CN112309571A - Screening method of prognosis quantitative characteristics of digital pathological image - Google Patents
Screening method of prognosis quantitative characteristics of digital pathological image Download PDFInfo
- Publication number
- CN112309571A CN112309571A CN202011185907.1A CN202011185907A CN112309571A CN 112309571 A CN112309571 A CN 112309571A CN 202011185907 A CN202011185907 A CN 202011185907A CN 112309571 A CN112309571 A CN 112309571A
- Authority
- CN
- China
- Prior art keywords
- characteristic
- feature
- risk
- vector
- value
- 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.)
- Granted
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
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- 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
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
Landscapes
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Physics & Mathematics (AREA)
- Image Analysis (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention discloses a method for screening prognosis quantization characteristics of a digital pathological image, which comprises the step of obtaining quantization characteristics a(0)(ii) a Sequentially adopting a Cox proportional risk algorithm, a machine learning model and a Pearson correlation coefficient to a quantization characteristic a(0)Screening to obtain a characteristic vector a(3)(ii) a Using feature vectors a(3)Constructing a Cox proportional risk model, and updating the feature vector a by adopting the Cox proportional risk model(3)Then, the step is executed once again, and the obtained feature vector a is updated again(3)And inputting a Cox proportional risk model, outputting a characteristic risk coefficient, and calculating a proportional risk value. Discretizing the risk coefficient corresponding to each feature in the feature vector a to obtain a risk evaluation score of each feature; features with an evaluation score greater than the cutoff score are divided into groups a, anddividing into group b; the corresponding features in group a are used as prognostic quantification features.
Description
Technical Field
The invention relates to a medical image extraction technology, in particular to a method for screening prognosis quantization characteristics of a digital pathological image.
Background
The occurrence and development of cancer are the result of the interaction between cancer cells and the tumor microenvironment, and the change of the types, the numbers or the forms of the cells in the tumor stroma has important medical guidance significance. For example, lymphocyte infiltrates in breast cancer generally have a better prognosis, while the presence of tumor-associated fibroblasts suggests a poor prognosis. In routine pathological work, changes in the cellular components and extracellular matrix in the tumor stroma are generally described qualitatively. Based on digital pathological image analysis, different components in the interstitium can be automatically segmented, and quantitative or qualitative research can be carried out.
In quantitative studies, using morphological quantitative characterization of pathological images, researchers found that there was a significant difference in prognosis between patients with low and high nuclear areas of the cell nucleus. To extract a rich set of quantitative features in breast cancer epithelial cells and stroma (6642 features), researchers at Stanford university have developed the C-Path System (Computational Patholoist) for measuring standard morphological descriptors and higher-level context, relationships, and global image features that include image objects.
At present, when a medical researcher analyzes a patient prognosis digital pathological image, only morphological quantitative characteristics closely related to tumor recurrence need to be analyzed to search the cause of tumor recurrence; however, at present, a data set downloaded from a public database by a medical researcher often includes all morphological quantitative features, the data volume is large, if the morphological quantitative features closely related to tumor recurrence are manually searched and screened from the public database, the workload of the researcher is very large, and in manual identification, visual fatigue is generated due to the fact that a large amount of information is browsed, and the identification accuracy is difficult to guarantee.
