CN111986150A - Interactive marking refinement method for digital pathological image - Google Patents
Interactive marking refinement method for digital pathological image Download PDFInfo
- Publication number
- CN111986150A CN111986150A CN202010690711.1A CN202010690711A CN111986150A CN 111986150 A CN111986150 A CN 111986150A CN 202010690711 A CN202010690711 A CN 202010690711A CN 111986150 A CN111986150 A CN 111986150A
- Authority
- CN
- China
- Prior art keywords
- patch
- digital pathological
- pathological image
- data
- digital
- 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.)
- Granted
Links
- 230000001575 pathological effect Effects 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 32
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 12
- 230000003211 malignant effect Effects 0.000 claims abstract description 22
- 238000013145 classification model Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 15
- 238000004043 dyeing Methods 0.000 claims abstract description 13
- 238000010586 diagram Methods 0.000 claims abstract description 8
- 238000007670 refining Methods 0.000 claims abstract description 6
- 238000002372 labelling Methods 0.000 claims description 20
- 238000010186 staining Methods 0.000 claims description 19
- 230000008520 organization Effects 0.000 claims description 8
- 230000007170 pathology Effects 0.000 claims description 7
- 238000003708 edge detection Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 230000003902 lesion Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- IXKSXJFAGXLQOQ-XISFHERQSA-N WHWLQLKPGQPMY Chemical compound C([C@@H](C(=O)N[C@@H](CC=1C2=CC=CC=C2NC=1)C(=O)N[C@@H](CC(C)C)C(=O)N[C@@H](CCC(N)=O)C(=O)N[C@@H](CC(C)C)C(=O)N1CCC[C@H]1C(=O)NCC(=O)N[C@@H](CCC(N)=O)C(=O)N[C@@H](CC(O)=O)C(=O)N1CCC[C@H]1C(=O)N[C@@H](CCSC)C(=O)N[C@@H](CC=1C=CC(O)=CC=1)C(O)=O)NC(=O)[C@@H](N)CC=1C2=CC=CC=C2NC=1)C1=CNC=N1 IXKSXJFAGXLQOQ-XISFHERQSA-N 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 206010028980 Neoplasm Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 201000011510 cancer Diseases 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
-
- 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
- G06T2207/20092—Interactive image processing based on input by user
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Radiology & Medical Imaging (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides an interactive annotation refinement method of a digital pathological image, which is characterized by comprising the following steps of: constructing and training a Resnet weak supervision classification model; acquiring a digital pathological image input in real time, preprocessing the digital pathological image, and acquiring the size corresponding to the tissue area of the digital pathological image and the batch slice data after dyeing standardization; inputting Patch slice data into a trained Resnet weak supervision classification model to obtain the benign and malignant classification of each Patch slice, and generating XML vector diagram annotation on an original digital pathological image according to the benign and malignant classification of the Patch slices to obtain a pre-annotated focus area; and obtaining a refined marking outline on the pre-marking focus area. The invention provides an interactive annotation refining method for a digital pathological image, which improves the annotation efficiency of doctors through automatic pre-annotation and outline refining treatment.
Description
Technical Field
The invention relates to the field of digital pathological image processing, in particular to an interactive annotation refinement method for a digital pathological image.
Background
The digital pathological image is a tissue slice of a pathological part of a patient, the size of the digital pathological image obtained by using a WSI (wireless sensor interface) technology is very large, and the digital pathological image can directly reflect the pathological condition in a tissue, so that a clinician needs to search a tissue area and draw the position of the specific pathological tissue area under different multiplying powers to serve as an important basis for disease diagnosis. For example, in cancer diagnosis, it is necessary to extract living tissues of a lesion to prepare pathological sections, and observe digitized pathological images to determine pathological characteristics.
The number of professional pathological doctors has serious gaps, daily requirements cannot be met, and the labeling of digital pathological images also greatly increases the burden of doctors. At present, some related technologies are used for solving the problem, but the operation process is responsible for a great deal of preset value work of doctors, and the whole digital pathological image needs to be observed, so that the workload is large, and the pathological tissue area is not directly focused on. For example, chinese patent CN105608319B, the invention provides a labeling method for digital pathological section, which requires a user to select a labeling point and label graphic type information, and the observation object is the whole digital pathological image; chinese patent CN105404896B, the invention mainly uses similarity as the result of automatic detection and labeling, and therefore is not suitable for labeling digital pathological images.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the labeling of digital pathology images greatly increases the burden on the physician.
