Neuroblastoma recognition and labeling system

Abstract

The invention discloses a neuroblastoma identification and labeling system. According to the technical scheme provided by the invention, the system comprises a data acquisition module, a training test module and an identification labeling module, the data acquisition module is used for extracting pathological section images of neuroblastoma cases in a database to generate a first section set; the training and testing module is used for training and testing the first neuroblastoma recognition model by using the first slice set to obtain a second neuroblastoma recognition model; and the identification and labeling module is used for performing cell identification and labeling on the target slice image by using the second neuroblastoma identification model. According to the invention, various cells and / or tissues in the pathological section image can be identified and labeled by using the trained neural network model based on artificial intelligence, and a doctor is assisted to carry out pathological judgment, so that the workload of the doctor is greatly reduced, the working efficiency is improved, and extremely high identification speed and accuracy are ensured.

Description

Technical Field

[0001] The present invention relates to the field of pathological detection, and in particular to a neuroblastoma identification and labeling system. Background Art

[0002] Neuroblastoma is the most common malignant tumor of the peripheral nervous system in children. It originates from neuroblasts, a type of cell present in the early embryo that is responsible for forming various components of the nervous system. Neuroblastoma usually occurs in early childhood, especially between a few months after birth and 5 years old. Due to its high malignancy, neuroblastoma poses a serious threat to children's health and survival, which makes early diagnosis and treatment crucial to increase survival rates and improve the prognosis of children. In the pathological diagnosis of neuroblastoma, it is necessary to determine the type and grade of the tumor. The earlier the neuroblastoma can be identified, the more conducive it is to complete early treatment for children and improve their survival rate.

[0003] Currently, most neuroblastoma detection and identification methods involve medical workers observing pathological sections through a microscope and determining whether they are neuroblastomas by examining the morphology, arrangement, karyotype, and other characteristics of tumor cells. However, due to the complex morphological characteristics of neuroblastoma, the accuracy of manual pathological diagnosis has yet to be improved. In addition, the process of determining and labeling neuroblastoma requires a lot of manpower and time, and both the labor and time costs are high. The inefficiency of the entire detection process may also delay the child's subsequent treatment plan and miss the window for early treatment. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a neuroblastoma identification and annotation system and corresponding computing equipment and computer storage medium.

[0005] According to one aspect of the present invention, a neuroblastoma identification and annotation system is provided, comprising: a data acquisition module, a training and testing module, and an identification and annotation module; wherein:

[0006] The data acquisition module is used to extract pathological slice images from the pathological data of neuroblastoma cases in the hospital pathology database to generate a first slice set; wherein the first slice set includes a second slice subset and a third slice subset;

[0007] The training and testing module is used to train and test the first neuroblastoma recognition model based on artificial intelligence that has completed the basic settings using the first slice set, so as to obtain a second neuroblastoma recognition model;

[0008] The identification and labeling module is used to perform cell identification and labeling on the target slice image using the second neuroblastoma recognition model.

[0009] In the above solution, the data acquisition module is further used to:

[0010] Extracting pathological slice images from the first slice set according to a preset ratio to generate a second slice subset;

[0011] A third slice subset is generated according to the remaining pathological slice images in the first slice set except the pathological slice images included in the second slice subset.

[0012] In the above scheme, the first neuroblastoma recognition model uses an artificial intelligence-based convolutional network model to achieve semantic segmentation to complete cell recognition and labeling; the implementation of the basic settings at least includes pre-training the convolutional network model for image recognition;

[0013] The training and testing module is further used for:

[0014] The second slice subset is used as a training set to train the first neuroblastoma recognition model;

[0015] The third slice subset is used as a test set to test the first neuroblastoma recognition model that has been trained.

[0016] In the above scheme, the training and testing module further includes: a statistical evaluation submodule; wherein,

[0017] The statistical evaluation submodule is used to:

[0018] Dividing each pathological slice image in the third slice subset according to a preset area to generate a plurality of pathological image blocks;

[0019] Removing invalid image blocks without tissue images from multiple pathological image blocks to obtain valid image blocks;

[0020] Comparing the pathological data corresponding to each pathological slice in the third slice subset with the test results of the first neuroblastoma recognition model to obtain statistical results;

[0021] According to the statistical results, the true positive rate and false positive rate at different thresholds in the test results are calculated, the receiver operating characteristic curve is plotted according to the true positive rate and the false positive rate, and the area under the receiver operating characteristic curve is calculated as the model evaluation indicator.

[0022] In the above solution, the identification and labeling module is further used to:

[0023] Based on the second neuroblastoma recognition model, multiple elements in the target slice image are identified;

[0024] Determine the pathological properties of various elements based on the identification results and calculate the proportion of various elements;

[0025] The pathological properties and proportions of various elements are marked in the target slice image.