Disclosure of Invention
Aiming at the defects in the prior art, the method for screening the prognosis quantitative characteristics of the digital pathological image can quickly screen the morphological quantitative characteristics closely related to tumor recurrence.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that:
a screening method for prognosis quantitative characteristics of a digital pathological image is provided, which comprises the following steps:
s1, acquiring all morphological quantitative characteristics of cancer prognosis digital pathological image as quantitative characteristics a(0)；
S2, sequentially adopting a Cox proportional risk algorithm, a machine learning model and a Pearson correlation coefficient to quantization feature a(0)Screening to obtain a characteristic vector a(3)；
S3, converting the feature vector a(3)Inputting the follow-up time information and the follow-up event information corresponding to the follow-up time information and the follow-up event information into a Cox proportional risk algorithm to construct a Cox proportional risk model;
s4, judging the characteristic vector a output by the Cox proportional risk model(3)If the p value of the middle feature is less than the set threshold value, if so, the feature vector a(3)Otherwise, go to step S5;
s5, increasing the preset threshold value by the preset value, and judging whether the p value of the current feature is smaller than the updated set threshold value, if so, in the feature vector a(3)If not, deleting the current characteristics;
s6, based on the adjusted feature vector a(3)Repeating the steps S3 to S5 once to obtain the finally determined feature vector a;
s7, constructing a proportional risk value expression F-d according to the characteristic risk coefficient obtained by the Cox proportional risk model in the characteristic input step S6 in the characteristic vector aTA, d are column vectors formed by all the characteristic risk coefficients, T is transposition, and a is a characteristic vector a;
s8, discretizing the characteristic risk coefficient of each characteristic in the characteristic vector a calculated by the comparative example risk value expression to obtain the evaluation score of each characteristic;
s9, dividing the characteristics with the evaluation scores larger than the cutoff values into a group a, and dividing the rest into a group b; the corresponding features in group a are used as prognostic quantification features.
The invention has the beneficial effects that: according to the scheme, dimension reduction can be performed on the screened quantitative features through a machine learning model, and morphological quantitative features closely related to cancer prognosis recurrence can be selected from a plurality of morphological quantitative features through a Cox proportional risk algorithm/model and evaluation scores. The medical researchers can analyze and research the incentive and the pathological changes of corresponding tissues when the subsequent cancers relapse based on the morphological quantitative characteristics obtained by the scheme.
By adopting the method for screening, a medical researcher does not need to manually classify numerous morphological quantitative characteristics, and after a large amount of data of each type are analyzed, closely related morphological quantitative characteristics are searched, so that the working intensity of the researcher is reduced, and the accuracy can be ensured.
Drawings
Fig. 1 is a flowchart of a screening method for prognostic quantitative features of a digital pathology image.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
Referring to fig. 1, fig. 1 shows a flow chart of a screening method of prognostic quantification characteristics of a digital pathology image; as shown in fig. 1, the method S includes steps S1 to S9.
In step S1, all morphological quantitative features of the cancer prognosis digital pathology image are acquired as the quantitative feature a(0)Quantifying the feature a(0)May be downloaded from a public medical database.
In step S2, the Cox proportional hazards algorithm, the machine learning model and the pearson correlation coefficient are used in sequence to quantify the feature a(0)Screening to obtain a characteristic vector a(3)。
In an embodiment of the present invention, the step S2 further includes:
s21, mixingQuantization feature a(0)The characteristic of the Cox proportional risk model is input into a Cox proportional risk algorithm one by one to construct a Cox proportional risk model, and a p value smaller than a threshold value T is selected according to 0 hypothesis test of the Cox proportional risk modelpForming a feature vector a(1)；
S22, converting the feature vector a(1)Inputting the machine learning model to obtain a characteristic vector a(1)The importance score of each feature in the feature vector a is selected, and the feature with the importance score larger than a preset score is selected to form a feature vector a(2)；
S23, calculating a characteristic vector a according to the Pearson correlation coefficient(2)The correlation coefficient between any two characteristics is screened out, and the characteristic vector a is formed by the characteristic pair with the correlation coefficient smaller than the preset coefficient(3)；
S24, for the feature pairs with the correlation coefficient more than or equal to the preset coefficient, updating the feature easy to understand in each pair to the feature vector a(3)To obtain the final feature vector a(3)。
Where readily understandable primarily refers to variables that are clinically readily understandable, such as,
the variance of the area of the cell nucleus is represented,
representing the weighted Hu5 moment, wherein
Is difficult to interpret and selects to reserve
To get rid of
In practice, the present scheme preferably provides that the training method of the machine learning model includes steps a1 to a 5:
a1, acquiring all morphological quantitative characteristics of the digital pathological image of cancer prognosis;
step A2, inputting the morphological quantitative characteristics into a Cox proportional risk model one by one to obtain characteristic coefficients, and adopting the morphological quantitative characteristics of which the characteristic coefficients are smaller than a preset threshold value to form a data set a;
step A3, dividing a data set a into a training set and a verification set a, and dividing morphological quantitative features corresponding to the same patient into the same training set or verification set;
a4, dividing a training set a into a training set b and a verification set b by adopting a cross verification method, and searching for the optimized hyperparameter by adopting a machine learning method based on the survival analysis and based on the Bayesian hyperparameter search package hyperopt training;
the machine learning method based on the survival analysis is XGboost algorithm or random forest generation.
