CN108305253A - A kind of pathology full slice diagnostic method based on more multiplying power deep learnings - Google Patents
A kind of pathology full slice diagnostic method based on more multiplying power deep learnings Download PDFInfo
- Publication number
- CN108305253A CN108305253A CN201810189171.1A CN201810189171A CN108305253A CN 108305253 A CN108305253 A CN 108305253A CN 201810189171 A CN201810189171 A CN 201810189171A CN 108305253 A CN108305253 A CN 108305253A
- Authority
- CN
- China
- Prior art keywords
- pathological
- cnn
- image
- pathological image
- multiplying power
- 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
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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- 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
Pathological image is indicated in RGB color in calculating pathological section scanning to electronics, obtains RGB pathological images by a kind of pathology full slice diagnostic method based on more multiplying power deep learnings；Pathological image is marked, more multiplying power pathological image training datasets are established；More multiplying power CNN pathological image disaggregated models are established, deep learning training is carried out to more times of CNN pathological image disaggregated models using RGB pathological images；CNN pathological images disaggregated model after being trained using deep learning carries out cancerous area detection to pathology full slice.The method of the present invention reduces the calculation amount of the existing cancerous region detection algorithm based on more multiplying power pathological images, while ensure that auxiliary diagnosis precision.
Description
Technical field
The present invention relates to image procossings and deep learning field, refer in particular to a kind of disease based on more multiplying power deep learnings
Manage full slice diagnostic method.
Background technology
Digital pathology full slice is to be converted to pathological section by dedicated scanning imaging system to show for computer
The high magnification large scale digital image for showing, transmitting and handling.As digital pathological image imaging technique is constantly ripe, based on number
The computer aided diagnosing method of word pathological image is quickly grown.In recent years, with convolutional Neural metanetwork (convolutional
Neural network, abbreviation CNN) be representative deep learning method become pathological image classification main stream approach.Wherein
One important research is to carry out canceration/non-cancer to all areas of pathology full slice using bis- disaggregated models of CNN to become classification, into
And the region of doubtful canceration in pathology full slice is found out, reach the auxiliary diagnosis effect of cancerous region detection.
According to existing research, the CNN models that high magnification digital pathological image (such as 20 times) is established and low range (such as 2 times)
The CNN models of lower foundation have complementary relationship in cancerous region detection performance.In order to obtain higher accuracy of detection, have
Scholar proposes to establish CNN model inspection cancerous regions under different multiplying, then carries out the testing result obtained under each multiplying power
Fusion, obtains final detection result.But carrying out repeating to detect under more high magnifications needs more computing resources；Especially
It is to be judged all areas in pathology full slice that very high calculation amount will be generated under high magnification；Problem above will cause
Clinical demand is not achieved in the auxiliary diagnosis time.
Invention content
The technical problem to be solved in the present invention is to provide a kind of pathology full slice diagnosis sides based on more multiplying power deep learnings
Method can play the advantage for the CNN pathological image cancerous region detection models established based on different multiplying, reduce existing base
In the calculation amount of the cancerous region detection algorithm of more multiplying power pathological images, while meeting in clinical application and cancerous region is detected
Requirement in auxiliary diagnosis software accuracy and speed.
In order to solve the above-mentioned technical problem, the technical scheme is that：A kind of pathology based on more multiplying power deep learnings
Full slice diagnostic method, which is characterized in that include the following steps：
Step 1：During pathological section scanning is calculated to electronics, pathological image is indicated in RGB color,
Obtain RGB pathological images；
Step 2：Pathological image is marked, more multiplying power pathological image training datasets are established, with set X20×It indicates；
Step 3：More multiplying power CNN pathological image disaggregated models are established, using the obtained RGB pathological images of step 1 to more
Times CNN pathological image disaggregated models carry out deep learning training；
Step 4：CNN pathological images disaggregated model after being trained using deep learning carries out cancer area to pathology full slice
It detects in domain.
Further, the step 3 includes the following steps：
Step 3.1:It includes multiple CNN networks to establish more multiplying power CNN pathological image disaggregated models, is expressed as CNN2×、
CNN5×、CNN20×, multiple CNN networks are trained successively according to the sequence of multiplying power from low to high；
Step 3.2：From more multiplying power pathological image training dataset X20×In down-sampled 10 times of every pathological image, generate packet
Training dataset containing 2 times of pathological imagesIn i-th of sample be expressed as (Ii,2×,yi), using describedInstruction
Practice sorter network CNN2×；
Step 3.3：Use CNN2×To X20×In all pathological images predicted, X20×In include i-th of pathology figure
The prediction result of picture is expressed as pi,2×, wherein pi,2×∈ [0,1] is represented general comprising cancerous region in i-th of pathological image
Rate, prediction result pi,2×Storage；
Step 3.4：Definition is for screeningIn difficult sample threshold value T2×∈ (0,1), by the difficult sample composition 2
Difficult sample data set under timesIt is formulated as follows：
Step 3.5：In the 20 haplotype data collection X20×Described in middle extractionIn include pathological image it is 20 times corresponding
Pathological image, and down-sampled 4 times, form 5 times of lower training book data sets X5×In i-th of sample be expressed as (Ii,5×,
yi)；
Step 3.6：Using describedTraining sorter network CNN5×；After the completion of training, CNN is used5×It is rightIn institute
There is pathological image to be predicted, it willIn include the prediction result of i-th of pathological image be expressed as pi,5×, wherein pi,5×
∈ [0,1], prediction result pi,5×Storage；
Step 3.7：Definition is for screeningIn difficult sample threshold value T5×∈ (0,1), by the difficult sample composition 5
Difficult sample data set under timesIt is formulated as follows：
Step 3.8：In the 20 haplotype data collection X20×Described in middle extractionIn include pathological image it is 20 times corresponding
Pathological image forms 20 times of lower training datasets
Step 3.9：Using describedTraining sorter network CNN20×；After the completion of training, deep learning training is completed.
Further, the step 4 includes the following steps：
Step 4.1：Pathology full slice to be diagnosed is cut under 2 times with sliding window method and training setIncluded in
The equal-sized subgraph of image forms 2 times of pathological image test sets, usesIt indicates,
Middle N2×It indicatesIn include subgraph quantity；
Step 4.2：Use the CNN trained2×Model pairIn image predicted, obtain prediction result collection, use
P2×={ pk,2×| k=1,2 ..., N2×Indicate；
Step 4.3：With threshold value T2×With set P2×Judge above-mentionedIn difficult sample, and acquire difficult sample in full slice
5 times of pathological images of middle corresponding position form 5 times of pathological image test sets, it is as follows to use formula：
Use the CNN trained5×Model pairIn image predicted, obtain prediction result collection, use P5×=
{pk,5×| k=1,2 ..., N5×Indicate, wherein N5×It representsIn include pathological image quantity；
Step 4.4：With threshold value T5×With set P5×Judge above-mentionedIn difficult sample, and acquire difficult sample in full slice
20 times of pathological images of middle corresponding position form 20 times of pathological image test sets, it is as follows to use formula：
Use the CNN trained20×Model pairIn image predicted, obtain prediction result collection, use P20×=
{pk,20×| k=1,2 ..., N20×Indicate, wherein N20×It representsIn include pathological image quantity；
Step 4.5：Given threshold T20×Judge Ik,20×Whether include cancerous region, it is as follows to use formula：
Step 4.6：Full slice cancerous region detection terminates, using above-mentioned prediction result judge be sliced in subregion whether
Including cancerous region, achievees the effect that full slice cancerous region detects, it is as follows to use formula：
Further, the step 1 is specially：Pathological section is scanned and stored in computer by slice scanner
In, the pixel value by the channel decomposition acquisition each pixel of image in tri- channels R, G, B, to obtain the number of image
It is believed that breath, it is as follows to use formula：
Wherein Ir(x,y)、Ig(x,y)、Ib(x, y) indicates three Color Channels of RGB of I (x, y) respectively.
Further, more multiplying power pathological image training dataset X described in step 220×It is suitble to by pathologist screening
The case of aided diagnosis method is developed, and the lesion region in the pathological section for including to case is labeled；It is special according to pathology
The mark of family, cuts the subregion pathological image of tape label, composition data collection in 20 times of pathological sections.
Sequence detection cancerous region of the present invention according to full slice multiplying power from small to large, judges in 2 times of full slice images
It is no longer participate in follow-up calculating for the region comprising cancer, is determined as that the region comprising cancer is no longer joined in 5 times of full slice images
It is calculated with follow-up so that be only sliced medium and small subregion and 20 times of images of dependence is needed to be judged, to greatly reduce
Calculation amount shortens Diagnostic Time.
Description of the drawings
Specific embodiments of the present invention will be described in further detail below in conjunction with the accompanying drawings.
Fig. 1 is flow chart of the present invention；
Fig. 2 is the training stage schematic diagram of the method for the invention；
Fig. 3 is the detection-phase schematic diagram of side's method of the present invention.
Specific implementation mode
With reference to the accompanying drawings and detailed description, the present invention will be further described.
Fig. 1 is to be specifically included the present invention provides a kind of pathology full slice diagnostic method based on more multiplying power deep learnings
Following steps：
Step 1：During pathological section scanning is calculated to electronics, pathological image is indicated in RGB color,
Obtain RGB pathological images；
Step 2：Pathological image is marked, more multiplying power pathological image training datasets are established, with set X20×It indicates；
Step 3：More multiplying power CNN pathological image disaggregated models are established, using the obtained RGB pathological images of step 1 to more
Times CNN pathological image disaggregated models carry out deep learning training；；
Step 4：CNN pathological images disaggregated model after being trained using deep learning carries out cancer area to pathology full slice
It detects in domain.
Specifically, step 1 is that pathological section is scanned and stored in computer by slice scanner, pass through channel point
Solution obtains pixel value of each pixel of image in tri- channels R, G, B and uses public affairs to obtain the data information of image
Formula is as follows：
Wherein Ir(x,y)、Ig(x,y)、Ib(x, y) indicates three Color Channels of RGB of I (x, y) respectively.Step 1 is logical
Cross channel decomposition obtain each pixel of image tri- channels R, G, B pixel value, to obtain image data letter
Breath, basis is provided for auxiliary diagnosis.
Step 2：Described in more multiplying power pathological image training dataset X20×It is screened by pathologist and is suitble to exploitation auxiliary
The case of diagnostic method, and the lesion region in the pathological section for including to case is labeled；According to the mark of pathologist
Note, cuts the subregion pathological image of tape label, composition data collection in 20 times of pathological sections.The X20×In i-th of sample
Originally it is expressed as (Ii,20×,yi), wherein Ii,20×Indicate a sub-regions pathological image, yi∈ { 0,1 } indicates Ii,20×Corresponding mark
Label.Specifically, yi=1 indicates pathological image IiIn include cancerous region, yi=0 indicates pathological image IiIn do not include canceration area
Domain.
As shown in Fig. 2, step 3 specifically comprises the following steps：
Step 3.1:It includes multiple CNN networks to establish more multiplying power CNN pathological image disaggregated models, is expressed as CNN2×、
CNN5×、CNN20×, multiple CNN networks are trained successively according to the sequence of multiplying power from low to high；
Step 3.2：From more multiplying power pathological image training dataset X20×In down-sampled 10 times of every pathological image, generate packet
Training dataset containing 2 times of pathological imagesIn i-th of sample be expressed as (Ii,2×,yi), using describedInstruction
Practice sorter network CNN2×；
Step 3.3：Use CNN2×To X20×In all pathological images predicted, X20×In include i-th of pathology figure
The prediction result of picture is expressed as pi,2×, wherein pi,2×∈ [0,1] is represented general comprising cancerous region in i-th pathological image
Rate, prediction result pi,2×Storage；
Step 3.4：Definition is for screeningIn difficult sample threshold value T2×∈ (0,1), by the difficult sample composition 2
Difficult sample data set under timesIt is formulated as follows：
Step 3.5：In the 20 haplotype data collection X20×Described in middle extractionIn include pathological image it is 20 times corresponding
Pathological image, and down-sampled 4 times, form 5 times of lower training book data setsX5×In i-th of sample be expressed as (Ii,5×,
yi)；
Step 3.6：Using describedTraining sorter network CNN5×；After the completion of training, CNN is used5×It is rightIn institute
There is pathological image to be predicted, it willIn include the prediction result of i-th of pathological image be expressed as pi,5×, wherein pi,5×
∈ [0,1], prediction result pi,5×Storage；
Step 3.7：Definition is for screeningIn difficult sample threshold value T5×∈ (0,1), by the difficult sample composition 5
Difficult sample data set under timesIt is formulated as follows：
Step 3.8：In the 20 haplotype data collection X20×Described in middle extractionIn include pathological image it is 20 times corresponding
Pathological image forms 20 times of lower training datasets
Step 3.9：Using describedTraining sorter network CNN20×；After the completion of training, deep learning training is completed.
In training CNN5×When, include only CNN in training set2×The sample for difficulty of classifying, makes CNN in training process5×Model
Differentiation is absorbed in for CNN2×For more difficult sample, do not influenced by simple sample, to obtain better classification capacity.
Similarly, in training CNN20×When, include only CNN in training set5×The sample for difficulty of classifying, makes CNN in training process20×Model
Differentiation is absorbed in for CNN5×For more difficult sample, do not influenced by simple sample, to obtain better classification capacity.
Such as Fig. 3, step 4 includes the following steps：
Step 4.1：Pathology full slice to be diagnosed is cut under 2 times with sliding window method and training setIncluded in
The equal-sized subgraph of image forms 2 times of pathological image test sets, usesIt indicates,
Middle N2×It indicatesIn include subgraph quantity；
Step 4.2：Use the CNN trained2×Model pairIn image predicted, obtain prediction result collection, use
P2×={ pk,2×| k=1,2 ..., N2×Indicate；
Step 4.3：With threshold value T2×With set P2×Judge above-mentionedIn difficult sample, and acquire difficult sample in full slice
5 times of pathological images of middle corresponding position form 5 times of pathological image test sets, it is as follows to use formula：
Use the CNN trained5×Model pairIn image predicted, obtain prediction result collection, use P5×=
{pk,5×| k=1,2 ..., N5×Indicate, wherein N5×It representsIn include pathological image quantity；
Step 4.4：With threshold value T5×With set P5×Judge above-mentionedIn difficult sample, and acquire difficult sample in full slice
20 times of pathological images of middle corresponding position form 20 times of pathological image test sets, it is as follows to use formula：
Use the CNN trained20×Model pairIn image predicted, obtain prediction result collection, use P20×=
{pk,20×| k=1,2 ..., N20×Indicate, wherein N20×It representsIn include pathological image quantity；
Step 4.5：Given threshold T20×Judge Ik,20×Whether include cancerous region, it is as follows to use formula：
Step 4.6：Full slice cancerous region detection terminates, using above-mentioned prediction result judge be sliced in subregion whether
Including cancerous region, achievees the effect that full slice cancerous region detects, it is as follows to use formula：
This method is verified on mammary gland Lymph Node Metastasis canceration image, given threshold T in verification2×=0.92, T5×=
0.85, T20×=0.5.Using the nicety of grading of " whether slice subgraph includes cancerous region " this classification problem as index,
The average time analyzed a digital pathology full slice image (about 50000 × 50000 pixel under 20 times), which is used as, to be calculated
Complexity index, the results are shown in Table 1.
1 feature extraction network performance of table compares
As it can be seen that the inventive method is compared with the side for diagnosing simultaneously fusion diagnosis result respectively under more multiplying powers (2 times, 5 times, 20 times)
Method speed promotes nearly 10 times, and diagnostic accuracy only declines 0.67%, disclosure satisfy that clinical application demand.
Although specifically showing and describing the present invention in conjunction with preferred embodiment, those skilled in the art should
Understand, do not departing from the spirit and scope of the present invention defined by the appended claims, in the form and details to this
Invention makes a variety of changes, and is protection scope of the present invention.
Claims (5)
1. a kind of pathology full slice diagnostic method based on more multiplying power deep learnings, which is characterized in that include the following steps：
Step 1：During pathological section scanning is calculated to electronics, pathological image is indicated in RGB color, is obtained
RGB pathological images；
Step 2：Pathological image is marked, more multiplying power pathological image training datasets are established, with set X20×It indicates；
Step 3：More multiplying power CNN pathological image disaggregated models are established, using the obtained RGB pathological images of step 1 to more times of CNN
Pathological image disaggregated model carries out deep learning training；；
Step 4：CNN pathological images disaggregated model after being trained using deep learning carries out cancerous area inspection to pathology full slice
It surveys.
2. a kind of pathology full slice diagnostic method based on more multiplying power deep learnings according to claim 1, feature exist
In the step 3 includes the following steps：
Step 3.1:It includes multiple CNN networks to establish more multiplying power CNN pathological image disaggregated models, is expressed as CNN2×、CNN5×、
CNN20×, multiple CNN networks are trained successively according to the sequence of multiplying power from low to high；
Step 3.2：From more multiplying power pathological image training dataset X20×In down-sampled 10 times of every pathological image, generate include 2 times
The training dataset of pathological image In i-th of sample be expressed as (Ii,2×,yi), using describedTraining classification net
Network CNN2×；
Step 3.3：Use CNN2×To X20×In all pathological images predicted, X20×In include i-th of pathological image it is pre-
It surveys result and is expressed as pi,2×, wherein pi,2×∈ [0,1] represents the probability for including cancerous region in i-th of pathological image, prediction knot
Fruit pi,2×Storage；
Step 3.4：Definition is for screeningIn difficult sample threshold value T2×∈ (0,1) forms the difficult sample under 2 times
Difficult sample data setIt is formulated as follows：
Step 3.5：In the 20 haplotype data collection X20×Described in middle extractionIn include the corresponding 20 times of pathology figures of pathological image
Picture, and down-sampled 4 times, form 5 times of lower training book data setsX5×In i-th of sample be expressed as (Ii,5×,yi)；
Step 3.6：Using describedTraining sorter network CNN5×；After the completion of training, CNN is used5×It is rightIn all pathology
Image is predicted, is incited somebody to actionIn include the prediction result of i-th of pathological image be expressed as pi,5×, wherein pi,5×∈ [0,1],
Prediction result pi,5×Storage；
Step 3.7：Definition is for screeningIn difficult sample threshold value T5×∈ (0,1), it is difficult under 5 times that the difficult sample is formed
Sample data setIt is formulated as follows：
Step 3.8：In the 20 haplotype data collection X20×Described in middle extractionIn include the corresponding 20 times of pathology figures of pathological image
Picture forms 20 times of lower training datasets
Step 3.9：Using describedTraining sorter network CNN20×；After the completion of training, deep learning training is completed.
3. according to claim 1 or a kind of pathology full slice diagnostic method based on more multiplying power deep learnings, feature
It is, the step 4 includes the following steps：
Step 4.1：Pathology full slice to be diagnosed is cut under 2 times with sliding window method and training setIncluded in image it is big
Small equal subgraph forms 2 times of pathological image test sets, usesIt indicates, wherein N2×Table
ShowIn include subgraph quantity；
Step 4.2：Use the CNN trained2×Model pairIn image predicted, obtain prediction result collection, use P2×=
{pk,2×| k=1,2 ..., N2×Indicate；
Step 4.3：With threshold value T2×With set P2×Judge above-mentionedIn difficult sample, and it is corresponding in full slice to acquire difficult sample
5 times of pathological images of position form 5 times of pathological image test sets, it is as follows to use formula：
Use the CNN trained5×Model pairIn image predicted, obtain prediction result collection, use P5×={ pk,5×| k=
1,2,...,N5×Indicate, wherein N5×It representsIn include pathological image quantity；
Step 4.4：With threshold value T5×With set P5×Judge above-mentionedIn difficult sample, and it is corresponding in full slice to acquire difficult sample
20 times of pathological images of position form 20 times of pathological image test sets, it is as follows to use formula：
Use the CNN trained20×Model pairIn image predicted, obtain prediction result collection, use P20×={ pk,20×|k
=1,2 ..., N20×Indicate, wherein N20×It representsIn include pathological image quantity；
Step 4.5：Given threshold T20×Judge Ik,20×Whether include cancerous region, it is as follows to use formula：
Step 4.6：The detection of full slice cancerous region terminates, and judges whether subregion includes cancer in slice using above-mentioned prediction result
Become region, achievees the effect that full slice cancerous region detects, it is as follows to use formula：
K-th of subgraph corresponds to the position in slice：
4. a kind of pathology full slice diagnostic method based on more multiplying power deep learnings according to claim 1, feature exist
In the step 1 is specially：Pathological section is scanned and stored in computer by slice scanner, passes through channel decomposition
It obtains pixel value of each pixel of image in tri- channels R, G, B and uses formula such as to obtain the data information of image
Under：
Wherein Ir(x,y)、Ig(x,y)、Ib(x, y) indicates three Color Channels of RGB of I (x, y) respectively.
5. a kind of pathology full slice diagnostic method based on more multiplying power deep learnings according to claim 1, feature exist
In more multiplying power pathological image training dataset X described in step 220×It is screened by pathologist and is suitble to exploitation aided diagnosis method
Case, and the lesion region in the pathological section for including to case is labeled；According to the mark of pathologist, in 20 times of diseases
The subregion pathological image of tape label, composition data collection are cut in reason slice.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810189171.1A CN108305253B (en) | 2018-03-08 | 2018-03-08 | Pathological image classification method based on multiple-time rate deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810189171.1A CN108305253B (en) | 2018-03-08 | 2018-03-08 | Pathological image classification method based on multiple-time rate deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108305253A true CN108305253A (en) | 2018-07-20 |
| CN108305253B CN108305253B (en) | 2021-04-06 |
Family
ID=62849654
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810189171.1A Active CN108305253B (en) | 2018-03-08 | 2018-03-08 | Pathological image classification method based on multiple-time rate deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108305253B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109903284A (en) * | 2019-03-04 | 2019-06-18 | 武汉大学 | A kind of other method and system of HER2 immunohistochemistry image automatic judging |
| CN111814893A (en) * | 2020-07-17 | 2020-10-23 | 首都医科大学附属北京胸科医院 | Lung full-scan image EGFR mutation prediction method and system based on deep learning |
| CN112070726A (en) * | 2020-08-17 | 2020-12-11 | 清华大学 | Grape fetus slice image processing method and device based on deep learning |
| CN113053512A (en) * | 2019-12-27 | 2021-06-29 | 无锡祥生医疗科技股份有限公司 | Evolution learning method, system and storage medium suitable for ultrasonic diagnosis |
| CN113109327A (en) * | 2021-03-09 | 2021-07-13 | 杭州市林业科学研究院 | Prediction method of dry rot of hickory nut |
| CN113269747A (en) * | 2021-05-24 | 2021-08-17 | 浙江大学医学院附属第一医院 | Pathological picture liver cancer diffusion detection method and system based on deep learning |
| CN114638292A (en) * | 2022-03-10 | 2022-06-17 | 中国医学科学院北京协和医院 | Artificial intelligence pathology auxiliary diagnosis system based on multi-scale analysis |
| EP3874447A4 (en) * | 2018-10-30 | 2022-08-17 | Perimeter Medical Imaging, Inc. | Methods and systems for medical image processing using a convolutional neural network (cnn) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102566035A (en) * | 2010-12-03 | 2012-07-11 | 索尼公司 | Image processing method, image processing apparatus, and image processing program |
| WO2012096992A1 (en) * | 2011-01-10 | 2012-07-19 | Rutgers, The State University Of New Jersey | Boosted consensus classifier for large images using fields of view of various sizes |
| US20140180977A1 (en) * | 2012-12-21 | 2014-06-26 | Nec Laboratories America, Inc. | Computationally Efficient Whole Tissue Classifier for Histology Slides |
| EP2854100A1 (en) * | 2013-08-21 | 2015-04-01 | Sectra AB | Methods, systems and circuits for generating magnification-dependent images suitable for whole slide images |
| CN105938243A (en) * | 2016-06-29 | 2016-09-14 | 华南理工大学 | Multi-magnification microscope fast focusing method applied to TFT-LCD detection |
-
2018
- 2018-03-08 CN CN201810189171.1A patent/CN108305253B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102566035A (en) * | 2010-12-03 | 2012-07-11 | 索尼公司 | Image processing method, image processing apparatus, and image processing program |
| WO2012096992A1 (en) * | 2011-01-10 | 2012-07-19 | Rutgers, The State University Of New Jersey | Boosted consensus classifier for large images using fields of view of various sizes |
| US20140180977A1 (en) * | 2012-12-21 | 2014-06-26 | Nec Laboratories America, Inc. | Computationally Efficient Whole Tissue Classifier for Histology Slides |
| EP2854100A1 (en) * | 2013-08-21 | 2015-04-01 | Sectra AB | Methods, systems and circuits for generating magnification-dependent images suitable for whole slide images |
| CN105938243A (en) * | 2016-06-29 | 2016-09-14 | 华南理工大学 | Multi-magnification microscope fast focusing method applied to TFT-LCD detection |
Non-Patent Citations (1)
| Title |
|---|
| BARIS GECER: "Detection and Classification of Breast Cancer in Whole Slide Histopathology Images Using Deep Convolutional Networks", 《 SEMANTIC SCHOLAR COMPUTER SCIENCE》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3874447A4 (en) * | 2018-10-30 | 2022-08-17 | Perimeter Medical Imaging, Inc. | Methods and systems for medical image processing using a convolutional neural network (cnn) |
| CN109903284A (en) * | 2019-03-04 | 2019-06-18 | 武汉大学 | A kind of other method and system of HER2 immunohistochemistry image automatic judging |
| CN109903284B (en) * | 2019-03-04 | 2022-07-05 | 武汉大学 | HER2 immunohistochemical image automatic discrimination method and system |
| CN113053512A (en) * | 2019-12-27 | 2021-06-29 | 无锡祥生医疗科技股份有限公司 | Evolution learning method, system and storage medium suitable for ultrasonic diagnosis |
| CN113053512B (en) * | 2019-12-27 | 2024-04-09 | 无锡祥生医疗科技股份有限公司 | Evolutionary learning method, system and storage medium suitable for ultrasonic diagnosis |
| CN111814893A (en) * | 2020-07-17 | 2020-10-23 | 首都医科大学附属北京胸科医院 | Lung full-scan image EGFR mutation prediction method and system based on deep learning |
| CN112070726A (en) * | 2020-08-17 | 2020-12-11 | 清华大学 | Grape fetus slice image processing method and device based on deep learning |
| CN113109327A (en) * | 2021-03-09 | 2021-07-13 | 杭州市林业科学研究院 | Prediction method of dry rot of hickory nut |
| CN113109327B (en) * | 2021-03-09 | 2023-11-17 | 杭州市林业科学研究院 | Method for predicting dry rot of hickory |
| CN113269747A (en) * | 2021-05-24 | 2021-08-17 | 浙江大学医学院附属第一医院 | Pathological picture liver cancer diffusion detection method and system based on deep learning |
| CN114638292A (en) * | 2022-03-10 | 2022-06-17 | 中国医学科学院北京协和医院 | Artificial intelligence pathology auxiliary diagnosis system based on multi-scale analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108305253B (en) | 2021-04-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108305253A (en) | A kind of pathology full slice diagnostic method based on more multiplying power deep learnings | |
| CN109344736B (en) | Static image crowd counting method based on joint learning | |
| CN110428432B (en) | Deep neural network algorithm for automatically segmenting colon gland image | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| US9741112B2 (en) | Generating image-based diagnostic tests by optimizing image analysis and data mining of co-registered images | |
| CN107665492B (en) | Colorectal panoramic digital pathological image tissue segmentation method based on depth network | |
| CN110889852B (en) | Liver segmentation method based on residual error-attention deep neural network | |
| US9159129B2 (en) | Generating image-based diagnostic tests by optimizing image analysis and data mining of co-registered images | |
| CN110503630A (en) | A kind of cerebral hemorrhage classification, positioning and prediction technique based on three dimensional depth learning model | |
| US11250568B2 (en) | Techniques for determining tissue characteristics using multiplexed immunofluorescence imaging | |
| CN114565761B (en) | Deep learning-based method for segmenting tumor region of renal clear cell carcinoma pathological image | |
| CN108629772A (en) | Image processing method and device, computer equipment and computer storage media | |
| CN109871875A (en) | A kind of building change detecting method based on deep learning | |
| CN115909006B (en) | Mammary tissue image classification method and system based on convolution transducer | |
| CN109360191B (en) | Image significance detection method based on variational self-encoder | |
| CN112750132A (en) | White blood cell image segmentation method based on dual-path network and channel attention | |
| IL301650A (en) | A method of processing an image of tissue and a system for processing an image of tissue | |
| CN115546605A (en) | Training method and device based on image labeling and segmentation model | |
| CN113657449A (en) | Traditional Chinese medicine tongue picture greasy classification method containing noise labeling data | |
| CN114140437A (en) | Fundus hard exudate segmentation method based on deep learning | |
| CN116912240B (en) | Mutation TP53 immunology detection method based on semi-supervised learning | |
| CN112927215A (en) | Automatic analysis method for digestive tract biopsy pathological section | |
| CN116884597A (en) | Pathological image breast cancer molecular typing method and system based on self-supervision pre-training and multi-example learning | |
| Pan et al. | A review of machine learning approaches, challenges and prospects for computational tumor pathology | |
| CN113822252A (en) | Pathological image cell robust detection method under microscope |
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 |