CN110705565A - Lymph node tumor region identification method and device - Google Patents
Lymph node tumor region identification method and device Download PDFInfo
- Publication number
- CN110705565A CN110705565A CN201910857766.4A CN201910857766A CN110705565A CN 110705565 A CN110705565 A CN 110705565A CN 201910857766 A CN201910857766 A CN 201910857766A CN 110705565 A CN110705565 A CN 110705565A
- Authority
- CN
- China
- Prior art keywords
- tumor
- lymph node
- region
- tumor region
- area
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30096—Tumor; Lesion
Abstract
The invention discloses a lymph node tumor region identification method and a lymph node tumor region identification device. The method comprises the steps of obtaining a plurality of local slices of a lymph node image to be detected, inputting each local slice into a trained deep learning neural network model for classification prediction to obtain tumor probability characteristic vectors of the local slices, inputting the tumor probability characteristic vectors into a conditional random field model to obtain tumor confidence coefficients of the local slices, mapping the local slices into the lymph node image to be detected, obtaining a tumor region heat map of a lymph node region to be detected according to the tumor confidence coefficients, generating a tumor region mask image according to the tumor region heat map, extracting a tumor region outline in the tumor region mask image through edge detection, and calculating the area occupation ratio of the tumor region area in the lymph node. The method has the advantages that the tumor area on the lymph nodes is automatically identified, the tumor area proportion of each lymph node can be rapidly and accurately calculated, meanwhile, the burden of doctors is reduced, and the time cost and the labor cost are saved.
Description
Technical Field
The invention relates to the field of image processing, in particular to a lymph node tumor region identification method and a lymph node tumor region identification device.
Background
Lymph node metastasis is a common form of tumor metastasis, and when cancer cells escape from a tumor, they can reach other parts of the body through the blood or lymphatic system, and can reach distant organs through the blood, and if they spread through the lymphatic system, they may end up in the lymph nodes, and grow out of the same tumor as a center.
At present, doctors' treatment of cancer is based on the type of cancer a person suffers from and the stage of the cancer, for example, doctors select a staging standard to assign the corresponding stage of development to the cancer, the most common staging system at present is the TNM system, where T stands for tumor, M for metastasis and N for lymph nodes. The factor of lymph node metastasis is seen to have great influence on the stage of cancer, so that the classification of lymph node metastasis can provide guarantee for accurately evaluating the prognosis of cancer patients. For example, the international union for anticancer proposes that lymph node involvement should be classified according to the number of metastatic lymph nodes.
However, the observation of the number of metastatic lymph nodes of a cancer patient by a doctor is time-consuming and labor-consuming, the doctor cannot accurately estimate the area ratio of a tumor region on the whole lymph node, the diagnosis accuracy rate may be reduced due to fatigue when a large number of histopathological images are read in a short period, and meanwhile, the problems of large workload, strong subjectivity, low efficiency and the like exist in the manual diagnosis process. Therefore, there is a need to provide a method for identifying tumor regions on lymph nodes by using a deep learning method, and calculating the area ratio of tumor metastasis on lymph nodes to reduce the workload of doctors and improve the identification accuracy.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art. Therefore, the invention aims to provide a method for identifying a tumor region on a lymph node by using a deep learning method, and simultaneously calculating the area ratio of tumor metastasis on the lymph node to reduce the workload of doctors and improve the identification accuracy.
The technical scheme adopted by the invention is as follows:
in a first aspect, the present invention provides a lymph node tumor region identification method, comprising:
acquiring a lymph node mask image of a lymph node image to be detected and cutting the lymph node mask image to generate a plurality of local sections;
inputting each local slice into a trained deep learning neural network model for classification prediction to obtain tumor probability feature vectors of the local slices;
inputting the tumor probability feature vector into a conditional random field model to obtain the tumor confidence of the local slice;
mapping the local section to the lymph node image to be detected, and obtaining a tumor region heat map of the lymph node region to be detected according to the tumor confidence;
generating a tumor region mask image according to the tumor region heat map, extracting a tumor region outline in the tumor region mask image through edge detection, and calculating the area ratio of the tumor region area in the lymph node.
Further, the classification prediction process is as follows: dividing the input local slice into a plurality of small slices with the same size, acquiring a central pixel of the small slices, and obtaining a tumor probability feature vector according to the category of the central pixel.
