Device and method for detecting and quantifying liver lobule inflammation in liver pathological slice images

Abstract

The present invention discloses a device and method for detecting and quantifying inflammation in liver lobule of liver pathological slice images. The device of the present invention can realize automatic identification of inflammatory foci in panoramic digitized images of liver pathological slices through modules such as target detection and search for the most intensively inflamed areas, and perform quantitative statistics of their number and position, providing automated assistance for doctors to quantitatively diagnose the degree of inflammation in liver lobule of pathological slices. The present invention uses artificial intelligence methods and image processing related algorithms to objectively and accurately provide quantitative data on the degree of inflammation in liver lobule. The present invention is fast, accurate, highly automated, highly repeatable, and has broad application prospects.

Description

Technical Field

[0001] The present invention belongs to the field of medical image analysis and artificial intelligence, and in particular relates to a device and method for detecting and quantifying liver lobule inflammation in liver pathological slice images. Background Art

[0002] Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease in the world. Among ordinary adults, the prevalence of NAFLD is between 6.3% and 45%, of which 10% to 30% is non-alcoholic steatohepatitis (NASH). With the continuous improvement of the quality of life, the excess nutritional intake has caused the incidence of this disease to show an upward trend.

[0003] Liver pathology sections are the clinical gold standard for pathologists to diagnose whether patients have NASH. Among them, judging the degree of liver lobular inflammation through liver pathology section images requires doctors to have rich diagnostic experience; therefore, generally under the standard clinical diagnosis process, only trained and certified doctors can issue diagnostic reports. In the diagnosis process, due to long-term, large-scale repetitive manual recognition operations of film reading, these experienced doctors face judgment fatigue, resulting in increased deviation and instability of diagnostic results, which will seriously lead to misdiagnosis. Doctors without training and certification in other secondary hospitals are even less able to provide pathological diagnosis services, resulting in a serious lack of medical resources in the institution. At present, the manual and accurate diagnosis of liver pathology section images by doctors alone is unstable and low-throughput, and there is a serious lack of auxiliary tools to assist doctors in objective, high-throughput analysis and diagnosis of liver lobular inflammation.

[0004] In recent years, computer vision has been used to detect, identify and segment targets in digital media materials such as images and videos. As a key research direction for the development of medical image processing technology, medical image segmentation tools can assist doctors in making reasonable assessments of patients' pathological sections, which is of great value in assisting doctors' clinical diagnosis and disease assessment as well as subsequent corresponding medication decisions. At present, there are invasive pathological section image diagnosis methods and non-invasive serum-related inflammatory biomarker detection methods for diagnosing liver lobular inflammation, such as biochemical indicators such as alanine aminotransferase (ALT), aspartate aminotransferase (AST), glutamyl transpeptidase (GGT / L-γ) and fibrinogen (FBG), but the specificity is very poor, so liver pathological sections are still the clinical gold standard for diagnosing the degree of liver lobular inflammation. At present, neural network model methods in the field of deep learning have been used to analyze medical image information to achieve the grading of the degree of liver lobular inflammation, but there is no quantitative determination method for the degree of liver lobular inflammation and the number and location of inflammatory foci. Summary of the invention

[0005] The object of the present invention is to provide a device and method for detecting and quantifying liver lobule inflammation in digitized images of liver pathological sections.

[0006] The first aspect of the present invention provides a device for detecting and quantifying liver lobular inflammation in liver pathological slice images, characterized in that it comprises:

[0007] (1) Image acquisition and storage module: used to acquire and store full-scan digital images of liver tissue pathological sections, and obtain panoramic digital images of the sections;

[0008] (2) Image segmentation module: used to segment the panoramic digitized image into small images without overlapping areas, and record the coordinate positions of the small images relative to the panoramic digitized image;

[0009] (3) Image labeling and classification module: The pathologist labels all inflammatory foci in all small images, discards small images without inflammatory foci, and assembles all labeled small images into a dataset, which is then divided into a training set and a test set.

