensity is based on the lightness channel in the HSV color space, which is converted and statistically analyzed for grayscale values. A threshold range is set in combination with the typical dark brown staining characteristics of lymphoma-positive cells. Cells in the initial screening area are matched with multi-dimensional parameters, and cells with an area smaller than 50 μm are removed. 2 It can remove fragmented impurities, irregularly shaped cells with a roundness of less than 0.6, and weakly positive interference areas with staining intensity below the lower limit of the threshold, thereby improving the specificity of true positive cell recognition.

10. The method for auxiliary immunohistochemical localization analysis of lymphoma positive cells according to claim 7, characterized in that, In step five, the interactive visualization interface supports multi-level zooming and panning of WSI images. It can visually display the spatial distribution density of positive cells through heatmap mode, present the morphological parameter clustering characteristics of individual cells through scatter plot mode, and support filtering and highlighting by parameters such as staining intensity and nucleocytoplasmic ratio. The standardized test report includes basic patient information, sample number, test date, total number of positive cells, positive rate, morphological parameter statistical histogram, and typical positive cell example image. The report format conforms to the requirements of the "Standard for Quantitative Analysis Report of Clinical Pathology". The manual correction function supports adjusting the cell classification status by rectangular selection, polygonal outline, or point selection, and records the manual intervention operation log. The corrected results are synchronized to the location report and statistical data in real time to ensure data traceability and repeatability.

Images