[0026] In the above solution, the data acquisition module further includes: a sample extraction submodule; wherein,

[0027] The sample extraction submodule is used to:

[0028] Determine the pathological slice images in the first slice set as the sampling range, determine the preset ratio as the random sampling ratio, and generate a random sampling rule;

[0029] Write the random sampling rules into the task table in the extraction record database, and create a temporary sample snapshot table in the extraction record database;

[0030] The pathological slice images in the first slice set are numbered so that each pathological slice image has its own corresponding ID number, and the ID numbers of all the pathological slice images are written into the temporary sample snapshot table;

[0031] Obtain the random seed number and write the random seed number into the task table; randomly extract the ID number from the temporary sample snapshot table based on the preset random extraction algorithm and the random seed number to obtain the ID number extraction result;

[0032] According to the correspondence between the ID number and the pathological slice image, the extracted pathological slice image is determined to generate a second slice subset.

[0033] In the above solution, the identification and marking module further includes: a target display submodule; wherein,

[0034] The target display submodule is used for:

[0035] Convert the target slice image into a grayscale image;

[0036] The grayscale image is segmented into the target area and the background area by using the segmentation algorithm;

[0037] Extracting the image of the target area according to the pathological properties of various elements, extracting the elements with malignant pathological properties as the first target image, and performing dilation and stroking processing on the first target image to obtain the second target image;

[0038] The second target image and the target slice image are superimposed to achieve highlighting of the malignant area in the target slice image.

[0039] In the above solution, the target display submodule is further used for:

[0040] Performing a morphological dilation operation on the first target image;

[0041] Performing edge enhancement on the first target image after the dilation operation is completed;

[0042] Within a preset range, grayscale stretching is performed on the edge-enhanced first target image to obtain a second target image.

[0043] In the above scheme, the identification and annotation module further includes: an annotation evaluation submodule; wherein,

[0044] The annotation evaluation submodule is used to:

[0045] Obtain annotations of the target slice image;

[0046] Construct an evaluation dataset based on the annotation of target slice images;

[0047] Randomly divide the evaluation data set into k evaluation sub-data sets;

[0048] Determine a labeling standard of any evaluation sub-dataset based on the labeling of the target slice image in any evaluation sub-dataset;

[0049] Based on the annotation standard of any evaluation sub-dataset, the similarity of the annotation standard of any evaluation sub-dataset with other evaluation sub-datasets is calculated one by one;

[0050] Based on the annotation standard similarity, the average value of the annotation standard similarity is calculated for any evaluation sub-dataset and it is determined whether the average value of the annotation standard similarity is less than the similarity average threshold. If the average value of the annotation standard similarity is less than the similarity average threshold, the annotation of the target slice image in any evaluation sub-dataset is re-annotated, otherwise no processing is performed. In the above scheme, the recognition and annotation module further includes: an annotation movement sub-module; wherein,

[0051] Annotate mobile submodules for:

[0052] Display the annotations of the target slice image for the user to view;

[0053] Perform annotation classification based on the annotation of the target slice image to determine a number of different types of annotated slice image sets;

[0054] Obtaining a user's viewing instruction for annotated slice image sets, dividing any type of annotated slice image sets into a plurality of viewed annotated slice image sets and a plurality of currently viewed annotated slice image sets;

[0055] When the user drags any currently viewed annotated slice image set, the drag direction and drag speed of the currently viewed annotated slice image set dragged by the user are obtained;

[0056] Adjust the position of any annotation of the currently viewed annotated slice image set dragged by the user so that the annotation position adjustment condition is met;

[0057] Among them, the conditions for adjusting the marking position include:

[0058] All annotations of any currently viewed annotated slice image set dragged by the user fall into the target sector;

[0059] The center of the target sector is the first center of the first display area of ​​all the viewed annotated slice image sets, and the line between the second center of the second display area of ​​the currently viewed annotated slice image set and the first center bisects the target sector;

[0060] The sector angle of the target sector is the angle corresponding to the drag speed in the preset speed-sector angle comparison table.

[0061] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0062] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute the operations executed by the neuroblastoma identification and annotation system as described above.

[0063] According to another aspect of the present invention, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to execute the operations executed by the neuroblastoma identification and annotation system as described above.