And step A5, training to obtain the optimal machine learning model by using the searched optimization hyperparameters.
In step S3, the feature vector a is added(3)And inputting the follow-up time information and follow-up event information (probability of certain specific event, such as death and disease recurrence, at the observation time point) corresponding to the follow-up time information and the follow-up event information into a Cox proportional risk algorithm to construct a Cox proportional risk model.
In step S4, the feature vector a of the Cox proportional hazards model output is determined(3)If the p value of the middle feature is less than the set threshold value, if so, the feature vector a(3)Otherwise, the process proceeds to step S5.
In step S5, the preset threshold is increased by a preset value, and it is determined whether the p value of the current feature is smaller than the updated set threshold, if yes, in the feature vector a(3)The current feature is retained, otherwise it is deleted.
In step S6, the feature vector a is adjusted based on(3)And repeating the steps S3 to S5 once to obtain the finally determined feature vector a.
In step S7, a proportional risk value expression F ═ d is constructed from the feature risk coefficients obtained by the Cox proportional risk model in the feature input step S6 in the feature vector aTA, d is the structure of all characteristic risk coefficientsForming a column vector, wherein T is a transposition, and a is a characteristic vector a;
in one embodiment of the present invention, the method for obtaining the cutoff value includes:
determining the number T of correctly predicted positive samples in preset time by adopting an ROC curve method based on Hegerty time dependence according to the proportional risk value, the visit time information corresponding to the characteristic vector a and the follow-up event informationPAnd number of mispredicted positive samples FP；
The number of positive samples T is correctly predictedPAnd number of mispredicted positive samples FPCalculating Yuden index J ═ maxc{TP+FP-1} and let cutoff be c*And c is a cutoff value; c. C*Is the optimum cutoff value.
In step S8, the feature risk coefficient of each feature in the feature vector a calculated by the comparative example risk value expression is discretized to obtain an evaluation score of each feature:
Q＝hz*10/(cutoff*5)
wherein Q is an intermediate coefficient; hz is a proportional risk value; score is the assessment score.
In step S9, the features whose evaluation scores are greater than the cutoff value are classified into group a, and the rest into group b; the corresponding features in group a are used as prognostic quantification features.
In conclusion, the screening method adopting the scheme can reduce the work intensity of classifying morphological quantitative features by medical researchers, and can screen out morphological quantitative features closely related to cancer prognosis recurrence so as to be used for the medical researchers to research the cause of cancer recurrence and related tissue pathological changes.