Further, the conditional random field model satisfies Gibbs distribution, and similarity of features in the tumor probability feature vector is judged through cosine similarity.
Further, the energy function of the conditional random field model is represented as:
wherein the content of the first and second substances,
representing a univariate potential for measuring small slices xiClassification as label yiThe cost of (a) of (b),
representing potential pairwise for measuring small slices xiAnd small slice xjCost, w, of classifying as one and the same tagijRepresents the training control parameter when yi＝yjWhen the condition [ mu ] (y) is satisfiedi＝yj)＝1。
Further, the process of extracting the tumor region contour in the tumor region mask image through edge detection and calculating the area ratio of the tumor region area in the lymph node specifically comprises the following steps:
extracting pixel coordinates of a lymph node outline and pixel coordinates of a tumor region outline in the lymph node mask image and the tumor region mask image;
respectively fitting to obtain a tumor region contour and a lymph node region contour;
and counting the number of nonzero pixel points in the contour of the tumor region, and calculating to obtain the area ratio of the tumor region in the lymph node region.
Further, still include: and when the lymph node outline area in the lymph node mask image is detected to be smaller than a preset threshold value, excluding the lymph node area corresponding to the lymph node outline.
Further, when the deep learning neural network model is trained, model pre-training is carried out through the ImageNet data set.
In a second aspect, the present invention also provides a lymph node tumor region identification apparatus comprising:
the acquisition module is used for acquiring a lymph node mask image of a lymph node image to be detected and cutting the lymph node mask image to generate a plurality of local sections;
a classification prediction module: the local section is input into a trained deep learning neural network model for classification prediction, and a tumor probability feature vector of the local section is obtained;
obtaining a tumor confidence module: the tumor probability feature vector is input into a conditional random field model to obtain the tumor confidence of the local slice;
generating a tumor region heat map module: the local section is mapped to the lymph node image to be detected, and a tumor region heat map of the lymph node region to be detected is obtained according to the tumor confidence;
the area ratio calculation module: the tumor area mask image is generated according to the tumor area heat map, the tumor area contour in the tumor area mask image is extracted through edge detection, and the area ratio of the tumor area in the lymph node is calculated.
In a third aspect, the present invention provides a lymph node tumor region identification apparatus comprising:
at least one processor, and a memory communicatively coupled to the at least one processor;
wherein the processor is adapted to perform the method of any of the first aspects by invoking a computer program stored in the memory.
In a fourth aspect, the present invention provides a computer-readable storage medium having stored thereon computer-executable instructions for causing a computer to perform the method of any of the first aspects.
The invention has the beneficial effects that:
the method comprises the steps of obtaining a plurality of local slices of a lymph node image to be detected, inputting each local slice into a trained deep learning neural network model for classification prediction to obtain tumor probability feature vectors of the local slices, inputting the tumor probability feature vectors into a conditional random field model to obtain tumor confidence of the local slices, mapping the local slices into the lymph node image to be detected, obtaining a tumor region heat map of the lymph node region to be detected according to the tumor confidence, finally generating a tumor region mask image according to the tumor region heat map, extracting a tumor region outline in the tumor region mask image through edge detection, and calculating the area occupation ratio of the tumor region area in the lymph node. According to the invention, the tumor region heat map is obtained by inputting the digitized histopathology image into the deep learning neural network model and the conditional random field model, the tumor region is automatically identified, the tumor region proportion of each lymph node can be rapidly and accurately calculated, meanwhile, the burden of a doctor is reduced, and the time cost and the labor cost are saved.
Can be widely applied to the field of medical pathological image processing.
Drawings
FIG. 1 is a flow chart of an implementation of a lymph node tumor area identification method according to an embodiment of the present invention;
fig. 2 is a block diagram showing the structure of a lymph node tumor region identification apparatus according to an embodiment of the present invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following description will be made with reference to the accompanying drawings. It is obvious that the drawings in the following description are only some examples of the invention, and that for a person skilled in the art, other drawings and embodiments can be derived from them without inventive effort.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
The first embodiment is as follows:
fig. 1 is a flowchart illustrating an implementation of a method for identifying a lymph node tumor region according to an embodiment of the present invention, and as shown in fig. 1, the method includes the following steps:
s1: and acquiring a lymph node mask image of the lymph node image to be detected and cutting to generate a plurality of local sections.
