CN109036556B - Method for diagnosing keratoconus case based on machine learning - Google Patents

Method for diagnosing keratoconus case based on machine learning Download PDF

Info

Publication number
CN109036556B
CN109036556B CN201810991993.1A CN201810991993A CN109036556B CN 109036556 B CN109036556 B CN 109036556B CN 201810991993 A CN201810991993 A CN 201810991993A CN 109036556 B CN109036556 B CN 109036556B
Authority
CN
China
Prior art keywords
cornea
value
keratoconus
feature
model
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
Application number
CN201810991993.1A
Other languages
Chinese (zh)
Other versions
CN109036556A (en
Inventor
王雁
季书帆
张琳
徐佳慧
王书航
裴乐琪
崔彤
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.)
Individual
Original Assignee
Individual
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 Individual filed Critical Individual
Priority to CN201810991993.1A priority Critical patent/CN109036556B/en
Publication of CN109036556A publication Critical patent/CN109036556A/en
Application granted granted Critical
Publication of CN109036556B publication Critical patent/CN109036556B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Public Health (AREA)
  • Theoretical Computer Science (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Biology (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Pathology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Eye Examination Apparatus (AREA)
  • Instructional Devices (AREA)

Abstract

The invention relates to a method for diagnosing a keratoconus case based on machine learning, which applies a support vector machine-recursive feature screening algorithm (SVM-REF) and a gradient lifting tree (GBDT) algorithm in machine learning to the accurate diagnosis of the keratoconus case for the first time, and carries out effective overall scheme design, flow design and algorithm parameter setting aiming at specific application examples. The diagnostic accuracy of the method is significantly improved and has been substantially satisfactory for clinical applications through a large number of clinical example tests.

Description

Method for diagnosing keratoconus case based on machine learning
Technical Field
The invention belongs to the field of ophthalmic medical diagnosis, relates to a machine learning technology, and particularly relates to a method for diagnosing keratoconus cases based on machine learning.
Background
Keratoconus refers to a keratotic disease with the progressive thinning and conic bulging of the cornea in the central or lateral central area of the cornea, which is mostly caused in young people of about 20 years old, especially in young male patients, and often causes high irregular astigmatism and visual impairment of different degrees, the prognosis is poor, and even the cornea transplantation and other treatments are finally needed. Because the conical cornea has great influence on vision and visual function, early screening and early treatment intervention are key points; but earlier intervention, with greater diagnostic challenges, it is more challenging to accurately identify early corneal ectasia changes than to identify mid-late disease. At present, the early screening of the keratoconus has no uniform standard and has larger dispute, the clinician is greatly confused by different parameter differences, and the relatively accurate diagnosis can be given after the detailed consultation analysis is carried out by an expert with abundant experience; meanwhile, a great number of cases, limited experts and complicated corneal parameters all add great difficulty to the early diagnosis of the keratoconus. Early, efficient and relatively accurate screening of keratoconus has become a critical issue that needs to be addressed.
In recent years, machine learning has been widely used in the medical field, and has achieved a series of achievements. Gradient boosting trees (GBDTs) are currently gaining increasing attention and application as one of the mainstream classification algorithms. The gradient lifting tree is a decision tree integrated classifier algorithm based on iterative accumulation, and the defect that a single decision tree is easy to be over-fitted is overcome by constructing a series of weak classifiers, setting different weights for each weak classifier and accumulating the results of a plurality of decision trees (weak classifiers) to serve as a final prediction result. In view of the advantages, the gradient lifting tree is adopted to carry out learning modeling and prediction diagnosis on the keratoconus case.
Before modeling diagnosis is carried out, denoising and redundancy removing are generally carried out on characteristic values of a training sample, and characteristics which are relevant to problems and have category distinguishing capacity are screened out so as to optimize the diagnosis effect. The support vector machine-recursive feature screening subtraction (SVM-RFE) is one of feature selection methods with good performance and strong generalization capability, has good performance in a plurality of machine learning applications, and can be well adapted to high-dimensional data. Therefore, the invention adopts SVM-RFE to select the characteristics of the corneal sample data.
From the clinical test result, the method can efficiently and accurately identify the keratoconus case, and effectively improves the efficiency and accuracy of the clinical diagnosis of doctors.
Disclosure of Invention
The invention provides a method for diagnosing a keratoconus case based on machine learning, which provides a reliable auxiliary tool for ophthalmologists to clinically diagnose the keratoconus case.
The technical scheme for realizing the purpose of the invention is as follows:
step 1: collecting a large amount of cornea case sample data labeled by an ophthalmologist, wherein the labeling labels comprise keratoconus, preclinical keratoconus (suspected keratoconus), and normal cornea;
step 2: carrying out characteristic value normalization processing on the case sample data to map the characteristic value of the case sample data to be between [0 and 1] to obtain training sample data;
and step 3: performing feature selection on training sample data by adopting a support vector machine-recursive feature screening subtraction method (SVM-RFE) to obtain an optimal feature subset as training features;
and 4, step 4: constructing a gradient lifting tree (GBDT) diagnosis model based on the training characteristics of the sample data;
and 5: the new case is diagnosed by using the GBDT diagnosis model and is judged as keratoconus, or preclinical keratoconus (suspected keratoconus), or normal cornea.
Preferably, in step 2, the case sample data is normalized by the feature value, so that the feature value is mapped to [0,1], and the specific formula is as follows:
Figure BDA0001780984530000021
wherein x is the original characteristic value, xmaxFor the maximum value of the class feature, xminIs the minimum value, x, of the class feature*And taking the value of the feature after normalization.
Preferably, in step 3, a support vector machine-recursive feature screening subtraction (SVM-RFE) is adopted to perform feature selection on training sample data to obtain an optimal feature subset, and the 17 features serving as training features include: central corneal astigmatism (Cornea Front tilting) of anterior surface of Cornea and central corneal astigmatism (Cornea Back A) of posterior surface of Corneastig), thinnest point posterior surface height (B Ele Th), ART index (ART max), anterior surface variation (Df), posterior surface variation (Db), integrated global variation (D), temporal aspheric Q (Temp 6mm Front) within 6mm diameter of anterior corneal surface, temporal aspheric Q (Temp 7mm Front) within 7mm diameter of anterior corneal surface, nasal aspheric Q (Nas 6mm Back) within 6mm diameter of posterior corneal surface, temporal aspheric Q (Temp 8mm Back) within 8mm diameter of posterior corneal surface, temporal aspheric Q (Temp 9mm Back) within 9mm diameter of posterior corneal surface, superior aspheric Q (Sup 7mm Back) within 7mm diameter of posterior corneal surface, surface variation Index (IVA), anterior corneal surface Z9 within 6mm diameter
Figure BDA0001780984530000022
(WFA FRONT 9), aberration Z9 of the posterior surface of the cornea within 6mm in diameter
Figure BDA0001780984530000023
(WFA Back 9), Total corneal aberration Z14 within 6mm diameter
Figure BDA0001780984530000024
(WFA Cornea 14)。
Advantages and advantageous effects of the invention
1. The method for diagnosing the keratoconus case based on machine learning provided by the invention is characterized in that a keratoconus case diagnosis model is trained and constructed by collecting a large number of cornea case samples labeled by ophthalmologists and applying advanced machine learning methods (SVM-REF and GBDT), so that accurate diagnosis of the keratoconus case is realized. The method is widely applied and popularized, can help clinicians to diagnose the keratoconus more accurately, screen out suspicious cases relatively accurately, effectively improve the efficiency and accuracy of the clinical diagnosis of the keratoconus, and has great clinical significance and clinical application value particularly for the screening and early warning of early keratoconus cases
2. The invention uses a support vector machine-recursive feature screening and reducing algorithm (SVM-REF) to screen out pathological features which have important influence on the determination of the keratoconus and the keratoconus in the early clinical period (suspected keratoconus), and has profound influence on the clinical research of the disease.
3. The method applies a support vector machine-recursive feature screening algorithm (SVM-REF) and a gradient lifting tree (GBDT) algorithm in machine learning to the accurate diagnosis of the keratoconus case for the first time, and performs effective overall scheme design, flow design and algorithm parameter setting aiming at specific application examples. The diagnostic accuracy of the method is significantly improved and has been substantially satisfactory for clinical applications through a large number of clinical example tests.
Drawings
FIG. 1 is a flow chart of the method.
Detailed Description
In order to make the purpose, technical solution and innovation point of the embodiments of the present invention clearer, the technical solution 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. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all 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.
Fig. 1 is a schematic flow chart of the method for diagnosing keratoconus case based on machine learning according to the present invention. The method comprises the following steps:
step 1: a large amount of cornea case sample data labeled by an ophthalmologist are collected, and labeling labels of the sample data comprise keratoconus, preclinical keratoconus (suspected keratoconus) and normal cornea.
Step 2: carrying out characteristic value normalization processing on the case sample data to map the characteristic value of the case sample data to [0,1] to obtain training sample data, wherein the normalization formula is as follows:
Figure BDA0001780984530000031
wherein x is the original characteristic value, xmaxFor the maximum value of the class feature, xminIs the minimum value, x, of the class feature*And taking the value of the feature after normalization.
And step 3: the method comprises the following steps of (1) selecting features of training sample data by adopting a support vector machine-recursive feature screening subtraction method (SVM-RFE) to obtain an optimal feature subset as training features:
step 3.1: initializing a feature subset F and an optimal feature subset F _ best to enable the feature subset F and the optimal feature subset F _ best to contain all case features;
step 3.2: inputting the feature subset F of the training sample into a Support Vector Machine (SVM) for training to obtain the classification weight of each feature on a class label (keratoconus, keratoconus before clinical period, normal cornea), and deleting the feature with the minimum classification weight from F;
step 3.3: calculating prediction accuracy P of training data in SVM (support vector machine) model based on F and F _ best respectivelyFAnd PF-best；
Step 3.4: if P isFGreater than PF-bestUpdating the optimal feature subset F _ best to F;
step 3.5: and repeating the steps 3.2-3.4 until the feature subset F is empty, and outputting the optimal feature subset F _ best. The results are shown below (17 features in total):
corneal anterior surface central corneal astigmatism (corneal Front astigmatism), corneal posterior surface central corneal astigmatism (corneal Back astigmatism), thinnest point posterior surface height (B Ele Th), ART index (ART max), anterior surface variation (Df), posterior surface variation (Db), integrated global variation (D), temporal aspheric Q (Temp 6mm Front) within a 6mm diameter range of the corneal anterior surface, temporal aspheric Q (Temp 7mm Front) within a 7mm diameter range of the corneal anterior surface, nasal aspheric Q (Nas 6mm Back) within a 6mm diameter range of the corneal posterior surface, temporal aspheric Q (Temp 8mm Back) within a 8mm diameter range of the corneal posterior surface, temporal aspheric Q (Temp 9mm Back) within a 9mm diameter range of the corneal posterior surface, supra aspheric Q (Temp 7mm Back) within a 7mm diameter range of the corneal posterior surface, surface variation Index (IVA), anterior corneal surface aberration Z9 within 6mm diameter
Figure BDA0001780984530000041
(WFA FRONT 9), aberration Z9 of the posterior surface of the cornea within 6mm in diameter
Figure BDA0001780984530000042
(WFA Back 9), Total corneal aberration Z14 within 6mm diameter
Figure BDA0001780984530000043
(WFA Cornea 14)。
And 4, step 4: the method comprises the following steps of constructing a gradient lifting tree (GBDT) diagnosis model based on training characteristics of sample data:
step 4.1: initialization model F0(x) Estimating a constant value γ that minimizes the loss function using a log-likelihood loss function L; wherein the data set
Figure BDA0001780984530000044
In, xiIs the ith sample, N is the sample size, yi∈{-1,+1},
Figure BDA0001780984530000045
Step 4.2: and establishing a model in the descending direction of the gradient of the last model loss function. The method comprises the following specific steps:
step 4.2.1: calculating the value of the negative gradient of the loss function in the round model, and taking the value as a residual error rmiEstimated value of (a):
Figure BDA0001780984530000046
m is the round of current model iteration
Step 4.2.2: will residual error rmiAs input, a fitted CART regression tree is solved, and the corresponding leaf node region is RjmJ1, 2, said, J, wherein J is a regression tree hmThe number of leaf nodes of (2);
step 4.2.3: estimating the value of leaf node region by linear search to minimize the loss function and obtain the step length betajmI.e. the weight of the regression tree:
Figure BDA0001780984530000051
beta is the residual error of the current model fitting data
Step 4.2.4: updating model Fm：
Figure BDA0001780984530000052
Step 4.3: repeating the step 4.2 until the iteration number reaches a set iteration number M;
step 4.4: obtaining a final classification model FM：
Figure BDA0001780984530000053
And I is an indication function, 1 is returned when the condition is met, and 0 is not returned.
The parameters related to the GBDT classifier Boosting framework are set as follows: the number of iterations is set to 150, the weight reduction factor of the weak learner is set to 0.05, the no-return sampling ratio is set to 0.8, the maximum feature number of the regression tree is set to the total number of features, the maximum depth of the tree is set to 5, the minimum number of samples required for internal node subdivision is set to 50, the minimum number of samples of leaf nodes is set to 10, and the maximum number of leaf nodes is not limited.
And 5: the new case is diagnosed by using the GBDT diagnosis model and is judged as keratoconus, or preclinical keratoconus (suspected keratoconus), or normal cornea.
In the embodiment of the invention, the training characteristics of the newly-added case are input into the trained gradient lifting tree GBDT diagnosis model, so that the diagnosis label of the case can be obtained: keratoconus, preclinical keratoconus (suspected keratoconus), normal cornea.
The correctness of the patient diagnosed by the method of the present invention is illustrated by three specific cases.
Figure BDA0001780984530000054
Figure BDA0001780984530000061
The above disclosure is only for a few specific embodiments of the present invention, but the present invention is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present invention.

Claims (5)

1. A method for diagnosing a keratoconus case based on machine learning is characterized in that: the method comprises the following steps:
step 1: collecting a large amount of cornea case sample data labeled by an ophthalmologist, wherein the labeling labels comprise keratoconus, suspected keratoconus and normal cornea;
step 2: carrying out characteristic value normalization processing on the case sample data to map the characteristic value of the case sample data to be between [0 and 1] to obtain training sample data;
and step 3: performing feature selection on training sample data by adopting an SVM-RFE (support vector machine-frequency extraction) to obtain an optimal feature subset as a training feature;
and 4, step 4: constructing a GBDT diagnosis model based on the training characteristics of the sample data;
and 5: and (4) diagnosing the label of the new case by using the GBDT diagnosis model, and judging the new case as the keratoconus or the suspected keratoconus or the normal cornea.
2. The method for machine learning-based diagnosis of keratoconus cases according to claim 1, wherein: the specific formula of the normalization processing in the step 2 is as follows:
Figure FDA0001780984520000011
wherein x is the original characteristic value, xmaxIs thatMaximum value of class feature, xminIs the minimum value, x, of the class feature*And taking the value of the feature after normalization.
3. The method for machine learning-based diagnosis of keratoconus cases according to claim 1, wherein: the 17 features as training features include: central corneal astigmatism of anterior surface of cornea, central corneal astigmatism of posterior surface of cornea, posterior surface height of thinnest point, ART index, anterior surface variation value, posterior surface variation value, integrated global variation value, temporal aspheric Q value within 6mm diameter range of anterior surface of cornea, temporal aspheric Q value within 7mm diameter range of anterior surface of cornea, nasal aspheric Q value within 6mm diameter range of posterior surface of cornea, temporal aspheric Q value within 8mm diameter range of posterior surface of cornea, temporal aspheric Q value within 9mm diameter range of posterior surface of cornea, upper aspheric Q value within 7mm diameter range of posterior surface of cornea, surface variation index, anterior surface aberration of cornea within 6mm diameter range of cornea, and anterior surface aberration of cornea within 6mm diameter range of cornea
Figure FDA0001780984520000012
Posterior surface aberration of cornea within 6mm diameter
Figure FDA0001780984520000013
Full corneal aberration within 6mm diameter
Figure FDA0001780984520000014
4. The method for machine learning-based diagnosis of keratoconus cases according to claim 1, wherein: the method for selecting the characteristics of training sample data by adopting the SVM-RFE specifically comprises the following steps:
step 1: initializing a feature subset F and an optimal feature subset F _ best to enable the feature subset F and the optimal feature subset F _ best to contain all case features;
step 2: inputting the feature subset F of the training sample into a Support Vector Machine (SVM) for training to obtain the classification weight of each feature on the class label, and deleting the feature with the minimum classification weight from the F;
and step 3: calculating prediction accuracy P of training data in SVM (support vector machine) model based on F and F _ best respectivelyFAnd PF-best；
And 4, step 4: if P isFGreater than PF-bestUpdating the optimal feature subset F _ best to F;
and 5: and repeating the steps 2-4 until the feature subset F is empty, and outputting the optimal feature subset F _ best.
5. The method for machine learning-based diagnosis of keratoconus cases according to claim 1, wherein: the GBDT diagnosis model construction method comprises the following steps:
step 1: initialization model F0(x) Estimating a constant value γ that minimizes the loss function using a log-likelihood loss function L; x is the number ofiIs the ith sample, N is the sample size, yi∈{-1,+1},
Figure FDA0001780984520000021
Step 2: establishing a model in the gradient descending direction of the last model loss function, and specifically comprising the following steps of:
step 2.1: calculating the value of the negative gradient of the loss function in the round model, and taking the value as a residual error rmiEstimated value of (a):
Figure FDA0001780984520000022
m is the round of current model iteration
Step 2.2: will residual error rmiAs input, a fitted CART regression tree is solved, and the corresponding leaf node region is RjmJ1, 2, said, J, wherein J is a regression tree hmThe number of leaf nodes of (2);
step 2.3: estimating the value of leaf node region by linear search to minimize the loss function and obtain the step length betajmI.e. the weight of the regression tree:
Figure FDA0001780984520000023
beta is the residual error of the current model fitting data
Step 2.4: updating model Fm：
Figure FDA0001780984520000024
I is an indication function, 1 is returned when the condition is met, and 0 is not used
And step 3: repeatedly executing the step 2 until the iteration times reach a set model iteration number M;
and 4, step 4: obtaining a final classification model FM：
Figure FDA0001780984520000025
The parameters related to the GBDT classifier Boosting framework are set as follows: the iteration number is set to 150, the weight reduction coefficient of the weak learner is set to 0.05, the unreplacing sampling ratio is set to 0.8, the maximum feature number of the regression tree is set to the total number of features, the maximum depth of the tree is set to 5, the minimum sample number required by internal node subdivision is set to 50, the minimum sample number of the leaf nodes is set to 10, and the maximum leaf node number is not limited;
and 5: and diagnosing the label of the new case by using the GBDT diagnosis model, and judging the new case as the keratoconus or the suspected keratoconus or the normal cornea.
CN201810991993.1A 2018-08-29 2018-08-29 Method for diagnosing keratoconus case based on machine learning Active CN109036556B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810991993.1A CN109036556B (en) 2018-08-29 2018-08-29 Method for diagnosing keratoconus case based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810991993.1A CN109036556B (en) 2018-08-29 2018-08-29 Method for diagnosing keratoconus case based on machine learning

Publications (2)

Publication Number Publication Date
CN109036556A CN109036556A (en) 2018-12-18
CN109036556B true CN109036556B (en) 2021-06-04

Family

ID=64625894

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810991993.1A Active CN109036556B (en) 2018-08-29 2018-08-29 Method for diagnosing keratoconus case based on machine learning

Country Status (1)

Country Link
CN (1) CN109036556B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110968693A (en) * 2019-11-08 2020-04-07 华北电力大学 Multi-label text classification calculation method based on ensemble learning
CN111626977A (en) * 2019-11-25 2020-09-04 梁帆 Steel rail surface detection method based on image recognition
CN111024557B (en) * 2019-12-20 2021-06-04 广州博依特智能信息科技有限公司 Soft measurement method for water absorption of paper sheet
CN112036448B (en) * 2020-08-11 2021-08-20 上海鹰瞳医疗科技有限公司 Keratoconus identification method and device
CN112806957B (en) * 2021-04-22 2021-08-31 浙江大学 Keratoconus and subclinical keratoconus detection system based on deep learning
CN113593697B (en) * 2021-07-31 2023-10-13 吾征智能技术(北京)有限公司 System, equipment and medium for judging diseases based on breath odor
CN113901999B (en) * 2021-09-29 2023-09-29 国网四川省电力公司电力科学研究院 Fault diagnosis method and system for high-voltage shunt reactor

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106650314A (en) * 2016-11-25 2017-05-10 中南大学 Method and system for predicting amino acid mutation
CN106650768A (en) * 2016-09-27 2017-05-10 北京航空航天大学 Gaussian image model-based brain network modeling and mode classification method
CN107924384A (en) * 2015-03-11 2018-04-17 阿雅斯迪公司 For the system and method using study model prediction result is predicted
CN108108592A (en) * 2017-12-29 2018-06-01 北京聚道科技有限公司 A kind of construction method of machine learning model for the pathogenic marking of hereditary variation
CN108389626A (en) * 2018-02-09 2018-08-10 上海长江科技发展有限公司 Cerebral apoplexy screening method based on artificial intelligence and system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107924384A (en) * 2015-03-11 2018-04-17 阿雅斯迪公司 For the system and method using study model prediction result is predicted
CN106650768A (en) * 2016-09-27 2017-05-10 北京航空航天大学 Gaussian image model-based brain network modeling and mode classification method
CN106650314A (en) * 2016-11-25 2017-05-10 中南大学 Method and system for predicting amino acid mutation
CN108108592A (en) * 2017-12-29 2018-06-01 北京聚道科技有限公司 A kind of construction method of machine learning model for the pathogenic marking of hereditary variation
CN108389626A (en) * 2018-02-09 2018-08-10 上海长江科技发展有限公司 Cerebral apoplexy screening method based on artificial intelligence and system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
A Comparative Analysis of SVM，Naive Bayes and GBDT for data Fault Detection in WSNs;Ye Yuan,et al.;《2018 IEEE International Conference on Software Quality，Reliability and Security Companion》;20180813;第394-399页 *
基于学习的圆锥角膜辅助诊断;高瑜鸿.;《中国优秀硕士学位论文全文数据库》;20180115(第01期);第E073-130页 *

Also Published As

Publication number Publication date
CN109036556A (en) 2018-12-18

Similar Documents

Publication Publication Date Title
CN109036556B (en) Method for diagnosing keratoconus case based on machine learning
Sungheetha et al. Design an early detection and classification for diabetic retinopathy by deep feature extraction based convolution neural network
US9445716B2 (en) Obtaining data for automatic glaucoma screening, and screening and diagnostic techniques and systems using the data
US10818398B2 (en) System and method for AI-based eye condition determinations
BR112021001576A2 (en) system and method for eye condition determinations based on ia
CN110111884B (en) CMKMC-based man-machine cooperative intelligent medical aid decision-making system
Acharya et al. Identification of cataract and post-cataract surgery optical images using artificial intelligence techniques
JP2011520503A (en) Automatic concave nipple ratio measurement system
CN110881969A (en) Stacking ensemble learning-based heart failure early warning method
CN110310282B (en) System for intelligently analyzing corneal nerve fibers by using in-vivo confocal microscope images
WO2021184802A1 (en) Blood pressure classification prediction method and apparatus
Zhang et al. MRMR optimized classification for automatic glaucoma diagnosis
CN113470816A (en) Machine learning-based diabetic nephropathy prediction method, system and prediction device
Das et al. Automated classification of retinal OCT images using a deep multi-scale fusion CNN
CN111986814A (en) Modeling method of lupus nephritis prediction model of lupus erythematosus patient
Reddy et al. Discovering optimal algorithm to predict diabetic retinopathy using novel assessment methods
Eladawi et al. Early diagnosis of diabetic retinopathy in octa images based on local analysis of retinal blood vessels and foveal avascular zone
Hong et al. A novel hierarchical deep learning framework for diagnosing multiple visual impairment diseases in the clinical environment
EP4365829A1 (en) Binocular keratoconus diagnosis method based on multi-modal data
CN112155550A (en) Alzheimer's disease detection device based on support vector machine
CN112419256A (en) Method for grading fundus images of diabetes mellitus based on fuzzy graph neural network
Li et al. Retinal OCT image classification based on domain adaptation convolutional neural networks
CN116721760B (en) Biomarker-fused multitasking diabetic retinopathy detection algorithm
CN116403717B (en) Myopia prediction method and device based on deep learning, terminal equipment and medium
Krishna et al. Glaucoma diagnosis using state of art image analysis techniques: a comprehensive survey

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