In order to solve the above technical problem, a technical solution of the present invention is to provide an interactive annotation refinement method for a digital pathological image, which is characterized by comprising the following steps:
step 2, labeling classification labels on the preprocessed Patch slice data according to the type of the sample digital pathological image, combining a plurality of Patch slice data with the same classification into Patch packets, wherein the type label of each Patch packet is the label of the single type of Patch slice data contained in the Patch packet;
step 3, inputting all Patch slice data in the Patch package into a Resnet weak supervision classification model, firstly performing feature extraction through a Resnet network, then classifying the Patch slices through a full-connection network, obtaining the probability that each Patch slice is malignant after passing through a Sigmoid activation function, selecting the Patch slice with the highest probability in a Patch slice set through the Resnet weak supervision classification model, if the probability that the Patch slice is malignant is greater than 0.5, determining the category of the whole Patch package to be malignant, otherwise, determining the category of the whole Patch package to be benign, and calculating loss with a label;
step 5, inputting the Patch slice data obtained in the step 4 into the Resnet weak supervision classification model trained in the step 3, obtaining the benign and malignant classification of each Patch slice, and generating XML vector diagram labels on the original digital pathological image according to the benign and malignant classification of the Patch slices to obtain a pre-labeled lesion area;
and 6, on the pre-marked focus area, the doctor carries out manual marking again, then sets areas of N pixels on two sides of the edge line of the manually marked focus as edge convergence candidate areas, calculates gradient values among all pixels in the edge convergence candidate areas, and finally selects the position with the maximum gradient as the corrected focus edge to obtain a refined marked outline.
Preferably, in step 1, the preprocessing of the sample digital pathology image comprises the following steps:
step 101, segmenting an organization region on an input sample digital pathological image by using an Otsu method on a thumbnail, and recording the coordinate position of the organization region part corresponding to an original sample digital pathological image;
step 102, cutting the sample digital pathological image according to the coordinate position recorded in the step 101 according to a certain size to obtain Patch slice data;
103, selecting a plurality of digital pathological images dyed in the same hospital as a data source of a standard dyeing space, converting all the digital pathological images into an LAB color space, counting the mean variance, clustering the mean variances of L, A, B channels serving as a feature vector by using K-means, and selecting the clustering center of the maximum class as the standard dyeing space;
104, standardizing the staining of the Patch section data to a standard staining space by using a Reinhard algorithm;
and 105, performing over-sampling data balance on the Patch slice data and performing data enhancement in a mode of random rotation, inversion and noise increase.
Preferably, in step 3, an alpha factor and a gamma factor are added to the loss function of the Resnet weakly supervised classification model, wherein the alpha factor is used for balancing positive samples with class labels being benign and the class labels being benignAnd if the malignant negative sample data size and the gamma factor are used for adjusting the size of the simple sample loss weight, the loss function L of the Resnet weak supervision classification model is as follows:
in the formula, y represents the true label of the sample, and y' represents the prediction probability of the sample.
Preferably, in step 4, the preprocessing of the digital pathology image comprises the following steps:
step 401, segmenting an organization region from the input digital pathological image on a thumbnail by using an Otsu method, and recording the coordinate position of the organization region part corresponding to the original sample digital pathological image;
step 402, cutting the digital pathological image according to the coordinate position recorded in step 401 according to a certain size to obtain Patch slice data;
step 403, selecting a plurality of digital pathological images dyed in the same hospital as a data source of a standard dyeing space, converting all the digital pathological images into an LAB color space, counting the mean variance, clustering the mean variances of L, A, B channels as a feature vector by using K-means, and selecting the clustering center of the maximum class as the standard dyeing space;
step 404, staining of the Patch section data is normalized to a standard staining space using Reinhard algorithm.