In the embodiment, a high-resolution (for example, 40 times of magnification) whole slice scanning device is used for scanning pathological slices into a digital tiff format histopathology image, so that the tumor region of the pathology image is observed and labeled. And then, scaling the high-resolution image by corresponding multiple to obtain a down-sampled original image.
Obtaining a lymph node mask image according to a lymph node region in an original image, and distinguishing the lymph node region, wherein in an optional mode, the lymph node region is marked to be white, then each pixel point coordinate of the white part is mapped to a histopathological section amplified by 20 times, and local section segmentation is carried out by taking the coordinate as a center, namely, the pixel point of each lymph node region is subjected to local section segmentation once to obtain a local section by taking the pixel point as the center, and optionally, the size of the local section is set to be 768x768 by considering the identification efficiency and the accuracy.
S2: and inputting each local slice into the trained deep learning neural network model for classification and prediction to obtain the tumor probability characteristic vector of the local slice.
S3: and inputting the tumor probability feature vector into a conditional random field model to obtain the tumor confidence of the local slice, wherein the conditional random field model meets Gibbs distribution, and the similarity of the features in the tumor probability feature vector is judged through cosine similarity.
S4: and mapping the local section to a lymph node image to be detected, and obtaining a tumor region heat map of the lymph node region to be detected according to the tumor confidence.
S5: generating a tumor area mask image according to the tumor area heat map, extracting a tumor area outline in the tumor area mask image through edge detection, and calculating the area ratio of the tumor area in the lymph node.
Specifically, in step S2, before performing the recognition, the deep learning neural network model needs to be trained to obtain the deep learning neural network model with optimized parameters, and optionally, a convolutional neural network, such as a resnet-18 model, a resnet-34 model, and the like, but not limited thereto, the deep learning models capable of implementing the classification recognition function in this embodiment all belong to the protection scope of the present invention, and the deep learning neural network model training process is described below.
Generating a sample set for training the model, in a specific embodiment, training and verifying the deep learning neural network model by using 900 marked pathological sections, namely WSIs (white slide images whole-field digital pathological sections), randomly selecting 500 WSIs with tumor metastasis and 200 WSIs without tumor metastasis as training sets for training, and using the other 200 WSIs as verification sets for verifying network performance.
The procedure, which first excluded the background region of each WSI, i.e., the non-lymph node region, based on the obtained lymph node mask image, and then performed local slice segmentation on each pathological section at 20 × magnification in order, resulting in 768 × 768 pixel local slices, yielded 761753 local slices in total, of which 633862 local slices were used as the training set and 127891 local slices were used as the test set.
In an optional mode, model pre-training is carried out through an ImageNet data set, network parameters are initialized, meanwhile, a cross entropy loss function is selected, the loss is calculated by using the cross entropy between the prediction probability and an actual label, the momentum is 0.9, the weight of 1e-4 is attenuated by using a random gradient descent mode, the batch size is set to be 1024 during actual training, the initial learning rate of the network is set to be 0.001, the iteration times are 80000 times, and the learning rate is attenuated along with the iteration times in the training process.
Further, in the verification process and the test process, 0.5 is used as a classification threshold value to distinguish a tumor region from a normal region in a lymph node region, and meanwhile, the recall rate, the accuracy and the model evaluation index AUC are calculated, so that in the actual verification process, the efficiency and the accuracy of model classification identification on a verification set are very high, the recall rate can reach 96.4%, the accuracy can reach 96.3%, and the AUC can reach 0.990.
In this embodiment, the local slice is input into the trained deep learning neural network model for classification prediction to obtain a tumor probability feature vector of the local slice, and the classification prediction is used to extract features of the local slice, and the process is as follows: dividing an input local slice into a plurality of small slices with the same size, obtaining central pixels of the small slices, obtaining a tumor probability feature vector according to the type of the central pixels, optionally setting the size of the small slices to be 256x256, namely dividing the local slices with the size of 768x768 into the small slices with the size of 256x256 of a 3x3 grid, wherein the corresponding tumor probability feature vector is a matrix of 3x3, and the feature of the matrix represents that the central pixels of the current small slices are labels corresponding to tumors or labels corresponding to normal lymph nodes.
