WO2022007337A1 - 肿瘤细胞含量评估方法、系统、计算机设备及存储介质 - Google Patents

肿瘤细胞含量评估方法、系统、计算机设备及存储介质 Download PDF

Info

Publication number
WO2022007337A1
WO2022007337A1 PCT/CN2020/137029 CN2020137029W WO2022007337A1 WO 2022007337 A1 WO2022007337 A1 WO 2022007337A1 CN 2020137029 W CN2020137029 W CN 2020137029W WO 2022007337 A1 WO2022007337 A1 WO 2022007337A1
Authority
WO
WIPO (PCT)
Prior art keywords
area
tumor cell
pathological
preset
tumor
Prior art date
Application number
PCT/CN2020/137029
Other languages
English (en)
French (fr)
Inventor
车拴龙
罗丕福
李映华
刘斯
丘伟松
吴涛
哈正蓬
吕海辉
Original Assignee
广州金域医学检验中心有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 广州金域医学检验中心有限公司 filed Critical 广州金域医学检验中心有限公司
Publication of WO2022007337A1 publication Critical patent/WO2022007337A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30024—Cell structures in vitro; Tissue sections in vitro

Definitions

  • the present application relates to the technical field of image processing, and in particular, to a method, system, computer equipment and storage medium for evaluating tumor cell content.
  • tumor tissue samples need to be provided for molecular pathological examination.
  • the pathological samples obtained from clinical surgical biopsy are affected by multiple factors such as patient factors, tumor factors, and biopsy sampling by clinical surgeons.
  • the tissue content and tumor content fluctuated in a wide range in each pathological sample submitted for inspection.
  • samples with low tissue content or low tumor content for the molecular pathology detection technology platform, it is often impossible to detect, resulting in false negatives, or the patient sends the two samples before and after for molecular pathology examination, and the test results are serious. the problem of inconsistency.
  • a method for evaluating tumor cell content comprising:
  • a preset evaluation rule is used to determine the tumor cell content of the digital pathological slice image.
  • a tumor cell content assessment system includes:
  • an area determination module used for acquiring a digital pathological slice image, and determining an effective pathological area according to the digital pathological slice image
  • a tumor cell identification module for identifying the tumor cell region corresponding to the effective pathological region through a deep learning-based pathological image classifier
  • the content evaluation module is configured to determine the tumor cell content of the digital pathological slice image by using a preset evaluation rule according to the tumor cell area.
  • a computer device includes a memory and a processor, the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, causes the processor to perform the following steps:
  • a preset evaluation rule is used to determine the tumor cell content of the digital pathological slice image.
  • One or more non-volatile readable storage media storing computer-readable instructions, wherein the computer-readable instructions, when executed by one or more processors, cause the one or more processors to execute:
  • a preset evaluation rule is used to determine the tumor cell content of the digital pathological slice image.
  • the above-mentioned tumor cell content assessment method, system, computer equipment and storage medium by acquiring digital pathological slice images, and determining an effective pathological area according to the digital pathological slice images;
  • the tumor cell area corresponding to the area; according to the tumor cell area, the tumor cell content of the digital pathological slice image is determined by using a preset evaluation rule, and the pathological slice image is identified by using a deep learning method.
  • the obtained cell area is comprehensively evaluated by a variety of evaluation methods.
  • the recognition technology of machine vision ensures the accuracy and objectivity of tumor cell identification.
  • the comprehensive evaluation using various evaluation methods further improves the accuracy of tumor cell content evaluation. .
  • FIG. 1 is a flowchart of a method for evaluating tumor cell content in one embodiment
  • FIG. 2 is a flowchart of a method for determining an effective pathological region in one embodiment
  • 3 is a comparison diagram of a digital pathological slice image and a grayscale image in one embodiment
  • FIG. 4 is a flowchart of a method for evaluating tumor cell content in another embodiment
  • FIG. 5 is a flowchart of a method for evaluating tumor cell content in yet another embodiment
  • FIG. 6 is a flowchart of a method for evaluating tumor cell content in yet another embodiment
  • FIG. 7 is a flowchart of a method for evaluating tumor cell content in yet another embodiment
  • Fig. 8 is a structural block diagram of a tumor cell content assessment system in one embodiment
  • FIG. 9 is a structural block diagram of a computer device in one embodiment.
  • a method for evaluating tumor cell content is provided, and the method for evaluating tumor cell content can be applied to both a terminal and a server.
  • application to a server is used as an example.
  • the tumor cell content assessment method specifically includes the following steps:
  • step 102 a digital pathological slice image is acquired, and an effective pathological area is determined according to the digital pathological slice image.
  • the digital pathological section image refers to the pathological section image in which the pathological section is displayed and saved in the form of a digital image. Specifically, it can be obtained by digitally scanning the pathological section of the biopsy sample with a digital pathological scanner.
  • the effective pathological area refers to the area containing cells in the pathological section, and the cells include normal and/or diseased cells.
  • the effective pathological area is determined by segmenting the digital pathological slice images, and the image segmentation methods include but are not limited to image binarization processing methods, image segmentation models based on machine learning or edge detection algorithms.
  • the binarization processing method is adopted, that is, the average gray value of the cell area and the non-cellular area is inconsistent, and a reasonable gray threshold is set to filter out a part of the non-cellular area, so as to extract the Effective pathological area, this method is simpler and faster than machine learning image segmentation models or edge detection algorithms.
  • Step 104 Identify the tumor cell area corresponding to the effective pathological area through the deep learning-based pathological image classifier.
  • Deep learning is a method based on representation learning of data in machine learning. Observations (such as images) can be represented in a variety of ways, as a vector of intensity values for each pixel, or more abstractly as a series of edges, regions of a particular shape, etc. Instead, it is easier to learn tasks (e.g. face recognition) from examples using some specific representation. Deep learning is used to build and simulate the neural network of the human brain for analysis and learning. It imitates the mechanism of the human brain to interpret data, such as images, sounds and texts. Unsupervised or semi-supervised feature learning and hierarchical feature extraction are efficient Algorithms can replace manual acquisition of features, which can help improve the objectivity and accuracy of prediction results.
  • the pathological image classifier is a classifier that can learn a machine learning algorithm model with classification ability through samples.
  • the pathological image classifier in this embodiment is used to divide different effective pathological regions into tumor cell regions and non-tumor cell regions. one type.
  • a classifier that can perform classification using at least one machine learning model may be one or more of the following: neural networks (eg, convolutional neural networks, BP neural networks, etc.), logistic regression models, support vector machines, decision trees, random forests, perceptrons, and other machine learning models Model.
  • the training input is images corresponding to various effective pathological regions, and through training, a relationship classifier corresponding to the effective pathological regions and tumor cell regions or non-tumor cell regions is established.
  • a relationship classifier corresponding to the effective pathological regions and tumor cell regions or non-tumor cell regions is established.
  • the pathological image classifier is a binary classifier, that is, two classification results are obtained, that is, a tumor cell area or a non-tumor cell area.
  • the non-tumor cell area refers to the area where normal cells are located in the effective pathological area
  • the tumor cell area refers to the area where the diseased cells are located in the effective pathological area, which is used by doctors to analyze the self-explosive content of the area to achieve disease diagnosis. It is understandable that by identifying the tumor cell region in the effective pathological region by means of deep learning, the automatic detection of the tumor cell region is realized, and the accuracy and objectivity of the tumor cell identification are improved.
  • Step 106 according to the tumor cell area, using a preset evaluation rule to determine the tumor cell content of the digital pathological slice image.
  • the preset evaluation rule refers to a preset evaluation method or index for evaluating the content of tumor cells, for example, the evaluation can be performed according to the area of the tumor cell area, the number of cells, the proportion of tumor, and the ratio of tumor stroma. Understandably, by using the preset evaluation rules for evaluation, the automatic evaluation of the tumor cell content of the digital pathological slice images from multiple angles is realized, the comprehensiveness and objectivity of the evaluation are ensured, and the tumor cell content is improved. The accuracy and consistency of the evaluation can be avoided to avoid subjective bias in manual analysis, and it is impossible to obtain a unified and accurate evaluation result.
