CN113538349A - Small sample diabetic retinopathy classification system based on model independent learning - Google Patents
Small sample diabetic retinopathy classification system based on model independent learning Download PDFInfo
- Publication number
- CN113538349A CN113538349A CN202110730369.8A CN202110730369A CN113538349A CN 113538349 A CN113538349 A CN 113538349A CN 202110730369 A CN202110730369 A CN 202110730369A CN 113538349 A CN113538349 A CN 113538349A
- Authority
- CN
- China
- Prior art keywords
- meta
- diabetic retinopathy
- training
- model
- testing
- 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
- 206010012689 Diabetic retinopathy Diseases 0.000 title claims abstract description 94
- 238000012549 training Methods 0.000 claims abstract description 89
- 238000013145 classification model Methods 0.000 claims abstract description 49
- 230000006870 function Effects 0.000 claims abstract description 28
- 238000000034 method Methods 0.000 claims abstract description 28
- 230000008447 perception Effects 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims description 75
- 238000013527 convolutional neural network Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 6
- 201000007917 background diabetic retinopathy Diseases 0.000 description 4
- 230000003902 lesion Effects 0.000 description 4
- 201000007914 proliferative diabetic retinopathy Diseases 0.000 description 4
- 201000003772 severe nonproliferative diabetic retinopathy Diseases 0.000 description 4
- 238000005457 optimization Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 201000004569 Blindness Diseases 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 230000007170 pathology Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004256 retinal image Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N3/045—Combinations of 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a small sample diabetic retinopathy classification system based on model independent meta learning, which comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is configured to: acquiring a diabetic retinopathy image to be classified; an image classification module configured to: classifying the diabetic retinopathy image to be classified based on a small sample diabetic retinopathy classification model; the classification model is obtained by training through a model-independent meta-learning method, and a loss function in the model is a difficult task perception loss function. The method effectively reduces the dependence of the model on the number of samples and improves the classification accuracy of the small-sample diabetic retinopathy.
Description
Technical Field
The invention belongs to the field of machine learning, and particularly relates to a small sample diabetic retinopathy classification system based on a difficult task perception model independent learning.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Diabetic retinopathy is one of the leading causes of blindness. The classification of the color fundus images has important significance for preventing and treating diabetic retinopathy. The effective computer-aided classification technology can greatly save the diagnosis time of ophthalmologists and improve the efficiency and the accuracy of the classification of diabetic retinopathy.
In recent years, the research of diabetic retinopathy classification based on deep learning has been greatly advanced. However, deep learning approaches tend to require large amounts of training data to train the model. Tagged data is often limited in real life, and for diabetic retinopathy, it is very difficult to collect a large number of images of various degrees of pathology. Further, due to interference of factors such as light brightness in image formation, a lesion in a retinal image may be less greatly distinguished from a background, and it is more difficult to classify the degree of the lesion.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method for classifying the small-sample diabetic retinopathy based on the independent learning of the difficult task perception model is provided, and high accuracy is achieved in a new type of diabetic retinopathy classification task through optimization of the small-sample diabetic retinopathy classification model.
In order to achieve the above object, one or more embodiments of the present invention provide the following technical solutions:
a small sample diabetic retinopathy classification system based on model independent learning, comprising:
a data acquisition module configured to: acquiring a diabetic retinopathy image to be classified;
an image classification module configured to: classifying the diabetic retinopathy image to be classified based on a small sample diabetic retinopathy classification model;
the classification model is obtained by training through a model-independent meta-learning method, and a loss function in the model is a difficult task perception loss function:
where μ is the adjustment factor and σ is satisfiedIs the smallest positive integer of (a) to (b),is a cross entropy loss function.
Further, the training method of the small sample diabetic retinopathy classification model comprises the following steps:
acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally, a small sample diabetic retinopathy classification model is obtained.
Further, after a meta-training set and a meta-testing set are constructed, data are extracted from the meta-training set and the meta-testing set respectively to construct a meta-training task and a meta-testing task;
performing meta-training on a pre-constructed convolutional neural network by adopting a plurality of meta-training tasks to obtain a small sample diabetic retinopathy classification model;
and testing the classification model by adopting a plurality of element testing tasks and adjusting parameters.
Further, the meta-test task comprises a support set and a query set, fine tuning is carried out on the support set of the plurality of meta-test tasks, and testing is carried out on the query set.
Further, the method comprises the following steps:
acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally obtaining a small sample diabetic retinopathy classification model;
the loss function in the model adopts a difficult task perception loss function:
where μ is the adjustment factor and σ is satisfiedIs the smallest positive integer of (a) to (b),is a cross entropy loss function.
Further, after a meta-training set and a meta-testing set are constructed, data are extracted from the meta-training set and the meta-testing set respectively to construct a meta-training task and a meta-testing task;
performing meta-training on a pre-constructed convolutional neural network by adopting a plurality of meta-training tasks to obtain a small sample diabetic retinopathy classification model;
and testing the classification model by adopting a plurality of element testing tasks and adjusting parameters.
Further, the meta-test task comprises a support set and a query set, fine tuning is carried out on the support set of the plurality of meta-test tasks, and testing is carried out on the query set.
One or more embodiments provide a small sample diabetic retinopathy classification model training system based on model independent meta learning, comprising:
a dataset acquisition module configured to: acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
a meta-learning module configured to: performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally obtaining a small sample diabetic retinopathy classification model;
the loss function in the model adopts a difficult task perception loss function:
where μ is the adjustment factor and σ is satisfiedIs the smallest positive integer of (a) to (b),is a cross entropy loss function.
One or more embodiments provide an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the classification model training method when executing the program.
One or more embodiments provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the classification model training method described herein.
One or more technical schemes have the following technical effects:
the model obtains priori knowledge through training on a meta-training task, so that the small-sample diabetic retinopathy classification model can be quickly adapted and has a good classification effect when facing a new small-sample classification task;
model independent learning is applied to solve the classification problem of the small sample diabetic retinopathy, and the influence caused by insufficient sample data is relieved;
the difficult task perception is provided to optimize the meta-training process, the meta-learner can replay larger weight on the difficult task by adjusting the loss function, and the rapid adaptability and the classification accuracy of the small-sample diabetic retinopathy classification model under limited data are improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a schematic diagram of a small sample diabetic retinopathy classification system based on a difficult task perception model independent of meta-learning in accordance with one or more embodiments of the present invention;
FIG. 2 is an architecture diagram of a small sample diabetic retinopathy classification system based on model independent meta learning in accordance with one or more embodiments of the present invention;
FIG. 3 is a flowchart of a method for training a small sample diabetic retinopathy classification model based on a difficult task perception model independent of meta-learning in one or more embodiments of the invention.
Detailed Description
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
The meta-learning can well solve the problem of small sample classification. Its essence is to learn a priori knowledge from a plurality of similar tasks and adapt quickly to the new task. Model independent meta-learning, one of the most classical meta-learning methods, aims at optimizing the model. The core idea of model independent element learning is that the optimal initialization parameters of the model are obtained through gradient descent in the element training process, and when the method is applied to a new task with limited data, the optimal performance can be achieved through gradient updating in several steps. In addition, the model independent meta-learning algorithm has model independence and universality and is very suitable for classifying the small-sample diabetic retinopathy.
Example one
The embodiment provides a diabetic retinopathy classification system based on model independent meta-learning, which utilizes a small-sample diabetic retinopathy classification model to acquire priori knowledge through learning of a plurality of similar tasks, and utilizes the priori knowledge to realize rapid and accurate small-sample diabetic retinopathy classification in a meta-testing stage.
With reference to fig. 1 and 2, the system specifically includes:
a data acquisition module configured to: acquiring a diabetic retinopathy image to be classified;
an image classification module configured to: classifying the diabetic retinopathy image to be classified based on a small sample diabetic retinopathy classification model;
the small sample diabetic retinopathy classification model is obtained based on a model training module, and the model training module specifically comprises:
a dataset acquisition module configured to: acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories, wherein the meta-training set and the meta-testing set are respectively used for meta-training and meta-testing;
the embodiment acquires various categories of diabetic retinopathy data based on an open diabetic retinopathy detection dataset provided by the asia-pacific remote ophthalmology society, and specifically includes five categories: healthy, mild non-proliferative diabetic retinopathy, moderate non-proliferative diabetic retinopathy, severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy. The number of these five categories of color fundus images is 1805, 370, 999, 193, and 295, respectively.
The meta-training set is three pathological changes of healthy, mild non-proliferative diabetic retinopathy and moderate non-proliferative diabetic retinopathy and is used for meta-training; the meta-test set is used for the meta-test of two lesion categories of severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy.
A model independent meta-learning module configured to include:
a meta-task building submodule configured to: extracting data from the meta-training set and the meta-testing set respectively, and constructing a meta-training task and a meta-testing task, wherein each meta-training task or meta-testing task comprises a support set and a query set;
the meta training task and the meta testing task are respectively composed of two categories of a meta training set and a meta testing set which are randomly sampled. In this embodiment, a support set is set to K samples of each category, a query set is set to 20 samples of each category, and K is set to 3 and 5 in the N-way K-shot manner according to the 2-way K-shot manner.
A model training submodule configured to: performing meta-training on a pre-constructed convolutional neural network by adopting a plurality of meta-training tasks to obtain a small sample diabetic retinopathy classification model;
in this embodiment, in order to obtain a small sample diabetic retinopathy classification model by training, a convolutional neural network model is pre-constructed, which includes 4 convolutional layers, a maximum pooling layer, a full-link layer, and an output layer, and hyper-parameters of the model, including an inner-loop learning rate, an outer-loop learning rate, a batch size, gradient update steps, an optimizer, iteration times, and the like, are set; specifically, the inner loop learning rate is set to 0.001, the outer loop learning rate is 0.01, the batch size is 3, the gradient updating step number is 5, Adam optimization is adopted, and the iteration number is 6000. The network structure parameters are shown in table 1.
TABLE 1 network architecture parameters
And performing gradient descent on the support set and the query set of the meta-training task to update the parameters of the specific task.
The small sample diabetic retinopathy classification model carries out inner loop updating on the support set in the meta-training stage, and the specific expression is as follows:
initial parameter theta of model in formula in meta-training task TiUpdating the parameters toWhere α is the inner loop learning rate, ε denotes gradient descent ε times, ε ∈ [0, m],Is a function of the loss as a function of,is a gradient, wherein the loss function is a cross entropy loss function, and the expression is:
in the formula, x(j)Is a color fundus image, y(j)Is the corresponding category.
In general, model independent meta-learning is treated equally to randomly sampled meta-tasks, and difficult tasks are always difficult to learn. The difficult task perception method can improve the effectiveness of the meta-training stage by changing the weight of the difficult task. In this embodiment, a loss function is adjusted, a difficult task perception function is constructed, and an expression is as follows:
The updated meta-learning parameter expression is as follows:
where β is the outer loop learning rate, and the meta learning parameter θ can gradually approach the optimal initialization parameter after several outer loops.
A model test submodule configured to: and carrying out fine adjustment on a support set of the meta-test task, and carrying out test on an inquiry set, thereby finally realizing the classification of the small sample diabetic retinopathy irrelevant to meta-learning based on a difficult task perception model. Specifically, the small-sample diabetic retinopathy classification model finely adjusts a support set on a meta-test task including severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy by using optimal initialization parameters, and completes small-sample diabetic retinopathy classification independent of meta-learning based on a difficult task perception model on a query set. Referring to the table 2, under the 2-way 3-shot experiment and the 2-way 5-shot experiment, the highest accuracy rate of 0.713 and 0.761 can be achieved, and the classification of the new type of small sample diabetic retinopathy is realized.
TABLE 2 comparison of accuracy of classification models for small samples of diabetic retinopathy under different parameters
The small sample diabetic retinopathy classification model finely adjusts a support set on a meta-test task containing severe non-proliferative diabetic retinopathy and proliferative diabetic retinopathy by using optimal initialization parameters, and completes small sample diabetic retinopathy classification based on the difficulty task perception model independent meta-learning on an inquiry set.
Example two
The present embodiment aims to provide a method for training a small sample diabetic retinopathy classification model based on model independent meta learning, as shown in fig. 3, comprising the following steps:
acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally, a small sample diabetic retinopathy classification model is obtained.
EXAMPLE III
The present embodiment aims to provide a small sample diabetic retinopathy classification model training system based on model independent meta learning, which includes:
a dataset acquisition module configured to: acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
a meta-learning module configured to: performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally, a small sample diabetic retinopathy classification model is obtained.
Example four
The embodiment aims at providing an electronic device.
An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method according to embodiment two when executing the program.
EXAMPLE five
An object of the present embodiment is to provide a computer-readable storage medium.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to embodiment two.
The steps or modules related to the second to fifth embodiments correspond to those of the first embodiment, and the detailed description thereof can be found in the relevant description of the first embodiment. The term "computer-readable storage medium" should be taken to include a single medium or multiple media containing one or more sets of instructions; it should also be understood to include any medium that is capable of storing, encoding or carrying a set of instructions for execution by a processor and that cause the processor to perform any of the methods of the present invention.
In summary, the present invention employs the public diabetes visual network lesion detection data set provided by the asia-pacific remote ophthalmology society to construct the meta-training task and the meta-testing task for the meta-training and the meta-testing, respectively. The method constructs the small-sample diabetic retinopathy classification model under the framework of model independent meta-learning, obtains the priori knowledge about classification tasks while constructing the initialization parameters of the difficult task perception function optimization model in the meta-training process of the model, and can realize the rapid and accurate classification of the new class of the small-sample diabetic retinopathy on a small number of samples by using the priori knowledge in the meta-testing stage. The method effectively reduces the dependence of the model on the number of samples and improves the classification accuracy of the small-sample diabetic retinopathy.
Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using general purpose computer means, or alternatively, they can be implemented using program code that is executable by computing means, such that they are stored in memory means for execution by the computing means, or they are separately fabricated into individual integrated circuit modules, or multiple modules or steps of them are fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A small sample diabetic retinopathy classification system based on model independent learning, comprising:
a data acquisition module configured to: acquiring a diabetic retinopathy image to be classified;
an image classification module configured to: classifying the diabetic retinopathy image to be classified based on a small sample diabetic retinopathy classification model;
the classification model is obtained by training through a model-independent meta-learning method, and a loss function in the model is a difficult task perception loss function:
2. The small-sample diabetic retinopathy classification system of claim 1 wherein the small-sample diabetic retinopathy classification model training method is:
acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally, a small sample diabetic retinopathy classification model is obtained.
3. The small-sample diabetic retinopathy classification system of claim 2, wherein after the meta training set and the meta test set are constructed, data are further extracted from the meta training set and the meta test set, respectively, to construct a meta training task and a meta test task;
performing meta-training on a pre-constructed convolutional neural network by adopting a plurality of meta-training tasks to obtain a small sample diabetic retinopathy classification model;
and testing the classification model by adopting a plurality of element testing tasks and adjusting parameters.
4. The small-sample diabetic retinopathy classification system of claim 3 wherein the meta-test tasks include a support set and a query set, fine-tuning on the support set of the plurality of meta-test tasks, and testing on the query set.
5. A small sample diabetic retinopathy classification model training method based on model independent learning is characterized by comprising the following steps:
acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally obtaining a small sample diabetic retinopathy classification model;
the loss function in the model adopts a difficult task perception loss function:
6. The small-sample diabetic retinopathy classification model training method of claim 5, wherein after the meta-training set and the meta-testing set are constructed, data are extracted from the meta-training set and the meta-testing set respectively to construct a meta-training task and a meta-testing task;
performing meta-training on a pre-constructed convolutional neural network by adopting a plurality of meta-training tasks to obtain a small sample diabetic retinopathy classification model;
and testing the classification model by adopting a plurality of element testing tasks and adjusting parameters.
7. The small sample diabetic retinopathy classification model training method of claim 6 wherein the meta-test tasks include a support set and a query set, and wherein the fine-tuning is performed on the support set of the plurality of meta-test tasks and the testing is performed on the query set.
8. A small sample diabetic retinopathy classification model training system based on model independent meta-learning is characterized by comprising:
a dataset acquisition module configured to: acquiring diabetic retinopathy data comprising multiple categories, constructing a meta-training set based on the data of one of the categories, and constructing a meta-testing set based on the data of the other categories;
a meta-learning module configured to: performing element training on a pre-constructed convolutional neural network model by adopting an element training set; testing the model by adopting a meta-test set, and adjusting parameters; finally obtaining a small sample diabetic retinopathy classification model;
the loss function in the model adopts a difficult task perception loss function:
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the classification model training method according to any one of claims 5 to 8 when executing the program.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the classification model training method according to any one of claims 5 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110730369.8A CN113538349A (en) | 2021-06-29 | 2021-06-29 | Small sample diabetic retinopathy classification system based on model independent learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110730369.8A CN113538349A (en) | 2021-06-29 | 2021-06-29 | Small sample diabetic retinopathy classification system based on model independent learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113538349A true CN113538349A (en) | 2021-10-22 |
Family
ID=78097170
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110730369.8A Pending CN113538349A (en) | 2021-06-29 | 2021-06-29 | Small sample diabetic retinopathy classification system based on model independent learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113538349A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190147298A1 (en) * | 2017-11-14 | 2019-05-16 | Magic Leap, Inc. | Meta-learning for multi-task learning for neural networks |
CN110210570A (en) * | 2019-06-10 | 2019-09-06 | 上海延华大数据科技有限公司 | The more classification methods of diabetic retinopathy image based on deep learning |
CN111476292A (en) * | 2020-04-03 | 2020-07-31 | 北京全景德康医学影像诊断中心有限公司 | Small sample element learning training method for medical image classification processing artificial intelligence |
CN112861995A (en) * | 2021-03-15 | 2021-05-28 | 中山大学 | Unsupervised few-sample image classification method and system based on model independent meta learning and storage medium |
-
2021
- 2021-06-29 CN CN202110730369.8A patent/CN113538349A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190147298A1 (en) * | 2017-11-14 | 2019-05-16 | Magic Leap, Inc. | Meta-learning for multi-task learning for neural networks |
CN110210570A (en) * | 2019-06-10 | 2019-09-06 | 上海延华大数据科技有限公司 | The more classification methods of diabetic retinopathy image based on deep learning |
CN111476292A (en) * | 2020-04-03 | 2020-07-31 | 北京全景德康医学影像诊断中心有限公司 | Small sample element learning training method for medical image classification processing artificial intelligence |
CN112861995A (en) * | 2021-03-15 | 2021-05-28 | 中山大学 | Unsupervised few-sample image classification method and system based on model independent meta learning and storage medium |
Non-Patent Citations (1)
Title |
---|
董雪媛: "基于深度元学习的糖尿病视网膜病变自动分类模型研究", 《中国优秀硕士学位论文全文数据库医药卫生科技辑 》, 15 February 2023 (2023-02-15) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10460231B2 (en) | Method and apparatus of neural network based image signal processor | |
CN105844285B (en) | Cucumber disease recognition methods and device based on image information | |
CN109934089B (en) | Automatic multi-stage epilepsia electroencephalogram signal identification method based on supervised gradient raiser | |
US20160379352A1 (en) | Label-free non-reference image quality assessment via deep neural network | |
Chen et al. | The fruit classification algorithm based on the multi-optimization convolutional neural network | |
CN106779064A (en) | Deep neural network self-training method based on data characteristics | |
CN107563406B (en) | Image fine classification method for autonomous learning | |
EP3620982B1 (en) | Sample processing method and device | |
CN113723451A (en) | Retinal image classification model training method, system, storage medium and device | |
CN111179216A (en) | Crop disease identification method based on image processing and convolutional neural network | |
CN111127423B (en) | Rice pest and disease identification method based on CNN-BP neural network algorithm | |
CN106503047B (en) | A kind of image crawler optimization method based on convolutional neural networks | |
CN111626969B (en) | Corn disease image processing method based on attention mechanism | |
Zhang et al. | Feature-transfer network and local background suppression for microaneurysm detection | |
CN110096976A (en) | Human behavior micro-Doppler classification method based on sparse migration network | |
US20220253751A1 (en) | Human identification method based on expert feedback mechanism | |
CN110991554B (en) | Improved PCA (principal component analysis) -based deep network image classification method | |
Stuchi et al. | A frequency-domain approach with learnable filters for image classification | |
Kshirsagar et al. | Recognition of diabetic retinopathy with ground truth segmentation using fundus images and neural network algorithm | |
CN111340829B (en) | Improved DME edema regional neural network segmentation model construction method | |
CN113538349A (en) | Small sample diabetic retinopathy classification system based on model independent learning | |
CN109993076A (en) | A kind of white mouse behavior classification method based on deep learning | |
CN116342919A (en) | Rice disease identification method based on attention and gating mechanism | |
CN115762578A (en) | Interpretable heart sound abnormity identification method and system based on fractional domain Fourier transform | |
CN113902672A (en) | Rare skin lesion classification system based on space transformation optimization element learning |
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 |