In step S3, a conditional random field model (denoted as CRF) is used to perform spatial correlation modeling on the local slice, where the conditional random field model is a markov random field of a random variable Y given a random variable X, Y is an output variable representing a marker sequence and X is an input variable representing an observation sequence in a conditional probability model P (Y/X). The method is used for carrying out global normalization on all the features, namely converting the local features into global features, thereby obtaining a global optimal solution.
In this example, a conditional random field satisfying a Gibbs distribution is used to model a conditional distribution P (Y | X), where Y is a label for each random variable X, and a fixed-length tumor probability feature vector encoded as a 3X3 grid is input, and the energy function is expressed as:
wherein the content of the first and second substances,
representing a univariate potential for measuring small slices xiClassification as label yiThe cost of (a) of (b),representing potential pairwise for measuring small slices xiAnd small slice xjCost, w, of classifying as one and the same tagijRepresents the training control parameter when yi＝yjWhen the condition [ mu ] (y) is satisfiedi＝yj) Otherwise, it is 0.
The distance between two features is measured by cosine similarity, corresponding punishment is applied to small slices with the same label and lower similarity, so as to encourage the small slices with the same label to have higher similarity, in addition, the value of E (Y, X) is obtained through the unitary potential and the paired potential calculation, so that the required value of P (Y | X) is obtained, finally, the P (Y | X) is maximized through the average field inference process, and the output of Y when the P (Y | X) is maximized is obtained as the final tumor confidence coefficient, so that the probability that the local slice belongs to the tumor region is obtained.
In step S4, the local slices are mapped to the lymph node image to be detected, and are mapped to the original image according to the coordinate information of each local slice, the tumor confidence of the small slice located at the center after each local slice is divided is selected as the tumor probability of the pixel point corresponding to the local slice in the original image, so as to obtain a probability map of the tumor region in the original image, and then different colors are assigned according to different probabilities to generate a tumor region heat map, for example, red represents that the region is high in tumor probability, and blue represents that the region is low in tumor probability. Similar to the way that a doctor checks the peripheral region of a tumor to assist in classification, the spatial correlation of the adjacent small slices is established through the conditional random field model, and the classification accuracy of the central small slice is improved.
In step S5, a tumor area mask image is first generated according to the tumor area heat map, and optionally, 0.5 is used as a classification threshold to mark the tumor area as white and other areas as black, and edge detection is performed to extract a tumor area contour in the tumor area mask image, and an area ratio of the tumor area in the lymph node is calculated.
The edge detection process is as follows: firstly, extracting pixel coordinates of a lymph node contour and pixel coordinates of a tumor region contour in a lymph node mask image and a tumor region mask image, and respectively fitting to obtain the tumor region contour and the lymph node region contour, namely extracting coordinate points along the periphery of the edge of a target object, wherein the periphery of the edge comprises: the selected coordinate points can be coordinate points on the edge of the target object or coordinate points of the area near the edge, the more the selected coordinate points are, the closer the fitted area is to the target object, and then the lymph node area and the tumor area in the lymph node area are drawn on the original image according to pixel coordinates.
Calculating to obtain the tumor area ratio of each lymph node by using a mask phase and phase mode, namely counting the number of non-zero pixel points in the outline of the tumor area and the outline of the lymph node area, dividing to obtain the tumor area ratio of each lymph node, and marking the value on the corresponding lymph node in the original image.
Further, when the area of the lymph node outline in the lymph node mask image is detected to be smaller than a preset threshold value, the lymph node area corresponding to the lymph node outline is excluded, so that data which do not have calculation significance are removed, and the detection efficiency and accuracy are improved.
Clinical research shows that the larger the area ratio of the tumor regions on the lymph nodes, the smaller the postoperative survival rate of the patient is, for example, a patient with a severe cancer stage, and if the area ratio of the tumor regions with lymph node metastasis is smaller, the survival rate of the patient is statistically much higher than that of a patient with a lighter stage but a larger tumor region, so that the area ratio of the tumor regions with lymph node metastasis has a negative correlation influence on the survival stage of the patient, and the area ratio of the tumor regions with lymph node metastasis can quantify the survival rate of the patient to a certain extent, so that the method has a very important application prospect in the aspects of tumor treatment and prognosis.
In the embodiment, the tumor region heat map is obtained by inputting the digitized histopathology image into the deep learning neural network model and the conditional random field model, the tumor region is automatically identified, the tumor region proportion of each lymph node can be rapidly and accurately calculated, the burden of a doctor is reduced, and the time cost and the labor cost are saved.
Example two:
the present embodiment provides a lymph node tumor area identification apparatus for performing the method according to the first embodiment. As shown in fig. 2, a block diagram of a lymph node tumor region identification apparatus according to the present embodiment includes:
the acquisition module 10 is configured to acquire a lymph node mask image of a lymph node image to be detected and cut the lymph node mask image to generate a plurality of local sections;
the classification prediction module 20: the system comprises a deep learning neural network model, a local section and a local section, wherein the deep learning neural network model is used for training the local section;
obtain tumor confidence module 30: the tumor probability feature vector is input into the conditional random field model to obtain the tumor confidence of the local slice;
generating a tumor area heat map module 40: the system is used for mapping the local section to a lymph node image to be detected and obtaining a tumor region heat map of the lymph node region to be detected according to the tumor confidence;
the calculate area ratio module 50: the tumor area mask image is generated according to the tumor area heat map, the tumor area contour in the tumor area mask image is extracted through edge detection, and the area ratio of the tumor area in the lymph node is calculated.
In addition, the present invention also provides a lymph node tumor region identification apparatus comprising:
at least one processor, and a memory communicatively coupled to the at least one processor;
wherein the processor is configured to perform the method according to embodiment one by calling the computer program stored in the memory.
In addition, the present invention also provides a computer-readable storage medium, which stores computer-executable instructions for causing a computer to perform the method according to the first embodiment.
The method comprises the steps of obtaining a plurality of local slices of a lymph node image to be detected, inputting each local slice into a trained deep learning neural network model for classification prediction to obtain tumor probability feature vectors of the local slices, inputting the tumor probability feature vectors into a conditional random field model to obtain tumor confidence of the local slices, mapping the local slices into the lymph node image to be detected, obtaining a tumor region heat map of the lymph node region to be detected according to the tumor confidence, finally generating a tumor region mask image according to the tumor region heat map, extracting a tumor region outline in the tumor region mask image through edge detection, and calculating the area occupation ratio of the tumor region area in the lymph node. Can be widely applied to the field of medical pathological image processing.
The above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same, although the present invention is described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; such modifications and substitutions do not depart from the spirit and scope of the present invention, and they should be construed as being included in the following claims and description.
Claims (10)
1. A lymph node tumor region identification method, comprising:
acquiring a lymph node mask image of a lymph node image to be detected and cutting the lymph node mask image to generate a plurality of local sections;
inputting each local slice into a trained deep learning neural network model for classification prediction to obtain tumor probability feature vectors of the local slices;
inputting the tumor probability feature vector into a conditional random field model to obtain the tumor confidence of the local slice;
mapping the local section to the lymph node image to be detected, and obtaining a tumor region heat map of the lymph node region to be detected according to the tumor confidence;
generating a tumor region mask image according to the tumor region heat map, extracting a tumor region outline in the tumor region mask image through edge detection, and calculating the area ratio of the tumor region area in the lymph node.
2. The method of claim 1, wherein the classification prediction process comprises: dividing the input local slice into a plurality of small slices with the same size, acquiring a central pixel of the small slices, and obtaining a tumor probability feature vector according to the category of the central pixel.
3. The method of claim 2, wherein the conditional random field model satisfies Gibbs distribution, and determines similarity of features in the tumor probability feature vector by cosine similarity.
4. The lymph node tumor region identification method according to claim 3, wherein the energy function of the conditional random field model is expressed as:
wherein the content of the first and second substances,
representing a unitary potential for measuring the microcuttingSheet xiClassification as label yiThe cost of (a) of (b),
representing potential pairwise for measuring small slices xiAnd small slice xjCost, w, of classifying as one and the same tagijRepresents the training control parameter when yi＝yjWhen the condition [ mu ] (y) is satisfiedi＝yj)＝1。
5. The method for identifying a lymph node tumor region according to claim 1, wherein the process of extracting the tumor region contour in the tumor region mask image by edge detection and calculating the area ratio of the tumor region area in the lymph node is specifically as follows:
extracting pixel coordinates of a lymph node outline and pixel coordinates of a tumor region outline in the lymph node mask image and the tumor region mask image;
respectively fitting to obtain a tumor region contour and a lymph node region contour;
and counting the number of nonzero pixel points in the contour of the tumor region, and calculating to obtain the area ratio of the tumor region in the lymph node region.
