CN109256207A - A method of based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case - Google Patents
A method of based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case Download PDFInfo
- Publication number
- CN109256207A CN109256207A CN201810992036.0A CN201810992036A CN109256207A CN 109256207 A CN109256207 A CN 109256207A CN 201810992036 A CN201810992036 A CN 201810992036A CN 109256207 A CN109256207 A CN 109256207A
- Authority
- CN
- China
- Prior art keywords
- sample
- keratoconus
- xgboost
- feature
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
-
- 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
- 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
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Primary Health Care (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Epidemiology (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Quality & Reliability (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Eye Examination Apparatus (AREA)
Abstract
The present invention provides a kind of method based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case, include the following steps: that the cornea for acquiring ophthalmic patient checks data, it is that each cornea sample marks a class label keratoconus, doubtful keratoconus, normal cornea by ophthalmologist, oculist, as training sample data；Each category feature of diagonal membrane sample data carries out characteristic value normalization processing respectively, is mapped between section [0,1]；Feature enlarging is carried out to sample data using XGBoost, using the feature set after enlarging as the training characteristics of sample；Training characteristics based on sample data, training building SVM diagnostic model；Diagnosis prediction is carried out to new case using diagnostic model.Experiments have shown that the diagnosis effect of this method has met clinical application.Using this method to keratoconus, the screening of especially doubtful keratoconus, the dependence to medical expert's diagnosis can be reduced, and diagnosis efficiency, accuracy rate can be promoted substantially.
Description
Technical field
The invention belongs to ophthalmic medical diagnostic fields, are related to machine learning techniques, especially a kind of to be based on XGBoost+SVM
The method of hybrid machine Learner diagnosis keratoconus illness.
Background technique
Keratoconus (keratoconus) is one kind characterized by kerectasis, causes Central corneal or other central area forward
Protrusion is in cone and generates the primary corneal degeneration disease of highly irregular myopic astigmatism and different inpairment of vision, can be with
It is a kind of independent disease, is also possible to the component part of a variety of syndromes.After it mostly occurs in preadolescence, without inflammation
Disease.Advanced stage will appear acute corneal edema, form scar, and eyesight is badly damaged.Apparent keratoconus is easy to make a definite diagnosis.Work as appearance
And when not being true to type seen in slit-lamp, early diagnose more difficult.Most efficient method is corneal topography inspection.But nonetheless,
The diagnostic criteria of early stage keratoconus is not completely unified, mostly need to be rich by experience for relative complex doubtful keratoconus case
Rich expert, which carries out consultation of doctors discussion in detail, can just give relatively accurate diagnosis；A large amount of case, limited expert, complexity simultaneously
Corneal parameters all examined to the early stage of keratoconus and increase very big difficulty.Therefore the early screening of keratoconus has become urgently
Critical issue to be solved.
Extreme value gradient lift scheme (XGBoost) algorithm being related in the present invention is on present industrial using most machines
One of device learning model, base classifier can arbitrarily be selected from decision tree (gbtree) and linear (gblinear) kernel function
It selects.Include a large amount of decision regression tree inside XGBoost, carrys out lift scheme using residual error, and joined regularization to prevent
Over-fitting guarantees the robustness of model.The present invention expands new feature using training sample training XGBoost model first,
Method are as follows: the structure feature based on the decision tree that XGBoost iteration generates can recorde every training sample data in each decision
Position in leaf child node carries out One-Hot coding to the training sample according to this, as the new spy of the training data
Sign is realized and is expanded the feature of training data.Since the new feature that One-Hot coding generates can preferably describe numerous weak typings
Device (decision tree that XGBoost iteration generates) is to the categorised decision of the sample, and therefore, these new features expanded will be helpful to mention
The class discrimination degree of sample, the i.e. predictive value of training for promotion sample are risen, and then promotes the precision of prediction model.Based on sample
Feature after expansion, the present invention are being solved using instantly popular support vector machine method (SVM) training building diagnostic model, SVM
Certainly small sample, show many distinctive advantages in high dimensional feature pattern recognition problem, final objective function will be by minority
Supporting vector determined, be not dependent on the dimension of sample space, avoid dimension disaster in a way, and have compared with
Good robustness and forecasting efficiency.
Diagnose clinical early stage keratoconus (doubtful keratoconus) and keratoconus case using upper, the present invention passes through
The feature of original sample is expanded using extreme value gradient lift scheme (XGBoost) model, and combination supporting vector machine
(SVM) prediction model carries out sieving and diagnosis, possesses preferable clinical effectiveness, can effectively assist eye doctor, to the clinical patient's condition
It makes efficiently and accurately diagnoses.
Summary of the invention
The present invention provides a kind of method based on machine learning diagnosis keratoconus case, carries out keratoconus, doubtful circle
The judgement of cone angle film and normal cornea provides reliable auxiliary tool for eye doctor's clinical diagnosis.
Realize the technical solution of the object of the invention are as follows:
Step 1: the cornea for acquiring ophthalmic patient checks data, is that each cornea sample marks a classification by ophthalmologist, oculist
Label (keratoconus, doubtful keratoconus, normal cornea), as training sample data；
Step 2: each category feature of diagonal membrane sample data carries out characteristic value normalization processing respectively, is mapped to section
[0,1] between；
Step 3: feature enlarging being carried out to sample data using extreme value gradient lift scheme (XGBoost), after enlarging
Training characteristics of the feature set as sample, the specific steps are as follows:
Step 3.1: based on sample data by XGBoost training N gradient boosted tree of building, i.e. XGBoost model iteration
Number is set as N；
Step 3.2: position of each training sample in N gradient boosted tree is subjected to 1,0 coding (also known as One- respectively
Hot coding, wherein 1 is used to indicate sample default position for recording sample position, 0), and the N group One-Hot of generation is compiled
Code, as new feature, is incorporated in sample primitive character；
Step 4: the training characteristics based on sample data, training building support vector machines (SVM) diagnostic model；
Step 5: diagnosis prediction being carried out to new case using diagnostic model, new case is determined as keratoconus, doubtful circle
Cone angle film or normal cornea, the specific steps are as follows:
Step 5.1: characteristic value normalization processing being carried out to each category feature of new case's sample data respectively, is mapped to
Between section [0,1]；
Step 5.2: spy being carried out to normalized new case's sample with the extreme value gradient lift scheme (XGBoost) constructed
Sign enlarging, i.e., map new samples on (Map) to the N gradient boosted tree built, record the sample and promoted in N gradient
Then position on tree is encoded using One-Hot, carry out feature enlarging to new case's sample；
Step 5.3: the feature after the enlarging of new case's sample being inputted into the SVM diagnostic model constructed, determines that its label is
Keratoconus, doubtful keratoconus or normal cornea.
Preferably, described in step 2, each category feature of diagonal membrane sample data carries out characteristic value normalization processing respectively, makes
It is mapped between section [0,1], specific formula is as follows:
Wherein, x is primitive character value, xmaxFor category profile maxima, xminFor category feature minimum value, x*For this
Value after feature normalization.
The advantages of the present invention
1, a kind of method based on machine learning diagnosis keratoconus case proposed by the present invention, has largely been marked by analysis
The cornea of note detects case sample, proposes to learn prediction model using XGBoost+SVM hybrid machine, identification keratoconus is doubted
Like the method for keratoconus and normal cornea.
2, the present invention expands the primitive character of cornea case sample using XGBoost method, and carries out One-Hot
Coding obtains to the higher new feature of sample discrimination, improves the classification performance of sample primary nonlinear feature.
3, experiments have shown that, the diagnosis effect of this method has met clinical application.Using this method to keratoconus, especially
The screening of doubtful keratoconus can reduce the dependence to medical expert's diagnosis, and can promote diagnosis efficiency, accuracy rate substantially.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is the One-Hot coding that extreme value gradient promotes decision tree (XGBoost).
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described.Obviously, described embodiment is
A part of the embodiments of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Method includes the following steps:
Step 1: the cornea for acquiring ophthalmic patient checks data, is that each cornea sample marks a classification by ophthalmologist, oculist
Label (keratoconus, doubtful keratoconus, normal cornea), as training sample data；
Step 2: each category feature of diagonal membrane sample data carries out characteristic value normalization processing respectively, is mapped to section
Between [0,1], specific formula is as follows:
Wherein, x is primitive character value, xmaxThe maximum value of category feature thus, xminThe minimum value of category feature thus, x*To return
Value after one change.
Step 3: feature enlarging being carried out to sample data using extreme value gradient lift scheme (XGBoost), specific steps are such as
Under:
Step 3.1: the objective function obj by solving minimization(1), initialize first post-class processing f1(xi):
Wherein, xiFor i-th of training sample, yiFor xiCorresponding label value, N sample size, L are loss function
Softmax, Ω (f1(xi)) it is decision tree f1(xi) regularization term.
Step 3.2: the objective function obj by solving minimization(t), obtain the t post-class processing ft(xi):
Wherein, obj(t)The objective function of t round, IjSample set in j-th of leaf of presentation class regression tree,Indicate that the single order of i-th of sample object function is led,Indicate i-th
The second order of a sample object function is led；J be post-class processing leaf node index value, t by training regression tree it is affiliated
Number is taken turns, γ, λ are punishment dynamics, and T is leaf number.
Step 3.3: iteration executes step 3.2 200 times totally, successively obtains post-class processing ft(xi), t=2,3,4 ...,
200, and record node location of each sample in each tree.
Wherein, the parameter that building XGBoost classifier is related to includes: that learning rate is set as 0.1, and the depth capacity of tree is set
It being set to 3, minimum leaf node sample weights and is set as 1, least disadvantage function drop-out value needed for node split is set as 0,
The maximum step-length of limitation each tree weight changes is set as no constraint, and random sample sampling ratio is set as 1, controls each of tree
The division each time of grade, the columns accounting of every stochastical sampling are set as 1, and the measure of data is set as AUC, generate it is pseudo- with
The seed of machine number is set as 10.
Step 3.4: position of each training sample in 200 gradient boosted trees is subjected to 1,0 coding (also known as respectively
One-Hot coding, wherein 1, for indicating sample default position, sees Fig. 2 example for recording sample position, 0), and by generation
200 groups of One-Hot codings are incorporated in sample primitive character as new feature；
Example: assuming that position of the data x in N=3 tree is as shown in Fig. 2, so according to x in leaf child node
Position (black node in Fig. 2), the new feature that it is constructed are 100000010010.
Step 4: the training characteristics based on sample data, training building support vector machines (SVM) diagnostic model, specific steps
It is as follows:
Step 4.1: original training sample is divided into the training sample of 3 two classification.Division mode is with one type mark
Signing (keratoconus, doubtful keratoconus, normal cornea) is positive sample, other two classes are that (such as: keratoconus is positive sample negative sample
This, doubtful keratoconus and normal cornea are negative sample)；
Step 4.2: using keratoconus as positive sample, doubtful keratoconus and normal cornea are negative sample training categorised decision
Function.Solve the objective function of separating hyperplance:
0≤αi≤ C, i=1,2 ..., N
Wherein, x is training sample feature, and y indicates positive and negative sample label, and N is number of samples, and punishment parameter C=1, i, j are
Training sample number, αi、αjIndicate Lagrange multiplier, the seed for generating pseudo random number is set as 10.
According to objective function, optimal solution is acquiredAndSelect α*One
ComponentIt calculatesObtain categorised decision function:
F1(x)=ω*·x+b*
Step 4.3: being negative sample by positive sample, keratoconus and normal cornea of doubtful keratoconus according to step 4.2
Construct decision function F2(x)；
Step 4.4: being negative sample by positive sample, keratoconus and doubtful keratoconus of normal cornea according to step 4.2
Construct decision function F3(x)；
Step 4.5: input diagnostic sample compares F1(x), F2(x), F3(x) value size, by the classification mark of maximum value
Label, as final prediction label, i.e. F (x)=Max [F1(x), F2(x), F3(x)]。
Step 5: diagnosis prediction being carried out to new case using diagnostic model, new case is determined as keratoconus, doubtful circle
Cone angle film or normal cornea, the specific steps are as follows:
Step 5.1: characteristic value normalization processing being carried out to each category feature of new case's sample data respectively, is mapped to
Between section [0,1]；
Step 5.2: spy being carried out to normalized new case's sample with the extreme value gradient lift scheme (XGBoost) constructed
Sign enlarging, i.e., map new samples on (Map) to the 200 gradient boosted trees built, record the sample in 200 gradients
Then position in boosted tree is encoded using One-Hot, carry out feature enlarging to new case's sample；
Step 5.3: the feature after the enlarging of new case's sample being inputted into the SVM diagnostic model constructed, determines that its label is
Keratoconus, doubtful keratoconus or normal cornea.
Illustrate the correctness for the sufferer being diagnosed to be by the method for the invention below by three specific cases.
Disclosed above is only several specific embodiments of the invention, and still, the present invention is not limited to this, any ability
What the technical staff in domain can think variation should all fall into protection scope of the present invention.
Claims (4)
1. a kind of method based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case, it is characterised in that: including such as
Lower step:
Step 1: the cornea for acquiring ophthalmic patient checks data, is that each cornea sample marks a classification mark by ophthalmologist, oculist
Label, respectively keratoconus, doubtful keratoconus, normal cornea, as training sample data；
Step 2: each category feature of diagonal membrane sample data carries out characteristic value normalization processing respectively, be mapped to section [0,
1] between；
Step 3: feature enlarging is carried out to sample data using XGBoost, the feature set after enlarging is special as the training of sample
Sign；
Step 4: the training characteristics based on sample data, training building SVM diagnostic model；
Step 5: diagnosis prediction being carried out to new case using diagnostic model, new case is determined as keratoconus, doubtful coning angle
Film or normal cornea.
2. the method according to claim 1 based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case,
Be characterized in that: step 2 normalized specific formula is as follows:
Wherein: x is primitive character value, xmaxFor category profile maxima, xminFor category feature minimum value, x*For this feature
Value after normalization.
3. the method according to claim 1 based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case,
Be characterized in that: carrying out feature enlarging to sample data using XGBoost, specific step is as follows:
Step 1: based on sample data by XGBoost training N gradient boosted tree of building, the setting of XGBoost model the number of iterations
For N；
Step 2: position of each training sample in N gradient boosted tree being subjected to 1,0 coding respectively, and by the N group of generation
One-Hot coding, as new feature, is incorporated in sample primitive character.
4. the method according to claim 1 based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case,
Be characterized in that: specific step is as follows for step 5:
Step 1: characteristic value normalization processing being carried out to each category feature of new case's sample data respectively, is mapped to section
[0,1] between；
Step 2: feature enlarging being carried out to normalized new case's sample with the XGBoost model constructed, i.e., is reflected new samples
It is mapped in the N gradient boosted tree built, records sample position in N gradient boosted tree, then use One-Hot
Coding carries out feature enlarging to new case's sample；
Step 3: the feature after the enlarging of new case's sample being inputted into the SVM diagnostic model constructed, determines its label for coning angle
Film, doubtful keratoconus or normal cornea.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810992036.0A CN109256207A (en) | 2018-08-29 | 2018-08-29 | A method of based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810992036.0A CN109256207A (en) | 2018-08-29 | 2018-08-29 | A method of based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109256207A true CN109256207A (en) | 2019-01-22 |
Family
ID=65049474
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810992036.0A Pending CN109256207A (en) | 2018-08-29 | 2018-08-29 | A method of based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109256207A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109920551A (en) * | 2019-01-24 | 2019-06-21 | 华东师范大学 | Autism children social action performance characteristic analysis system based on machine learning |
| CN111984872A (en) * | 2020-09-09 | 2020-11-24 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
| CN112465657A (en) * | 2021-02-02 | 2021-03-09 | 北京淇瑀信息科技有限公司 | Risk assessment method and device based on tree model feature derivation and electronic equipment |
| CN112700863A (en) * | 2020-12-28 | 2021-04-23 | 天津市眼科医院 | Method for accurately evaluating diopter based on Scheimpflug anterior segment morphology and application |
| CN113611404A (en) * | 2021-07-09 | 2021-11-05 | 哈尔滨智吾康软件开发有限公司 | Plasma sample cancer early screening method based on ensemble learning |
| CN114463014A (en) * | 2022-02-23 | 2022-05-10 | 河南科技大学 | SVM-Xgboost-based mobile payment risk early warning method |
| CN114548845A (en) * | 2022-04-27 | 2022-05-27 | 北京智芯微电子科技有限公司 | Distribution network management method, device and system |
| WO2023272876A1 (en) * | 2021-06-28 | 2023-01-05 | 上海美沃精密仪器股份有限公司 | Binocular keratoconus diagnosis method based on multi-modal data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107582037A (en) * | 2017-09-30 | 2018-01-16 | 深圳前海全民健康科技有限公司 | Method based on pulse wave design medical product |
| 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 |
-
2018
- 2018-08-29 CN CN201810992036.0A patent/CN109256207A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107582037A (en) * | 2017-09-30 | 2018-01-16 | 深圳前海全民健康科技有限公司 | Method based on pulse wave design medical product |
| 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 (8)
| Title |
|---|
| 余锦华等: "正常角膜和圆锥角膜的特征提取", 《光学 精密工程》 * |
| 俞高: "基于XGBoost算法的病例分型系统设计", 《中国数字医学》 * |
| 刘玉尧: "基于Gradient Boosting算法的海量健康数据挖掘研究与应用", 《中国优秀博硕士学位论文全文数据库(硕士)医药卫生科技辑（月刊）》 * |
| 吴良斌: "《SAR图像处理与目标识别》", 31 January 2013, 北京：航空工业出版社 * |
| 贾文慧等: "基于XGBoost模型的股骨颈骨折手术预后质量评分预测", 《太原理工大学学报》 * |
| 金升剑: "《深度学习 基于MATLAB的设计实例》", 30 April 2018, 北京航空航天大学出版社 * |
| 高瑜鸿: "基于学习的圆锥角膜辅助诊断", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
| 魏勋等: "基于大规模不平衡数据集的糖尿病诊断研究", 《计算机系统应用》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109920551A (en) * | 2019-01-24 | 2019-06-21 | 华东师范大学 | Autism children social action performance characteristic analysis system based on machine learning |
| CN111984872A (en) * | 2020-09-09 | 2020-11-24 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
| CN111984872B (en) * | 2020-09-09 | 2021-03-16 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
| CN112700863A (en) * | 2020-12-28 | 2021-04-23 | 天津市眼科医院 | Method for accurately evaluating diopter based on Scheimpflug anterior segment morphology and application |
| CN112465657A (en) * | 2021-02-02 | 2021-03-09 | 北京淇瑀信息科技有限公司 | Risk assessment method and device based on tree model feature derivation and electronic equipment |
| WO2023272876A1 (en) * | 2021-06-28 | 2023-01-05 | 上海美沃精密仪器股份有限公司 | Binocular keratoconus diagnosis method based on multi-modal data |
| US11717151B2 (en) | 2021-06-28 | 2023-08-08 | Shanghai MediWorks Precision Instruments Co., Ltd. | Method for early diagnosis of keratoconus based on multi-modal data |
| JP2023540651A (en) * | 2021-06-28 | 2023-09-26 | 上海美沃精密儀器股▲フン▼有限公司 | Binocular keratoconus testing method based on multimodal data |
| JP7454679B2 (en) | 2021-06-28 | 2024-03-22 | 上海美沃精密儀器股▲フン▼有限公司 | Binocular keratoconus evaluation system based on multimodal data |
| CN113611404A (en) * | 2021-07-09 | 2021-11-05 | 哈尔滨智吾康软件开发有限公司 | Plasma sample cancer early screening method based on ensemble learning |
| CN114463014A (en) * | 2022-02-23 | 2022-05-10 | 河南科技大学 | SVM-Xgboost-based mobile payment risk early warning method |
| CN114548845A (en) * | 2022-04-27 | 2022-05-27 | 北京智芯微电子科技有限公司 | Distribution network management method, device and system |
| CN114548845B (en) * | 2022-04-27 | 2022-07-12 | 北京智芯微电子科技有限公司 | Distribution network management method, device and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109256207A (en) | A method of based on XGBoost+SVM hybrid machine Learner diagnosis keratoconus case | |
| Ran et al. | Cataract detection and grading based on combination of deep convolutional neural network and random forests | |
| CN112132817B (en) | Retina blood vessel segmentation method for fundus image based on mixed attention mechanism | |
| Fang et al. | Adam challenge: Detecting age-related macular degeneration from fundus images | |
| CN104809480A (en) | Retinal vessel segmentation method of fundus image based on classification and regression tree and AdaBoost | |
| Paul et al. | Octx: Ensembled deep learning model to detect retinal disorders | |
| CN105957047A (en) | Supervised multimodal brain image fusion method | |
| CN106709421A (en) | Cell image recognition and classification method based on transform domain characteristics and CNN (Convolutional Neural Network) | |
| CN112580580A (en) | Pathological myopia identification method based on data enhancement and model fusion | |
| CN111028230A (en) | Fundus image optic disc and macula lutea positioning detection algorithm based on YOLO-V3 | |
| Firke et al. | Convolutional neural network for diabetic retinopathy detection | |
| Sharma et al. | Harnessing the Strength of ResNet50 to Improve the Ocular Disease Recognition | |
| CN111598866A (en) | Lens key feature positioning method based on eye B-ultrasonic image | |
| CN104978569A (en) | Sparse representation based incremental face recognition method | |
| Choudhary et al. | ANN glaucoma detection using cup-to-disk ratio and neuroretinal rim | |
| Ji et al. | An image diagnosis algorithm for keratitis based on deep learning | |
| Krishnamoorthi et al. | Hybrid feature vector based detection of Glaucoma | |
| US11017323B2 (en) | Method and apparatus for improving a profile analysis of an interpretive framework based on digital measurement of the production of and responses to visual stimuli | |
| Guo et al. | DSLN: Dual-tutor student learning network for multiracial glaucoma detection | |
| CN110189299A (en) | A kind of cerebrovascular events automatic identifying method and system based on MoileNet | |
| Karthik et al. | Design and Implementation of Multi-Retinal Disease Classification Using Deep Neural Network | |
| Sangamesh et al. | A New Approach to Recognize a Patient with Diabetic Retinopathy using Pre-trained Deep Neural Network EfficientNetB0 | |
| Rayavel et al. | Comparative analysis of deep learning classifiers for diabetic retinopathy identification and detection | |
| Wang et al. | Dilation-supervised Learning: A Novel Strategy for Scale Difference in Retinal Vessel Segmentation | |
| George et al. | Convolutional Neural Network for automatic segmentation of EDI OCT images |
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: 20190122 |