Claims (6)
1. The method for screening the prognosis quantitative characteristics of the digital pathological image is characterized by comprising the following steps:
s1, obtaining cancer prognosis numberAll morphological quantitative features of pathological images are taken as quantitative features a(0)；
S2, sequentially adopting a Cox proportional risk algorithm, a machine learning model and a Pearson correlation coefficient to quantization feature a(0)Screening to obtain a characteristic vector a(3)；
S3, converting the feature vector a(3)Inputting the follow-up time information and the follow-up event information corresponding to the follow-up time information and the follow-up event information into a Cox proportional risk algorithm to construct a Cox proportional risk model;
s4, judging the characteristic vector a output by the Cox proportional risk model(3)If the p value of the middle feature is less than the set threshold value, if so, the feature vector a(3)Otherwise, go to step S5;
s5, increasing the preset threshold value by the preset value, and judging whether the p value of the current feature is smaller than the updated set threshold value, if so, in the feature vector a(3)If not, deleting the current characteristics;
s6, based on the adjusted feature vector a(3)Repeating the steps S3 to S5 once to obtain the finally determined feature vector a;
s7, constructing a proportional risk value expression F-d according to the characteristic risk coefficient obtained by the Cox proportional risk model in the characteristic input step S6 in the characteristic vector aTA, d are column vectors formed by all the characteristic risk coefficients, T is transposition, and a is a characteristic vector a;
s8, discretizing the characteristic risk coefficient of each characteristic in the characteristic vector a calculated by the comparative example risk value expression to obtain the evaluation score of each characteristic;
s9, dividing the characteristics with the evaluation scores larger than the cutoff values into a group a, and dividing the rest into a group b; the corresponding features in group a are used as prognostic quantification features.
2. The method for screening prognostic quantization characteristics of digital pathology images according to claim 1, wherein said step S2 further includes:
s21, quantizing the feature a(0)The characteristic of the Cox ratio is input into a Cox ratio risk algorithm one by one to construct a Cox ratioSelecting p value smaller than threshold T according to 0 hypothesis test of Cox proportional risk modelpForming a feature vector a(1)；
S22, converting the feature vector a(1)Inputting the machine learning model to obtain a characteristic vector a(1)The importance score of each feature in the feature vector a is selected, and the feature with the importance score larger than a preset score is selected to form a feature vector a(2)；
S23, calculating a characteristic vector a according to the Pearson correlation coefficient(2)The correlation coefficient between any two characteristics is screened out, and the characteristic vector a is formed by the characteristic pair with the correlation coefficient smaller than the preset coefficient(3)；
S24, for the feature pairs with the correlation coefficient more than or equal to the preset coefficient, updating the feature easy to understand in each pair to the feature vector a(3)To obtain the final feature vector a(3)。
3. The method for screening prognostic quantitative features of digital pathological images according to claim 2, wherein the training method of machine learning model includes:
acquiring all morphological quantitative characteristics of a digital pathological image for cancer prognosis;
inputting the morphological quantitative characteristics into a Cox proportional risk model one by one to obtain characteristic coefficients, and adopting the morphological quantitative characteristics of which the characteristic coefficients are smaller than a preset threshold value to form a data set a;
dividing a data set a into a training set and a verification set a, and dividing morphological quantitative features corresponding to the same patient into the same training set or verification set;
dividing the training set a into a training set b and a verification set b by adopting a cross validation method, and searching for the optimized hyperparameter by adopting a machine learning method based on survival analysis and based on Bayesian hyperparameter search package hyperopt training;
and training to obtain the optimal machine learning model by using the searched optimized hyper-parameter.
4. The method for screening prognostic quantification characteristics of digital pathological images according to claim 3, wherein the machine learning method based on survival analysis is XGboost algorithm or random forest generation.
5. The method for screening prognostic quantification characteristics of digital pathology images according to claim 1, characterized in that said method for obtaining the cutoff value comprises:
determining the number T of correctly predicted positive samples in preset time by adopting an ROC curve method based on Hegerty time dependence according to the proportional risk value, the visit time information corresponding to the characteristic vector a and the follow-up event informationPAnd number of mispredicted positive samples FP；
The number of positive samples T is correctly predictedPAnd number of mispredicted positive samples FPCalculating Yuden index J ═ maxc{TP+FP-1} and let cutoff be c*And c is a cutoff value; c. C*Is the optimum cutoff value.