6. The method of identifying a lymph node tumor region according to claim 5, further comprising: and when the lymph node outline area in the lymph node mask image is detected to be smaller than a preset threshold value, excluding the lymph node area corresponding to the lymph node outline.
7. The method according to claim 1, wherein the deep learning neural network model is trained by using ImageNet data set for model pre-training.
8. A lymph node tumor region identification apparatus, comprising:
the acquisition module is used for acquiring a lymph node mask image of a lymph node image to be detected and cutting the lymph node mask image to generate a plurality of local sections;
a classification prediction module: the local section is input into a trained deep learning neural network model for classification prediction, and a tumor probability feature vector of the local section is obtained;
obtaining a tumor confidence module: the tumor probability feature vector is input into a conditional random field model to obtain the tumor confidence of the local slice;
generating a tumor region heat map module: the local section is mapped to the lymph node image to be detected, and a tumor region heat map of the lymph node region to be detected is obtained according to the tumor confidence;
the area ratio calculation module: the tumor area mask image is generated according to the tumor area heat map, the tumor area contour in the tumor area mask image is extracted through edge detection, and the area ratio of the tumor area in the lymph node is calculated.
9. A lymph node tumor region identification apparatus, comprising:
at least one processor; and a memory communicatively coupled to the at least one processor;
wherein the processor is operable to perform the method of any one of claims 1 to 7 by invoking a computer program stored in the memory.
10. A computer-readable storage medium having stored thereon computer-executable instructions for causing a computer to perform the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910857766.4A CN110705565A (en) | 2019-09-09 | 2019-09-09 | Lymph node tumor region identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910857766.4A CN110705565A (en) | 2019-09-09 | 2019-09-09 | Lymph node tumor region identification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110705565A true CN110705565A (en) | 2020-01-17 |
Family
ID=69194726
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910857766.4A Pending CN110705565A (en) | 2019-09-09 | 2019-09-09 | Lymph node tumor region identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110705565A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111340128A (en) * | 2020-03-05 | 2020-06-26 | 上海市肺科医院（上海市职业病防治院） | Lung cancer metastatic lymph node pathological image recognition system and method |
| CN111784637A (en) * | 2020-06-04 | 2020-10-16 | 复旦大学附属中山医院 | Prognostic characteristic visualization method, system, equipment and storage medium |
| CN111862021A (en) * | 2020-07-13 | 2020-10-30 | 中山大学 | Deep learning-based automatic head and neck lymph node and drainage area delineation method |
| CN111967465A (en) * | 2020-07-07 | 2020-11-20 | 广州金域医学检验中心有限公司 | Method, system, computer device and storage medium for evaluating tumor cell content |
| CN112102332A (en) * | 2020-08-30 | 2020-12-18 | 复旦大学 | Cancer WSI segmentation method based on local classification neural network |
| CN112233123A (en) * | 2020-09-21 | 2021-01-15 | 广州金域医学检验中心有限公司 | Method and device for classifying tumor stroma in pathological picture, computer equipment and storage medium |
| CN112801939A (en) * | 2020-12-31 | 2021-05-14 | 杭州迪英加科技有限公司 | Method for improving index accuracy of pathological image KI67 |
| CN114140639A (en) * | 2021-11-04 | 2022-03-04 | 杭州医派智能科技有限公司 | Deep learning-based renal blood vessel extreme urine pole classification method in image, computer equipment and computer readable storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110060245A (en) * | 2019-04-15 | 2019-07-26 | 青岛大学 | A kind of gastric cancer lymph node dyeing pathological image automatic identifying method based on deep neural network |
-
2019
- 2019-09-09 CN CN201910857766.4A patent/CN110705565A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110060245A (en) * | 2019-04-15 | 2019-07-26 | 青岛大学 | A kind of gastric cancer lymph node dyeing pathological image automatic identifying method based on deep neural network |
Non-Patent Citations (3)
| Title |
|---|
| 《实体瘤病理数据集建设和数据标注质量控制专家共识》筹备组: ""实体瘤病理数据集建设和数据标注质量控制专家意见（2019）"", 《第二军医大学学报》 * |