  • the above tumor cell content assessment method by acquiring digital pathological slice images, determines the effective pathological area according to the digital pathological slice images; identifying the tumor cell area corresponding to the effective pathological area through a deep learning-based pathological image classifier; according to the tumor cell area,
  • the pre-set evaluation rules are used to determine the tumor cell content of the digital pathological slice images, which realizes the automatic evaluation of the tumor cell content in the digital pathological slice images, and improves the accuracy and objectivity of the evaluation of the tumor cell content.
  • the effective pathological area is determined according to the digital pathological slice image, including:
  • Step 102A performing binarization processing on the digital pathological slice image to obtain a grayscale image
  • Step 102B extracting a region with a grayscale value smaller than a preset grayscale threshold from the grayscale image as an effective pathological region.
  • the color of the effective pathological area is the area stained by red and blue dyeing reagents
  • the grayscale of the background area is colorless or white
  • the grayscale value is larger than the grayscale value of the effective pathological area.
  • the grayscale of the pathological area determines the preset grayscale threshold. Therefore, the digital pathological slice image is subjected to binarization processing, wherein the binarization processing includes but is not limited to global binarization, the optimal threshold method based on histogram, or The OTSU Otsu threshold method based on clustering converts the color values of color pictures into grayscale images.
  • the comparison diagram of the digital pathological slice image and the grayscale image is shown in FIG. 3 .
  • a region with a grayscale value smaller than a preset grayscale threshold is extracted from the grayscale image, thereby extracting an effective pathological region for subsequent further processing of the effective pathological region.
  • a preset evaluation rule is used to determine the tumor cell content of the digital pathological slice image, including:
  • Step 106A determining the first area of the tumor cell area, and determining the second area of the effective pathological area
  • Step 106B Calculate the tumor proportion and tumor-stroma ratio of the digital pathological slice image according to the first area and the second area.
  • the first area can be calculated according to the tumor cell area
  • the second area can be calculated according to the effective pathological area.
  • the following two formulas were used to calculate the tumor proportion and tumor-stroma ratio, respectively:
  • S1 represents the first area
  • S2 represents the second area
  • P1 represents the tumor proportion
  • P2 represents the tumor-stromal ratio.
  • the tumor proportion and the tumor-stroma ratio are used as indicators for evaluating the content of tumor cells, so that the doctor can diagnose the disease according to the tumor proportion and the tumor-stroma ratio.
  • the method further includes:
  • Step 108 obtaining multiple test diameters corresponding to multiple tumor cells, and performing mean calculation on the multiple test diameters to obtain the average diameter of a single tumor cell;
  • Step 110 determining the average area of a single tumor cell according to the average diameter of the single tumor cell
  • Step 112 Calculate the number of tumor cells per unit area according to the average area of a single tumor cell
  • Step 114 calculating a first number of tumor cells according to the first area and the number of tumor cells in a unit area.
  • a tumor cell such as a lung cancer cell
  • a plurality of tumor cells such as 100
  • 100 test diameters are obtained, and the average value of the plurality of test diameters is calculated to obtain a single tumor cell
  • the number of tumor cells per unit area is obtained, and the first number of tumor cells can be calculated by multiplying the first area by the number of tumor cells per unit area.
  • the doctor can diagnose and analyze the disease based on the first number.
  • the method further includes:
  • the tumor cell region is cell-segmented by a cell-segmentation algorithm to determine the second number of tumor cells.
  • the cell segmentation algorithm is a method for segmenting cells, including but not limited to the cell segmentation algorithm of the FCN fully convolutional neural network, the cell segmentation algorithm of the U-Net model, or the small segmentation algorithm based on deep learning, etc. .
  • the second number is obtained, so that the doctor can diagnose and analyze the disease based on the first number.
  • the second quantity in this embodiment is calculated based on the characteristics of the cells, and is more accurate than the first quantity in step 114 . Therefore, it is suitable for more testing or disease diagnosis occasions.
  • Step 116 obtaining a training sample set, where the training sample set includes training pathological regions and corresponding training cell types;
  • Step 118 take the training pathological region as the input of the preset classifier, and use the training cell type as the expected output, and train the preset classifier to obtain a trained pathological image classifier.
  • the samples of the tumor cell area and the non-tumor cell area marked by the doctor are obtained, the training pathological area is used as the input of the preset classifier, the training cell type is used as the expected output, and the preset classifier is trained, A cell type corresponding to the training pathological region in the training sample set can be generated, so as to train a preset classifier according to the expected output corresponding to the current training pathological region, and obtain a trained pathological image classifier.
  • the training sample set includes a tumor cell area and a non-tumor cell area, which ensures the comprehensiveness of the training sample set.
  • the cell types trained by using such a training sample set can learn more comprehensive and accurate cell type classification rules.
  • the efficiency of training machine learning preset classifiers can be further improved, so that the efficiency of tumor cell region recognition can be further improved.
  • the training pathological region is used as the input of the preset classifier, and the training cell type is used as the expected output, and the preset classifier is trained to obtain the trained pathological image.
  • the classifier also include:
  • Step 120 obtaining a test sample set, the test sample includes the test effective area and the corresponding test cell type;
  • Step 122 input the test valid area into the preset classifier, and obtain the output verification cell type
  • Step 124 Acquire the error between the verification cell type and the test cell type, and determine that the training of the preset classifier is completed when the error is less than the preset error;
  • a test sample set is obtained, the test sample includes the test effective area and the corresponding test cell type, and the trained machine learning classifier is used to predict the test sample set, and the known classification result of the test sample set, that is, the expected cell is obtained.
  • Type compare the predicted prediction results with the known classification results, get the classification prediction accuracy rate of the corresponding machine learning classifier, and obtain the error between the verified cell type and the expected cell type, if the error is less than the preset error.
  • the parameter values used by the classifier otherwise, continue to train the machine learning classifier using the obtained parameter values and the test sample set.
  • test sample set using the test sample set to roughly locate the parameter value, by obtaining the error between the verification cell type and the expected environment type, or the number of training times, the most suitable parameter value can be found as much as possible, so that the parameter can be used to obtain the most suitable parameter value.
  • Values and test sample sets are used for training, and the trained machine learning classifier can achieve higher accuracy in distinguishing cell types.
  • a system for evaluating tumor cell content which includes:
  • an area determination module 802 configured to acquire a digital pathological slice image, and determine an effective pathological area according to the digital pathological slice image
  • a tumor cell identification module 804 configured to identify the tumor cell region corresponding to the effective pathological region through a deep learning-based pathological image classifier;
  • the content evaluation module 806 is configured to determine the tumor cell content of the digital pathological slice image by using a preset evaluation rule according to the tumor cell region.
  • the area determination module includes:
  • a binarization processing unit configured to perform binarization processing on the digital pathological slice image to obtain a grayscale image
  • a region determination unit configured to extract a region with a grayscale value smaller than a preset grayscale threshold from the grayscale image as an effective pathological region.
  • the content assessment module includes:
  • an area determination unit configured to determine a first area of the tumor cell area and a second area of the effective pathological area
  • a content evaluation unit configured to calculate and obtain the tumor proportion and tumor-stroma ratio of the digital pathological slice image according to the first area and the second area.
  • the tumor cell content assessment system further includes:
  • a diameter acquisition module used for acquiring multiple test diameters corresponding to multiple tumor cells, and performing mean calculation on the multiple test diameters to obtain the average diameter of a single tumor cell;
  • an area calculation module for determining the average area of a single tumor cell according to the average diameter of the single tumor cell
  • a cell number calculation module for calculating the number of tumor cells per unit area according to the average area of the single tumor cell
  • a first number calculation module configured to calculate and obtain a first number of tumor cells according to the first area and the number of tumor cells in the unit area.