Preferably, in step 6, the edge convergence candidate region is subjected to edge detection using laplacian: second order partial derivatives of images in the x-direction and y-direction, respectively
After merging, there are:
in the formula, f (x, y) represents a pixel value at the position where the image coordinate is (, y), a gradient map of an edge convergence candidate region is obtained by performing convolution operation on a Laplacian operator and the edge convergence candidate region, so that a gradient value between each pixel in the edge convergence candidate region is calculated, and finally, a position with the maximum gradient is selected as a corrected focus edge to obtain a refined labeling outline.
The invention provides an interactive annotation refining method for a digital pathological image, which improves the annotation efficiency of doctors through automatic pre-annotation and outline refining treatment.
Drawings
FIG. 1 is a schematic diagram of an interactive annotation refinement method for digital pathological images according to the present invention;
FIG. 2 is a flow diagram of a data preprocessing module of the present invention;
FIG. 3 is a MIL network model training flow diagram of the present invention;
FIG. 4 is a flowchart of a label refinement processing module according to the present invention;
FIG. 5 is a schematic diagram of an edge detection operator used in the present invention.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
As shown in fig. 1, the interactive annotation refinement method for digital pathological images provided by the present invention includes the following steps:
step S1, data preprocessing step
Preprocessing the input pathological image to realize the standardization and normalization of the pathological image;
step S2, automatic pre-labeling step of weak supervision model
Training a neural network by using an MIL training method to obtain a pre-labeled area;
manually marking after the weak supervision model automatic pre-marking step is carried out, and carrying out focus outline rough drawing after a doctor manually confirms a pre-marked region to obtain a focus outline rough drawing region;
step 3, marking result refinement processing
Inputting the rough outline drawing area of the focus into a fine processing module to obtain a fine marking result.
The technical scheme comprises three parts: the first part is a data preprocessing process, the second part is an automatic pre-labeling process of the weak supervision model, and the third part is a refining process of coarse labeling of a doctor based on a pre-labeling result. The above three parts are detailed as follows:
as shown in fig. 2, the data preprocessing process in step S1 includes classification labeling of the whole slice, automatic extraction of the tissue portion of the pathological section, and staining normalization based on Reinhard and cluster statistics. The method specifically comprises the following steps:
step S101: and (3) segmenting the tissue area on the input pathological image by using the Otsu method on the thumbnail, and recording the coordinate position of the tissue area part corresponding to the original pathological image.
Step S102: patch slices were obtained from the original pathology image by cropping at the tissue region locations recorded at the previous step in the size of 224X 224.
Step S103: in order to ensure that the standard staining space counts staining of enough tissues and enough average staining, 3000 pieces of digital pathological images stained in the same hospital are selected as data sources of the standard staining space, all 3000 pieces of digital pathological images are converted into an LAB color space, and the mean variance is counted. Since 3000 digital pathological images do not always meet the staining standard, there may be staining nonstandard or tissues constituting special digital pathological images, the invention does not directly average the mean variance of the LAB space to obtain the standard staining space, but uses the mean variances of L, A, B three channels as a feature vector and uses K-means clustering, and selects the clustering center of the maximum class as the standard staining space.
Step S104: the staining of the input image was normalized to a standard staining space using the Reinhard algorithm.
When the Resnet weak supervision model is trained, the method further comprises the following steps
Step S105: and carrying out oversampled data balance on the Patch slice data and carrying out data enhancement in a mode of random rotation, inversion and noise increase. Step S105 is only used when a Resnet weak supervision model is trained, and the interactive annotation process does not do the step.
As shown in fig. 3, the automatic pre-labeling process of the weakly supervised model includes pre-labeling using a Resnet weakly supervised classification model trained in a Multiple Instance Learning (MIL) manner, and the expert labels the data set of the entire digital pathological image category. When a Resnet weak supervision classification model is trained by using a multi-instance learning mode, a Patch slice data set of a trained digital pathological image is only labeled by a good or malignant class of the whole image, and no lesion position is labeled.
The method for training the Resnet weak supervision classification model by using the multi-instance learning mode comprises the following steps:
the first step is as follows: and labeling the Patch section obtained after preprocessing according to the category of the whole digital pathological image. Several identically classified Patch slices are combined into Patch packages, each Patch package having a class label that is a label of a single kind of Patch slice contained therein.
