CN114663679A - Blood coagulation index abnormity classification method based on feature fusion meta-learning - Google Patents

Blood coagulation index abnormity classification method based on feature fusion meta-learning Download PDF

Info

Publication number
CN114663679A
CN114663679A CN202210571838.0A CN202210571838A CN114663679A CN 114663679 A CN114663679 A CN 114663679A CN 202210571838 A CN202210571838 A CN 202210571838A CN 114663679 A CN114663679 A CN 114663679A
Authority
CN
China
Prior art keywords
learning
meta
feature fusion
image
coagulation index
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
Application number
CN202210571838.0A
Other languages
Chinese (zh)
Inventor
李登旺
高祝敏
黄浦
陆华
洪亭轩
王醒
李玉玲
周顺风
赵本靖
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Normal University
Original Assignee
Shandong Normal University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Normal University filed Critical Shandong Normal University
Priority to CN202210571838.0A priority Critical patent/CN114663679A/en
Publication of CN114663679A publication Critical patent/CN114663679A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a blood coagulation index abnormity classification method based on feature fusion meta-learning, which comprises the following steps of: acquiring curve images of samples PT and TT, extracting characteristics of the curve images, and creating a training set and a testing set; inputting the training set image into a resnet network, and outputting a processed image; inputting the image to a meta-training model for gradient descent updating of a parameter theta, and constructing an abnormal coagulation index classification model for feature fusion meta-learning; setting a hyper-parameter of an abnormal blood coagulation index classification model; and performing iterative test and fine adjustment on the trained model by using the test set, inputting the test set into the abnormal coagulation index classification model subjected to iterative test and fine adjustment, and classifying by using the abnormal coagulation index classification model subjected to feature fusion element learning. The invention utilizes meta-learning to promote rapid adaptation and generalization based on deep neural networks to identify abnormal indexes with less annotation data and improve classification performance.

Description

Blood coagulation index abnormity classification method based on feature fusion meta-learning
Technical Field
The invention relates to a blood coagulation index abnormity classification method based on feature fusion element learning, and belongs to the technical field of machine learning.
Background
At present, along with the development of medical science, people have more and more profound understanding on the occurrence and development of hemostasis and thrombosis, which are related to the occurrence and development of various diseases in clinic and have close relation with clinical treatment and prognosis. The detection of blood coagulation is the most important application of in vitro diagnosis of thrombosis and hemostasis. In recent years, research results and clinical practice show that prothrombin time PT (prothrombin time) and activated partial thromboplastin time APTT (activated partial thromboplastin time) have more practical significance for the assessment of the blood coagulation function state of a preoperative patient.
The purpose of meta-learning is to realize fast learning, and the key to realizing fast learning lies in the accuracy and rapidness of gradient descent of the neural network model. The neural network is used for learning and predicting the gradient by utilizing the previous task, so that in the face of a new task, the learning is fast as long as the gradient prediction is accurate. But currently there is no method to make the gradient of the neural network model decline faster and more accurate.
Disclosure of Invention
The invention aims to provide a blood coagulation index abnormality classification method based on feature fusion meta-learning, which utilizes the meta-learning to promote rapid adaptation and generalization based on a deep neural network so as to identify abnormal indexes with less annotation data and improve classification performance.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a blood coagulation index abnormity classification method based on feature fusion element learning comprises the following steps:
the method comprises the following steps: acquiring a sample prothrombin time and thrombin time curve image, extracting characteristics of the curve image, converting the curve image into a characteristic vector of 1 multiplied by 39, and creating a training set and a testing set;
step two: inputting a training set image into a residual error network, wherein the residual error network comprises four groups of blocks, each group comprises 3, 4, 6 and 3 blocks, and outputting a processed image;
step three: inputting the image output in the step two to a meta-training model for gradient descent updating of a parameter theta, and constructing an abnormal coagulation index classification model for feature fusion meta-learning;
step four: setting hyper-parameters of the abnormal coagulation index classification model, including meta-learning rate, batch size, gradient updating step number, optimizer and iteration number;
step five: and (3) carrying out fine adjustment and iterative test on the trained model by using the element test set, inputting the test set into the abnormal coagulation index classification model subjected to iterative test and fine adjustment, and realizing classification by using the abnormal coagulation index classification model subjected to feature fusion element learning.
Preferably, the training set creating process is as follows:
performing feature fusion on the sample prothrombin time and thrombin time curve images, comparing the gray values of the feature fused images in 8 directions of the same position, namely upper, lower, left, right, upper left, lower left, upper right and lower right, calculating the average value of the gray values in 8 directions as the gray value of a new image at the position, traversing all points in the images to obtain a feature fusion image, and forming a training set;
the test set creation process is as follows:
calculating the gradient energy sum of the areas of the upper, lower, left and right points around the main line layout of each pixel point of the two images of the prothrombin time and the thrombin time and the five points per se, and taking the gray value of the point with high energy as the gray value of the point of a new image to form a test set.
Preferably, the residual network adopts a residual convolution-50 network structure, and 50 layers have weights, including a convolution layer and a full-link layer, and excluding a pooling layer and a batch normalization layer.
Preferably, the meta-training model is further constraint-trained using the following loss function:
Figure 389170DEST_PATH_IMAGE001
in the formula:
Figure 303906DEST_PATH_IMAGE002
which represents the input of an image, is,
Figure 589393DEST_PATH_IMAGE003
the weight representing the class of the image,
Figure 732930DEST_PATH_IMAGE004
representing the weight of the corresponding class image.
Preferably, the formula for updating the parameter θ in the meta-training model is as follows:
Figure 787474DEST_PATH_IMAGE005
=
Figure 482897DEST_PATH_IMAGE006
⊙exp(γ
Figure 114736DEST_PATH_IMAGE007
⊙
Figure 186597DEST_PATH_IMAGE008
)
Figure 236592DEST_PATH_IMAGE009
=
Figure 837338DEST_PATH_IMAGE010
−η
Figure 213263DEST_PATH_IMAGE005
⊙
Figure 823236DEST_PATH_IMAGE008
in the formula: t denotes time, η denotes a step size, g denotes a gradient, V denotes a non-negative variable, and an exclusive OR operation.
Preferably, the magnitude of the non-negative variable V is equal to the parameter θ.
Preferably, in the hyper-parameters, the meta-learning rate is 0.001, the batch size is 5, the optimizer is an adaptive moment estimation optimizer, the iteration times are 10000 times, and the gradient updating step number is 5.
The invention has the advantages that: the invention uses the small sample element to learn the space conversion, enhances the anti-interference capability of the discriminator, and simultaneously can make the discriminator have the rotation invariance and the scale invariance. In addition, the invention utilizes meta-learning to promote rapid adaptation and generalization based on a deep neural network so as to identify abnormal indexes with less annotation data, improve the classification performance and enable the abnormal indexes to be more accurate and rapid.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention.
FIG. 1 is a schematic diagram of the principle of the abnormal coagulation index classification method model based on feature fusion element learning according to the present invention.
FIG. 2 is a schematic diagram of the multi-feature fusion processing of the abnormal coagulation index classification method based on feature fusion element learning according to the present invention.
FIG. 3 is a schematic diagram of a prothrombin time curve image of the abnormal coagulation index classification method based on feature fusion element learning.
FIG. 4 is a schematic diagram of an image of the activated partial thromboplastin time curve of the abnormal coagulation index classification method based on feature fusion element learning according to the present invention.
FIG. 5 is a schematic diagram of a fibrinogen curve image of the abnormal coagulation index classification method based on feature fusion element learning according to the present invention.
FIG. 6 is a schematic diagram of a thrombin time curve image of the abnormal coagulation index classification method based on feature fusion element learning according to the present invention.
The fig. 3-6 are the input images of fig. 2.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A blood coagulation index abnormity classification method based on feature fusion element learning comprises the following steps:
1) in this test, 500 sets of activated partial thromboplastin time APTT and fibrinogen FIB (fibrinogen) curve images were subjected to feature-level extraction (4 images per set). The images are converted into 1 x 39 feature vectors for creating a training set, a test set. And performing characteristic fusion on the sample prothrombin time PT (prothrombin time) and thrombin time TT (prothrombin time) curve images, namely prothrombin time and thrombin time images to form a training set and a testing set.
The training set creation process is as follows:
performing feature fusion on the sample prothrombin time and thrombin time curve images, comparing the gray values of the feature fused images in 8 directions of the same position, namely upper, lower, left, right, upper left, lower left, upper right and lower right, calculating the average value of the gray values in 8 directions as the gray value of a new image at the position, traversing all points in the images to obtain a feature fusion image, and forming a training set;
the test set creation process is as follows:
calculating the gradient energy sum of the areas of the upper, lower, left and right points around the main line layout of each pixel point of the two images of the prothrombin time and the thrombin time and the five points per se, and taking the gray value of the point with high energy as the gray value of the point of a new image to form a test set.
2) Inputting the training set image into a residual error network, wherein the residual error network comprises four groups of blocks, each group comprises 3, 4, 6 and 3 blocks, and outputting the processed image.
Specifically, the residual network adopts a residual network-50 (ResNet-50) network structure, the pooling layer is defined as average pooling, the network depth is 50, 50 layers with weights are specified, the network comprises a convolution layer and a full connection layer, the pooling layer and a BN layer are not included, and the BN layer is fully called Batch Normalization, namely a Batch Normalization layer. The residual error network and the variant network series thereof are excellent for general image recognition tasks, and specific network structure improvement is required according to specific skin disease characteristic information, such as weight increase, activation function change and the like. But the classification of the abnormal indexes is carried out by considering the gradient problem and the result which can be obtained by the previous training of the training residual error network.
3) And (3) inputting the image output in the step (2) to a meta-training model to perform gradient descent updating on a parameter theta, and constructing an abnormal coagulation index classification model for feature fusion meta-learning.
The meta-training model is constraint-trained using the following loss function:
Figure 586792DEST_PATH_IMAGE001
Figure 499385DEST_PATH_IMAGE011
the parameter theta in the meta-training model is updated according to the formula as follows:
Figure 246761DEST_PATH_IMAGE005
=
Figure 785058DEST_PATH_IMAGE006
⊙exp(γ
Figure 403122DEST_PATH_IMAGE007
⊙
Figure 752195DEST_PATH_IMAGE008
)
Figure 721288DEST_PATH_IMAGE009
=
Figure 797697DEST_PATH_IMAGE010
−η
Figure 801425DEST_PATH_IMAGE005
⊙
Figure 55820DEST_PATH_IMAGE008
in the formula: t denotes time, η denotes step size, g denotes gradient, V denotes a non-negative variable, and an, indicates an OR operation. Calculating adaptive parameters by gradient descent method
Figure 512209DEST_PATH_IMAGE012
When classifying abnormal blood coagulation images in the face of a new task, parameters are updated after several gradient reductions, and model parameters are updated
Figure 392309DEST_PATH_IMAGE012
Become into
Figure 250544DEST_PATH_IMAGE013
. The gradient reduction in the whole model is accelerated by the formula, so that a local minimum value can be obtained in the optimization process by a method of randomly selecting a part of data sets, overfitting can be effectively prevented by using a small part of data sets, the calculated amount in the training process is greatly reduced, and the classification speed and the accuracy can be improved.
Specifically, the formula derivation process is as follows:
the learning rate of the model is adjusted through meta-learning, the gradient of the objective function L is recorded as g from the general gradient decline, and then the updating formula is
Figure 800474DEST_PATH_IMAGE014
=
Figure 619525DEST_PATH_IMAGE015
−
Figure 913103DEST_PATH_IMAGE016
(1)
In order to ensure that each component adjusts the learning rate, a non-negative variable V is introduced with the same size as the parameter, and the formula is updated as follows:
Figure 487829DEST_PATH_IMAGE014
=
Figure 474239DEST_PATH_IMAGE015
−
Figure 780587DEST_PATH_IMAGE017
⊙
Figure 612277DEST_PATH_IMAGE018
(2)
to minimize L, so the update rule for V should also be a gradient descent, using an exponential gradient descent for V, resulting in the following equation:
Figure 179524DEST_PATH_IMAGE019
(3)
from equation (2), at time t
Figure 461470DEST_PATH_IMAGE015
=
Figure 379747DEST_PATH_IMAGE020
−
Figure 890494DEST_PATH_IMAGE021
⊙
Figure 312248DEST_PATH_IMAGE022
So, according to the chain rule, there are:
Figure 765095DEST_PATH_IMAGE023
（4）
substituting into the updated formula (3) of V to obtain
Figure 905089DEST_PATH_IMAGE024
=
Figure 485106DEST_PATH_IMAGE025
(5)
Both gamma and eta represent step length, and the gamma and eta are combined into a parameter gamma, so that the updating formula of the whole model is as follows:
Figure 26946DEST_PATH_IMAGE024
=
Figure 260481DEST_PATH_IMAGE026
(6)
Figure 12406DEST_PATH_IMAGE027
(7)
the learning rate adjusting idea of the system is as follows: if the gradients of two adjacent steps of a certain component are always in the same sign, the accumulation result of the corresponding item is positive, and the learning rate can be properly expanded; if the gradients of two adjacent steps are frequently opposite in sign, the accumulation result of the corresponding terms is likely to be negative, and the learning rate can be reduced appropriately.
If V is initialized to all 1 s, then there will be
Figure 989589DEST_PATH_IMAGE028
(8)
4) Setting hyper-parameters of an abnormal coagulation index classification model, wherein the specific meta-learning rate is 0.001, the batch size is 5, the optimizer is an adaptive moment estimation optimizer, the iteration times are 10000, and the gradient updating step number is 5.
5) And performing iterative test and fine tuning on the trained model by using the test set, and training by using the optimal initialization parameters, the modified network and the data of the model to enable the parameters to adapt to the test data.
And inputting the test set into the abnormal coagulation index classification model subjected to iterative test and fine tuning, and realizing classification by using the abnormal coagulation index classification model subjected to feature fusion element learning.
The proportion of false acceptance, the false rejection rate value and the generalized autoregression of the abnormal coagulation index classification system based on feature fusion meta-learning are shown in the following table:
proportion of false acceptances 0.2%
False rejection rate 0.2%
Generalized autoregression 99.8%

Claims (7)

1. The blood coagulation index abnormity classification method based on feature fusion element learning is characterized by comprising the following steps of:
the method comprises the following steps: acquiring a prothrombin time and thrombin time curve image of a sample, extracting characteristics of the curve image, converting the curve image into a characteristic vector of 1 x 39, and creating a training set and a testing set;
step two: inputting a training set image into a residual error network, wherein the residual error network comprises four groups of blocks, each group comprises 3, 4, 6 and 3 blocks respectively, and outputting a processed image;
step three: inputting the image output in the step two to a meta-training model for gradient descent updating of a parameter theta, and constructing an abnormal coagulation index classification model for feature fusion meta-learning;
step four: setting hyper-parameters of the abnormal coagulation index classification model, including meta-learning rate, batch size, gradient updating step number, optimizer and iteration number;
step five: and (3) carrying out fine adjustment and iterative test on the trained model by using the element test set, inputting the test set into the abnormal coagulation index classification model subjected to iterative test and fine adjustment, and realizing classification by using the abnormal coagulation index classification model subjected to feature fusion element learning.
2. The method for classifying blood coagulation index abnormality based on feature fusion element learning according to claim 1, wherein the training set creation process is as follows:
performing feature fusion on the sample prothrombin time and thrombin time curve images, comparing the gray values of the feature fused images in 8 directions of the same position, namely upper, lower, left, right, upper left, lower left, upper right and lower right, calculating the average value of the gray values in 8 directions as the gray value of a new image at the position, traversing all points in the images to obtain a feature fusion image, and forming a training set;
the test set creation process is as follows:
calculating the gradient energy sum of the areas of the upper, lower, left and right points around the main line layout of each pixel point of the two images of the prothrombin time and the thrombin time and the five points per se, and taking the gray value of the point with high energy as the gray value of the point of a new image to form a test set.
3. The coagulation index abnormality classification method based on feature fusion element learning of claim 1, wherein the residual network adopts a residual convolution-50 network structure, and 50 layers have weights, including a convolution layer and a full connection layer, and excluding a pooling layer and a batch normalization layer.
4. The coagulation index abnormality classification method based on feature fusion meta-learning according to claim 1, wherein the meta-training model is further subjected to constraint training using a loss function as follows:
Figure 61839DEST_PATH_IMAGE001
in the formula:
Figure 731854DEST_PATH_IMAGE002
which represents the input of an image, is,
Figure 971075DEST_PATH_IMAGE003
the weight representing the class of the image,
Figure 751949DEST_PATH_IMAGE004
representing the weight of the corresponding class image.
5. The method for classifying blood coagulation index abnormality based on feature fusion meta-learning according to claim 1, wherein the parameter θ in the meta-training model is updated according to the formula:
Figure 2802DEST_PATH_IMAGE005
=
Figure 719085DEST_PATH_IMAGE006
⊙exp(γ
Figure 320967DEST_PATH_IMAGE007
⊙
Figure 30166DEST_PATH_IMAGE008
)
Figure 135526DEST_PATH_IMAGE009
=
Figure 22710DEST_PATH_IMAGE010
−η
Figure 846310DEST_PATH_IMAGE005
⊙
Figure 373848DEST_PATH_IMAGE008
in the formula: t denotes time, η denotes step size, g denotes gradient, V denotes a non-negative variable, and an, indicates an OR operation.
6. The method for classifying coagulation index abnormality based on feature fusion element learning according to claim 5, wherein the size of the non-negative variable V is equal to a parameter θ.
7. The method for classifying blood coagulation index abnormalities based on feature fusion meta-learning of claim 1, wherein in the hyper-parameters, the meta-learning rate is 0.001, the batch size is 5, the optimizer is an adaptive moment estimation optimizer, the number of iterations is 10000, and the number of gradient update steps is 5.
CN202210571838.0A 2022-05-25 2022-05-25 Blood coagulation index abnormity classification method based on feature fusion meta-learning Pending CN114663679A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210571838.0A CN114663679A (en) 2022-05-25 2022-05-25 Blood coagulation index abnormity classification method based on feature fusion meta-learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210571838.0A CN114663679A (en) 2022-05-25 2022-05-25 Blood coagulation index abnormity classification method based on feature fusion meta-learning

Publications (1)

Publication Number Publication Date
CN114663679A true CN114663679A (en) 2022-06-24

Family

ID=82038311

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210571838.0A Pending CN114663679A (en) 2022-05-25 2022-05-25 Blood coagulation index abnormity classification method based on feature fusion meta-learning

Country Status (1)

Country Link
CN (1) CN114663679A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112613556A (en) * 2020-09-01 2021-04-06 电子科技大学 Low-sample image emotion classification method based on meta-learning
CN113139536A (en) * 2021-05-12 2021-07-20 哈尔滨工业大学（威海） Text verification code identification method and equipment based on cross-domain meta learning and storage medium
CN113902672A (en) * 2021-09-02 2022-01-07 山东师范大学 Rare skin lesion classification system based on space transformation optimization element learning
CN113907710A (en) * 2021-09-29 2022-01-11 山东师范大学 Skin lesion classification system based on model-independent image enhancement meta-learning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112613556A (en) * 2020-09-01 2021-04-06 电子科技大学 Low-sample image emotion classification method based on meta-learning
CN113139536A (en) * 2021-05-12 2021-07-20 哈尔滨工业大学（威海） Text verification code identification method and equipment based on cross-domain meta learning and storage medium
CN113902672A (en) * 2021-09-02 2022-01-07 山东师范大学 Rare skin lesion classification system based on space transformation optimization element learning
CN113907710A (en) * 2021-09-29 2022-01-11 山东师范大学 Skin lesion classification system based on model-independent image enhancement meta-learning

Similar Documents

Publication Publication Date Title
CN111261282A (en) Sepsis early prediction method based on machine learning
CN103714261B (en) Intelligent auxiliary medical treatment decision supporting method of two-stage mixed model
Aslan et al. Multi-classification deep CNN model for diagnosing COVID-19 using iterative neighborhood component analysis and iterative ReliefF feature selection techniques with X-ray images
Jatav An algorithm for predictive data mining approach in medical diagnosis
Inan et al. A hybrid probabilistic ensemble based extreme gradient boosting approach for breast cancer diagnosis
CN116153495A (en) Prognosis survival prediction method for immunotherapy of esophageal cancer patient
Chakradeo et al. Malaria parasite detection using deep learning methods
Darapureddy et al. Implementation of optimization algorithms on Wisconsin Breast cancer dataset using deep neural network
US6941288B2 (en) Online learning method in a decision system
CN112861626A (en) Fine-grained expression classification method based on small sample learning
Sathiya et al. Random Forest Classifier based detection of Parkinson’s disease
Wang et al. Early Diagnosis of Parkinson's Disease with Speech Pronunciation Features Based on XGBoost Model
CN114663679A (en) Blood coagulation index abnormity classification method based on feature fusion meta-learning
CN116030902A (en) Drug interaction prediction method and system based on graph attention sampling
Ono et al. Introduction to supervised machine learning in clinical epidemiology
Çetin et al. Detection of lung cancer with enhanced feed forward backpropagation artificial neural networks
Rajmohan et al. G-Sep: A Deep Learning Algorithm for Detection of Long-Term Sepsis Using Bidirectional Gated Recurrent Unit
CN116226629B (en) Multi-model feature selection method and system based on feature contribution
Ayoub et al. A Genetic Approach Wrapped Support Vector Machine for Feature Selection Applied to Parkinson's Disease Diagnosi
Zhong et al. Gestational Diabetes Mellitus Prediction Based on Two Classification Algorithms
Jhajharia et al. An amalgamated prediction model for breast cancer detection using fuzzy features
Suthagar et al. Analysis of breast cancer classification using various algorithms
Muthulakshmi et al. Prediction of Heart Disease using Ensemble Learning
CN116452910B (en) scRNA-seq data characteristic representation and cell type identification method based on graph neural network
Sen et al. Analysis and Prediction of Parkinson's Disease using Machine Learning Algorithms

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20220624