| F. G. ZANJANI: ""Cancer detection in histopathology whole-slide images using conditional random fields on deep embedded spaces"", 《SPIE MEDICAL IMAGING》 * |
| YI LI: ""Cancer Metastasis Detection With Neural Conditional Random Field"", 《ARXIV:1806.07064V1》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111340128A (en) * | 2020-03-05 | 2020-06-26 | 上海市肺科医院（上海市职业病防治院） | Lung cancer metastatic lymph node pathological image recognition system and method |
| CN111784637A (en) * | 2020-06-04 | 2020-10-16 | 复旦大学附属中山医院 | Prognostic characteristic visualization method, system, equipment and storage medium |
| CN111967465A (en) * | 2020-07-07 | 2020-11-20 | 广州金域医学检验中心有限公司 | Method, system, computer device and storage medium for evaluating tumor cell content |
| CN111862021A (en) * | 2020-07-13 | 2020-10-30 | 中山大学 | Deep learning-based automatic head and neck lymph node and drainage area delineation method |
| CN111862021B (en) * | 2020-07-13 | 2022-06-24 | 广州柏视医疗科技有限公司 | Deep learning-based automatic head and neck lymph node and drainage area delineation method |
| CN112102332A (en) * | 2020-08-30 | 2020-12-18 | 复旦大学 | Cancer WSI segmentation method based on local classification neural network |
| CN112233123A (en) * | 2020-09-21 | 2021-01-15 | 广州金域医学检验中心有限公司 | Method and device for classifying tumor stroma in pathological picture, computer equipment and storage medium |
| CN112801939A (en) * | 2020-12-31 | 2021-05-14 | 杭州迪英加科技有限公司 | Method for improving index accuracy of pathological image KI67 |
| CN112801939B (en) * | 2020-12-31 | 2022-07-22 | 杭州迪英加科技有限公司 | Method for improving index accuracy of pathological image KI67 |
| CN114140639A (en) * | 2021-11-04 | 2022-03-04 | 杭州医派智能科技有限公司 | Deep learning-based renal blood vessel extreme urine pole classification method in image, computer equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110705565A (en) | Lymph node tumor region identification method and device | |
| CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
| CN108229509B (en) | Method and device for identifying object class and electronic equipment | |
| CN109886179B (en) | Image segmentation method and system of cervical cell smear based on Mask-RCNN | |
| CN107506761B (en) | Brain image segmentation method and system based on significance learning convolutional neural network | |
| CN110348319B (en) | Face anti-counterfeiting method based on face depth information and edge image fusion | |
| CN106570505B (en) | Method and system for analyzing histopathological images | |
| CN109919230B (en) | Medical image pulmonary nodule detection method based on cyclic feature pyramid | |
| CN108564085B (en) | Method for automatically reading of pointer type instrument | |
| CN111462042A (en) | Cancer prognosis analysis method and system | |
| JP6945253B2 (en) | Classification device, classification method, program, and information recording medium | |
| CN107909102A (en) | A kind of sorting technique of histopathology image | |
| CN105389821B (en) | It is a kind of that the medical image cutting method being combined is cut based on cloud model and figure | |
| CN107273824A (en) | Face identification method based on multiple dimensioned multi-direction local binary patterns | |
| CN111325750A (en) | Medical image segmentation method based on multi-scale fusion U-shaped chain neural network | |
| CN113269257A (en) | Image classification method and device, terminal equipment and storage medium | |
| Chen et al. | Identifying metastases in sentinel lymph nodes with deep convolutional neural networks | |
| CN112396619A (en) | Small particle segmentation method based on semantic segmentation and internally complex composition | |
| Limper et al. | Mesh Saliency Analysis via Local Curvature Entropy. | |
| CN106919884A (en) | Human facial expression recognition method and device | |
| CN113160185A (en) | Method for guiding cervical cell segmentation by using generated boundary position | |
| CN114445356A (en) | Multi-resolution-based full-field pathological section image tumor rapid positioning method | |
| Paeng et al. | A unified framework for tumor proliferation score prediction in breast histopathology | |
| Valkonen et al. | Dual structured convolutional neural network with feature augmentation for quantitative characterization of tissue histology | |
| CN111666813B (en) | Subcutaneous sweat gland extraction method of three-dimensional convolutional neural network based on non-local information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |