CN117893875A - Construction method and system of pancreatic cancer recognition model based on convolutional neural network - Google Patents
Construction method and system of pancreatic cancer recognition model based on convolutional neural network Download PDFInfo
- Publication number
- CN117893875A CN117893875A CN202410003838.XA CN202410003838A CN117893875A CN 117893875 A CN117893875 A CN 117893875A CN 202410003838 A CN202410003838 A CN 202410003838A CN 117893875 A CN117893875 A CN 117893875A
- Authority
- CN
- China
- Prior art keywords
- image
- pancreatic cancer
- convolutional neural
- neural network
- training
- 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
- 206010061902 Pancreatic neoplasm Diseases 0.000 title claims abstract description 78
- 208000015486 malignant pancreatic neoplasm Diseases 0.000 title claims abstract description 78
- 201000002528 pancreatic cancer Diseases 0.000 title claims abstract description 78
- 208000008443 pancreatic carcinoma Diseases 0.000 title claims abstract description 78
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 47
- 238000010276 construction Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 62
- 238000012795 verification Methods 0.000 claims abstract description 47
- 210000004923 pancreatic tissue Anatomy 0.000 claims abstract description 35
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 238000012502 risk assessment Methods 0.000 claims description 59
- 238000001514 detection method Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 27
- 230000005540 biological transmission Effects 0.000 claims description 25
- 238000013524 data verification Methods 0.000 claims description 21
- 238000000034 method Methods 0.000 claims description 16
- 206010028980 Neoplasm Diseases 0.000 description 8
- 230000001575 pathological effect Effects 0.000 description 8
- 210000001519 tissue Anatomy 0.000 description 7
- 238000013473 artificial intelligence Methods 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 238000012805 post-processing Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000013179 statistical model Methods 0.000 description 2
- 230000009469 supplementation Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000003292 glue Substances 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 238000012333 histopathological diagnosis Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000011347 resin Substances 0.000 description 1
- 229920005989 resin Polymers 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a construction method and a construction system of a pancreatic cancer recognition model based on a convolutional neural network, wherein the construction method comprises the following steps: acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue; verifying the authenticity of the first image; when the verification is passed, preprocessing the first image and the second image to obtain a third image; training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence. The construction method of pancreatic cancer identification model based on convolutional neural network realizes accurate construction of identification model.
Description
Technical Field
The invention relates to the technical field of model construction, in particular to a construction method and a construction system of pancreatic cancer recognition models based on convolutional neural networks.
Background
In recent years, with the proposal of the concept of artificial intelligence +, the trend of improving or even subverting the traditional industry by using information technology means is more and more obvious, and the medical industry is the key point and the difficulty in the concept. Intelligent medical treatment is a new model capable of realizing resource, service and experience sharing in the medical industry through latest information technology means such as Internet of things, cloud computing and artificial intelligent analysis.
The establishment of an intelligent medical system depends on three technical supports, namely the combination of big data and artificial intelligence, the popularization of mobile medical treatment and the connection of a cloud platform of an island. Wherein, artificial intelligence will exert the effect of well flowing column in wisdom medical treatment. In addition, in medical big data, more than 80% of the data come from medical images, and a more efficient and accurate technical means is objectively required for reading a large amount of image data, and the artificial intelligence just can meet the requirements.
In clinical medical practice, histopathological diagnosis is recognized as a gold standard for tumor diagnosis, and can provide indispensable reference information for subsequent treatment, and is very important. However, even for the same kind of tumor only, different tissue areas, different stages of disease course, different environmental influences, different physical factors, etc., can cause a great deal of difference in the morphology of tissue cells in pathological sections.
By means of pathological section scanning equipment, hospitals can digitize a large number of sections, which lays a solid foundation for implementing pathological image intelligent analysis. In addition, with the rapid development of artificial intelligence technology, the performance of intelligent algorithms in fields such as picture classification, man-machine play, automatic driving, etc. has reached or even exceeded the human level. The diagnosis is assisted by the artificial intelligence technology, so that the diagnosis is not a very unreachable and highly infeasible matter.
In order to better realize the assistance of diagnosis of doctors, firstly, how to accurately construct an artificial intelligent identification model is to be solved.
Disclosure of Invention
The invention aims to provide a construction method of a pancreatic cancer recognition model based on a convolutional neural network, which realizes accurate construction of the recognition model.
The embodiment of the invention provides a construction method of a pancreatic cancer recognition model based on a convolutional neural network, which comprises the following steps:
acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue;
verifying the authenticity of the first image;
when the verification is passed, preprocessing the first image and the second image to obtain a third image;
training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence.
Preferably, acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue includes:
acquiring an image uploaded by a detection terminal through the detection terminal authenticated by the system;
analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a first image containing pancreatic cancer;
and/or the number of the groups of groups,
and analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a second image of normal pancreatic tissues.
Preferably, acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue includes:
extracting a first image from a pancreatic cancer picture library;
and extracting a second image from the normal pancreas tissue picture library.
Preferably, the verifying of the authenticity of the first image comprises:
acquiring terminal information of a shooting terminal of a first image;
determining whether the first image is directly uploaded by the shooting terminal;
if so, determining whether the shooting terminal exists in a preset terminal record library or not based on the terminal information; when present, the verification passes;
when not, acquiring a transmission path of the first image; determining a risk value of a first image based on a preset risk analysis library, a transmission path and a shooting terminal; when the risk value is smaller than or equal to a preset risk threshold value, the verification is passed;
the method for determining the risk value of the first image based on the preset risk analysis library, the transmission path and the shooting terminal comprises the following steps:
extracting characteristics of terminal information of a shooting terminal to obtain a plurality of first characteristic values;
extracting features of path information of the transmission path to obtain a plurality of second feature values;
constructing a risk analysis parameter set based on the plurality of first feature values and the plurality of second feature values;
matching the risk analysis parameter set with the standard parameter set correspondingly associated with each risk analysis result in the risk analysis library;
extracting risk analysis results correspondingly associated with the standard parameter sets matched with the risk analysis parameter sets;
analyzing the risk analysis result and determining a risk value.
Preferably, training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence includes:
extracting the characteristics of each third image, and constructing a characteristic parameter set corresponding to the third image based on the extracted characteristic values;
grouping each third image based on the characteristic parameter set of each third image to obtain a plurality of groups;
determining the packet number corresponding to each packet based on a preset packet number library;
constructing a data verification vector based on the group numbers and the image numbers corresponding to the groups;
verifying based on a preset data verification library and a data verification vector;
when the verification fails, the first image, the second image and the third image are acquired again;
when the verification is passed, dividing each group into preset parts of subgroups; constructing each training set and each testing set based on different subgroups of different groups;
based on the training set, training the initial convolutional neural network model, testing the network model through the testing set after training convergence, and obtaining the pancreatic cancer recognition model after the testing set passes.
The invention also provides a construction system of the pancreatic cancer recognition model based on the convolutional neural network, which comprises the following steps:
an acquisition module for acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue;
the verification module is used for verifying the authenticity of the first image;
the preprocessing module is used for preprocessing the first image and the second image after the verification is passed, and obtaining a third image;
the training module is used for training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence.
Preferably, the acquisition module acquires a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue, and performs the following operations:
acquiring an image uploaded by a detection terminal through the detection terminal authenticated by the system;
analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a first image containing pancreatic cancer;
and/or the number of the groups of groups,
and analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a second image of normal pancreatic tissues.
Preferably, acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue includes:
extracting a first image from a pancreatic cancer picture library;
and extracting a second image from the normal pancreas tissue picture library.
Preferably, the verification module performs authenticity verification on the first image, and performs the following operations:
acquiring terminal information of a shooting terminal of a first image;
determining whether the first image is directly uploaded by the shooting terminal;
if so, determining whether the shooting terminal exists in a preset terminal record library or not based on the terminal information; when present, the verification passes;
when not, acquiring a transmission path of the first image; determining a risk value of a first image based on a preset risk analysis library, a transmission path and a shooting terminal; when the risk value is smaller than or equal to a preset risk threshold value, the verification is passed;
the method for determining the risk value of the first image based on the preset risk analysis library, the transmission path and the shooting terminal comprises the following steps:
extracting characteristics of terminal information of a shooting terminal to obtain a plurality of first characteristic values;
extracting features of path information of the transmission path to obtain a plurality of second feature values;
constructing a risk analysis parameter set based on the plurality of first feature values and the plurality of second feature values;
matching the risk analysis parameter set with the standard parameter set correspondingly associated with each risk analysis result in the risk analysis library;
extracting risk analysis results correspondingly associated with the standard parameter sets matched with the risk analysis parameter sets;
analyzing the risk analysis result and determining a risk value.
Preferably, the training module trains the initial convolutional neural network model based on the third image, obtains the pancreatic cancer recognition model after the training converges, and executes the following operations:
extracting the characteristics of each third image, and constructing a characteristic parameter set corresponding to the third image based on the extracted characteristic values;
grouping each third image based on the characteristic parameter set of each third image to obtain a plurality of groups;
determining the packet number corresponding to each packet based on a preset packet number library;
constructing a data verification vector based on the group numbers and the image numbers corresponding to the groups;
verifying based on a preset data verification library and a data verification vector;
when the verification fails, the first image, the second image and the third image are acquired again;
when the verification is passed, dividing each group into preset parts of subgroups; constructing each training set and each testing set based on different subgroups of different groups;
based on the training set, training the initial convolutional neural network model, testing the network model through the testing set after training convergence, and obtaining the pancreatic cancer recognition model after the testing set passes.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention. In the drawings:
FIG. 1 is a schematic diagram of a construction method of a pancreatic cancer recognition model based on a convolutional neural network in an embodiment of the invention;
FIG. 2 is a schematic diagram of a convolutional neural network;
fig. 3 is a schematic diagram of a pancreatic cancer recognition model construction system based on a convolutional neural network according to an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
The embodiment of the invention provides a construction method of a pancreatic cancer recognition model based on a convolutional neural network, which is shown in fig. 1 and comprises the following steps:
step S1: acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue;
step S2: verifying the authenticity of the first image;
step S3: when the verification is passed, preprocessing the first image and the second image to obtain a third image;
step S4: training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence.
The working principle and the beneficial effects of the technical scheme are as follows:
before model training, the obtained first image which is used as a training basis and contains pancreatic cancer and the second image of normal pancreatic tissue are subjected to authenticity verification, so that the accuracy and the effectiveness of the data which are used as the training basis are ensured, and the accuracy and the effectiveness of the trained model are ensured. After the authenticity verification is passed, the first image and the second image are preprocessed, the preprocessing mainly aiming at flaws occurring during the production and scanning process, for example: imaging blur, unevenness, resin glue noise, and the like; the image quality is reduced, the learning and the prediction of the model are affected, algorithms in graphic images such as a color balance algorithm, a geometric affine transformation algorithm and the like can be comprehensively utilized, and the preprocessing flow of the image is designed to complete the preprocessing work of the image; the color balance algorithm, the geometric affine transformation algorithm and the like are all mature technologies, and are not described herein. Convolutional neural networks are the basis for constructing computer vision models for digital pathological section analysis, and in general, typical models belong to supervised learning models. When a model is constructed, the topological structure of the network, the link relation among layers, the number of parameters of each layer, the calculation method and the like are required to be determined. By setting these super parameters, the construction of the network is completed. Meanwhile, an optimization method, training parameters and a gradient generation mode of the network are required to be defined, so that the network can be trained. The specific training steps of the model are also known in the art and are not described in detail.
When the model is applied to a specific application after the model is built, the capability of distinguishing digital pathological sections is obtained by a supervised learning mode because of the model built by the method. As shown in fig. 2, since the model based on the convolutional neural network has mobility, it is possible to construct a model capable of identifying various tumor tissues and completing a plurality of diagnostic tasks. When the model goes through a certain task, it will be organized only at the level of abstraction after learning about it. When such a model is used to accomplish another similar task, the learned knowledge can be migrated through a design to solve the new task. Thus, the data and time required to learn a new task is greatly reduced. This is the greatest advantage of vision models based on convolutional neural networks. To take full advantage of this, a finer network design is needed, and the present application strives to make the model reusable as much as possible. Since the original image is too large to be directly processed by the computer, the original image must be diced and sampled, so the result of the model is for the sampled tiles (first image and second image) rather than for the original image. Further post-processing is necessary to obtain the final result for the complete image. And splicing the results of the corresponding image blocks through post-processing to obtain a tumor probability distribution result aiming at the complete image. Meanwhile, in order to facilitate understanding of doctors, the original image and the result image are integrated, so that the tumor area can be clearly displayed. Some key features may be missed due to the use of sampling. Through the post-processing process, samples of false positives and false negatives are collected, key samples which are missed during sampling can be filtered out, and the further optimization model provides data. In addition, further processing of the tumor region in the resulting image is required in order to reach clinically desirable diagnostic conclusions. The result is subjected to secondary mining to further extract morphological features such as: the number, the area, the perimeter and the like of the tumor communicating areas, and the relation between the characteristics and clinical diagnosis is established by constructing a statistical model, so that the prediction of the whole pathological picture is completed.
In summary, the present application uses computer graphics techniques to build supervised learning models using convolutional neural networks and to implement multitasking migration of the models. Finally, post-processing of model output data is carried out, morphological characteristics are extracted, a statistical model is built, and finally a digital pathological auxiliary diagnosis result (auxiliary graphic) is obtained.
In one embodiment, acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue comprises:
acquiring an image uploaded by a detection terminal through the detection terminal authenticated by the system;
analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a first image containing pancreatic cancer;
and/or the number of the groups of groups,
and analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a second image of normal pancreatic tissues.
The working principle and the beneficial effects of the technical scheme are as follows:
only the detection terminal authenticated by the system can upload the first image and the second image; the system mainly submits authentication information in advance, for example: detecting the model, brand, serial number, name of the unit, name of the responsible person, contact information, whether the unit is qualified, and the like. After the authentication is completed, when a worker of the unit to which the detection terminal belongs uploads the image by using the detection terminal, the description information is filled in an uploading interface of the uploading image, so that whether the uploading is the first image or the second image is determined; the rules of system authentication include: the detection terminal may be a detection device for pathological sections of the laboratory detection departments of the respective hospitals or a computer connected thereto.
In one embodiment, acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue comprises:
extracting a first image from a pancreatic cancer picture library;
and extracting a second image from the normal pancreas tissue picture library.
The working principle and the beneficial effects of the technical scheme are as follows:
a pancreatic cancer picture library and a normal pancreatic tissue picture library are built in a system of the platform and are used for recording a first image and a second image which are confirmed by each expert on the platform; to ensure the accuracy of the images used for training.
In one embodiment, a method for constructing a pancreatic cancer recognition model based on a convolutional neural network, in step S1: acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue; further comprises:
when receiving a model construction application of a user, outputting a preset data acquisition mode distribution interface;
receiving the total data amount configured on a data acquisition mode distribution interface by a user and the data duty ratio of each data source;
when executing the step S1, acquiring a first image and a second image based on the total data amount configured by the user and the data duty ratio of each data source;
the ratio between the first image and the second image in each data source can be configured in a specific configuration; the data sources may be individual units (hospitals), platform stores, and the like. The user-defined configuration can meet the autonomous requirements of different users, and the model is built.
In addition, credit evaluation is carried out on each data source and displayed on a data acquisition mode distribution interface, so that the data acquisition duty ratio configuration of a user is facilitated. The credit evaluation is specifically as follows: and determining the credit evaluation value as the credit evaluation of each data source by inquiring the corresponding table of the ratio of the image of the mistransmission of each data source to the total uploaded image and the corresponding ratio and the credit evaluation value. Wherein the mistransmitted image comprises: transmitting an image of normal pancreatic tissue as a first image; transmitting an image containing pancreatic cancer as a second image; transmitting an image containing other cancers as a first image or a second image; transmitting the image of the other tissue as the first image or the second image, and the like.
In one embodiment, verifying the authenticity of the first image comprises:
acquiring terminal information of a shooting terminal of a first image;
determining whether the first image is directly uploaded by the shooting terminal;
if so, determining whether the shooting terminal exists in a preset terminal record library or not based on the terminal information; when present, the verification passes;
when not, acquiring a transmission path of the first image; determining a risk value of a first image based on a preset risk analysis library, a transmission path and a shooting terminal; when the risk value is smaller than or equal to a preset risk threshold value, the verification is passed;
the method for determining the risk value of the first image based on the preset risk analysis library, the transmission path and the shooting terminal comprises the following steps:
extracting characteristics of terminal information of a shooting terminal to obtain a plurality of first characteristic values; the first characteristic value includes: a feature value indicating a brand-type of the terminal, a feature value indicating a setting position of the terminal, a feature value indicating a unit code of a unit of the terminal, and the like;
extracting features of path information of the transmission path to obtain a plurality of second feature values; the second characteristic value comprises a characteristic value representing the total number of nodes in the path, a characteristic value representing the type of each node, a characteristic value representing the risk level of each node and the like; wherein the risk score is evaluated by configuring evaluation rules in the system, for example: when a strange node, namely a node which has never uploaded data, the risk level of the strange node is configured to be the highest; the nodes subjected to information supplementation pre-configure risk levels according to the perfection degree of the information supplementation, and different risk levels are configured; and then, after the professional analyzes the supplementary information, the preconfigured risk level is adjusted.
Constructing a risk analysis parameter set based on the plurality of first feature values and the plurality of second feature values; sequentially arranging the first characteristic value and the second characteristic value to form a risk analysis parameter set;
matching the risk analysis parameter set with the standard parameter set correspondingly associated with each risk analysis result in the risk analysis library; the matching can be realized by calculating the similarity of the risk analysis parameter set and the standard parameter set, and when the calculated similarity is the largest and is larger than a preset similarity threshold (0.95), the matching of the two parameter sets is determined; the risk analysis results are associated with the standard parameter sets in a one-to-one correspondence manner in a risk analysis library;
extracting risk analysis results correspondingly associated with the standard parameter sets matched with the risk analysis parameter sets;
analyzing the risk analysis result and determining a risk value. The risk results include risk values.
The embodiment provides that risk analysis is required to be carried out according to the transmission path of the image for the image which is not directly shot and uploaded by shooting equipment, so that the safety, accuracy and reliability of the image are further ensured, and the accuracy and reliability of the model are further ensured.
In one embodiment, training the initial convolutional neural network model based on the third image, and after convergence of the training, obtaining a pancreatic cancer recognition model, including:
extracting the characteristics of each third image, and constructing a characteristic parameter set corresponding to the third image based on the extracted characteristic values; the characteristic values include: number of tissues in image, number of abnormal tissues (cells), abnormal tissue ratio, ratio of respective chromaticities, and the like
Grouping each third image based on the characteristic parameter set of each third image to obtain a plurality of groups; calculating the similarity between the characteristic parameter set of each third image and the standard parameter group number set corresponding to each group in the preset group number library, and dividing images with similarity greater than a preset group threshold (0.8) into the same group;
determining the packet number corresponding to each packet based on a preset packet number library; the group numbers in the group number library are associated with the standard group parameter sets in a one-to-one correspondence manner;
constructing a data verification vector based on the group numbers and the image numbers corresponding to the groups;
verifying based on a preset data verification library and a data verification vector; each standard vector in the data verification library is associated with a verification result in a one-to-one correspondence manner; the corresponding verification result is called through the matching of the standard vector and the data verification vector; in general, building a data verification library generally does not allow for large differences in the amount of data between packets and the more data packets are verified the greater the likelihood; for example: the number of the packets is more than 5, and the data volume of each packet is not less than 100, and the verification is passed; otherwise, not pass.
When the verification fails, the first image, the second image and the third image are acquired again;
when the verification is passed, dividing each group into preset parts of subgroups; constructing each training set and each testing set based on different subgroups of different groups; the preset number of copies may be set to 5; i.e. into 5 subgroups; each group corresponds to one of 4 training sets and one test set;
based on the training set, training the initial convolutional neural network model, testing the network model through the testing set after training convergence, and obtaining the pancreatic cancer recognition model after the testing set passes. In addition, the training set and the test set can be exchanged for training for multiple times, and a model with the optimal test result is extracted to be used as a final pancreatic cancer identification model, so that the accuracy of the model is further improved.
The invention also provides a construction system of the pancreatic cancer recognition model based on the convolutional neural network, which comprises the following steps:
an acquisition module 1 for acquiring a plurality of first images containing pancreatic cancer and a plurality of second images of normal pancreatic tissue;
a verification module 2, configured to perform authenticity verification on the first image;
the preprocessing module 3 is used for preprocessing the first image and the second image after the verification is passed, and obtaining a third image;
and the training module 4 is used for training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after the training converges.
In one embodiment, the acquisition module acquires a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue, performs the following:
acquiring an image uploaded by a detection terminal through the detection terminal authenticated by the system;
analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a first image containing pancreatic cancer;
and/or the number of the groups of groups,
and analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a second image of normal pancreatic tissues.
In one embodiment, acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue comprises:
extracting a first image from a pancreatic cancer picture library;
and extracting a second image from the normal pancreas tissue picture library.
In one embodiment, the verification module performs authenticity verification on the first image, performing the following:
acquiring terminal information of a shooting terminal of a first image;
determining whether the first image is directly uploaded by the shooting terminal;
if so, determining whether the shooting terminal exists in a preset terminal record library or not based on the terminal information; when present, the verification passes;
when not, acquiring a transmission path of the first image; determining a risk value of a first image based on a preset risk analysis library, a transmission path and a shooting terminal; when the risk value is smaller than or equal to a preset risk threshold value, the verification is passed;
the method for determining the risk value of the first image based on the preset risk analysis library, the transmission path and the shooting terminal comprises the following steps:
extracting characteristics of terminal information of a shooting terminal to obtain a plurality of first characteristic values;
extracting features of path information of the transmission path to obtain a plurality of second feature values;
constructing a risk analysis parameter set based on the plurality of first feature values and the plurality of second feature values;
matching the risk analysis parameter set with the standard parameter set correspondingly associated with each risk analysis result in the risk analysis library;
extracting risk analysis results correspondingly associated with the standard parameter sets matched with the risk analysis parameter sets;
analyzing the risk analysis result and determining a risk value.
In one embodiment, the training module trains the initial convolutional neural network model based on the third image, obtains the pancreatic cancer recognition model after training convergence, and performs the following operations:
extracting the characteristics of each third image, and constructing a characteristic parameter set corresponding to the third image based on the extracted characteristic values;
grouping each third image based on the characteristic parameter set of each third image to obtain a plurality of groups;
determining the packet number corresponding to each packet based on a preset packet number library;
constructing a data verification vector based on the group numbers and the image numbers corresponding to the groups;
verifying based on a preset data verification library and a data verification vector;
when the verification fails, the first image, the second image and the third image are acquired again;
when the verification is passed, dividing each group into preset parts of subgroups; constructing each training set and each testing set based on different subgroups of different groups;
based on the training set, training the initial convolutional neural network model, testing the network model through the testing set after training convergence, and obtaining the pancreatic cancer recognition model after the testing set passes.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (10)
1. The construction method of the pancreatic cancer identification model based on the convolutional neural network is characterized by comprising the following steps of:
acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue;
verifying the authenticity of the first image;
when the verification is passed, preprocessing the first image and the second image to obtain a third image;
training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence.
2. The method for constructing a pancreatic cancer recognition model based on a convolutional neural network according to claim 1, wherein acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue comprises:
acquiring an image uploaded by a detection terminal through the detection terminal authenticated by the system;
analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a first image containing pancreatic cancer;
and/or the number of the groups of groups,
and analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a second image of normal pancreatic tissues.
3. The method for constructing a pancreatic cancer recognition model based on a convolutional neural network according to claim 1, wherein acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue comprises:
extracting a first image from a pancreatic cancer picture library;
and extracting a second image from the normal pancreas tissue picture library.
4. The method for constructing a pancreatic cancer recognition model based on a convolutional neural network according to claim 1, wherein the performing of the authenticity verification on the first image comprises:
acquiring terminal information of a shooting terminal of a first image;
determining whether the first image is directly uploaded by the shooting terminal;
if so, determining whether the shooting terminal exists in a preset terminal record library or not based on the terminal information; when present, the verification passes;
when not, acquiring a transmission path of the first image; determining a risk value of a first image based on a preset risk analysis library, a transmission path and a shooting terminal; when the risk value is smaller than or equal to a preset risk threshold value, the verification is passed;
the method for determining the risk value of the first image based on the preset risk analysis library, the transmission path and the shooting terminal comprises the following steps:
extracting characteristics of terminal information of a shooting terminal to obtain a plurality of first characteristic values;
extracting features of path information of the transmission path to obtain a plurality of second feature values;
constructing a risk analysis parameter set based on the plurality of first feature values and the plurality of second feature values;
matching the risk analysis parameter set with the standard parameter set correspondingly associated with each risk analysis result in the risk analysis library;
extracting risk analysis results correspondingly associated with the standard parameter sets matched with the risk analysis parameter sets;
analyzing the risk analysis result and determining a risk value.
5. The method for constructing a pancreatic cancer recognition model based on a convolutional neural network according to claim 1, wherein training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after convergence of training, comprises:
extracting the characteristics of each third image, and constructing a characteristic parameter set corresponding to the third image based on the extracted characteristic values;
grouping each third image based on the characteristic parameter set of each third image to obtain a plurality of groups;
determining the packet number corresponding to each packet based on a preset packet number library;
constructing a data verification vector based on the group numbers and the image numbers corresponding to the groups;
verifying based on a preset data verification library and a data verification vector;
when the verification fails, the first image, the second image and the third image are acquired again;
when the verification is passed, dividing each group into preset parts of subgroups; constructing each training set and each testing set based on different subgroups of different groups;
based on the training set, training the initial convolutional neural network model, testing the network model through the testing set after training convergence, and obtaining the pancreatic cancer recognition model after the testing set passes.
6. A system for constructing a pancreatic cancer recognition model based on a convolutional neural network, comprising:
an acquisition module for acquiring a plurality of first images including pancreatic cancer and a plurality of second images of normal pancreatic tissue;
the verification module is used for verifying the authenticity of the first image;
the preprocessing module is used for preprocessing the first image and the second image after the verification is passed, and obtaining a third image;
the training module is used for training the initial convolutional neural network model based on the third image, and obtaining the pancreatic cancer recognition model after training convergence.
7. The system for constructing a model for identifying pancreatic cancer based on a convolutional neural network of claim 6, wherein the acquisition module acquires a plurality of first images containing pancreatic cancer and a plurality of second images of normal pancreatic tissue, and performs the following operations:
acquiring an image uploaded by a detection terminal through the detection terminal authenticated by the system;
analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a first image containing pancreatic cancer;
and/or the number of the groups of groups,
and analyzing the image description information synchronously uploaded when the detection terminal uploads the image, and determining whether the image is a second image of normal pancreatic tissues.
8. The convolutional neural network-based pancreatic cancer recognition model construction system of claim 6, wherein acquiring a plurality of first images comprising pancreatic cancer and a plurality of second images of normal pancreatic tissue comprises:
extracting a first image from a pancreatic cancer picture library;
and extracting a second image from the normal pancreas tissue picture library.
9. The system for constructing a pancreatic cancer recognition model based on a convolutional neural network of claim 6, wherein the verification module performs an authenticity verification on the first image by:
acquiring terminal information of a shooting terminal of a first image;
determining whether the first image is directly uploaded by the shooting terminal;
if so, determining whether the shooting terminal exists in a preset terminal record library or not based on the terminal information; when present, the verification passes;
when not, acquiring a transmission path of the first image; determining a risk value of a first image based on a preset risk analysis library, a transmission path and a shooting terminal; when the risk value is smaller than or equal to a preset risk threshold value, the verification is passed;
the method for determining the risk value of the first image based on the preset risk analysis library, the transmission path and the shooting terminal comprises the following steps:
extracting characteristics of terminal information of a shooting terminal to obtain a plurality of first characteristic values;
extracting features of path information of the transmission path to obtain a plurality of second feature values;
constructing a risk analysis parameter set based on the plurality of first feature values and the plurality of second feature values;
matching the risk analysis parameter set with the standard parameter set correspondingly associated with each risk analysis result in the risk analysis library;
extracting risk analysis results correspondingly associated with the standard parameter sets matched with the risk analysis parameter sets;
analyzing the risk analysis result and determining a risk value.
10. The pancreatic cancer recognition model construction system based on the convolutional neural network according to claim 6, wherein the training module trains the initial convolutional neural network model based on the third image, and after the training converges, obtains the pancreatic cancer recognition model, and performs the following operations:
extracting the characteristics of each third image, and constructing a characteristic parameter set corresponding to the third image based on the extracted characteristic values;
grouping each third image based on the characteristic parameter set of each third image to obtain a plurality of groups;
determining the packet number corresponding to each packet based on a preset packet number library;
constructing a data verification vector based on the group numbers and the image numbers corresponding to the groups;
verifying based on a preset data verification library and a data verification vector;
when the verification fails, the first image, the second image and the third image are acquired again;
when the verification is passed, dividing each group into preset parts of subgroups; constructing each training set and each testing set based on different subgroups of different groups;
based on the training set, training the initial convolutional neural network model, testing the network model through the testing set after training convergence, and obtaining the pancreatic cancer recognition model after the testing set passes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410003838.XA CN117893875A (en) | 2024-01-03 | 2024-01-03 | Construction method and system of pancreatic cancer recognition model based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410003838.XA CN117893875A (en) | 2024-01-03 | 2024-01-03 | Construction method and system of pancreatic cancer recognition model based on convolutional neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117893875A true CN117893875A (en) | 2024-04-16 |
Family
ID=90640427
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410003838.XA Pending CN117893875A (en) | 2024-01-03 | 2024-01-03 | Construction method and system of pancreatic cancer recognition model based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117893875A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107369151A (en) * | 2017-06-07 | 2017-11-21 | 万香波 | System and method are supported in GISTs pathological diagnosis based on big data deep learning |
CN113538435A (en) * | 2021-09-17 | 2021-10-22 | 北京航空航天大学 | Pancreatic cancer pathological image classification method and system based on deep learning |
US20220270244A1 (en) * | 2019-07-19 | 2022-08-25 | The Jackson Laboratory | Convolutional neural networks for classification of cancer histological images |
CN115954072A (en) * | 2023-01-09 | 2023-04-11 | 杭州数垚科技有限公司 | Intelligent clinical test scheme generation method and related device |
CN116416225A (en) * | 2023-03-21 | 2023-07-11 | 上海交通大学 | Pancreatic cancer diagnosis method, pancreatic cancer diagnosis system, pancreatic cancer medium and pancreatic cancer electronic device |
-
2024
- 2024-01-03 CN CN202410003838.XA patent/CN117893875A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107369151A (en) * | 2017-06-07 | 2017-11-21 | 万香波 | System and method are supported in GISTs pathological diagnosis based on big data deep learning |
US20220270244A1 (en) * | 2019-07-19 | 2022-08-25 | The Jackson Laboratory | Convolutional neural networks for classification of cancer histological images |
CN113538435A (en) * | 2021-09-17 | 2021-10-22 | 北京航空航天大学 | Pancreatic cancer pathological image classification method and system based on deep learning |
CN115954072A (en) * | 2023-01-09 | 2023-04-11 | 杭州数垚科技有限公司 | Intelligent clinical test scheme generation method and related device |
CN116416225A (en) * | 2023-03-21 | 2023-07-11 | 上海交通大学 | Pancreatic cancer diagnosis method, pancreatic cancer diagnosis system, pancreatic cancer medium and pancreatic cancer electronic device |
Non-Patent Citations (1)
Title |
---|
唐思源, 杨敏, 刘燕茹: "基于卷积神经网络的胰腺癌检测方法的研究", 《软件》, vol. 41, no. 5, 31 May 2020 (2020-05-31), pages 87 - 90 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110909780B (en) | Image recognition model training and image recognition method, device and system | |
CN108446730B (en) | CT pulmonary nodule detection device based on deep learning | |
CN110473192B (en) | Digestive tract endoscope image recognition model training and recognition method, device and system | |
CN110647875B (en) | Method for segmenting and identifying model structure of blood cells and blood cell identification method | |
CN109543526B (en) | True and false facial paralysis recognition system based on depth difference characteristics | |
CN113011485B (en) | Multi-mode multi-disease long-tail distribution ophthalmic disease classification model training method and device | |
CN111488921B (en) | Intelligent analysis system and method for panoramic digital pathological image | |
CN109544518B (en) | Method and system applied to bone maturity assessment | |
CN110647874B (en) | End-to-end blood cell identification model construction method and application | |
CN107680088A (en) | Method and apparatus for analyzing medical image | |
CN112949786A (en) | Data classification identification method, device, equipment and readable storage medium | |
CN107368859A (en) | Training method, verification method and the lesion pattern recognition device of lesion identification model | |
CN112581438B (en) | Slice image recognition method and device, storage medium and electronic equipment | |
CN110335668A (en) | Thyroid cancer cell pathological map auxiliary analysis method and system based on deep learning | |
CN110729045A (en) | Tongue image segmentation method based on context-aware residual error network | |
CN110731773A (en) | abnormal electrocardiogram screening method based on fusion of global and local depth features of electrocardiogram | |
Tobias et al. | CNN-based deep learning model for chest X-ray health classification using tensorflow | |
CN110415815A (en) | The hereditary disease assistant diagnosis system of deep learning and face biological information | |
CN111462082A (en) | Focus picture recognition device, method and equipment and readable storage medium | |
CN113903082A (en) | Human body gait monitoring algorithm based on dynamic time planning | |
CN110503636B (en) | Parameter adjustment method, focus prediction method, parameter adjustment device and electronic equipment | |
CN117237351B (en) | Ultrasonic image analysis method and related device | |
CN110175588A (en) | A kind of few sample face expression recognition method and system based on meta learning | |
CN112263220A (en) | Endocrine disease intelligent diagnosis system | |
Tobias et al. | Android Application for Chest X-ray Health Classification From a CNN Deep Learning TensorFlow Model |
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 |