6. The method for screening prognostic quantization characteristics of digital pathology images according to claim 5, wherein the formula for discretizing the comparative example risk value is:
Q＝hz*10/(cutoff*5)
wherein Q is an intermediate coefficient; hz is a proportional risk value; score is the assessment score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011185907.1A CN112309571B (en) | 2020-10-30 | 2020-10-30 | Screening method of prognosis quantitative characteristics of digital pathological image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011185907.1A CN112309571B (en) | 2020-10-30 | 2020-10-30 | Screening method of prognosis quantitative characteristics of digital pathological image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112309571A true CN112309571A (en) | 2021-02-02 |
| CN112309571B CN112309571B (en) | 2022-04-15 |
Family
ID=74331883
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011185907.1A Active CN112309571B (en) | 2020-10-30 | 2020-10-30 | Screening method of prognosis quantitative characteristics of digital pathological image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112309571B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662698A (en) * | 2022-02-11 | 2022-06-24 | 南京英锐祺科技有限公司 | Industrial internet multi-modal machine learning data processing method |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100297618A1 (en) * | 2006-06-23 | 2010-11-25 | Christopher Hutchison | Methods for determining a prognosis of colorectal cancer |
| CN105277718A (en) * | 2015-09-29 | 2016-01-27 | 上海知先生物科技有限公司 | Product for malignant tumor related screening and assessing, and application and method thereof |
| CN108962382A (en) * | 2018-05-31 | 2018-12-07 | 电子科技大学 | A kind of layering important feature selection method based on breast cancer clinic high dimensional data |
| CN109670510A (en) * | 2018-12-21 | 2019-04-23 | 万达信息股份有限公司 | A kind of gastroscopic biopsy pathological data screening system and method based on deep learning |
| CN110177886A (en) * | 2017-04-24 | 2019-08-27 | 洛博生物科技有限公司 | Cluster classification and prognosis prediction system based on gastric cancer biological property |
| CN110305964A (en) * | 2019-07-26 | 2019-10-08 | 安徽医科大学第一附属医院 | A kind of foundation of patients with prostate cancer prognosis recurrence risk profile mark tool and its risk evaluation model |
| CN110423816A (en) * | 2018-10-12 | 2019-11-08 | 天津医科大学肿瘤医院 | Prognosis in Breast Cancer quantitative evaluation system and application |
| CN110770849A (en) * | 2017-06-16 | 2020-02-07 | 南托米克斯有限责任公司 | PRAEGNANT prognostic indicators of poor outcome in a group of metastatic breast cancers |
| CN110993106A (en) * | 2019-12-11 | 2020-04-10 | 深圳市华嘉生物智能科技有限公司 | Liver cancer postoperative recurrence risk prediction method combining pathological image and clinical information |
| CN111094977A (en) * | 2017-07-13 | 2020-05-01 | 古斯塔夫·鲁西研究所 | Imaging tools based on imaging omics to monitor tumor lymphocyte infiltration and prognosis in anti-PD-1/PD-L1 treated tumor patients |
| CN111462042A (en) * | 2020-03-03 | 2020-07-28 | 西北工业大学 | Cancer prognosis analysis method and system |
-
2020
- 2020-10-30 CN CN202011185907.1A patent/CN112309571B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100297618A1 (en) * | 2006-06-23 | 2010-11-25 | Christopher Hutchison | Methods for determining a prognosis of colorectal cancer |
| CN105277718A (en) * | 2015-09-29 | 2016-01-27 | 上海知先生物科技有限公司 | Product for malignant tumor related screening and assessing, and application and method thereof |
| CN110177886A (en) * | 2017-04-24 | 2019-08-27 | 洛博生物科技有限公司 | Cluster classification and prognosis prediction system based on gastric cancer biological property |
| CN110770849A (en) * | 2017-06-16 | 2020-02-07 | 南托米克斯有限责任公司 | PRAEGNANT prognostic indicators of poor outcome in a group of metastatic breast cancers |
| CN111094977A (en) * | 2017-07-13 | 2020-05-01 | 古斯塔夫·鲁西研究所 | Imaging tools based on imaging omics to monitor tumor lymphocyte infiltration and prognosis in anti-PD-1/PD-L1 treated tumor patients |
| CN108962382A (en) * | 2018-05-31 | 2018-12-07 | 电子科技大学 | A kind of layering important feature selection method based on breast cancer clinic high dimensional data |
| CN110423816A (en) * | 2018-10-12 | 2019-11-08 | 天津医科大学肿瘤医院 | Prognosis in Breast Cancer quantitative evaluation system and application |
| CN109670510A (en) * | 2018-12-21 | 2019-04-23 | 万达信息股份有限公司 | A kind of gastroscopic biopsy pathological data screening system and method based on deep learning |
| CN110305964A (en) * | 2019-07-26 | 2019-10-08 | 安徽医科大学第一附属医院 | A kind of foundation of patients with prostate cancer prognosis recurrence risk profile mark tool and its risk evaluation model |
| CN110993106A (en) * | 2019-12-11 | 2020-04-10 | 深圳市华嘉生物智能科技有限公司 | Liver cancer postoperative recurrence risk prediction method combining pathological image and clinical information |
| CN111462042A (en) * | 2020-03-03 | 2020-07-28 | 西北工业大学 | Cancer prognosis analysis method and system |
Non-Patent Citations (2)
| Title |
|---|
| WEN LUO等: "hsa-mir-3199-2 and hsa-mir-1293 as Novel Prognostic Biomarkers of Papillary Renal Cell Carcinoma by COX Ratio Risk Regression Model Screening", 《JOURNAL OF CELLULAR BIOCHEMISTRY》 * |
| 罗治文等: "机器学习算法和COX列线图在肝细胞癌术后生存预测中的应用价值", 《中华消化外科杂志》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662698A (en) * | 2022-02-11 | 2022-06-24 | 南京英锐祺科技有限公司 | Industrial internet multi-modal machine learning data processing method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112309571B (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sahasrabudhe et al. | Self-supervised nuclei segmentation in histopathological images using attention | |
| CN110728224A (en) | Remote sensing image classification method based on attention mechanism depth Contourlet network | |
| CN112635063B (en) | Comprehensive lung cancer prognosis prediction model, construction method and device | |
| CN110738247B (en) | Fine-grained image classification method based on selective sparse sampling | |
| CN111462042B (en) | Cancer prognosis analysis method and system | |
| CN103714261A (en) | Intelligent auxiliary medical treatment decision supporting method of two-stage mixed model | |
| Ferlaino et al. | Towards deep cellular phenotyping in placental histology | |
| CN112543934A (en) | Method for determining degree of abnormality, corresponding computer readable medium and distributed cancer analysis system | |
| CN111028944A (en) | Cerebrovascular disease nerve function damage degree prediction model based on kernel principal component analysis and polynomial characteristics | |
| CN114445356A (en) | Multi-resolution-based full-field pathological section image tumor rapid positioning method | |
| CN112309571B (en) | Screening method of prognosis quantitative characteristics of digital pathological image | |
| Dessouky et al. | Feature extraction of the Alzheimer’s disease images using different optimization algorithms | |
| EP4138091A1 (en) | Cancer prognosis | |
| Wolcott et al. | Automated classification of estrous stage in rodents using deep learning | |
| CN107480441A (en) | A kind of modeling method and system of children's septic shock prognosis prediction based on SVMs | |
| CN114580501A (en) | Bone marrow cell classification method, system, computer device and storage medium | |
| CN114187979A (en) | Data processing, model training, molecular prediction and screening method and device thereof | |
| CN112489689B (en) | Cross-database voice emotion recognition method and device based on multi-scale difference countermeasure | |
| CN116228759B (en) | Computer-aided diagnosis system and apparatus for renal cell carcinoma type | |
| CN112861881A (en) | Honeycomb lung recognition method based on improved MobileNet model | |
| Qattous et al. | PaCMAP-embedded convolutional neural network for multi-omics data integration | |
| Pham | Time-frequency time-space long short-term memory networks for image classification of histopathological tissue | |
| CN108319935B (en) | Face group identification method based on region sparsity | |
| Alfyani | Comparison of Naïve Bayes and KNN algorithms to understand hepatitis | |
| US20220044762A1 (en) | Methods of assessing breast cancer using machine learning systems |
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 |