The second step is that: inputting all Patch slices in the Patch package into a Resnet weak supervision classification model, firstly performing feature extraction by a Resnet network, then classifying the Patch slices by a full-connection network, obtaining the probability that each Patch slice is malignant after a Sigmoid activation function, selecting the Patch slice with the highest probability in a Patch slice set by the Resnet weak supervision classification model, if the probability that the Patch slice is malignant is more than 0.5, determining the category of the whole Patch package to be malignant, otherwise, determining the category of the whole Patch package to be benign, and calculating loss with a label.
Patch bag with uneven positive and negative data volume due to digital pathological imageAnd 4, calculating the weighting loss according to the data ratio and performing back propagation. The specific contents are as follows: the loss function L of the Resnet weakly supervised classification model is:
in the formula, y represents the true label of the sample, and y' represents the prediction probability of the sample. In order to balance the positive and negative sample data volumes and solve the problems of simple and difficult samples, an alpha factor and a gamma factor are added into a loss function of a Resnet weak supervision classification model, wherein the alpha factor can balance the positive and negative sample data volumes, the gamma factor can adjust the size of the loss weight of a simple sample, and the improved loss function L is as follows:
pre-labeling the preprocessed patch slices by using the Resnet weak supervision classification model obtained by training in the steps to obtain the benign and malignant classification of each patch slice, and generating XML vector diagram labels on the original digital pathological images according to the benign and malignant classification of the patch slices to be used as references of doctor labels.
As shown in fig. 4, the label refinement process includes the following steps: and on a pre-marked focus area of the Resnet weak supervision classification model, the doctor carries out manual marking, then sets areas of N pixels on two sides of an edge line of the manually marked focus as edge convergence candidate areas, calculates gradient values among all pixels in the candidate areas, and finally selects a position with the maximum gradient as a corrected focus edge to obtain a refined marked outline. The method specifically comprises the following steps:
step S301: candidate region selection
And (3) taking N pixel values at two sides of a focus edge line manually marked by a doctor to form a candidate region.
Step S302: edge detection is performed by using a Laplacian operator, and second-order partial derivatives are obtained in the x direction and the y direction of the image
After merging:
in the formula, f (x, y) represents a pixel value at which the image coordinates are (x, y). Therefore, as shown in fig. 5, the laplacian for edge detection is used to perform convolution operation on the laplacian and the candidate region to obtain a gradient map of the candidate region, and the position with the largest gradient is selected as the optimized boundary.
Claims (5)
1. An interactive labeling and refining method for a digital pathological image is characterized by comprising the following steps:
step 1, obtaining digital pathological images of samples, preprocessing the digital pathological images of the samples, and obtaining Patch slice data with the size corresponding to a tissue area of each digital pathological image of the samples and after dyeing standardization, wherein the Patch slice data only label the benign and malignant categories of the digital pathological images of the samples and do not label the positions of focuses;
step 2, labeling classification labels on the preprocessed Patch slice data according to the type of the sample digital pathological image, combining a plurality of Patch slice data with the same classification into Patch packets, wherein the type label of each Patch packet is the label of the single type of Patch slice data contained in the Patch packet;
step 3, inputting all Patch slice data in the Patch package into a Resnet weak supervision classification model, firstly performing feature extraction through a Resnet network, then classifying the Patch slices through a full-connection network, obtaining the probability that each Patch slice is malignant after passing through a Sigmoid activation function, selecting the Patch slice with the highest probability in a Patch slice set through the Resnet weak supervision classification model, if the probability that the Patch slice is malignant is greater than 0.5, determining the category of the whole Patch package to be malignant, otherwise, determining the category of the whole Patch package to be benign, and calculating loss with a label;
step 4, acquiring a digital pathological image input in real time, preprocessing the digital pathological image, and acquiring the size corresponding to the tissue area of the digital pathological image and the batch slice data after dyeing standardization;
step 5, inputting the Patch slice data obtained in the step 4 into the Resnet weak supervision classification model trained in the step 3, obtaining the benign and malignant classification of each Patch slice, and generating XML vector diagram labels on the original digital pathological image according to the benign and malignant classification of the Patch slices to obtain a pre-labeled lesion area;
and 6, on the pre-marked focus area, the doctor carries out manual marking again, then sets areas of N pixels on two sides of the edge line of the manually marked focus as edge convergence candidate areas, calculates gradient values among all pixels in the edge convergence candidate areas, and finally selects the position with the maximum gradient as the corrected focus edge to obtain a refined marked outline.