[0064] According to the technical solution provided by the present invention, the neuroblastoma identification and annotation system includes: a data acquisition module, a training and testing module, and an identification and annotation module; wherein the data acquisition module is used to extract pathological slice images from the pathological data of neuroblastoma cases in the hospital pathology database to generate a first slice set; wherein the first slice set includes a second slice subset and a third slice subset; the training and testing module is used to use the first slice set to train and test the first neuroblastoma identification model based on artificial intelligence that has completed the basic settings to obtain a second neuroblastoma identification model; the identification and annotation module is used to use the second neuroblastoma identification model to perform cell identification and annotation on the target slice image. Based on deep learning, the relevant pathological slice images extracted from the pathology database are used as training and testing data to establish a neural network model, and the trained neural network model is used to identify and annotate various types of cells and / or tissues in the pathological slice image. Since the neuroblastoma recognition model completes the corresponding training based on deep learning, it can ensure a high recognition speed and accuracy for pathological slice images, greatly reducing the workload of pathologists, improving recognition efficiency, and reducing labor costs and time costs; at the same time, the neuroblastoma recognition model selects training sets and test sets from the pathology database by fast random extraction, which can not only speed up the model training speed, but also ensure the scientific nature of the training process through the randomness of the training set; in addition, by highlighting the malignant area in the target slice image, medical workers can observe the malignant area more clearly, greatly avoiding the omissions that may occur in the manual observation process, and improving the accuracy of the final determination of neuroblastoma. In summary, this method is conducive to the rapid and accurate pathological identification and related information annotation of children, so that medical workers can quickly and accurately obtain pathological information, improve accuracy and efficiency, and reduce the workload of medical staff, so that children can receive treatment as soon as possible, which helps to improve the cure rate and prognosis of children.

[0065] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0066] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0068] Figure 1 A structural block diagram of a neuroblastoma identification and annotation system according to an embodiment of the present invention is shown;

[0069] Figure 2 A schematic diagram of data acquisition and test results according to an embodiment of the present invention is shown;

[0070] Figure 3 A schematic diagram of a label legend for cell identification and labeling according to an embodiment of the present invention is shown;

[0071] Figure 4 A structural block diagram of a neuroblastoma identification and annotation system according to another embodiment of the present invention is shown;

[0072] Figure 5 A schematic diagram of a receiver operating characteristic curve according to an embodiment of the present invention is shown;

[0073] Figure 6 A structural block diagram of a neuroblastoma identification and annotation system according to another embodiment of the present invention is shown;

[0074] Figure 7 A schematic diagram of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention is shown;

[0075] Figure 8 A schematic diagram of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention is shown;

[0076] Fig. 9 A schematic diagram of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention is shown;

[0077] Fig.10 A schematic diagram of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention is shown;

[0078] Fig.11 A schematic diagram of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention is shown;

[0079] Fig.12 A schematic diagram of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention is shown;

[0080] Fig.13 A schematic structural diagram of a computing device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0081] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0082] Figure 1 FIG. 4 shows a structural block diagram of a neuroblastoma identification and annotation system according to an embodiment of the present invention. Figure 1 As shown, the system includes: a data acquisition module 101, a training and testing module 102, and a recognition and annotation module 103; wherein,

[0083] The data acquisition module 101 is used to extract pathological slice images from the pathological data of neuroblastoma cases in the hospital pathology database to generate a first slice set; wherein the first slice set includes a second slice subset and a third slice subset.

[0084] Specifically, the data acquisition module 101 is further used to:

[0085] Extracting pathological slice images from the first slice set according to a preset ratio to generate a second slice subset;

[0086] A third slice subset is generated according to the remaining pathological slice images in the first slice set except the pathological slice images included in the second slice subset.

[0087] Preferably, for neuroblastoma, a pathological slice image is extracted from each case of pathological data of neuroblastoma positive cases collected by the hospital from the pathological database, and the pathological slice images extracted from all cases are collected to generate a first slice set.

[0088] Preferably, the preset ratio may be 80%, that is, 80% of the first slice set is selected as the second slice subset, and the remaining 20% ​​is selected as the third slice subset.

[0089] The training and testing module 102 is used to train and test the first neuroblastoma recognition model based on artificial intelligence that has completed the basic settings using the first slice set, so as to obtain a second neuroblastoma recognition model.

[0090] Preferably, the first neuroblastoma recognition model uses an artificial intelligence-based convolutional network model to achieve semantic segmentation to complete cell recognition and labeling.

[0091] Preferably, implementing the basic settings at least includes pre-training the convolutional network model for image recognition; wherein, through pre-training, the convolutional network model can preliminarily realize basic image recognition functions.

[0092] Specifically, the training and testing module 102 is further used to:

[0093] The second slice subset is used as a training set to train the first neuroblastoma recognition model;

[0094] The third slice subset is used as a test set to test the first neuroblastoma recognition model that has been trained.

[0095] Preferably, the identifiable cells and / or tissues in the target slice image include at least: non-tumor, ganglion cells, Schwann cell stroma, mature ganglion cells, nested neuroblasts, small round cells, HW rosettes (Homer-Wright pseudorosettes), MKI (Mitosis-Karyorrhexis Index, usually the number of neuroblasts with nuclear division and nuclear fragmentation in 5000 neuroblasts), nodules, nerve fiber filaments and poor quality tissue (i.e., tissue that cannot be clearly displayed in the image). The identified cells and / or tissues are annotated by various annotation tags.

[0096] The identified tissues can be divided into non-tumor, benign, borderline, malignant and excluded (corresponding to poor quality tissues) according to their pathological properties.

[0097] Preferably, the case slice images in the training set and the test set are segmented into 320×320 pixels to generate image blocks, and the image blocks without tissue are removed. By identifying the cells and / or tissues in each image block and determining their pathological properties, the training and testing results for the first neuroblastoma recognition model and the second neuroblastoma recognition model after training are obtained.

[0098] Figure 2 FIG. 2 shows a schematic diagram of data acquisition and test results according to an embodiment of the present invention. Figure 2 As shown, relevant cases of Beijing Children's Hospital were obtained from the pathology database for model training and testing, with a total of 237 cases; the data set was divided according to the preset ratio, and the case slice images corresponding to 80% of the cases were used as the training set, with a total of 192 cases; the case slice images corresponding to the remaining 20% ​​of the cases were used as the test set, with a total of 46 cases. Correspondingly, there were 1,451,220 image blocks obtained by segmenting the pathology slice images in the training set, and 340,940 image blocks obtained by segmenting the pathology slice images in the test set.

[0099] Among them, for the case slice images in the training set and the test set, the annotation labels corresponding to the identified cells and / or tissues include: non-tumor, ganglion cells, Schwann cell stroma, mature ganglion cells, nested neuromasts, small round cells, HW rosettes, MKI, nodules, nerve fiber filaments, and poor quality tissue. At the same time, for each annotation label, the corresponding pathological properties, the corresponding number of image blocks, and the percentage of the number of such image blocks in the total number of image blocks are displayed. For example, the ganglion cells identified and annotated have a benign pathological property, and the number of such image blocks is 614436, accounting for 34.28% of the total number of image blocks; the Schwann cell stroma has a benign pathological property, and the number of such image blocks is 5209, accounting for 0.19% of the total number of image blocks; the nodules have a malignant pathological property, and the number of such image blocks is 44126, accounting for 2.46% of the total number of image blocks.

[0100] The identification and labeling module 103 is used to perform cell identification and labeling on the target slice image using the second neuroblastoma recognition model.

[0101] Specifically, based on the neuroblastoma II recognition model, multiple elements in the target slice image are identified; the pathological properties of various elements are determined according to the recognition results, and the proportions of various elements are counted; the pathological properties and proportions of various elements are marked in the target slice image.

[0102] Preferably, a label legend for cell identification and annotation is provided for the target slice image, such as Figure 3 As shown, Figure 3 A schematic diagram of a label legend for cell identification and labeling according to an embodiment of the present invention is shown.

[0103] According to the neuroblastoma identification and annotation system provided in this embodiment, it includes: a data acquisition module, a training and testing module, and an identification and annotation module; wherein the data acquisition module is used to extract pathological slice images from the pathological data of neuroblastoma cases in the hospital pathology database to generate a first slice set; wherein the first slice set includes a second slice subset and a third slice subset; the training and testing module is used to use the first slice set to train and test the first neuroblastoma identification model based on artificial intelligence that has completed the basic settings to obtain a second neuroblastoma identification model; the identification and annotation module is used to use the second neuroblastoma identification model to perform cell identification and annotation on the target slice image. Based on deep learning, the relevant pathological slice images extracted from the pathology database are used as training and testing data to establish a neural network model, and the trained neural network model is used to identify and annotate various types of cells and / or tissues in the pathological slice images. Since the neuroblastoma recognition model is trained based on deep learning, it can ensure high recognition speed and accuracy for pathological slice images, greatly reducing the workload of pathologists, improving recognition efficiency, and reducing labor costs and time costs; by marking the malignant areas in the target slice images, medical workers can observe the malignant areas more clearly, greatly avoiding possible omissions in the manual observation process, and improving the accuracy of the final determination of neuroblastoma. In summary, this method is conducive to the rapid and accurate pathological identification and related information labeling of children, so that medical workers can quickly and accurately obtain pathological information, improve accuracy and efficiency, and reduce the workload of medical staff, so that children can receive treatment as early as possible, which helps to improve the cure rate and prognosis of children.

[0104] Figure 4 FIG. 4 shows a structural block diagram of a neuroblastoma identification and annotation system according to another embodiment of the present invention. Figure 4 As shown, the system includes: a data acquisition module 401, a training and testing module 402, and a recognition and annotation module 403; wherein the training and testing module 402 further includes: a statistical evaluation submodule 404, which is used to:

[0105] Dividing each pathological slice image in the third slice subset according to a preset area to generate a plurality of pathological image blocks;

[0106] Removing invalid image blocks without tissue images from multiple pathological image blocks to obtain valid image blocks;

[0107] Comparing the pathological data corresponding to each pathological slice in the third slice subset with the test results of the first neuroblastoma recognition model to obtain statistical results;

[0108] According to the statistical results, the true positive rate and false positive rate at different thresholds in the test results were calculated, the receiver operating characteristic (ROC) curve was plotted according to the true positive rate and the false positive rate, and the area under the receiver operating characteristic curve (AUC) was calculated as the model evaluation indicator.

[0109] like Figure 5 As shown, Figure 5 FIG. 4 shows a schematic diagram of a receiver operating characteristic curve according to an embodiment of the present invention; wherein:

[0110] According to the statistical results, the accuracy, sensitivity and precision of the training set and test set are calculated respectively.

[0111] The receiver operating characteristic curve ROC is plotted with the true positive rate as the ordinate and the false positive rate as the abscissa; the area under the receiver operating characteristic curve (AUC) corresponding to the ROC curve in the figure is 0.918923, which is used to characterize the classification and recognition ability of the neuroblastoma recognition model.

[0112] According to the neuroblastoma identification and labeling system in this embodiment, the functions implemented by the data acquisition module 401, the training and testing module 402, and the identification and labeling module 403 are the same as those of the corresponding modules in the above embodiments, and will not be repeated here.

[0113] According to the above-mentioned neuroblastoma identification and annotation system, the results of previous training and testing can be compared with the actual results of pathological sections to complete the statistics of training and testing results, and based on this, the ROC curve can be drawn and the AUC area can be calculated. The performance of the neuroblastoma identification model is reflected in an intuitive way of graphics and data, which is convenient for medical workers to clearly and easily understand the neuroblastoma identification model, and helps medical workers to better use the neuroblastoma identification and annotation system to assist doctors in the diagnosis process and improve work efficiency.

[0114] Figure 6 FIG. 4 shows a structural block diagram of a neuroblastoma identification and annotation system according to another embodiment of the present invention. Figure 6 As shown, the system includes: a data acquisition module 601, a training and testing module 602, and a recognition and annotation module 603; wherein,

[0115] The training and testing module 602 further includes: a statistical evaluation submodule 604;

[0116] The data acquisition module 601 further includes a sample extraction submodule 605;

[0117] The identification and marking module 603 further includes a target display submodule 606 .

[0118] According to the neuroblastoma identification and labeling system in this embodiment, the functions implemented by the data acquisition module 601, the training and testing module 602, the identification and labeling module 603 and the statistical evaluation submodule 604 are the same as those of the corresponding modules in the above embodiments, and will not be repeated here.

[0119] Specifically, the sample extraction submodule 605 is used to:

[0120] Determine the pathological slice images in the first slice set as the sampling range, determine the preset ratio as the random sampling ratio, and generate a random sampling rule;

[0121] Write the random sampling rules into the task table in the extraction record database, and create a temporary sample snapshot table in the extraction record database;

[0122] The pathological slice images in the first slice set are numbered so that each pathological slice image has its own corresponding ID number, and the ID numbers of all the pathological slice images are written into the temporary sample snapshot table;

[0123] Obtain the random seed number and write the random seed number into the task table; randomly extract the ID number from the temporary sample snapshot table based on the preset random extraction algorithm and the random seed number to obtain the ID number extraction result;

[0124] According to the correspondence between the ID number and the pathological slice image, the extracted pathological slice image is determined to generate a second slice subset.

[0125] Specifically, the target display submodule 606 is used to:

[0126] Convert the target slice image into a grayscale image;

[0127] The grayscale image is segmented into the target area and the background area by using the segmentation algorithm;

[0128] Extracting the image of the target area according to the pathological properties of various elements, extracting the elements with malignant pathological properties as the first target image, and performing dilation and stroking processing on the first target image to obtain the second target image;

[0129] The second target image and the target slice image are superimposed to achieve highlighting of the malignant area in the target slice image.

[0130] According to the above neuroblastoma identification and annotation system, training sets and test sets can be selected from the pathology database by fast random extraction, which can not only speed up the model training speed, but also ensure the scientific nature of the training process through the randomness of the training set; in addition, by highlighting the malignant area in the target slice image, medical workers can observe the malignant area more clearly, which greatly avoids the omissions that may occur in the manual observation process and improves the accuracy of the final determination of neuroblastoma.

[0131] Furthermore, the neuroblastoma recognition and annotation system is used to identify and annotate the target pathological sections, and the result obtained is as follows: Figures 7 to 12 As shown, Figures 7 to 12 A total of 6 schematic diagrams of recognition results of a neuroblastoma recognition and annotation system according to an embodiment of the present invention are shown; specifically, Figure 7 , Figure 8 , Fig. 9 , Fig.10 , Fig.11 , Fig.12 , respectively showing schematic diagrams of identification and labeling for 6 different pathological section images.

[0132] The recognition results are annotated according to the label legend, showing the type and pathological properties of cells and / or tissues, and highlighting the malignant areas.

[0133] The present invention also provides a non-volatile computer storage medium, which stores at least one executable instruction, and the executable instruction can execute the neuroblastoma identification and labeling method in any of the above method embodiments.

[0134] Specifically, the identification and annotation module 603 also includes an annotation evaluation submodule, which is used to:

[0135] Obtain annotations of the target slice image;

[0136] Construct an evaluation dataset based on the annotation of target slice images;

[0137] Randomly divide the evaluation data set into k evaluation sub-data sets;

[0138] Determine a labeling standard of any evaluation sub-dataset based on the labeling of the target slice image in any evaluation sub-dataset;

[0139] Based on the annotation standard of any evaluation sub-dataset, the similarity of the annotation standard of any evaluation sub-dataset with other evaluation sub-datasets is calculated one by one;

[0140] Based on the annotation standard similarity, the average value of the annotation standard similarity is calculated for any evaluation sub-dataset and it is determined whether the average value of the annotation standard similarity is less than the similarity average threshold. If the average value of the annotation standard similarity is less than the similarity average threshold, the annotation of the target slice image in any evaluation sub-dataset is re-annotated, otherwise no processing is performed.

[0141] This embodiment obtains an annotated target slice image, and constructs an evaluation data set, and randomly divides the evaluation data set into k evaluation sub-data sets. Select any one of the evaluation sub-data sets, and this embodiment first evaluates the first evaluation sub-data set, and determines the annotation standard of the first evaluation sub-data set according to the annotation of the target slice image in the first evaluation sub-data set, for example, the size of the malignant area of ​​neuroblastoma. Then, the annotation standards of the other k-1 evaluation sub-data sets are determined in turn. First, the annotation standard of the first evaluation sub-data set is calculated with the annotation standard of the other k-1 evaluation sub-data sets to determine the similarity of the annotation standard of the first evaluation sub-data set and the other k-1 evaluation sub-data sets, and the k-1 annotation standard similarity values ​​are averaged to obtain the average value of the annotation standard similarity of the first evaluation sub-data set. The average value of the annotation standard similarity of the first evaluation sub-data set is judged. If the average value of the annotation standard similarity of the first evaluation sub-data set is less than the similarity average threshold, it means that the annotation standard of the first evaluation sub-data set is different from that of the others, and there is an error, and the first evaluation sub-data set needs to be re-annotated, otherwise no processing.

[0142] This embodiment evaluates and self-checks the annotations of the target slice images, thereby improving the accuracy and consistency of the annotations of the target slice images and reducing erroneous annotations of the target slice images.

[0143] Specifically, the identification and marking module 603 further includes a marking movement submodule, which is used to:

[0144] Display the annotations of the target slice image for the user to view;

[0145] Perform annotation classification based on the annotation of the target slice image to determine a number of different types of annotated slice image sets;

[0146] Obtaining a user's viewing instruction for annotated slice image sets, dividing any type of annotated slice image sets into a plurality of viewed annotated slice image sets and a plurality of currently viewed annotated slice image sets;

[0147] When the user drags any currently viewed annotated slice image set, the drag direction and drag speed of the currently viewed annotated slice image set dragged by the user are obtained;

[0148] Adjust the position of any annotation of the currently viewed annotated slice image set dragged by the user so that the annotation position adjustment condition is met;

[0149] Among them, the conditions for adjusting the marking position include:

[0150] All annotations of any currently viewed annotated slice image set dragged by the user fall into the target sector;

[0151] The center of the target sector is the first center of the first display area of ​​all the viewed annotated slice image sets, and the line between the second center of the second display area of ​​the currently viewed annotated slice image set and the first center bisects the target sector;

[0152] The sector angle of the target sector is the angle corresponding to the drag speed in the preset speed-sector angle comparison table.

[0153] In this embodiment, the annotations of the target slice image are displayed for the user to view. The annotations of the target slice image are classified based on the annotations, and several different types of annotated slice image sets are determined. For example, according to the location of the neuroblastoma, it is classified into an annotated slice image set of the adrenal gland, a neck annotated slice image set, a waist annotated slice image set, etc. According to the user's viewing of the annotated slice image set, a certain type of annotated slice image set is divided into a plurality of viewed annotated slice image sets and a plurality of currently viewed annotated slice image sets. For example, if the user wants to view the information in the neck annotated slice image set, the neck annotated slice image set is divided into a plurality of viewed annotated slice image sets and a plurality of currently viewed annotated slice image sets according to the user's viewing of the annotated slice image set. If the annotations of the currently viewed annotated image set overlap with the plurality of viewed annotated slice image sets, it is easy to affect the user's viewing of the annotations of the currently viewed annotated image set. Therefore, it is necessary to adjust the position of the annotations of the currently viewed annotated image set and separate it from the plurality of viewed annotated slice image sets, so as to facilitate the user to view the annotations of the currently viewed annotated image set. If the user drags a currently viewed annotated slice image set, the annotation position of the currently viewed annotated slice image set is adjusted so that the annotation position of the currently viewed annotated slice image set does not overlap with the viewed annotated slice image set, thereby facilitating user viewing. When the user drags a currently viewed annotated slice image set, the dragging direction and dragging speed are obtained. The position of the annotation of the currently viewed annotated slice image set dragged by the user is adjusted so that the annotation position adjustment condition is met. The annotation position adjustment condition includes: placing the annotation of a currently viewed annotated slice image set A within the target sector. According to the second center O of the second display area of ​​the currently viewed annotated slice image set A dragged by the user and the first center of the first display area of ​​the viewed annotated slice image set B The target sector C is determined by marking, that is, the center of the target sector C is the first center of the first display area of ​​all the viewed annotated slice image sets. The line between the second center and the first center of the second display area of ​​the currently viewed annotated slice image set The target sector C is bisected. According to the relationship between the drag speed and the sector angle, a speed-sector angle comparison table is preset. According to the drag speed, the corresponding angle of the target sector C is searched in the preset speed-sector angle comparison table.

[0154] The present invention adjusts the position of the annotations of the currently viewed target slice image based on user dragging, so that the annotations of the currently viewed annotated slice image set do not overlap with the previously viewed annotated slice image set, making it convenient for the user to view the annotations of the currently viewed annotated slice image set, thereby facilitating medical observation and judgment.

[0155] like Figure 8 As shown, the computing device may include: a processor (processor) 802 , a communication interface (Communications Interface) 804 , a memory (memory) 806 , and a communication bus 808 .

[0156] in:

[0157] The processor 802 , the communication interface 804 , and the memory 806 communicate with each other via a communication bus 808 .

[0158] The communication interface 804 is used to communicate with other devices such as clients or other servers.

[0159] The processor 802 is used to execute the program 810, and specifically can execute the relevant steps in the above-mentioned neuroblastoma identification and labeling method embodiment.

[0160] Specifically, the program 810 may include program codes, and the program codes include computer operation instructions.

[0161] The processor 802 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0162] The memory 806 is used to store the program 810. The memory 806 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0163] The program 810 can be specifically used to enable the processor 802 to execute the neuroblastoma identification and labeling method in any of the above-mentioned method embodiments. The specific implementation of each step in the program 810 can refer to the corresponding descriptions in the corresponding steps and units in the above-mentioned neuroblastoma identification and labeling method embodiments, which will not be repeated here. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described devices and modules can refer to the corresponding process description in the above-mentioned method embodiments, which will not be repeated here.

[0164] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0165] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0166] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

[0167] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0168] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims, any one of the claimed embodiments may be used in any combination.

[0169] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., computer program and computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0170] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A neuroblastoma identification and annotation system, comprising: Data acquisition module, training and testing module, and recognition and annotation module; among them, The data acquisition module is used to extract pathological slice images from the pathological data of neuroblastoma cases in the hospital pathology database to generate a first slice set; wherein the first slice set includes a second slice subset and a third slice subset; The training and testing module is used to train and test the first neuroblastoma recognition model based on artificial intelligence that has completed the basic settings using the first slice set, so as to obtain a second neuroblastoma recognition model; The identification and labeling module is used to perform cell identification and labeling on the target slice image using the second neuroblastoma recognition model.

2. The system according to claim 1, characterized in that The data acquisition module is further used for: Extracting pathological slice images from the first slice set according to a preset ratio to generate a second slice subset; A third slice subset is generated according to the remaining pathological slice images in the first slice set except the pathological slice images included in the second slice subset.

3. The system according to claim 1, characterized in that The first neuroblastoma recognition model uses an artificial intelligence-based convolutional network model to achieve semantic segmentation to complete cell recognition and labeling; the implementation of the basic settings at least includes pre-training the convolutional network model for image recognition; The training and testing module is further used for: The second slice subset is used as a training set to train the first neuroblastoma recognition model; The third slice subset is used as a test set to test the first neuroblastoma recognition model that has been trained.

4. The system according to claim 1 or 3, characterized in that: The training and testing module further includes: a statistical evaluation submodule; wherein, The statistical evaluation submodule is used to: Dividing each pathological slice image in the third slice subset according to a preset area to generate a plurality of pathological image blocks; Removing invalid image blocks without tissue images from multiple pathological image blocks to obtain valid image blocks; Comparing the pathological data corresponding to each pathological slice in the third slice subset with the test results of the first neuroblastoma recognition model to obtain statistical results; According to the statistical results, the true positive rate and false positive rate at different thresholds in the test results are calculated, the receiver operating characteristic curve is plotted according to the true positive rate and the false positive rate, and the area under the receiver operating characteristic curve is calculated as the model evaluation indicator.

5. The system according to claim 1, characterized in that The identification and annotation module is further used to: Based on the second neuroblastoma recognition model, multiple elements in the target slice image are identified; Determine the pathological properties of various elements based on the identification results and calculate the proportion of various elements; The pathological properties and proportions of various elements are marked in the target slice image.

6. The system according to claim 2, characterized in that The data acquisition module further includes: a sample extraction submodule; wherein, The sample extraction submodule is used to: Determine the pathological slice images in the first slice set as the sampling range, determine the preset ratio as the random sampling ratio, and generate a random sampling rule; Write the random sampling rules into the task table in the extraction record database, and create a temporary sample snapshot table in the extraction record database; The pathological slice images in the first slice set are numbered so that each pathological slice image has its own corresponding ID number, and the ID numbers of all the pathological slice images are written into the temporary sample snapshot table; Obtain the random seed number and write the random seed number into the task table; randomly extract the ID number from the temporary sample snapshot table based on the preset random extraction algorithm and the random seed number to obtain the ID number extraction result; According to the correspondence between the ID number and the pathological slice image, the extracted pathological slice image is determined to generate a second slice subset.

7. The system according to claim 1, characterized in that The identification and marking module further includes: a target display submodule; wherein: The target display submodule is used for: Convert the target slice image into a grayscale image; The grayscale image is segmented into the target area and the background area by using the segmentation algorithm; Extracting the image of the target area according to the pathological properties of various elements, extracting the elements with malignant pathological properties as the first target image, and performing dilation and stroking processing on the first target image to obtain the second target image; The second target image and the target slice image are superimposed to achieve highlighting of the malignant area in the target slice image.

8. The system according to claim 7, characterized in that The target display submodule is further used for: Performing a morphological dilation operation on the first target image; Performing edge enhancement on the first target image after the dilation operation is completed; Within a preset range, grayscale stretching is performed on the edge-enhanced first target image to obtain a second target image.

9. The system according to claim 1, characterized in that The recognition and annotation module further includes: an annotation evaluation submodule; wherein, The annotation evaluation submodule is used to: Obtain annotations of target slice images; Construct an evaluation dataset based on the annotation of target slice images; Randomly divide the evaluation data set into k evaluation sub-data sets; Determine a labeling standard of any evaluation sub-dataset based on the labeling of the target slice image in any evaluation sub-dataset; Based on the annotation standard of any evaluation sub-dataset, the similarity of the annotation standard of any evaluation sub-dataset with other evaluation sub-datasets is calculated one by one; Based on the annotation standard similarity, the average value of the annotation standard similarity is calculated for any evaluation sub-dataset and it is determined whether the average value of the annotation standard similarity is less than the similarity average threshold. If the average value of the annotation standard similarity is less than the similarity average threshold, the annotation of the target slice image in any evaluation sub-dataset is re-annotated, otherwise no processing is performed.

10. The system according to claim 1, characterized in that The identification and marking module further includes: a marking movement submodule; wherein, Annotate mobile submodules for: Display the annotations of the target slice image for the user to view; Perform annotation classification based on the annotation of the target slice image to determine a number of different types of annotated slice image sets; Obtaining a user's viewing instruction for annotated slice image sets, dividing any type of annotated slice image sets into a plurality of viewed annotated slice image sets and a plurality of currently viewed annotated slice image sets; When the user drags any currently viewed annotated slice image set, the drag direction and drag speed of the currently viewed annotated slice image set dragged by the user are obtained; Adjust the position of any annotation of the currently viewed annotated slice image set dragged by the user so that the annotation position adjustment condition is met; Among them, the conditions for adjusting the marking position include: All annotations of any currently viewed annotated slice image set dragged by the user fall into the target sector; The center of the target sector is the first center of the first display area of ​​all the viewed annotated slice image sets, and the line between the second center of the second display area of ​​the currently viewed annotated slice image set and the first center bisects the target sector; The sector angle of the target sector is the angle corresponding to the drag speed in the preset speed-sector angle comparison table.

Images