  • the tumor cell content assessment system further includes: a second quantity determination module, configured to perform cell segmentation on the tumor cell region through a cell segmentation algorithm to determine the second quantity of tumor cells.
  • the tumor cell content assessment system further includes:
  • a training sample acquisition module used for acquiring a training sample set, the training sample set includes training pathological regions and corresponding training cell types;
  • a classifier training module used for taking the training pathological region as the input of the preset classifier, using the training cell type as the expected output, training the preset classifier, and obtaining the pathological image after training Classifier.
  • the tumor cell content assessment system further includes:
  • test sample acquisition module used for acquiring a test sample set, the test sample includes a test effective area and a corresponding test cell type;
  • a classification test module for inputting the test effective area into a preset classifier to obtain the output verification cell type
  • the verification module is used to obtain the error between the verification cell type and the test cell type, and when the error is less than the preset error, it is determined that the preset classifier has been trained;
  • the training times corresponding to the preset classifier are acquired, and when the training times reaches the maximum preset times, it is determined that the preset classifier has been trained.
  • Figure 9 shows an internal structure diagram of a computer device in one embodiment.
  • the computer device may specifically be a server, and the server includes but is not limited to high-performance computers and high-performance computer clusters.
  • the computer device includes a processor, memory and a network interface connected by a system bus.
  • the memory includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium of the computer device stores an operating system and also stores computer-readable instructions, which, when executed by the processor, can cause the processor to implement a method for evaluating tumor cell content.
  • Computer-readable instructions may also be stored in the internal memory, and when executed by the processor, the computer-readable instructions may cause the processor to execute the method for assessing tumor cell content.
  • FIG. 9 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.
  • the method for assessing tumor cell content provided by the present application may be implemented in the form of computer-readable instructions, and the computer-readable instructions may be executed on a computer device as shown in FIG. 9 .
  • Various program templates constituting the tumor cell content assessment system can be stored in the memory of the computer device. For example, the region determination module 802 , the tumor cell identification module 804 , and the content assessment module 806 .
  • a computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, the processor executing the computer-readable instructions to perform the steps of: obtaining a digital For pathological slice images, an effective pathological area is determined according to the digital pathological slice images; a tumor cell area corresponding to the effective pathological area is identified through a deep learning-based pathological image classifier; according to the tumor cell area, a preset An evaluation rule determines the tumor cell content of the digital pathology slide image.
  • determining an effective pathological area according to the digital pathological slice image includes: performing a binarization process on the digital pathological slice image to obtain a grayscale image; extracting a grayscale value from the grayscale image The area smaller than the preset grayscale threshold is regarded as the effective pathological area.
  • using a preset evaluation rule to determine the tumor cell content of the digital pathological slice image comprising: determining a first area of the tumor cell region, and determining the effective pathology The second area of the region; the tumor proportion and tumor-stroma ratio of the digital pathological slice image are calculated according to the first area and the second area.
  • the method further includes: acquiring multiple test diameters corresponding to multiple tumor cells, Calculate the average value of a plurality of the test diameters to obtain the average diameter of a single tumor cell; determine the average area of a single tumor cell according to the average diameter of the single tumor cell; calculate the unit according to the average area of the single tumor cell The number of tumor cells in the area; the first number of tumor cells is calculated according to the first area and the number of tumor cells in the unit area.
  • the method further includes: performing a cell segmentation algorithm on the tumor cell region by using a cell segmentation algorithm. Segment, determine a second number of tumor cells.
  • the method for evaluating tumor cell content further includes: acquiring a training sample set, the training sample set including a training pathological region and a corresponding training cell type; using the training pathological region as an input of a preset classifier , taking the training cell type as the expected output, and training the preset classifier to obtain the pathological image classifier that has been trained.
  • the training pathological region is used as the input of a preset classifier, and the training cell type is used as the desired output, and the preset classifier is trained to obtain the trained classifier.
  • the pathological image classifier Before the pathological image classifier, it also includes: acquiring a test sample set, the test sample includes a test effective area and a corresponding test cell type; inputting the test effective area into a preset classifier, and acquiring the output verification cell type; acquiring Verify the error between the cell type and the test cell type, and in the case where the error is less than the preset error, determine that the preset classifier is trained; or, obtain the training times corresponding to the preset classifier, and in the When the number of times of training reaches the maximum preset number of times, it is determined that the preset classifier has been trained.
  • One or more non-volatile readable storage media storing computer readable instructions
  • the one or more non-volatile readable storage media storing computer readable instructions
  • one or more processors are caused to perform the following steps: acquiring a digital pathological slice image, and determining an effective pathological area according to the digital pathological slice image; The tumor cell area corresponding to the area; according to the tumor cell area, a preset evaluation rule is used to determine the tumor cell content of the digital pathological slice image.
  • determining an effective pathological area according to the digital pathological slice image includes: performing a binarization process on the digital pathological slice image to obtain a grayscale image; extracting a grayscale value from the grayscale image The area smaller than the preset grayscale threshold is regarded as the effective pathological area.
  • using a preset evaluation rule to determine the tumor cell content of the digital pathological slice image comprising: determining a first area of the tumor cell region, and determining the effective pathology The second area of the region; the tumor proportion and tumor-stroma ratio of the digital pathological slice image are calculated according to the first area and the second area.
  • the method further includes: acquiring multiple test diameters corresponding to multiple tumor cells, Calculate the average value of a plurality of the test diameters to obtain the average diameter of a single tumor cell; determine the average area of a single tumor cell according to the average diameter of the single tumor cell; calculate the unit according to the average area of the single tumor cell The number of tumor cells in the area; the first number of tumor cells is calculated according to the first area and the number of tumor cells in the unit area.
  • the method further includes: performing a cell segmentation algorithm on the tumor cell region by using a cell segmentation algorithm. Segment, determine a second number of tumor cells.
  • the method for evaluating tumor cell content further includes: acquiring a training sample set, the training sample set including a training pathological region and a corresponding training cell type; using the training pathological region as an input of a preset classifier , taking the training cell type as the expected output, and training the preset classifier to obtain the pathological image classifier that has been trained.
  • the training pathological region is used as the input of a preset classifier, and the training cell type is used as the desired output, and the preset classifier is trained to obtain the trained classifier.
  • the pathological image classifier Before the pathological image classifier, it also includes: acquiring a test sample set, the test sample includes a test effective area and a corresponding test cell type; inputting the test effective area into a preset classifier, and acquiring the output verification cell type; acquiring Verify the error between the cell type and the test cell type, and in the case where the error is less than the preset error, determine that the preset classifier is trained; or, obtain the training times corresponding to the preset classifier, and in the When the number of times of training reaches the maximum preset number of times, it is determined that the preset classifier has been trained.
  • Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Road (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM) and so on.
  • SRAM static RAM
  • DRAM dynamic RAM
  • SDRAM synchronous DRAM
  • DDRSDRAM double data rate SDRAM
  • ESDRAM enhanced SDRAM
  • SLDRAM synchronous chain Road (Synchlink) DRAM
  • SLDRAM synchronous chain Road (Synchlink) DRAM
  • Rambus direct RAM
  • DRAM direct memory bus dynamic RAM
  • RDRAM memory bus dynamic RAM