2. The method for interactive annotation refinement of digital pathology image of claim 1, wherein the step 1 of preprocessing the sample digital pathology image comprises the steps of:
step 101, segmenting an organization region on an input sample digital pathological image by using an Otsu method on a thumbnail, and recording the coordinate position of the organization region part corresponding to an original sample digital pathological image;
step 102, cutting the sample digital pathological image according to the coordinate position recorded in the step 101 according to a certain size to obtain Patch slice data;
103, selecting a plurality of digital pathological images dyed in the same hospital as a data source of a standard dyeing space, converting all the digital pathological images into an LAB color space, counting the mean variance, clustering the mean variances of L, A, B channels serving as a feature vector by using K-means, and selecting the clustering center of the maximum class as the standard dyeing space;
104, standardizing the staining of the Patch section data to a standard staining space by using a Reinhard algorithm;
and 105, performing over-sampling data balance on the Patch slice data and performing data enhancement in a mode of random rotation, inversion and noise increase.
3. The method according to claim 1, wherein in step 3, α and γ factors are added to the loss function of the Resnet weakly supervised classification model, where α is used to balance the amount of positive samples with class labels being benign and the amount of negative samples with class labels being malignant, and γ is used to adjust the magnitude of the loss weight of the simple samples, so that the loss function L of the Resnet weakly supervised classification model is:
in the formula, y represents the true label of the sample, and y' represents the prediction probability of the sample.
4. The method as claimed in claim 1, wherein the step 4 of preprocessing the digital pathology image comprises the steps of:
step 401, segmenting an organization region from the input digital pathological image on a thumbnail by using an Otsu method, and recording the coordinate position of the organization region part corresponding to the original sample digital pathological image;
step 402, cutting the digital pathological image according to the coordinate position recorded in step 401 according to a certain size to obtain Patch slice data;
step 403, selecting a plurality of digital pathological images dyed in the same hospital as a data source of a standard dyeing space, converting all the digital pathological images into an LAB color space, counting the mean variance, clustering the mean variances of L, A, B channels as a feature vector by using K-means, and selecting the clustering center of the maximum class as the standard dyeing space;
step 404, staining of the Patch section data is normalized to a standard staining space using Reinhard algorithm.
5. The method of claim 1, wherein in step 6, the laplacian is used to perform edge detection on the edge convergence candidate region: second order partial derivatives of images in the x-direction and y-direction, respectively
After merging, there are:
in the formula, f (x, y) represents a pixel value at the position where the image coordinate is (x, y), a gradient map of an edge convergence candidate region is obtained by performing convolution operation on a laplacian operator and the edge convergence candidate region, so that a gradient value between each pixel in the edge convergence candidate region is calculated, and finally, a position with the maximum gradient is selected as a corrected focus edge, so that a refined labeling outline is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010690711.1A CN111986150B (en) | 2020-07-17 | 2020-07-17 | The method comprises the following steps of: digital number pathological image Interactive annotation refining method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010690711.1A CN111986150B (en) | 2020-07-17 | 2020-07-17 | The method comprises the following steps of: digital number pathological image Interactive annotation refining method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111986150A true CN111986150A (en) | 2020-11-24 |
| CN111986150B CN111986150B (en) | 2024-02-09 |
Family
ID=73437888
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010690711.1A Active CN111986150B (en) | 2020-07-17 | 2020-07-17 | The method comprises the following steps of: digital number pathological image Interactive annotation refining method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111986150B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112331314A (en) * | 2020-11-25 | 2021-02-05 | 中山大学附属第六医院 | Image annotation method and device, storage medium and electronic equipment |
| CN112446881A (en) * | 2021-02-01 | 2021-03-05 | 北京小白世纪网络科技有限公司 | Pathological image segmentation system and method |
| CN112488234A (en) * | 2020-12-10 | 2021-03-12 | 武汉大学 | End-to-end histopathology image classification method based on attention pooling |
| CN112884724A (en) * | 2021-02-02 | 2021-06-01 | 广州智睿医疗科技有限公司 | Intelligent judgment method and system for lung cancer histopathological typing |
| CN113299372A (en) * | 2021-05-14 | 2021-08-24 | 深圳大学 | Photoacoustic pathological image processing method, storage medium and terminal device |
| CN113469972A (en) * | 2021-06-30 | 2021-10-01 | 沈阳东软智能医疗科技研究院有限公司 | Method, device, storage medium and electronic equipment for labeling medical slice image |
| CN113628199A (en) * | 2021-08-18 | 2021-11-09 | 四川大学华西第二医院 | Pathological picture stained tissue area detection method, pathological picture stained tissue area detection system and prognosis state analysis system |
| CN113674288A (en) * | 2021-07-05 | 2021-11-19 | 华南理工大学 | Automatic segmentation method for non-small cell lung cancer digital pathological image tissues |
| CN114187281A (en) * | 2021-12-14 | 2022-03-15 | 数坤(北京)网络科技股份有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN114596298A (en) * | 2022-03-16 | 2022-06-07 | 华东师范大学 | Hyperspectral imaging-based automatic generation method of fine-labeled digital pathological data set |
| CN117576127A (en) * | 2024-01-17 | 2024-02-20 | 神州医疗科技股份有限公司 | Liver cancer area automatic sketching method based on pathological image |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102314479A (en) * | 2011-07-05 | 2012-01-11 | 万达信息股份有限公司 | Method for preventing repeated marking of slice map |
| US9176987B1 (en) * | 2014-08-26 | 2015-11-03 | TCL Research America Inc. | Automatic face annotation method and system |
| CN105608319A (en) * | 2015-12-21 | 2016-05-25 | 江苏康克移软软件有限公司 | Digital pathological section labeling method and device |
| US20170330059A1 (en) * | 2016-05-11 | 2017-11-16 | Xerox Corporation | Joint object and object part detection using web supervision |
| CN108062574A (en) * | 2017-12-31 | 2018-05-22 | 厦门大学 | A kind of Weakly supervised object detection method based on particular category space constraint |
| CN109378052A (en) * | 2018-08-31 | 2019-02-22 | 透彻影像(北京)科技有限公司 | The preprocess method and system of image labeling |
| CN109544561A (en) * | 2018-11-07 | 2019-03-29 | 杭州迪英加科技有限公司 | Cell mask method, system and device |
| CN109670489A (en) * | 2019-02-18 | 2019-04-23 | 广州视源电子科技股份有限公司 | Weakly supervised formula early-stage senile maculopathy classification method based on more case-based learnings |
| CN109766830A (en) * | 2019-01-09 | 2019-05-17 | 深圳市芯鹏智能信息有限公司 | A kind of ship seakeeping system and method based on artificial intelligence image procossing |
| CN109872803A (en) * | 2019-01-28 | 2019-06-11 | 透彻影像(北京)科技有限公司 | A kind of artificial intelligence pathology labeling system |
| CN110009679A (en) * | 2019-02-28 | 2019-07-12 | 江南大学 | A kind of object localization method based on Analysis On Multi-scale Features convolutional neural networks |
| CN110378885A (en) * | 2019-07-19 | 2019-10-25 | 王晓骁 | A kind of focal area WSI automatic marking method and system based on machine learning |
| WO2020064323A1 (en) * | 2018-09-26 | 2020-04-02 | Safran | Method and system for the non-destructive testing of an aerospace part by contour readjustment |
-
2020
- 2020-07-17 CN CN202010690711.1A patent/CN111986150B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102314479A (en) * | 2011-07-05 | 2012-01-11 | 万达信息股份有限公司 | Method for preventing repeated marking of slice map |
| US9176987B1 (en) * | 2014-08-26 | 2015-11-03 | TCL Research America Inc. | Automatic face annotation method and system |
| CN105608319A (en) * | 2015-12-21 | 2016-05-25 | 江苏康克移软软件有限公司 | Digital pathological section labeling method and device |
| US20170330059A1 (en) * | 2016-05-11 | 2017-11-16 | Xerox Corporation | Joint object and object part detection using web supervision |
| CN108062574A (en) * | 2017-12-31 | 2018-05-22 | 厦门大学 | A kind of Weakly supervised object detection method based on particular category space constraint |
| CN109378052A (en) * | 2018-08-31 | 2019-02-22 | 透彻影像(北京)科技有限公司 | The preprocess method and system of image labeling |
| WO2020064323A1 (en) * | 2018-09-26 | 2020-04-02 | Safran | Method and system for the non-destructive testing of an aerospace part by contour readjustment |
| CN109544561A (en) * | 2018-11-07 | 2019-03-29 | 杭州迪英加科技有限公司 | Cell mask method, system and device |
| CN109766830A (en) * | 2019-01-09 | 2019-05-17 | 深圳市芯鹏智能信息有限公司 | A kind of ship seakeeping system and method based on artificial intelligence image procossing |
| CN109872803A (en) * | 2019-01-28 | 2019-06-11 | 透彻影像(北京)科技有限公司 | A kind of artificial intelligence pathology labeling system |
| CN109670489A (en) * | 2019-02-18 | 2019-04-23 | 广州视源电子科技股份有限公司 | Weakly supervised formula early-stage senile maculopathy classification method based on more case-based learnings |
| CN110009679A (en) * | 2019-02-28 | 2019-07-12 | 江南大学 | A kind of object localization method based on Analysis On Multi-scale Features convolutional neural networks |
| CN110378885A (en) * | 2019-07-19 | 2019-10-25 | 王晓骁 | A kind of focal area WSI automatic marking method and system based on machine learning |
Non-Patent Citations (4)
| Title |
|---|
| RAMAZAN GOKBERK CINBIS等: "Weakly Supervised Object Localization with Multi-Fold Multiple Instance Learning", IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, vol. 39, no. 1, pages 189 - 203, XP011635517, DOI: 10.1109/TPAMI.2016.2535231 * |
| 于观贞等: "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, no. 5, pages 6 - 11 * |
| 崔磊: "病理图像精细化分析算法研究", 中国博士学位论文全文数据库 医药卫生科技辑, no. 1, pages 059 - 38 * |
| 徐小程: "基于弱监督的图像区域自动标注算法研究", 中国优秀硕士学位论文全文数据库 信息科技辑, no. 1, pages 138 - 567 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112331314A (en) * | 2020-11-25 | 2021-02-05 | 中山大学附属第六医院 | Image annotation method and device, storage medium and electronic equipment |
| CN112488234B (en) * | 2020-12-10 | 2022-04-29 | 武汉大学 | End-to-end histopathology image classification method based on attention pooling |
| CN112488234A (en) * | 2020-12-10 | 2021-03-12 | 武汉大学 | End-to-end histopathology image classification method based on attention pooling |
| CN112446881A (en) * | 2021-02-01 | 2021-03-05 | 北京小白世纪网络科技有限公司 | Pathological image segmentation system and method |
| CN112884724A (en) * | 2021-02-02 | 2021-06-01 | 广州智睿医疗科技有限公司 | Intelligent judgment method and system for lung cancer histopathological typing |
| CN113299372B (en) * | 2021-05-14 | 2023-02-03 | 深圳大学 | Photoacoustic pathological image processing method, storage medium and terminal device |
| CN113299372A (en) * | 2021-05-14 | 2021-08-24 | 深圳大学 | Photoacoustic pathological image processing method, storage medium and terminal device |
| CN113469972A (en) * | 2021-06-30 | 2021-10-01 | 沈阳东软智能医疗科技研究院有限公司 | Method, device, storage medium and electronic equipment for labeling medical slice image |
| CN113469972B (en) * | 2021-06-30 | 2024-04-23 | 沈阳东软智能医疗科技研究院有限公司 | Method and device for labeling medical slice image, storage medium and electronic equipment |
| CN113674288A (en) * | 2021-07-05 | 2021-11-19 | 华南理工大学 | Automatic segmentation method for non-small cell lung cancer digital pathological image tissues |
| CN113674288B (en) * | 2021-07-05 | 2024-02-02 | 华南理工大学 | Automatic segmentation method for digital pathological image tissue of non-small cell lung cancer |
| CN113628199A (en) * | 2021-08-18 | 2021-11-09 | 四川大学华西第二医院 | Pathological picture stained tissue area detection method, pathological picture stained tissue area detection system and prognosis state analysis system |
| CN114187281A (en) * | 2021-12-14 | 2022-03-15 | 数坤(北京)网络科技股份有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN114596298A (en) * | 2022-03-16 | 2022-06-07 | 华东师范大学 | Hyperspectral imaging-based automatic generation method of fine-labeled digital pathological data set |
| CN114596298B (en) * | 2022-03-16 | 2022-11-15 | 华东师范大学 | Hyperspectral imaging-based automatic generation method of fine-labeled digital pathological data set |
| US11763453B1 (en) | 2022-03-16 | 2023-09-19 | East China Normal University | Automatic generation method of fine-labeled digital pathological data set based on hyperspectral imaging |
| CN117576127A (en) * | 2024-01-17 | 2024-02-20 | 神州医疗科技股份有限公司 | Liver cancer area automatic sketching method based on pathological image |
| CN117576127B (en) * | 2024-01-17 | 2024-04-19 | 神州医疗科技股份有限公司 | Liver cancer area automatic sketching method based on pathological image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111986150B (en) | 2024-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111986150B (en) | The method comprises the following steps of: digital number pathological image Interactive annotation refining method | |
| Dundar et al. | Computerized classification of intraductal breast lesions using histopathological images | |
| Zhao et al. | Seens: Nuclei segmentation in pap smear images with selective edge enhancement | |
| Song et al. | A deep learning based framework for accurate segmentation of cervical cytoplasm and nuclei | |
| Fauzi et al. | Computerized segmentation and measurement of chronic wound images | |
| George et al. | Remote computer-aided breast cancer detection and diagnosis system based on cytological images | |
| Veta et al. | Detecting mitotic figures in breast cancer histopathology images | |
| Zhang et al. | Automated semantic segmentation of red blood cells for sickle cell disease | |
| Dov et al. | Thyroid cancer malignancy prediction from whole slide cytopathology images | |
| Apou et al. | Detection of lobular structures in normal breast tissue | |
| CN115063592B (en) | Multi-scale-based full-scanning pathological feature fusion extraction method and system | |
| Sornapudi et al. | Automated cervical digitized histology whole-slide image analysis toolbox | |
| Mäder et al. | Image-processing scheme to detect superficial fungal infections of the skin | |
| Saito et al. | Dawn of the digital diagnosis assisting system, can it open a new age for pathology? | |
| Shirazi et al. | Automated pathology image analysis | |
| CN111554381A (en) | Artificial intelligent pathological diagnosis method and diagnosis model for renal clear cell carcinoma based on deep learning | |
| Hameed et al. | Immunohistochemical analysis of oral cancer tissue images using support vector machine | |
| US20220366619A1 (en) | Systems and methods to process electronic images to adjust attributes of the electronic images | |
| CN111062909A (en) | Method and equipment for judging benign and malignant breast tumor | |
| CN114898862A (en) | Cervical cancer computer-aided diagnosis method based on convolutional neural network and pathological section image | |
| CN113870194A (en) | Deep layer characteristic and superficial layer LBP characteristic fused breast tumor ultrasonic image processing device | |
| Mouzai et al. | Xray-Net: Self-supervised pixel stretching approach to improve low-contrast medical imaging | |
| Chang et al. | Automatic analysis of HER-2/neu immunohistochemistry in breast cancer | |
| Naik et al. | Hybrid Feature Set based Mitotic Detection in Breast Histopathology Images | |
| Riana et al. | Comparison of Segmentation Analysis in Nucleus Detection with GLCM Features using Otsu and Polynomial Methods |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |