CN116824207B - Multidimensional pathological image classification and early warning method based on reinforcement learning mode - Google Patents
Multidimensional pathological image classification and early warning method based on reinforcement learning mode Download PDFInfo
- Publication number
- CN116824207B CN116824207B CN202310472529.2A CN202310472529A CN116824207B CN 116824207 B CN116824207 B CN 116824207B CN 202310472529 A CN202310472529 A CN 202310472529A CN 116824207 B CN116824207 B CN 116824207B
- Authority
- CN
- China
- Prior art keywords
- network
- module
- classification
- reinforcement learning
- early warning
- 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.)
- Active
Links
- 230000001575 pathological effect Effects 0.000 title claims abstract description 76
- 230000002787 reinforcement Effects 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000000605 extraction Methods 0.000 claims abstract description 15
- 230000009471 action Effects 0.000 claims description 14
- 239000013598 vector Substances 0.000 claims description 9
- 238000012502 risk assessment Methods 0.000 claims description 8
- 201000010099 disease Diseases 0.000 claims description 7
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 238000011160 research Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims 7
- 238000005728 strengthening Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000003044 adaptive effect Effects 0.000 abstract 2
- 238000013135 deep learning Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006806 disease prevention Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012545 processing Methods 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- 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
- G06N3/047—Probabilistic or stochastic 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
- G06N3/048—Activation functions
-
- 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
- G06N3/092—Reinforcement 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (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)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The multi-dimensional pathological image classification and early warning method based on the reinforcement learning mode comprises a multi-dimensional feature extraction module, a reinforcement learning classification module, an early warning module and an adaptive feature selection module, wherein the multi-dimensional feature extraction module is used for extracting multi-dimensional features from pathological images, the reinforcement learning classification module classifies the pathological images through a deep reinforcement learning network, the early warning module early warns the pathological images based on classification results, and the adaptive feature selection module automatically learns and selects different features in the pathological images. The invention has the beneficial effects that: the multi-dimensional features of the pathological images can be automatically extracted, and the importance of different features is learned, so that more accurate and reliable pathological classification and early warning are realized.
Description
Technical Field
The invention relates to the field of multi-dimensional pathological image classification and early warning, in particular to a multi-dimensional pathological image classification and early warning method based on a reinforcement learning mode.
Background
With the continuous development of medical imaging technology, the scale and complexity of medical image data are continuously increased, and automatic classification and early warning of medical images become an important research field. In multi-dimensional pathological image classification and early warning, how to improve classification accuracy and prediction performance is an important problem. Traditional image classification methods are mainly based on manually designed feature extraction algorithms, which typically require manual selection of features and do not necessarily enable efficient classification of all types of medical images. Furthermore, these methods may be limited by computational resources and time when processing large-scale data sets, and thus a more efficient and accurate method for image classification and prediction is needed.
In recent years, with the continuous development of machine learning and deep learning technologies, methods based on feature learning and end-to-end training have become mainstream. However, these methods typically require a large amount of labeling data, which is often limited due to the complexity and difficulty of data acquisition in the medical image classification and prediction field, which presents challenges to conventional machine learning and deep learning methods.
Reinforcement learning is a machine learning method that enables optimization of decisions through interactive learning. In medical image classification and prediction, reinforcement learning can discover an optimal decision strategy through interactive learning, so that classification accuracy and prediction performance are improved. In addition, reinforcement learning has the advantages of self-adaptability, robustness, universality and the like, and can cope with complexity and uncertainty in multi-dimensional pathological image classification and early warning.
Based on the background and the requirements, the patent provides a multidimensional pathological image classification and early warning method based on a reinforcement learning mode, which mainly comprises a multidimensional feature extraction module, a reinforcement learning classification module, a self-adaptive feature selection module and an early warning module. The method can adaptively select the optimal feature subset, classify and predict the feature subset through the reinforcement learning model, and improve classification accuracy and prediction performance.
Disclosure of Invention
Aiming at the problems, a multidimensional pathological image classification and early warning method based on a reinforcement learning mode is provided.
The aim of the invention is realized by the following technical scheme: the invention provides a multi-dimensional pathological image classification and early warning method based on a reinforcement learning mode, which comprises a multi-dimensional feature extraction module, a reinforcement learning classification module, an early warning module and a self-adaptive feature selection module, wherein the multi-dimensional feature extraction module is used for extracting multi-dimensional features from pathological images, converting information in the pathological images into feature vectors which can be processed by a deep reinforcement learning network, the reinforcement learning classification module classifies the pathological images through the deep reinforcement learning network, adopts an Actor-Critic algorithm to classify the multi-dimensional pathological images, realizes quick classification and accurate judgment of the pathological images by learning the relation and rules among the multi-dimensional features, the early warning module early warns the pathological images based on classification results, can timely find abnormal conditions in the pathological images, provides a specific treatment scheme, provides important references for clinicians, and the self-adaptive feature selection module automatically learns and selects different features in the pathological images, flexibly adjusts feature selection strategies according to different pathological image types and clinical requirements, and improves classification accuracy and robustness.
Further, the multidimensional feature extraction module converts information in the pathological image into feature vectors which can be processed by the deep reinforcement learning network, so that the classification and early warning accuracy of the pathological image are improved. Meanwhile, the collected patient disease and medical information data are mapped from the high-dimensional feature space to the low-dimensional feature space through the low-dimensional feature mapping module, so that the data have good separability, and the classification and early warning efficiency is improved.
Furthermore, the reinforcement learning classification module automatically learns an optimal decision strategy on the basis of the feature vectors obtained by the multi-dimensional feature extraction module, so that the accuracy and the efficiency of classification and early warning of the pathological images are improved to the greatest extent.
Further, the Actor-Critic algorithm is adopted to classify the multidimensional pathological images, and the method is as follows:
the Actor-Critic algorithm comprises a strategy network and a value network, and the state set of the Actor-Critic algorithm is recorded asWherein->State of time 1->State of time 2->State of t-th time is indicated,/->Indicate->Policy network for status of time of day->Calculating probability distribution, policy network->Comprises three elements of action, state and strategy network parameters, randomly sampling according to probability distribution>Action at time->To ensure unbiased samples, random decimation is performed>Reinforcement learning environment will provide ∈ ->Status of time->And discount factor->Will->Status of time->Reinforcement learning input, use of policy network +.>Calculating new probability distribution, obtaining new actions randomly>Here->Is a fictitious action, satisfy->=/>Wherein->Is->Is not performed by reinforcement learning network, and +.>Wherein->Is the policy network parameter at the time t, the output of the 2 times value network is +.>And->The invention expresses time difference error by adopting a mode of error square sum>：/>Wherein->For discount factors, the value network is then derived, recorded as +.>：
Wherein,for value network parameters, the value network parameters are scaled by using a time difference algorithm>Further updating:wherein->For learning factors, deriving the strategy network, recorded as +.>：
And further updating strategy network parameters by using a gradient ascent method:wherein->Monte Carlo approximation, which is a strategy gradient overall, in order to further reduce the variance, makes the algorithm converge faster, replacing +.>The obtained updated expected value of the strategy network parameter is unchanged, so that the variance is reduced, and the method is updated as follows:
wherein,for learning factor, here learning factor +.>Is->Independent of each other, and finally, the updated strategy network parameters and value network parameters are brought into a training set of a reinforcement learning Actor-Critic algorithm throughThe activation function outputs.
Further, the early warning module is a module for performing disease early warning and risk assessment according to the classification result of the pathological image and other related information on the basis of the pathological image classification module, and for the research period T, the early warning module is used for performing early warning and risk assessment byThe activation function outputs the reinforcement learning Actor-Critic algorithm to obtain: />Judging whether to alarm by calculating a center distance alarm module between an output value and a normal value:
wherein the method comprises the steps ofFor the order of->The early warning module is used for carrying out risk assessment and illness state prediction on pathological images in a normal value interval, and provides a more comprehensive and accurate clinical decision basis for doctors.
Furthermore, the self-adaptive feature selection module can automatically learn and select the optimal feature subset according to different pathological images so as to fully utilize feature information and improve classification accuracy and prediction performance, and meanwhile, the self-adaptive feature selection module automatically eliminates irrelevant and redundant features, avoids overfitting and noise influence and further improves classification accuracy and prediction.
The invention has the beneficial effects that: the invention adopts SLFE algorithm to extract the characteristic information of multi-dimensional pathological image, the innovation is that manifold similar matrix is utilized to replace traditional manifold matrix to make the matrix bounded and converged, the embedded coordinates of the middle point of low-dimensional space are updated in an incremental mode, and the characteristic value and the characteristic vector are calculated locally and linearly through inner product matrix, in addition, the multi-dimensional pathological image is classified by adopting the Actor-Critic algorithm, the innovation is that the original time difference error is updated, the error square sum is utilized to replace the time difference residual error, the supervision function of the abnormal value of the training set can be played, the value error deviating from the normal interval is amplified and is abandoned by the Actor-Critic training layer, in addition, when the strategy network parameters are updated, the invention uses the time difference error to replace the time difference errorCompared with the traditional deep learning-based method, the method can carry out intelligent adjustment according to actual conditions, can reduce classification errors and can improve classificationThe method has the advantages that the optimal feature subset is selected in a self-adaptive mode, the classification accuracy and the prediction performance are further improved, the method has wide application prospects in medical image classification and prediction, prediction and early warning of pathological images are achieved, abnormal conditions of the pathological images can be found early, and the method has important significance for prevention and treatment of diseases. Compared with the traditional deep learning-based method, the method can more accurately predict the development trend of the pathological image, thereby better guiding the treatment decision of doctors, improving the classification accuracy and the prediction performance, greatly improving the automation degree of pathological image classification and early warning, and simultaneously reducing the cost and time of manual intervention, so as to better guide the treatment and prevention of diseases.
Drawings
The invention will be further described with reference to the accompanying drawings, in which embodiments do not constitute any limitation on the invention, and other drawings can be obtained by one of ordinary skill in the art without undue effort from the following drawings.
Fig. 1 is a schematic diagram of the structure of the present invention.
Detailed Description
The invention will be further described with reference to the following examples.
The invention provides a multi-dimensional pathological image classification and early warning method based on a reinforcement learning mode, which comprises a multi-dimensional feature extraction module, a reinforcement learning classification module, an early warning module and a self-adaptive feature selection module, wherein the multi-dimensional feature extraction module is used for extracting multi-dimensional features from pathological images, converting information in the pathological images into feature vectors which can be processed by a deep reinforcement learning network, the reinforcement learning classification module classifies the pathological images through the deep reinforcement learning network, adopts an Actor-Critic algorithm to classify the multi-dimensional pathological images, realizes quick classification and accurate judgment of the pathological images by learning the relation and rules among the multi-dimensional features, the early warning module early warns the pathological images based on classification results, can timely find abnormal conditions in the pathological images, provides a specific treatment scheme, provides important references for clinicians, and the self-adaptive feature selection module automatically learns and selects different features in the pathological images, flexibly adjusts feature selection strategies according to different pathological image types and clinical requirements, and improves classification accuracy and robustness.
Specifically, the multidimensional feature extraction module converts information in the pathological image into feature vectors which can be processed by the deep reinforcement learning network, so that the classification and early warning accuracy of the pathological image are improved. Meanwhile, the collected patient disease and medical information data are mapped from the high-dimensional feature space to the low-dimensional feature space through the low-dimensional feature mapping module, so that the data have good separability, and the classification and early warning efficiency is improved.
Specifically, the reinforcement learning classification module automatically learns an optimal decision strategy on the basis of the feature vectors obtained by the multidimensional feature extraction module, so that the accuracy and the efficiency of classification and early warning of the pathological images are improved to the greatest extent.
Preferably, the Actor-Critic algorithm is adopted to classify the multidimensional pathological images, and the method is as follows:
the Actor-Critic algorithm comprises a strategy network and a value network, and the state set of the Actor-Critic algorithm is recorded asWherein->State of time 1->State of time 2->State of t-th time is indicated,/->Indicate->Policy network for status of time of day->Calculating probability distribution, policy network->Comprises three elements of action, state and strategy network parameters, randomly sampling according to probability distribution>Action at time->To ensure unbiased samples, random decimation is performed>Reinforcement learning environment will provide ∈ ->Status of time->And discount factor->Will->Status of time->Reinforcement learning input, use of policy network +.>Calculating new probability distribution, obtaining new actions randomly>Here->Is a fictitious action, satisfy->=/>Wherein->Is->Is not performed by reinforcement learning network, and +.>Wherein->Is the policy network parameter at the time t, the output of the 2 times value network is +.>And->The invention expresses time difference error by adopting a mode of error square sum>：/>Wherein->For discount factors, the value network is then derived, recorded as +.>：
Wherein,for value network parameters, the value network parameters are scaled by using a time difference algorithm>Further updating:wherein->For learning factors, deriving the strategy network, recorded as +.>：
And further updating strategy network parameters by using a gradient ascent method:wherein->Monte Carlo approximation, which is a strategy gradient overall, in order to further reduce the variance, makes the algorithm converge faster, replacing +.>The obtained updated expected value of the strategy network parameter is unchanged, so that the variance is reduced, and the method is updated as follows:
wherein,for learning factor, here learning factor +.>Is->Independent of each other, and finally, the updated strategy network parameters and value network parameters are brought into a training set of a reinforcement learning Actor-Critic algorithm throughThe activation function outputs.
Specifically, the early warning module is a module for performing disease early warning and risk assessment according to the classification result of the pathological image and other related information on the basis of the pathological image classification module, and for the research period T, the early warning module is used for performing early warning and risk assessment byThe activation function outputs the reinforcement learning Actor-Critic algorithm to obtain: />Judging whether to alarm by calculating a center distance alarm module between an output value and a normal value:
wherein the method comprises the steps ofFor the order of->The early warning module is used for carrying out risk assessment and illness state prediction on pathological images in a normal value interval, and provides a more comprehensive and accurate clinical decision basis for doctors.
Specifically, the self-adaptive feature selection module can automatically learn and select the optimal feature subset according to different pathological images so as to fully utilize feature information and improve classification accuracy and prediction performance, and meanwhile, the self-adaptive feature selection module automatically eliminates irrelevant and redundant features, avoids overfitting and noise influence and further improves classification accuracy and prediction.
Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the scope of the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications can be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims (3)
1. The multi-dimensional pathological image classification and early warning method based on the reinforcement learning mode comprises a multi-dimensional feature extraction module, a reinforcement learning classification module, an early warning module and a self-adaptive feature selection module, wherein the multi-dimensional feature extraction module is used for extracting multi-dimensional features from pathological images, converting information in the pathological images into feature vectors which can be processed by a deep reinforcement learning network, the reinforcement learning classification module classifies the pathological images through the deep reinforcement learning network, adopts an Actor-Critic algorithm to classify the multi-dimensional pathological images, realizes quick classification and accurate judgment of the pathological images by learning the relation and rules among the multi-dimensional features, the early warning module early warns the pathological images based on classification results, can timely find abnormal conditions in the pathological images, provides a specific treatment scheme, provides important references for clinicians, and the self-adaptive feature selection module automatically learns and selects different features in the pathological images, flexibly adjusts feature selection strategies according to different pathological image types and clinical requirements, and improves classification accuracy and robustness;
the reinforcement learning classification module classifies the multidimensional pathological images by adopting an Actor-Critic algorithm, and specifically comprises the following steps:
the Actor-Critic algorithm comprises a strategy network and a value network, and the state set of the Actor-Critic algorithm is recorded as S= { S 1 ,s 2 ,…,s t ,s t+1 … }, wherein s 1 State s representing time 1 2 State s representing time 2 t Representing the state at time t, s t+1 The state at time t+1 is represented by a policy network pi comprising three elements of action, state and policy network parameters, the action a at time t is randomly sampled according to the probability distribution t ToEnsuring unbiasedness of samples, performing random decimated a t The reinforcement learning environment provides a state s at time t+1 t+1 And discount factor r t State s at time t+1 t+1 Strengthening learning input, calculating new probability distribution by using strategy network pi, and obtaining new actions randomlyHere, theIs a fictitious action, satisfy->Wherein (1)>Is->Is not performed by reinforcement learning network, and +.>Wherein θ t Is the policy network parameter at the time t, the output of the 2 times value network is q respectively t ＝q(s t ,a t ；w t ) And->The invention expresses the time difference error delta by adopting a mode of error square sum t ：/>Wherein gamma is a discount factor, followed by deriving the value network, denoted D w ：
Wherein d w,1 Is the derivation of the value network at time 1, d w,2 Is the derivation of the value network at time 2,
d w,t is the value network at the t moment to conduct derivation, d w,t+1 Deriving the value network at the t+1 moment, wherein w is a value network parameter, and further updating the value network parameter w by using a time difference algorithm: w (w) t+1 ＝w t -αδ t d w,t Wherein alpha is a learning factor, deriving a strategy network, and recording as D θ ：
Wherein d θ,1 Is the derivation of the strategy network at the 1 st moment, d θ,2 Is the derivation of the policy network at time 2,
d θ,t is the derivation of the policy network at the t moment, d θ,t+1 The policy network at the t+1 moment is subjected to derivation, and the parameters of the policy network are further updated by using a gradient ascent method: θ t+1 ＝θ t +βq t d θ,t Wherein q is t d θ,t Monte Carlo approximation, which is a strategic gradient overall, allows the algorithm to converge faster in order to further reduce variance by using time differential error instead of q t The obtained updated expected value of the strategy network parameter is unchanged, so that the variance is reduced, and the method is updated as follows:
wherein, beta is a learning factor, the learning factor beta and the learning factor alpha are mutually independent, and finally, the updated strategy network parameters and value network parameters are brought into a training set of a reinforcement learning Actor-Critic algorithm and output through a softmax activation function; the early warning module is a pathological image classification moduleBased on the blocks, according to the classification result of the pathological images and other related information, the module for carrying out disease early warning and risk assessment outputs a reinforcement learning Actor-Critic algorithm through a softmax activation function for a research period T, so as to obtain: y=softmax (a T ,s T ,w T ,θ T ) Wherein a is T The action at the moment T is an action at the moment T,
s T is in the state of T moment, w T Value network parameter theta at time T T For the policy network parameter at the moment T, judging whether to alarm or not by calculating a center distance alarm module between an output value and a normal value:
wherein E () is mathematical expectation, k is order, common is normal value interval, and the early warning module is used for carrying out risk assessment and illness state prediction on pathological images, so as to provide more comprehensive and accurate clinical decision basis for doctors.
2. The reinforcement learning mode-based multidimensional pathological image classification and early warning method according to claim 1, wherein the multidimensional feature extraction module converts information in pathological images into feature vectors which can be processed by a deep reinforcement learning network so as to improve the accuracy of classification and early warning of the pathological images; meanwhile, the collected patient disease and medical information data are mapped from the high-dimensional feature space to the low-dimensional feature space through the low-dimensional feature mapping module, so that the data have good separability, and the classification and early warning efficiency is improved.
3. The multi-dimensional pathological image classification and early warning method based on the reinforcement learning mode according to claim 1, wherein the self-adaptive feature selection module can automatically learn and select the optimal feature subset according to different pathological images so as to fully utilize feature information and improve classification accuracy and prediction performance, and simultaneously, the self-adaptive feature selection module automatically eliminates irrelevant and redundant features, avoids overfitting and noise influence and further improves classification accuracy and predictability.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310472529.2A CN116824207B (en) | 2023-04-27 | 2023-04-27 | Multidimensional pathological image classification and early warning method based on reinforcement learning mode |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310472529.2A CN116824207B (en) | 2023-04-27 | 2023-04-27 | Multidimensional pathological image classification and early warning method based on reinforcement learning mode |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116824207A CN116824207A (en) | 2023-09-29 |
| CN116824207B true CN116824207B (en) | 2024-04-12 |
Family
ID=88119285
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310472529.2A Active CN116824207B (en) | 2023-04-27 | 2023-04-27 | Multidimensional pathological image classification and early warning method based on reinforcement learning mode |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116824207B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110349027A (en) * | 2019-07-19 | 2019-10-18 | 湘潭大学 | Pairs trade system based on deeply study |
| CN113553442A (en) * | 2021-07-13 | 2021-10-26 | 西安邮电大学 | Unsupervised event knowledge graph construction method and system |
| CN114491525A (en) * | 2021-12-16 | 2022-05-13 | 四川大学 | Android malicious software detection feature extraction method based on deep reinforcement learning |
| WO2022221979A1 (en) * | 2021-04-19 | 2022-10-27 | 华为技术有限公司 | Automated driving scenario generation method, apparatus, and system |
| CN115496201A (en) * | 2022-09-06 | 2022-12-20 | 上海电气泰雷兹交通自动化***有限公司 | Train accurate parking control method based on deep reinforcement learning |
| CN115840240A (en) * | 2022-11-29 | 2023-03-24 | 广东工业大学 | Automatic driving positioning method and system based on LSTM deep reinforcement learning |
-
2023
- 2023-04-27 CN CN202310472529.2A patent/CN116824207B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110349027A (en) * | 2019-07-19 | 2019-10-18 | 湘潭大学 | Pairs trade system based on deeply study |
| WO2022221979A1 (en) * | 2021-04-19 | 2022-10-27 | 华为技术有限公司 | Automated driving scenario generation method, apparatus, and system |
| CN113553442A (en) * | 2021-07-13 | 2021-10-26 | 西安邮电大学 | Unsupervised event knowledge graph construction method and system |
| CN114491525A (en) * | 2021-12-16 | 2022-05-13 | 四川大学 | Android malicious software detection feature extraction method based on deep reinforcement learning |
| CN115496201A (en) * | 2022-09-06 | 2022-12-20 | 上海电气泰雷兹交通自动化***有限公司 | Train accurate parking control method based on deep reinforcement learning |
| CN115840240A (en) * | 2022-11-29 | 2023-03-24 | 广东工业大学 | Automatic driving positioning method and system based on LSTM deep reinforcement learning |
Non-Patent Citations (1)
| Title |
|---|
| 基于CT图像的肺部GGO检测方法与***;周姣;《中国优秀硕士学位论文全文数据库医药卫生科技辑》;论文摘要、第13页、第51-52页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116824207A (en) | 2023-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110930418B (en) | Retina blood vessel segmentation method fusing W-net and conditional generation confrontation network | |
| Luo et al. | Retinal image classification by self-supervised fuzzy clustering network | |
| CN114898121B (en) | Automatic generation method for concrete dam defect image description based on graph attention network | |
| CN109583331B (en) | Deep learning-based accurate positioning method for positions of wrist vein and mouth of person | |
| CN110263804B (en) | Medical image segmentation method based on safe semi-supervised clustering | |
| CN112907604A (en) | Self-adaptive super-pixel FCM (pixel-frequency modulation) method for fundus velveteen speckle image segmentation | |
| CN111798439A (en) | Medical image quality interpretation method and system for online and offline fusion and storage medium | |
| CN117132849A (en) | Cerebral apoplexy hemorrhage transformation prediction method based on CT flat-scan image and graph neural network | |
| CN110929888B (en) | Grain crop disease prediction method based on random compact Gaussian kernel function | |
| CN116824207B (en) | Multidimensional pathological image classification and early warning method based on reinforcement learning mode | |
| CN116350231A (en) | Semi-supervised electrocardiogram myocardial infarction positioning method based on hierarchical different granularity class constraint | |
| CN112785559B (en) | Bone age prediction method based on deep learning and formed by mutually combining multiple heterogeneous models | |
| CN111631683B (en) | Electrocardiogram data or electroencephalogram data processing and detecting method and storage medium | |
| CN113781490A (en) | Fundus blood vessel image segmentation method based on self-supervision learning | |
| CN113380417A (en) | LR-N based cardiovascular disease prediction method | |
| Jestin et al. | Genetic algorithm for retinal image analysis | |
| CN111414286A (en) | Fault diagnosis method for unbalanced hard disk data based on deep learning | |
| CN116741372B (en) | Auxiliary diagnosis system and device based on double-branch characterization consistency loss | |
| CN117951632B (en) | PU contrast learning anomaly detection method and system based on multi-mode prototype network | |
| CN118072976B (en) | System and method for predicting respiratory tract diseases of children based on data analysis | |
| TWI845274B (en) | A cell quality prediction system and method, as well as an expert knowledge parameterization method. | |
| CN117314755B (en) | Multi-view plant generation method and device based on cross-modal image generation | |
| CN113408463B (en) | Cell image small sample classification system based on distance measurement | |
| CN112381766B (en) | Pulmonary CT image feature detection system based on small sample learning | |
| CN116012303A (en) | Self-supervision-based organoid abnormality detection method |
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 |