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Biology (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

一种肿瘤细胞含量评估方法，通过获取数字病理切片图像，根据数字病理切片图像确定有效病理区域（102）；通过基于深度学习的病理图像分类器，识别与有效病理区域对应的肿瘤细胞区域（104）；根据肿瘤细胞区域，采用预设的评估规则确定数字病理切片图像的肿瘤细胞含量（106），实现了对数字病理切片图像的肿瘤细胞含量的自动化评估，提高了对肿瘤细胞含量评估的准确性和客观性。此外，还提出了一种肿瘤细胞含量评估系统、计算机设备及存储介质。

Description

肿瘤细胞含量评估方法、系统、计算机设备及存储介质
本申请以2020年7月7日提交的申请号为202010644331.4，名称为“肿瘤细胞含量评估方法、系统、计算机设备及存储介质”的中国发明专利申请为基础，并要求其优先权。
技术领域
本申请涉及图像处理技术领域，尤其涉及一种肿瘤细胞含量评估方法、系统、计算机设备及存储介质。
 背景技术
肿瘤分子病理的基因检测及蛋白检测，目前已有大量的医疗市场需求，对于常见的肺癌、乳腺癌、肠癌等都有了相关的分子病理检测的方案。
 技术问题
目前对于需要精准治疗，如靶向治疗和免疫治疗的肿瘤患者，需要提供肿瘤组织样本进行分子病理检查。而由于临床手术活检取材的病理学样本，受患者因素、肿瘤自身因素及临床手术医生活检取样等多重因素的影响。造成了每一份送检的病理学样本中，组织含量及肿瘤含量存在较大范围的波动。对于组织含量少或者肿瘤含量少的样本，对于分子病理检测技术平台来说，经常造成无法检出，造成假阴性的现象，或者患者将前后两次的样本送检分子病理检查，检测结果出现严重的不一致性的问题。
 技术解决方案
基于此，有必要针对上述问题，提出一种肿瘤细胞含量评估方法、系统、计算机设备及存储介质，以提高对肿瘤细胞含量评估的客观性和准确性。
一种肿瘤细胞含量评估方法，所述方法包括：
获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
一种肿瘤细胞含量评估系统，所述系统包括：
区域确定模块，用于获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
肿瘤细胞识别模块，用于通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
含量评估模块，用于根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
一种计算机设备，包括存储器和处理器，所述存储器存储有计算机可读指令，所述计算机可读指令被所述处理器执行时，使得所述处理器执行以下步骤：
获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
一个或多个存储有计算机可读指令的非易失性可读存储介质，其特征在于，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行如下步骤：
获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
 有益效果
上述肿瘤细胞含量评估方法、系统、计算机设备及存储介质，通过获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量，通过采用深度学习的方法对病理切片图像进行肿瘤细胞的识别，对识别出的细胞区域采用多种评估方式进行全面评估，通过机器视觉的识别技术，保证了肿瘤细胞识别得准确性和客观性，采用多种评估方式进行全面评估进一步提高了肿瘤细胞含量评估得准确性。
 附图说明
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
其中：
图1为一个实施例中肿瘤细胞含量评估方法的流程图；
图2为一个实施例中有效病理区域确定方法的流程图；
图3为一个实施例中数字病理切片图像与灰度图像对比图；
图4为另一个实施例中肿瘤细胞含量评估方法的流程图；
图5为又一个实施例中肿瘤细胞含量评估方法的流程图；
图6为再一个实施例中肿瘤细胞含量评估方法的流程图；
图7为还一个实施例中肿瘤细胞含量评估方法的流程图；
图8为一个实施例中肿瘤细胞含量评估系统的结构框图；
图9为一个实施例中计算机设备的结构框图。
 本发明的实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
如图1所示，在一个实施例中，提供了一种肿瘤细胞含量评估方法，该肿瘤细胞含量评估方法既可以应用于终端，也可以应用于服务器，本实施例以应用于服务器举例说明。该肿瘤细胞含量评估方法具体包括以下步骤：
步骤102，获取数字病理切片图像，根据数字病理切片图像确定有效病理区域。
其中，数字病理切片图像是指将病理切片以数字图像的方式进行显示、保存的病理切片图像。具体地，可以通过数字病理扫描仪对活检样本的病理切片进行数字扫描后获取得到。有效病理区域，指病理切片中包含细胞的区域，其中的细胞包括正常和/或病变细胞。通过对数字病理切片图像进行分割，确定有效病理区域，其中的图像分割方法包括但不限于是图像二值化处理方法，基于机器学习的图像分割模型或者边缘检测算法等。作为本实施例的优选，采用二值化处理方式，即利用细胞区域和非细胞区域的平均灰度值不一致的特点，通过设置一个合理的灰度阈值，过滤掉一部分非细胞区域，从而提取出有效病理区域，该方法相较于机器学习的图像分割模型或者边缘检测算法，更加简单快速。
步骤104，通过基于深度学习的病理图像分类器，识别与有效病理区域对应的肿瘤细胞区域。
其中，深度学习是机器学习中一种基于对数据进行表征学习的方法。观测值(例如图像)可以使用多种方式来表示，如每个像素强度值的向量，或者更抽象地表示成一系列边、特定形状的区域等。而使用某些特定的表示方法更容易从实例中学习任务(例如人脸识别)。深度学习用于建立、模拟人脑进行分析学习的神经网络，它模仿人脑的机制来解释数据，例如图像，声音和文本，用非监督式或半监督式的特征学习和分层特征提取高效算法来替代手工获取特征，能够有利于提高预测结果的客观性和准确性。病理图像分类器是一种分类器，可通过样本学习具备分类能力的机器学习算法模型，本实施例的病理图像分类器用于将不同的有效病理区域划分到肿瘤细胞区域和非肿瘤细胞区域中的一类。具体地，可以利用至少一个机器学习模型进行分类的分类器。其中的机器学习模型可以是如下的一个或者多个：神经网络(例如，卷积神经网络、BP神经网络等)、逻辑回归模型、支持向量机、决策树、随机森林、感知器以及其它机器学习模型。作为这样的机器学习模型的训练的部分，训练输入是各种有效病理区域对应图像，通过训练，建立有效病理区域与肿瘤细胞区域或者非肿瘤细胞区域的对应的关系分类器。使得该病理图像分类器具备判断输入的有效病理区域对应的是肿瘤细胞区域还是非肿瘤细胞区域的能力。本实施例中，该病理图像分类器为二分类器，即得到两个分类结果，也即肿瘤细胞区域或者非肿瘤细胞区域。
非肿瘤细胞区域是指有效病理区域中正常细胞所在的区域，肿瘤细胞区域是指有效病理区域中发生病变的细胞所在的区域，用于医生对该区域的自爆含量进行分析实现疾病诊断。可以理解地，通过深度学习的方式对有效病理区域中的肿瘤细胞区域进行识别，实现了对肿瘤细胞区域的自动检测，提高了对肿瘤细胞识别的准确性和客观性。
步骤106，根据肿瘤细胞区域，采用预设的评估规则确定数字病理切片图像的肿瘤细胞含量。
其中，预设的评估规则是指预先设定的用于评估肿瘤细胞含量的评估方式或者指标，例如可以根据肿瘤细胞区域的面积，细胞数量、肿瘤占比和肿瘤间质比等进行评估。可以理解地，通过采用预设的评估规则进行评估，从而实现了从多个角度对数字病理切片图像的肿瘤细胞含量的自动化评估，保证了评估的全面性和客观性，提高了肿瘤细胞含量的评估得准确性和一致性，避免人工分析存在主观偏差，无法得到统一、精确的评估结果。
上述肿瘤细胞含量评估方法，通过获取数字病理切片图像，根据数字病理切片图像确定有效病理区域；通过基于深度学习的病理图像分类器，识别与有效病理区域对应的肿瘤细胞区域；根据肿瘤细胞区域，采用预设的评估规则确定数字病理切片图像的肿瘤细胞含量，实现了对数字病理切片图像的肿瘤细胞含量的自动化评估，提高了对肿瘤细胞含量评估的准确性和客观性。
如图2所示，在一个实施例中，根据数字病理切片图像确定有效病理区域，包括：
步骤102A，将数字病理切片图像进行二值化处理，得到灰度图像；
步骤102B，从灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
在这个实施例中，有效病理区域的颜色为红色和蓝色两种染色试剂所染区域，背景区域的灰度为无色或者白色，灰度值比有效病理区域的灰度值大，根据有效病理区域的灰度确定预设灰度阈值，因此，通过数字病理切片图像进行二值化处理，其中的二值化处理包括但不限于是全局二值化、基于直方图的最优阈值方法或者基于聚类的OTSU大津阈值方法，即将彩色图片颜色值转变为灰度图像。以肺癌的数字病理切片图像为例，如图3所示的数字病理切片图像与灰度图像对比图。从灰度图像中提取出灰度值小于预设灰度阈值的区域，从而提取出了有效病理区域，以便后续对有效病理区域进行进一步处理。
如图4所示，在一个实施例中，根据肿瘤细胞区域，采用预设的评估规则确定数字病理切片图像的肿瘤细胞含量，包括：
步骤106A，确定肿瘤细胞区域的第一面积，并确定有效病理区域的第二面积；
步骤106B，根据第一面积和第二面积计算得到数字病理切片图像的肿瘤占比和肿瘤间质比。
具体地，根据肿瘤细胞区域可以计算得到第一面积，根据有效病理区域可以计算得到第二面积。采用如下两个公式分别计算肿瘤占比和肿瘤间质比：
P1=S1/S2×100%；
P2=S1/[S2-S1]×100%；
公式中，S1表示为第一面积，S2表示为第二面积，P1表示为肿瘤占比，P2表示为肿瘤间质比。本实施例中，通过肿瘤占比和肿瘤间质比作为评估肿瘤细胞含量的指标，以便医生根据肿瘤占比和肿瘤间质比进行疾病诊断。
如图5所示，在一个实施例中，在根据肿瘤细胞区域，采用预设的评估规则确定数字病理切片图像的肿瘤细胞含量之后，还包括：
步骤108，获取多个肿瘤细胞对应的多个测试直径，并对多个测试直径进行均值计算，得到单个肿瘤细胞的平均直径；
步骤110，根据单个肿瘤细胞的平均直径确定单个肿瘤细胞的平均面积；
步骤112，根据单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数；
步骤114，根据第一面积和单位面积内肿瘤细胞数计算得到肿瘤细胞的第一数量。
具体地，通过预先测试一种肿瘤细胞（如肺癌细胞）的直径，例如测试多个肿瘤细胞（如100个），得到100个测试直径，并对多个测试直径进行均值计算，得到单个肿瘤细胞的平均直径，根据单个肿瘤细胞的平均直径计算单个肿瘤细胞的平均面积， S=π×（d/2)²，d表示为平均直径，S表示为平均面积根据单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数，将第一面积乘以单位面积内肿瘤细胞数，即可计算得到肿瘤细胞的第一数量。本实施例中，通过计算肿瘤细胞区域的细胞的第一数量，以便医生基于该第一数量进行疾病诊断和分析。
在一个实施例中，在根据肿瘤细胞区域，采用预设的评估规则确定数字病理切片图像的肿瘤细胞含量之后，还包括：
通过细胞分割算法对肿瘤细胞区域进行细胞分割，确定肿瘤细胞的第二数量。
其中，细胞分割算法是一种用于对细胞进行分割的方法，包括但不限于是FCN全卷积神经网络的细胞分割算法、U-Net模型的细胞分割算法或者基于深度学习的小分割算法等。通过将肿瘤细胞区域中的细胞分割开来，然后统计细胞的数量，级得到第二数量，以便医生基于该第一数量进行疾病诊断和分析。
值得说明的是，本实施例中第二数量由于是基于细胞本身特征进行计算得到，比步骤114中的第一数量更为精确。因此，适用于更多的检验或者疾病诊断场合。
如图6所示，在一个实施例中，还包括：
步骤116，获取训练样本集，训练样本集中包括训练病理区域和对应的训练细胞类型；
步骤118，将训练病理区域作为预设的分类器的输入，将训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的病理图像分类器。
具体地，获取通过医生进行标注的肿瘤细胞区域和非肿瘤细胞区域的样本，将训练病理区域作为预设的分类器的输入，训练细胞类型作为期望的输出，对预设的分类器进行训练，可生成与训练样本集中的训练病理区域相应的细胞类型，从而根据与当前训练病理区域对应的期望的输出，训练预设的分类器，得到训练完成的病理图像分类器。
本实施例中，训练样本集包括肿瘤细胞区域和非肿瘤细胞区域，保证了训练样本集的全面性，利用这样训练样本集训练出的细胞类型能够学习到更加全面准确的细胞类型分类规则，提高了训练机器学习预设分类器的效率，从而可以进一步提高对肿瘤细胞区域识别的效率。
如图7所示，在一个实施例中，在将训练病理区域作为预设的分类器的输入，将训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的病理图像分类器之前，还包括：
步骤120，获取测试样本集，测试样本包括测试有效区域和对应的测试细胞类型；
步骤122，将测试有效区域输入预设的分类器，获取输出的验证细胞类型；
步骤124，获取验证细胞类型与测试细胞类型之间的误差，在误差小于预设误差的情况下，确定预设的分类器训练完毕；
或，获取预设的分类器对应的训练次数，在训练次数达到最大预设次数的情况下，确定预设的分类器训练完毕。
具体地，获取测试样本集，测试样本包括测试有效区域和对应的测试细胞类型，并利用训练的机器学习的分类器对测试样本集进行预测，获取测试样本集已知的分类结果即期望的细胞类型，将预测得到的预测结果与已知的分类结果比较，得到相应机器学习分类器的分类预测正确率，获取验证细胞类型与期望的细胞类型之间的误差，在误差小于预设误差的情况下，确定预设的分类器训练完毕，或者，获取预设的分类器对应的训练次数，在训练次数达到最大预设次数的情况下，确定预设的分类器训练完毕，获取该机器学习的分类器所用的参数取值，否则，利用获取的参数取值以及测试样本集继续训练机器学习分类器。
本实施例中，利用测试样本集粗略定位参数取值，通过获取验证细胞类型与期望的环境类型之间的误差，或者训练次数，可以尽可能找到最合适的参数取值，从而利用该参数取值以及测试样本集进行训练，训练出的机器学习分类器对细胞类型进行区分可以达到更高的正确率。
如图8所示，在一个实施例中，提出了一种肿瘤细胞含量评估系统，所述系统包括：
区域确定模块802，用于获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
肿瘤细胞识别模块804，用于通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
含量评估模块806，用于根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
在一个实施例中，区域确定模块包括：
二值化处理单元，用于将所述数字病理切片图像进行二值化处理，得到灰度图像；
区域确定单元，用于从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
在一个实施例中，含量评估模块包括：
面积确定单元，用于确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；
含量评估单元，用于根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
在一个实施例中，该肿瘤细胞含量评估系统还包括：
直径获取模块，用于获取多个肿瘤细胞对应的多个测试直径，并对多个所述测试直径进行均值计算，得到单个肿瘤细胞的平均直径；
面积计算模块，用于根据所述单个肿瘤细胞的平均直径确定单个肿瘤细胞的平均面积；
细胞数计算模块，用于根据所述单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数；
第一数量计算模块，用于根据所述第一面积和所述单位面积内肿瘤细胞数计算得到肿瘤细胞的第一数量。
在一个实施例中，该肿瘤细胞含量评估系统还包括：第二数量确定模块，用于通过细胞分割算法对所述肿瘤细胞区域进行细胞分割，确定肿瘤细胞的第二数量。
在一个实施例中，该肿瘤细胞含量评估系统还包括：
训练样本获取模块，用于获取训练样本集，所述训练样本集中包括训练病理区域和对应的训练细胞类型；
分类器训练模块，用于将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器。
在一个实施例中，该肿瘤细胞含量评估系统还包括：
测试样本获取模块，用于获取测试样本集，所述测试样本包括测试有效区域和对应的测试细胞类型；
分类测试模块，用于将所述测试有效区域输入预设的分类器，获取输出的验证细胞类型；
验证模块，用于获取验证细胞类型与测试细胞类型之间的误差，在误差小于预设误差的情况下，确定所述预设的分类器训练完毕；
或，获取所述预设的分类器对应的训练次数，在所述训练次数达到最大预设次数的情况下，确定所述预设的分类器训练完毕。
图9示出了一个实施例中计算机设备的内部结构图。该计算机设备具体可以是服务器，所述服务器包括但不限于高性能计算机和高性能计算机集群。如图9所示，该计算机设备包括通过系统总线连接的处理器、存储器和网络接口。其中，存储器包括非易失性存储介质和内存储器。该计算机设备的非易失性存储介质存储有操作系统，还可存储有计算机可读指令，该计算机可读指令被处理器执行时，可使得处理器实现肿瘤细胞含量评估方法。该内存储器中也可储存有计算机可读指令，该计算机可读指令被处理器执行时，可使得处理器执行肿瘤细胞含量评估方法。本领域技术人员可以理解，图9中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
在一个实施例中，本申请提供的肿瘤细胞含量评估方法可以实现为一种计算机可读指令的形式，计算机可读指令可在如图9所示的计算机设备上运行。计算机设备的存储器中可存储组成肿瘤细胞含量评估系统的各个程序模板。比如，区域确定模块802，肿瘤细胞识别模块804，含量评估模块806。
一种计算机设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现如下步骤：获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
在一个实施例中，根据所述数字病理切片图像确定有效病理区域，包括：将所述数字病理切片图像进行二值化处理，得到灰度图像；从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
在一个实施例中，根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量，包括：确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
在一个实施例中，在所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量之后，还包括：获取多个肿瘤细胞对应的多个测试直径，并对多个所述测试直径进行均值计算，得到单个肿瘤细胞的平均直径；根据所述单个肿瘤细胞的平均直径确定单个肿瘤细胞的平均面积；根据所述单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数；根据所述第一面积和所述单位面积内肿瘤细胞数计算得到肿瘤细胞的第一数量。
在一个实施例中，在所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量之后，还包括：通过细胞分割算法对所述肿瘤细胞区域进行细胞分割，确定肿瘤细胞的第二数量。
在一个实施例中，该肿瘤细胞含量评估方法还包括：获取训练样本集，所述训练样本集中包括训练病理区域和对应的训练细胞类型；将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器。
在一个实施例中，在所述将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器之前，还包括：获取测试样本集，所述测试样本包括测试有效区域和对应的测试细胞类型；将所述测试有效区域输入预设的分类器，获取输出的验证细胞类型；获取验证细胞类型与测试细胞类型之间的误差，在误差小于预设误差的情况下，确定所述预设的分类器训练完毕；或，获取所述预设的分类器对应的训练次数，在所述训练次数达到最大预设次数的情况下，确定所述预设的分类器训练完毕。
一个或多个存储有计算机可读指令的非易失性可读存储介质，该一个或多个存储有计算机可读指令的非易失性可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行如下步骤：获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
在一个实施例中，根据所述数字病理切片图像确定有效病理区域，包括：将所述数字病理切片图像进行二值化处理，得到灰度图像；从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
在一个实施例中，根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量，包括：确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
在一个实施例中，在所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量之后，还包括：获取多个肿瘤细胞对应的多个测试直径，并对多个所述测试直径进行均值计算，得到单个肿瘤细胞的平均直径；根据所述单个肿瘤细胞的平均直径确定单个肿瘤细胞的平均面积；根据所述单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数；根据所述第一面积和所述单位面积内肿瘤细胞数计算得到肿瘤细胞的第一数量。
在一个实施例中，在所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量之后，还包括：通过细胞分割算法对所述肿瘤细胞区域进行细胞分割，确定肿瘤细胞的第二数量。
在一个实施例中，该肿瘤细胞含量评估方法还包括：获取训练样本集，所述训练样本集中包括训练病理区域和对应的训练细胞类型；将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器。
在一个实施例中，在所述将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器之前，还包括：获取测试样本集，所述测试样本包括测试有效区域和对应的测试细胞类型；将所述测试有效区域输入预设的分类器，获取输出的验证细胞类型；获取验证细胞类型与测试细胞类型之间的误差，在误差小于预设误差的情况下，确定所述预设的分类器训练完毕；或，获取所述预设的分类器对应的训练次数，在所述训练次数达到最大预设次数的情况下，确定所述预设的分类器训练完毕。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机可读指令来指令相关的硬件来完成，所述的程序可存储于一非易失性计算机可读取存储介质中，该程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink) DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对本申请专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种肿瘤细胞含量评估方法，其特征在于，包括：
    获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
    通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
    根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
  2. 根据权利要求1所述的肿瘤细胞含量评估方法，其特征在于，所述根据所述数字病理切片图像确定有效病理区域，包括：
    将所述数字病理切片图像进行二值化处理，得到灰度图像；
    从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
  3. 根据权利要求1所述的肿瘤细胞含量评估方法，其特征在于，所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量，包括：
    确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；
    根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
  4. 根据权利要求3所述的肿瘤细胞含量评估方法，其特征在于，在所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量之后，还包括：
    获取多个肿瘤细胞对应的多个测试直径，并对多个所述测试直径进行均值计算，得到单个肿瘤细胞的平均直径；
    根据所述单个肿瘤细胞的平均直径确定单个肿瘤细胞的平均面积；
    根据所述单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数；
    根据所述第一面积和所述单位面积内肿瘤细胞数计算得到肿瘤细胞的第一数量。
  5. 根据权利要求1所述的肿瘤细胞含量评估方法，其特征在于，在所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量之后，还包括：
    通过细胞分割算法对所述肿瘤细胞区域进行细胞分割，确定肿瘤细胞的第二数量。
  6. 根据权利要求1所述的肿瘤细胞含量评估方法，其特征在于，所述方法还包括：
    获取训练样本集，所述训练样本集中包括训练病理区域和对应的训练细胞类型；
    将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器。
  7. 根据权利要求6所述的肿瘤细胞含量评估方法，其特征在于，在所述将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器之前，还包括：
    获取测试样本集，所述测试样本包括测试有效区域和对应的测试细胞类型；
    将所述测试有效区域输入预设的分类器，获取输出的验证细胞类型；
    获取验证细胞类型与测试细胞类型之间的误差，在误差小于预设误差的情况下，确定所述预设的分类器训练完毕；
    或，获取所述预设的分类器对应的训练次数，在所述训练次数达到最大预设次数的情况下，确定所述预设的分类器训练完毕。
  8. 一种肿瘤细胞含量评估系统，其特征在于，所述肿瘤细胞含量评估系统包括：
    区域确定模块，用于获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
    肿瘤细胞识别模块，用于通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
    含量评估模块，用于根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
  9. 根据权利要求8所述的肿瘤细胞含量评估系统，其特征在于，所述区域确定模块包括：
    二值化处理单元，用于将所述数字病理切片图像进行二值化处理，得到灰度图像；
    区域确定单元，用于从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
  10. 根据权利要求8所述的肿瘤细胞含量评估系统，其特征在于，所述含量评估模块包括：
    面积确定单元，用于确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；
    含量评估单元，用于根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
  11. 根据权利要求1所述的肿瘤细胞含量评估系统，其特征在于，所述肿瘤细胞含量评估系统还包括：
    直径获取模块，用于获取多个肿瘤细胞对应的多个测试直径，并对多个所述测试直径进行均值计算，得到单个肿瘤细胞的平均直径；
    面积计算模块，用于根据所述单个肿瘤细胞的平均直径确定单个肿瘤细胞的平均面积；
    细胞数计算模块，用于根据所述单个肿瘤细胞的平均面积，计算得到单位面积内肿瘤细胞数；
    第一数量计算模块，用于根据所述第一面积和所述单位面积内肿瘤细胞数计算得到肿瘤细胞的第一数量。
  12. 根据权利要求8所述的肿瘤细胞含量评估系统，其特征在于，所述肿瘤细胞含量评估系统还包括：
    第二数量确定模块，用于通过细胞分割算法对所述肿瘤细胞区域进行细胞分割，确定肿瘤细胞的第二数量。
  13. 根据权利要求8所述的肿瘤细胞含量评估系统，其特征在于，所述肿瘤细胞含量评估系统还包括：
    训练样本获取模块，用于获取训练样本集，所述训练样本集中包括训练病理区域和对应的训练细胞类型；
    分类器训练模块，用于将所述训练病理区域作为预设的分类器的输入，将所述训练细胞类型作为期望的输出，对预设的分类器进行训练，得到训练完成的所述病理图像分类器。
  14. 根据权利要求13所述的肿瘤细胞含量评估系统，其特征在于，所述肿瘤细胞含量评估系统还包括：
    测试样本获取模块，用于获取测试样本集，所述测试样本包括测试有效区域和对应的测试细胞类型；
    分类测试模块，用于将所述测试有效区域输入预设的分类器，获取输出的验证细胞类型；
    验证模块，用于获取验证细胞类型与测试细胞类型之间的误差，在误差小于预设误差的情况下，确定所述预设的分类器训练完毕；
    或，获取所述预设的分类器对应的训练次数，在所述训练次数达到最大预设次数的情况下，确定所述预设的分类器训练完毕。
  15. 一种计算机设备，其特征在于，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令，其特征在于，所述处理器执行所述计算机可读指令时实现如下步骤：
    获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
    通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
    根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
  16. 根据权利要求15所述的计算机设备，其特征在于，所述根据所述数字病理切片图像确定有效病理区域，包括：
    将所述数字病理切片图像进行二值化处理，得到灰度图像；
    从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
  17. 根据权利要求15所述的计算机设备，其特征在于，所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量，包括：
    确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；
    根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
  18. 一个或多个存储有计算机可读指令的非易失性可读存储介质，其特征在于，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行如下步骤：
    获取数字病理切片图像，根据所述数字病理切片图像确定有效病理区域；
    通过基于深度学习的病理图像分类器，识别与所述有效病理区域对应的肿瘤细胞区域；
    根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量。
  19. 根据权利要求18所述的非易失性可读存储介质，其特征在于，所述根据所述数字病理切片图像确定有效病理区域，包括：
    将所述数字病理切片图像进行二值化处理，得到灰度图像；
    从所述灰度图像中提取出灰度值小于预设灰度阈值的区域作为有效病理区域。
  20. 根据权利要求18所述的非易失性可读存储介质，其特征在于，所述根据所述肿瘤细胞区域，采用预设的评估规则确定所述数字病理切片图像的肿瘤细胞含量，包括：
    确定所述肿瘤细胞区域的第一面积，并确定所述有效病理区域的第二面积；
    根据所述第一面积和所述第二面积计算得到所述数字病理切片图像的肿瘤占比和肿瘤间质比。
PCT/CN2020/137029 2020-07-07 2020-12-17 肿瘤细胞含量评估方法、系统、计算机设备及存储介质 WO2022007337A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010644331.4A CN111967465A (zh) 2020-07-07 2020-07-07 肿瘤细胞含量评估方法、系统、计算机设备及存储介质
CN202010644331.4 2020-07-07

Publications (1)

Publication Number Publication Date
WO2022007337A1 true WO2022007337A1 (zh) 2022-01-13

Family

ID=73361259

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/137029 WO2022007337A1 (zh) 2020-07-07 2020-12-17 肿瘤细胞含量评估方法、系统、计算机设备及存储介质

Country Status (2)

Country Link
CN (1) CN111967465A (zh)
WO (1) WO2022007337A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114743195A (zh) * 2022-04-13 2022-07-12 赛维森(广州)医疗科技服务有限公司 甲状腺细胞病理数字图像识别器训练方法、图像识别方法
CN114820502A (zh) * 2022-04-21 2022-07-29 济宁医学院附属医院 一种用于肠粘膜组织中蛋白激酶ck2的着色检测方法
CN116313115A (zh) * 2023-05-10 2023-06-23 浙江大学 基于线粒体动态表型和深度学习的药物作用机制预测方法
CN116580397A (zh) * 2023-07-12 2023-08-11 北京大学 病理图像识别方法、装置、设备及存储介质
CN116703917A (zh) * 2023-08-07 2023-09-05 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析系统

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111967465A (zh) * 2020-07-07 2020-11-20 广州金域医学检验中心有限公司 肿瘤细胞含量评估方法、系统、计算机设备及存储介质
CN112884724B (zh) * 2021-02-02 2022-06-03 广州智睿医疗科技有限公司 一种用于肺癌组织病理分型的智能判断方法及系统
CN113790946A (zh) * 2021-11-14 2021-12-14 梅傲科技(广州)有限公司 用于数字病理扫描分析系统的细胞间质染色试剂盒
CN114235539A (zh) * 2021-12-22 2022-03-25 宁波舜宇仪器有限公司 基于深度学习的pd-l1病理切片自动判读方法及系统
CN114581908B (zh) * 2022-02-21 2024-03-19 广州锟元方青医疗科技有限公司 一种pd-l1免疫组化评分方法、系统、装置与存储介质
CN115330949B (zh) * 2022-08-16 2024-04-05 中山大学附属第一医院 一种肝细胞癌侵袭边缘预测方法及系统
CN116580216B (zh) * 2023-07-12 2023-09-22 北京大学 病理图像匹配方法、装置、设备及存储介质
CN117593746B (zh) * 2024-01-18 2024-04-19 武汉互创联合科技有限公司 基于目标检测的细胞分裂均衡度评估系统及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017198790A1 (en) * 2016-05-18 2017-11-23 F. Hoffmann-La Roche Ag Tumor proximity measure
CN109063747A (zh) * 2018-07-16 2018-12-21 武汉大学人民医院（湖北省人民医院） 基于深度学习的肠道病理切片图像识别分析系统及方法
CN110060244A (zh) * 2019-04-15 2019-07-26 深圳市麦迪普科技有限公司 基于深度学习神经网络的细胞检测与分割的系统和方法
CN110111895A (zh) * 2019-05-15 2019-08-09 天津医科大学肿瘤医院 一种鼻咽癌远端转移预测模型的建立方法
CN110992303A (zh) * 2019-10-29 2020-04-10 平安科技（深圳）有限公司 一种异常细胞筛选方法、装置、电子设备和存储介质
CN111126162A (zh) * 2019-11-28 2020-05-08 东软集团股份有限公司 一种识别图像中炎症细胞的方法、装置及存储介质
CN111967465A (zh) * 2020-07-07 2020-11-20 广州金域医学检验中心有限公司 肿瘤细胞含量评估方法、系统、计算机设备及存储介质

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108717554A (zh) * 2018-05-22 2018-10-30 复旦大学附属肿瘤医院 一种甲状腺肿瘤病理组织切片图像分类方法及其装置
CN109271969B (zh) * 2018-10-29 2020-03-24 北京青燕祥云科技有限公司 一种脑胶质瘤分级评价方法及其装置
CN110632069B (zh) * 2019-08-20 2022-07-26 西人马大周(深圳)医疗科技有限公司 循环肿瘤细胞检测方法、装置、设备及介质
CN110517268A (zh) * 2019-08-29 2019-11-29 中国科学院深圳先进技术研究院 病理图像处理方法、装置、图像分析系统及存储介质
CN110705565A (zh) * 2019-09-09 2020-01-17 西安电子科技大学 淋巴结肿瘤区域识别方法与装置
CN111161212B (zh) * 2019-12-05 2024-01-16 广州金域医学检验中心有限公司 数字病理切片有丝核分裂象统计方法、装置、设备和介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017198790A1 (en) * 2016-05-18 2017-11-23 F. Hoffmann-La Roche Ag Tumor proximity measure
CN109063747A (zh) * 2018-07-16 2018-12-21 武汉大学人民医院（湖北省人民医院） 基于深度学习的肠道病理切片图像识别分析系统及方法
CN110060244A (zh) * 2019-04-15 2019-07-26 深圳市麦迪普科技有限公司 基于深度学习神经网络的细胞检测与分割的系统和方法
CN110111895A (zh) * 2019-05-15 2019-08-09 天津医科大学肿瘤医院 一种鼻咽癌远端转移预测模型的建立方法
CN110992303A (zh) * 2019-10-29 2020-04-10 平安科技（深圳）有限公司 一种异常细胞筛选方法、装置、电子设备和存储介质
CN111126162A (zh) * 2019-11-28 2020-05-08 东软集团股份有限公司 一种识别图像中炎症细胞的方法、装置及存储介质
CN111967465A (zh) * 2020-07-07 2020-11-20 广州金域医学检验中心有限公司 肿瘤细胞含量评估方法、系统、计算机设备及存储介质

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114743195A (zh) * 2022-04-13 2022-07-12 赛维森(广州)医疗科技服务有限公司 甲状腺细胞病理数字图像识别器训练方法、图像识别方法
CN114743195B (zh) * 2022-04-13 2022-12-09 赛维森(广州)医疗科技服务有限公司 甲状腺细胞病理数字图像识别器训练方法、图像处理方法
CN114820502A (zh) * 2022-04-21 2022-07-29 济宁医学院附属医院 一种用于肠粘膜组织中蛋白激酶ck2的着色检测方法
CN114820502B (zh) * 2022-04-21 2023-10-24 济宁医学院附属医院 一种用于肠粘膜组织中蛋白激酶ck2的着色检测方法
CN116313115A (zh) * 2023-05-10 2023-06-23 浙江大学 基于线粒体动态表型和深度学习的药物作用机制预测方法
CN116313115B (zh) * 2023-05-10 2023-08-15 浙江大学 基于线粒体动态表型和深度学习的药物作用机制预测方法
CN116580397A (zh) * 2023-07-12 2023-08-11 北京大学 病理图像识别方法、装置、设备及存储介质
CN116580397B (zh) * 2023-07-12 2023-11-10 北京大学 病理图像识别方法、装置、设备及存储介质
CN116703917A (zh) * 2023-08-07 2023-09-05 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析系统
CN116703917B (zh) * 2023-08-07 2024-01-26 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析系统

Also Published As

Publication number Publication date
CN111967465A (zh) 2020-11-20

Similar Documents

Publication Publication Date Title
WO2022007337A1 (zh) 肿瘤细胞含量评估方法、系统、计算机设备及存储介质
US11468564B2 (en) Systems and methods for automatic detection and quantification of pathology using dynamic feature classification
CN110120040B (zh) 切片图像处理方法、装置、计算机设备和存储介质
KR102583103B1 (ko) 계산 검출 방법들을 위해 전자 이미지들을 처리하기 위한 시스템들 및 방법들
WO2020182710A1 (en) Multiple instance learner for prognostic tissue pattern identification
CN109670510A (zh) 一种基于深度学习的胃镜活检病理数据筛查系统和方法
WO2023124562A1 (zh) 基于机器学习的细胞分割和分型方法、装置、设备及介质
Xu et al. Using transfer learning on whole slide images to predict tumor mutational burden in bladder cancer patients
CN114332132A (zh) 图像分割方法、装置和计算机设备
CN115546605A (zh) 一种基于图像标注和分割模型的训练方法及装置
CN114511523B (zh) 基于自监督学习的胃癌分子亚型分类方法及装置
RU2668699C1 (ru) Интеллектуальный способ диагностики и обнаружения новообразований в легких
US20220036558A1 (en) Method and system for predicting expression of biomarker from medical image
CN109800820A (zh) 一种基于超声造影图像均匀程度的分类方法
Silva et al. Development of a quantitative semi-automated system for intestinal morphology assessment in Atlantic salmon, using image analysis
CN114529554A (zh) 一种胃癌her2数字病理切片智能辅助判读方法
US20210319554A1 (en) Method and system for assessing fibrosis in a tissue sample
US11798163B2 (en) Systems and methods for quantitative phenotyping of fibrosis
Ganapathy Cancer detection using deep neural network differentiation of squamous carcinoma cells in oral pathology
US20240029247A1 (en) Systems and methods for quantitative phenotyping of biological fibrilar structures
Perumprath et al. Deep Learning for Segmentation of Brain Tumors using MR Images based on U-Net Architecture
CN118014945A (zh) 一种动态超声造影图像识别方法、装置、设备和介质
CN117524499A (zh) 一种晚期非小细胞肺癌免疫治疗疗效评估方法及系统
Musleh Machine learning framework for simulation of artifacts in paranasal sinuses diagnosis using CT images
CN111312372A (zh) 医学图像标准测试数据集的建立方法及装置

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 04.09.2023)

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20944542

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20944542

Country of ref document: EP

Kind code of ref document: A1