[0010] (4) Image training module: Use the training set images to input a Faster R-CNN model for training. The training makes the model loss error converge to 0.26 and save the model.

[0011] (5) Target detection module: All small images obtained from module (2) are sequentially input into the Faster R-CNN model trained in module (4) for target detection, identifying all inflammatory foci in each small image and obtaining their coordinate positions in the small image. By combining the coordinate positions of each small image in module (2) in each panoramic image, the absolute positions of all identified inflammatory foci in each panoramic digitized image can be obtained, and their absolute positions are stored in the position list;

[0012] (6) Search module: The absolute positions of all inflammatory foci obtained in module (5) in each panoramic image are used to search for the most densely inflamed area in the panoramic image:

[0013] (a) Select an inflammatory focus in the order of the inflammatory focus location list (X j ) and calculate all other inflammatory foci (Y i ) Relative to the Euclidean distance of the selected inflammatory focus, the calculation formula of the Euclidean distance is: i, j are the position list numbers and i≠j, n is the number of all inflammatory foci in a panoramic image. The field of view size based on the most densely populated area of ​​inflammatory foci is 1.6750×1.6750mm 2 The standard is set to record the situation where the Euclidean distance between two inflammatory foci is less than 0.8375 mm as the situation where the inflammatory foci are dense and store the records in a new list until the statistics are completed. j All Euclidean distance judgments;

[0014] (b) Complete all X according to step (a) j The Euclidean distance is calculated and the record is stored in each corresponding X j in the new list;

[0015] (c) Compare and find the list with the largest number of inflammatory foci, and identify the corresponding inflammatory foci X j The position is taken as the coordinate center of the most intensively inflamed area in the panoramic image;

[0016] (7) Module for estimating the degree of inflammation: Based on the most densely inflamed area calculated and located in module (6), the number of inflammatory foci in the most densely inflamed area can be obtained by counting, and the degree of inflammation can be estimated based on the following scoring criteria:

[0017] The number of inflammatory foci = 0: no inflammation; the number of inflammatory foci < 2: mild inflammation; the number of inflammatory foci = 2-4: moderate inflammation; the number of inflammatory foci > 4: severe inflammation;

[0018] (8) Debinarization processing module: Debinarize the original image of the small image obtained by module (2) and set its threshold. Count the number of positive pixel values ​​in the debinarized image to obtain the liver tissue area, and sum up the liver tissue areas of all small images in a panoramic image to obtain a complete liver tissue area;

[0019] (9) Module for calculating the density of inflammatory foci: The total number of inflammatory foci in the panoramic image obtained by module (5) is divided by the liver tissue area in the panoramic image obtained by module (8) to obtain the average number of inflammatory foci per unit area, i.e., the density of inflammatory foci.

[0020] Furthermore, in module (1), observation is performed under a microscope using a 5-fold objective lens magnification, and a camera is used to perform full scan digital image storage of liver tissue pathological sections.

[0021] Furthermore, the resolution of the image in module (1) reaches above 1.0952 micrometers per pixel.

[0022] Furthermore, in module (2), the size of the thumbnail cut is set to 1024×1024 pixels.

[0023] Furthermore, in module (3), 80% of the small images in the data set are randomly selected to form an artificial intelligence deep learning training set, and the remaining 20% ​​of the small images are used as a deep learning test set.

[0024] Furthermore, the threshold of debinarization in module (8) is 200.

[0025] The second aspect of the present invention discloses a method for detecting and quantifying hepatic lobular inflammation in liver pathological slice images, which is characterized by comprising:

[0026] (1) Observe under a microscope using a 5x objective lens magnification and use a camera to perform full scan digital image storage of liver tissue pathology sections to obtain a panoramic digital image of the section. In order to ensure the accuracy of identifying inflammatory foci, the image resolution should be above 1.0952 μm / pixel;

[0027] (2) Setting the image size to 1024 × 1024 pixels, dividing the panoramic digitized image into small images without overlapping areas, and recording the coordinate positions of the small images relative to the panoramic image;

[0028] (3) Segmenting all the panoramic digital images collected, collated and scanned into small images according to the method described in step (2), having a pathologist mark all the inflammatory foci in all the small images, discarding small images without inflammatory foci, and forming a data set with all the marked small images, randomly extracting 80% of the small images in the data set to form an artificial intelligence deep learning training set, and the remaining 20% ​​of the small images as a deep learning test set;

[0029] (4) Use the training set images to input a Faster R-CNN model for training. The training makes the model loss error converge to 0.26 and save the model.

[0030] (5) All the small images obtained in step (2) are sequentially input into the Faster R-CNN model trained in step (4) for target detection, and all the inflammatory foci in each small image are identified and their coordinate positions in the small image are obtained. Combined with the coordinate position of each small image in step (2) in each panoramic image, the absolute positions of all identified inflammatory foci in each panoramic digitized image can be obtained, and their absolute positions are stored in the position list;

[0031] (6) The absolute positions of all inflammatory foci obtained in step (5) in each panoramic image are used to search for the most densely inflamed area of ​​the panoramic image:

[0032] (a) Select an inflammatory focus in the order of the inflammatory focus location list (X j ) and calculate all other inflammatory foci (Y i ) Relative to the Euclidean distance of the selected inflammatory focus, the calculation formula of the Euclidean distance is: i, j are the position list numbers and i≠j, n is the number of all inflammatory foci in a panoramic image; the field of view size based on the most densely populated area of ​​inflammatory foci is 1.6750×1.6750mm 2 The standard is set to record the situation where the Euclidean distance between two inflammatory foci is less than 0.8375 mm as the situation where the inflammatory foci are dense and store the records in a new list until the statistics are completed. j All Euclidean distance judgments;

[0033] (b) Complete all X according to step (a) j The Euclidean distance is calculated and the record is stored in each corresponding X j New list of;

[0034] (c) Compare and find the list with the largest number of inflammatory foci, and identify the corresponding inflammatory foci X j The position is taken as the coordinate center of the most intensively inflamed area in the panoramic image;

[0035] (7) According to the calculation in step (6), the number of inflammatory foci in the most densely populated area of ​​inflammation can be obtained by counting, and the degree of inflammation can be estimated according to the following scoring criteria:

[0036] The number of inflammatory foci = 0: no inflammation; the number of inflammatory foci < 2: mild inflammation; the number of inflammatory foci = 2-4: moderate inflammation; the number of inflammatory foci > 4: severe inflammation;

[0037] (8) In order to obtain the area of ​​liver tissue for calculating the average number of inflammatory foci per unit area, the original image of the small image obtained in step (2) is debinarized, and the threshold is set to 200. The number of positive pixel values ​​in the debinarized image is counted to obtain the liver tissue area, and the liver tissue areas of all small images in a panoramic image are summed up to obtain a complete liver tissue area;

[0038] (9) By dividing the total number of inflammatory foci in the panoramic image obtained in step (5) by the liver tissue area in the panoramic image obtained in step (8), the average number of inflammatory foci per unit area, i.e., the density of inflammatory foci, can be obtained.

[0039] The third aspect of the present invention discloses the use of the above-mentioned device or method for liver pathological section image analysis.

[0040] The fourth aspect of the present invention discloses the use of the above-mentioned device or method for the detection and quantitative analysis of liver lobular inflammation.

[0041] Compared with the prior art, the present invention has the following advantages:

[0042] (1) The present invention uses artificial intelligence methods and image processing algorithms to objectively and accurately provide quantitative data on the degree of inflammation in the liver lobules. It also performs quantitative statistical methods on the number and location of the data, providing automated assistance to doctors in quantitatively diagnosing the degree of inflammation in the liver lobules in pathological sections.

[0043] (2) The present invention is fast, accurate, highly automated, and highly repeatable, and has broad application prospects.

[0044] (3) The present invention has innovatively developed a device and method for quantitatively measuring the degree of liver lobular inflammation and the number and location of inflammatory foci, filling the current research and technology gap in this field. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 To divide the panoramic image into several small images and record their coordinate positions in the panoramic image;

[0046] Figure 2 Mark the inflammatory foci in small images for doctors and divide the small images into training sets and test sets;

[0047] Figure 3 The Faster R-CNN model is trained and verified on the test set images, and its accuracy AP 50 65%;

[0048] Figure 4 To record the absolute coordinate position of the marked inflammatory focus in each panoramic digital image;

[0049] Figure 5 To search for the most intensively inflamed area in the panoramic image;

[0050] Figure 6 This is a graph showing the difference in the grading of inflammation levels of hepatic lobular inflammatory foci determined by doctors and identified by computers;

[0051] Figure 7 This is a graph showing the relationship between the number of inflammatory foci in the most intensively inflamed area and the grade of lobular inflammation;

[0052] Figure 8 This is a graph showing the relationship between the density of inflammatory foci and the grade of liver lobular inflammation;

[0053] Fig. 9 The present invention is a flow chart of the method for grading and quantitatively analyzing inflammation content. DETAILED DESCRIPTION

[0054] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples. The experimental methods in the following examples without specifying specific conditions are carried out according to conventional methods and conditions, or selected according to the product specifications.

[0055] Example 1

[0056] An artificial intelligence target detection device that learns and automatically identifies all hepatic lobular inflammatory foci in the panoramic digitized image of liver pathological sections, and performs quantitative statistics on the number and location of all inflammatory foci, and gives scores to grade the degree of inflammation. It specifically includes the following modules:

[0057] (1) Image acquisition and storage module: Scan liver tissue slices at a 5x objective magnification to obtain a panoramic digital image of the slice. To ensure the accuracy of identifying inflammatory foci, the image resolution should be above 1.0952 μm / pixel;

[0058] (2) Image segmentation module: segment the panoramic digitized image obtained in module (1) into a number of small images without overlapping areas, and record the coordinate positions of the small images relative to the panoramic image ( Figure 1 ). Set the small image cutting size to 1024×1024 pixels;

[0059] (3) Image labeling and classification module: The pathologist labels all inflammatory foci in the small images obtained in module (2), discards small images without inflammatory foci, and forms a data set with all the labeled small images. Small images in the data set are randomly selected and divided into an artificial intelligence deep learning training set (accounting for 80%) and a test set (accounting for 20%) ( Figure 2 );

[0060] (4) Image training module: Use the training set images obtained in module (3) to input a Faster R-CNN model for training. The model is trained and optimized, and the accuracy AP is verified on the test set. 50 65% ( Figure 3 );

[0061] (5) Target detection module: All small images obtained from module (2) are sequentially input into the Faster R-CNN model trained in module (4) for target detection, and the coordinate positions of all inflammatory foci in the small images are obtained. Combined with the coordinate positions of each small image in module (2) in each panoramic image, the absolute positions of all identified inflammatory foci in each panoramic digitized image can be obtained, and their absolute positions are stored in the position list ( Figure 4 );

[0062] (6) Search module: The absolute positions of all inflammatory foci obtained in module (5) in each panoramic image are used to search for the most densely inflamed area in the panoramic image:

[0063] (a) Select an inflammatory focus in the order of the inflammatory focus location list (X j ) and calculate all other inflammatory foci (Y i ) Relative to the Euclidean distance of the selected inflammatory focus, the calculation formula of the Euclidean distance is: i, j are the position list numbers and i≠j, n is the number of all inflammatory foci in a panoramic image. The field of view size based on the most densely populated area of ​​inflammatory foci is 1.6750×1.6750mm 2 The standard is set to record the situation where the Euclidean distance between two inflammatory foci is less than 0.8375 mm as the situation where the inflammatory foci are dense and store the records in a new list until the statistics are completed. j All Euclidean distance judgments;

[0064] (b) Complete all X according to step (a) j The Euclidean distance is calculated and the record is stored in each corresponding X j in the new list;

[0065] (c) Compare and find the list with the largest number of inflammatory foci, and identify the corresponding inflammatory foci X j The position of the most densely inflamed area in the panoramic image is taken as the coordinate center ( Figure 5 );

[0066] (7) Module for estimating the degree of inflammation: according to the most densely inflamed area located in module (6), the number of inflammatory foci in the most densely inflamed area is counted, and the degree of inflammation is estimated according to the following scoring criteria: the number of inflammatory foci = 0, <2, =2-4, >4 represents no inflammation, mild inflammation, moderate inflammation, and severe inflammation, respectively;

[0067] (8) Debinarization processing module: In order to obtain the area of ​​liver tissue for calculating the average number of inflammatory foci per unit area, the original image of the small image obtained by module (2) is debinarized, and the threshold is set to 200. The number of positive pixel values ​​in the debinarized image is counted to obtain the liver tissue area, and the liver tissue areas of all small images in a panoramic image are summed up to obtain a complete liver tissue area;

[0068] (9) Module for calculating the density of inflammatory foci: the total number of inflammatory foci in the panoramic image obtained by module (5) is divided by the liver tissue area in the panoramic image obtained by module (8), so as to obtain the average number of inflammatory foci per unit area, i.e., the density of inflammatory foci;

[0069] (10) The small image obtained by module (2) is sequentially subjected to the operations of modules (5), (6), (7), etc., to complete the grading of the degree of inflammation of the hepatic lobule inflammation foci in about 542 liver tissue pathological sections, and with reference to the grading of the hepatic lobule inflammation prejudged by the doctor, a relationship diagram between the grading of the degree of inflammation of the hepatic lobule inflammation foci prejudged by the doctor and identified by the computer is obtained ( Figure 6 );

[0070] (11) The small image obtained by module (2) was subjected to the operations of modules (5), (6), and (7) in sequence to obtain the number of inflammatory foci in the most intensively inflamed area in about 542 liver tissue pathological sections, and the hepatic lobular inflammation grade prejudged by the doctor was used as a reference for grouping, thereby obtaining a relationship diagram between the number of inflammatory foci in the most intensively inflamed area and the hepatic lobular inflammation grade ( Figure 7 );

[0071] (12) The small image obtained by module (2) was sequentially subjected to the operations of modules (5), (6), (7), (8), and (9) to obtain the density of inflammatory foci in approximately 542 liver tissue pathology sections, and the hepatic lobular inflammation grade prejudged by the doctor was used as a reference for grouping, thereby obtaining a relationship diagram between the density of inflammatory foci and the hepatic lobular inflammation grade ( Figure 8 ).

[0072] Example 2

[0073] An artificial intelligence target detection method learns and automatically identifies all hepatic lobular inflammatory foci in the panoramic digitized image of liver pathological sections, and performs quantitative statistics on the number and location of all inflammatory foci, and scores them to give a grade of inflammation. The method specifically includes the following steps:

[0074] (1) Scan the liver tissue slice at a 5x objective magnification to obtain a panoramic digital image of the slice. To ensure the accuracy of identifying inflammatory foci, the image resolution should be above 1.0952 μm / pixel;

[0075] (2) Segment the panoramic digitized image obtained in step (1) into a number of small images without overlapping areas, and record the coordinate positions of the small images relative to the panoramic image ( Figure 1 ). Set the small image cutting size to 1024×1024 pixels;

[0076] (3) A pathologist marks all inflammatory foci in the small images obtained in step (2), discards small images without inflammatory foci, and forms a data set with all the marked small images. Small images in the data set are randomly selected and divided into an artificial intelligence deep learning training set (accounting for 80%) and a test set (accounting for 20%) ( Figure 2 );

[0077] (4) Use the training set images obtained in step (3) to input a Faster R-CNN model for training. The accuracy AP of the model after training and optimization is verified on the test set. 50 65% ( Figure 3 );

[0078] (5) All the small images obtained in step (2) are sequentially input into the Faster R-CNN model trained in step (4) for target detection, and the coordinate positions of all inflammatory foci in the small images are obtained. Combined with the coordinate positions of each small image in step (2) in each panoramic image, the absolute positions of all identified inflammatory foci in each panoramic digitized image can be obtained, and their absolute positions are stored in the position list ( Figure 4 );

[0079] (6) The absolute positions of all inflammatory foci obtained in step (5) in each panoramic image are used to search for the most densely inflamed area of ​​the panoramic image:

[0080] (a) Select an inflammatory focus in the order of the inflammatory focus location list (X j ) and calculate all other inflammatory foci (Y i ) Relative to the Euclidean distance of the selected inflammatory focus, the calculation formula of the Euclidean distance is: i, j are the position list numbers and i≠j, n is the number of all inflammatory foci in a panoramic image. The field of view size based on the most densely populated area of ​​inflammatory foci is 1.6750×1.6750mm 2 The standard is set to record the situation where the Euclidean distance between two inflammatory foci is less than 0.8375 mm as the situation where the inflammatory foci are dense and store the records in a new list until the statistics are completed. j All Euclidean distance judgments;

[0081] (b) Complete all X according to step (a) j The Euclidean distance is calculated and the record is stored in each corresponding X j in the new list;

[0082] (c) Compare and find the list with the largest number of inflammatory foci, and identify the corresponding inflammatory foci X j The position of the most densely inflamed area in the panoramic image is taken as the coordinate center ( Figure 5 );

[0083] (7) According to the most densely inflamed area located in step (6), the number of inflammatory foci in the most densely inflamed area is counted, and the degree of inflammation is estimated according to the following scoring criteria: the number of inflammatory foci = 0, <2, =2-4, >4 represents no inflammation, mild inflammation, moderate inflammation, and severe inflammation, respectively;

[0084] (8) In order to obtain the area of ​​liver tissue for calculating the average number of inflammatory foci per unit area, the original image of the small image obtained in step (2) is debinarized, and the threshold is set to 200. The number of positive pixel values ​​in the debinarized image is counted to obtain the liver tissue area, and the liver tissue areas of all small images in a panoramic image are summed up to obtain a complete liver tissue area;

[0085] (9) Dividing the total number of inflammatory foci in the panoramic image obtained in step (5) by the liver tissue area in the panoramic image obtained in step (8) can obtain the average number of inflammatory foci per unit area, i.e., the density of inflammatory foci;

[0086] (10) The small image obtained in step (2) is sequentially subjected to the operations of steps (5), (6), (7), etc., to complete the grading of the inflammation degree of the hepatic lobule inflammation foci in about 542 liver tissue pathological sections, and with reference to the hepatic lobule inflammation grading prejudged by the doctor, a relationship diagram of the difference between the grading of the inflammation degree of the hepatic lobule inflammation foci prejudged by the doctor and identified by the computer is obtained ( Figure 6 );

[0087] (11) The small image obtained in step (2) is sequentially subjected to steps (5), (6), (7), etc., to obtain the number of inflammatory foci in the most intensively inflamed area in approximately 542 liver tissue pathological sections, and the hepatic lobular inflammation grade pre-judged by the doctor is used as a reference for grouping, thereby obtaining a relationship diagram between the number of inflammatory foci in the most intensively inflamed area and the hepatic lobular inflammation grade ( Figure 7 );

[0088] (12) The small image obtained in step (2) is sequentially subjected to steps (5), (6), (7), (8), (9), etc., to obtain the density of inflammatory foci in approximately 542 liver tissue pathology sections, and the hepatic lobular inflammation grade pre-judged by the doctor is used as a reference for grouping, thereby obtaining a relationship diagram between the density of inflammatory foci and the hepatic lobular inflammation grade ( Figure 8 ).

[0089] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A device for detecting and quantifying liver lobular inflammation in liver pathological slice images, characterized in that: include: (1) Image acquisition and storage module: used to acquire and store full-scan digital images of liver tissue pathological sections, and obtain panoramic digital images of the sections; (2) Image segmentation module: used to segment the panoramic digitized image into small images without overlapping areas, and record the coordinate positions of the small images relative to the panoramic digitized image; (3) Image labeling and classification module: Pathologists label all inflammatory foci in all small images, discard small images without inflammatory foci, and form a data set with all labeled small images. The data set is divided into a training set and a test set. (4) Image training module: Use the training set images to input a Faster R-CNN model for training. The training makes the model loss error converge to 0.26 and save the model. (5) Target detection module: All small images obtained from the image segmentation module are sequentially input into the Faster R-CNN model trained in the image training module for target detection. All inflammatory foci in each small image are identified and their coordinate positions in the small image are obtained. Combined with the coordinate positions of each small image in the image segmentation module in each panoramic image, the absolute positions of all identified inflammatory foci in each panoramic digitized image can be obtained and stored in the position list. (6) Search module: The absolute positions of all inflammatory foci obtained in each panoramic image by the target detection module are used to search for the most densely inflamed area in the panoramic image: (a) Select an inflammatory focus in the order of the inflammatory focus location list X j location, counting all other inflammatory foci Y i Relative to the Euclidean distance of the selected inflammatory focus, the calculation formula of the Euclidean distance is: dist(X j ,Y i=1→n-1 ) = , i , j Number the list of positions and i ≠ j , n is the number of all inflammatory foci in a panoramic image; the field of view based on the densest area of ​​inflammatory foci is 1.6750×1.6750 mm 2 The standard was set to record the situation where the Euclidean distance between two inflammatory foci was less than 0.8375 mm as the situation where the inflammatory foci were dense and store the records in a new list until the statistics were completed. X j All Euclidean distance judgments; (b) Complete all steps in step (a) X j The Euclidean distance is calculated and the records are stored in each corresponding X j in the new list; (c) Compare and find the list with the largest number of densely populated inflammatory foci and identify the corresponding inflammatory foci X j The position is taken as the coordinate center of the most intensively inflamed area in the panoramic image; (7) Module for estimating the degree of inflammation: Based on the most densely inflamed area calculated and located in the search module, the number of inflammatory foci in the most densely inflamed area can be counted, and the degree of inflammation can be estimated based on the following scoring criteria: The number of inflammatory foci = 0: no inflammation; the number of inflammatory foci < 2: mild inflammation; the number of inflammatory foci = 2-4: moderate inflammation; the number of inflammatory foci > 4: severe inflammation; (8) Debinarization processing module: Debinarizes the original image of the small image obtained by the image segmentation module and sets its threshold; The obtained debinarized image is used to count the number of positive pixel values ​​to obtain the liver tissue area, and the liver tissue areas of all small images in a panoramic image are summed up to obtain a complete liver tissue area; (9) Module for calculating the density of inflammatory foci: The total number of inflammatory foci in the panoramic image obtained by the target detection module is divided by the liver tissue area obtained by the debinarization processing module to obtain the average number of inflammatory foci per unit area, i.e., the density of inflammatory foci.

2. The device according to claim 1, wherein: In the image acquisition and storage module, observation is performed under a microscope using a 5x objective lens magnification, and a camera is used to perform full scan digital image storage of liver tissue pathology sections.

3. The device according to claim 1, wherein: The resolution of the image in the image acquisition and storage module reaches above 1.0952 microns / pixel.

4. The device according to claim 1, wherein: In the image segmentation module, the size of the small image cut is set to 1024×1024 pixels.

5. The device according to claim 1, wherein: In the image labeling and classification module, 80% of the small images in the data set are randomly selected to form the artificial intelligence deep learning training set, and the remaining 20% ​​of the small images are used as the deep learning test set.

6. The device according to claim 1, wherein: The debinarization threshold in the debinarization processing module is 200.

7. A method for detecting and quantifying hepatic lobular inflammation in liver pathological slice images, characterized in that: include: (1) Observe under a microscope using a 5x objective lens magnification and use a camera to perform full scan digital image storage of liver tissue pathology sections to obtain a panoramic digital image of the section. In order to ensure the accuracy of identifying inflammatory foci, the image resolution should be above 1.0952 μm / pixel; (2) Setting the image size to 1024 × 1024 pixels, dividing the panoramic digitized image into small images without overlapping areas, and recording the coordinate positions of the small images relative to the panoramic image; (3) All the panoramic digital images collected, collated and scanned are divided into small images according to the method described in step (2), and the pathologist marks all the inflammatory foci in all the small images, discards the small images without inflammatory foci, and forms a data set with all the marked small images. 80% of the small images in the data set are randomly selected to form an artificial intelligence deep learning training set, and the remaining 20% ​​of the small images are used as a deep learning test set; (4) Use the training set images to input a Faster R-CNN model for training. The training makes the model loss error converge to 0.26 and save the model. (5) All the small images obtained in step (2) are sequentially input into the Faster R-CNN model trained in step (4) for target detection, all the inflammatory foci in each small image are identified and their coordinate positions in the small image are obtained. Combined with the coordinate position of each small image in step (2) in each panoramic image, the absolute positions of all identified inflammatory foci in each panoramic digitized image can be obtained, and their absolute positions are stored in the position list; (6) The absolute positions of all inflammatory foci obtained in step (5) in each panoramic image are used to search for the most densely inflamed area of ​​the panoramic image: (a) Select an inflammatory focus in the order of the inflammatory focus location list X j location, counting all other inflammatory foci Y i Relative to the Euclidean distance of the selected inflammatory focus, the calculation formula of the Euclidean distance is: dist(X j ,Y i=1→n-1 ) = , i , j Number the list of positions and i ≠ j , n is the number of all inflammatory foci in a panoramic image; the field of view based on the densest area of ​​inflammatory foci is 1.6750×1.6750 mm 2 The standard was set to record the situation where the Euclidean distance between two inflammatory foci was less than 0.8375 mm as the situation where the inflammatory foci were dense and store the records in a new list until the statistics were completed. X j All Euclidean distance judgments; (b) Complete all steps according to step (a). X j The Euclidean distance is calculated and the records are stored in each corresponding X j New list of; (c) Compare and find the list with the largest number of inflammatory foci, and identify the corresponding inflammatory foci X j The position is taken as the coordinate center of the most intensively inflamed area in the panoramic image; (7) According to the calculation in step (6), the number of inflammatory foci in the most densely populated area of ​​inflammation can be obtained by counting, and the degree of inflammation can be estimated according to the following scoring criteria: The number of inflammatory foci = 0: no inflammation; the number of inflammatory foci < 2: mild inflammation; the number of inflammatory foci = 2-4: moderate inflammation; the number of inflammatory foci > 4: severe inflammation; (8) In order to obtain the area of ​​liver tissue for calculating the average number of inflammatory foci per unit area, the original image of the small image obtained in step (2) is debinarized, and the threshold is set to 200; the number of positive pixel values ​​in the debinarized image is counted to obtain the liver tissue area, and the liver tissue areas of all small images in a panoramic image are summed up to obtain a complete liver tissue area; (9) By dividing the total number of inflammatory foci in the panoramic image obtained in step (5) by the liver tissue area in the panoramic image obtained in step (8), the average number of inflammatory foci per unit area, i.e., the density of inflammatory foci, can be obtained.

8. The device according to any one of claims 1 to 6 or the method according to claim 7 is used in liver pathological section image analysis.

Images