CN109166105B - Tumor malignancy risk layered auxiliary diagnosis system based on artificial intelligent medical image - Google Patents
Tumor malignancy risk layered auxiliary diagnosis system based on artificial intelligent medical image Download PDFInfo
- Publication number
- CN109166105B CN109166105B CN201810866088.3A CN201810866088A CN109166105B CN 109166105 B CN109166105 B CN 109166105B CN 201810866088 A CN201810866088 A CN 201810866088A CN 109166105 B CN109166105 B CN 109166105B
- Authority
- CN
- China
- Prior art keywords
- model
- data
- malignant
- diagnosis
- image
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- 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
- 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
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
- G06T2207/30064—Lung nodule
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Biomedical Technology (AREA)
- Radiology & Medical Imaging (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Pathology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a tumor malignancy risk layered auxiliary diagnosis system of artificial intelligent medical images, which comprises: the system comprises a data acquisition module, a data preprocessing module, a model establishing module, a model verifying and optimizing module, a layered diagnosis module and a database platform. The tumor malignant risk layering auxiliary diagnosis system can realize successive layering of malignant risks of tumors based on an artificial intelligence technology, simulates a clinical diagnosis idea, automatically diagnoses space-occupying lesions with clear image characteristics by high-precision benign lesion detection and malignant tumor detection capabilities of an artificial intelligence model, thereby being capable of substantially assisting space-occupying lesion clinical management decisions, improving the existing working process of clinical diagnosis, increasing diagnosis confidence of doctors, relieving working pressure, reducing anxiety of low-malignant-risk lesion patients, greatly improving the diagnosis rate of benign lesions and malignant tumors, and being expected to realize the floor implementation of artificial intelligence clinical auxiliary diagnosis.
Description
Technical Field
The invention relates to a system and a method for auxiliary diagnosis of malignant tumor risk stratification, in particular to a system and a method for auxiliary diagnosis of malignant tumor risk stratification of artificial intelligent medical images; belongs to the technical field of artificial intelligence auxiliary diagnosis of medical images.
Background
Tumor refers to a new organism formed by local histiocyte hyperplasia under the action of various tumorigenic factors. Most benign lesions do not require surgical intervention when they do not have space occupying effects. The prognosis of early surgical intervention of malignant tumor is much better, while the prognosis of late surgical intervention or radiotherapy and chemotherapy is generally worse. At present, medical institutions all invest great investment to research the occurrence, development and treatment of tumors, and doctors mostly rely on their own experience and knowledge in the process of tumor diagnosis and refer to the medical guidelines of the existing medicine to diagnose and treat at the present stage, so that the problems of insufficient diagnosis confidence, low efficiency, fussy work and the like are caused.
Taking lung cancer as an example, it causes about 137 million deaths worldwide each year, accounting for 18% of all cancer deaths, and is the highest malignant tumor in global morbidity and mortality. With the development of the CT technology, the detection rate of lung nodules is rapidly increased, which not only causes the diagnosis workload of imaging physicians to sharply rise, but also brings great challenges to clinical decision and management of lung nodules. In order to assist imaging doctors in clinical diagnosis of space-occupying lesions, improve the working efficiency and reduce the working intensity, some artificial intelligence has been explored in recent years to assist in diagnosis of benign and malignant space-occupying lesions.
The invention patent with application number 201710737538.4 discloses a lesion benign and malignant risk layering auxiliary diagnosis system which comprises the steps of nodule detection, doctor guidance, semantic annotation generation, sample generation, online learning and the like, wherein the doctor guidance is required, namely the doctor marks an attention point, the prediction probability of a pulmonary nodule outline in the attention area and each pixel point in the outline is automatically calculated, and a pulmonary nodule annotation sample containing an accurate semantic outline is generated for the learning of a self-learning system.
In general, most of the current systems are trained by using an open data set or a single-center data set, and the generalization and robustness of the model need to be tested; meanwhile, the data quality is uneven, partial data exists or even no definite pathological diagnosis information exists, and the reliability of the data and the effective sample volume of the data are questioned; in addition, most of the current systems adopt a classification idea, and seek high accuracy or high AUC (area under ROC curve), but the classification mode based on malignancy does not consider the practical situation of clinical application, and cannot practically provide reference for clinical decision management of placeholder lesions.
In view of the above, there is a need to intensively study the artificial intelligent image-assisted diagnosis technology to help clinicians to reduce the working pressure and benefit patients.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide an artificial intelligent medical image tumor malignancy risk layering auxiliary diagnosis system, which improves the diagnosis confidence of clinicians and assists in clinical management decision of neoplastic lesions.
The malignant tumor risk stratification auxiliary diagnosis system of artificial intelligent medical image comprises:
a data acquisition module: the system is used for acquiring the existing data, including image images, clinical information, gene information and pathological benign and malignant labels, verifying and evaluating the data integrity and the data quality, performing data desensitization processing and sending the data desensitization processing to a data preprocessing module;
a data preprocessing module: carrying out noise reduction and data normalization processing on the acquired existing data, searching for a pathological change corresponding to a pathological result, and carrying out pathological change labeling and segmentation on the image;
a model building module: based on the lesions obtained by the data preprocessing module, carrying out feature extraction and feature selection on the region of interest, screening out low-redundancy high-correlation features, establishing a tumor malignant risk hierarchical model based on medical image images, and selecting model parameters and threshold values of a high-sensitivity benign lesion detection model and a high-specificity malignant lesion detection model according to different application scenes: for clinical screening application scenes, a high-sensitivity benign lesion detection threshold is selected, so that the negative predictive value/true benign rate is more than 99%; aiming at a definite diagnosis scene of malignant lesions, selecting a high-specificity malignant lesion detection threshold value to ensure that the positive predictive value/true malignancy rate is more than 99 percent, and layering extremely high-risk lesions;
a model verification and optimization module: the method is used for evaluating the effectiveness, the generalization and the robustness of the model under different parameters and different scenes, iterating the performance of the model by adopting a reward and punishment mechanism, gradually modifying the parameters of the model on the premise of ensuring high sensitivity and high specificity, and improving the diagnosis capability and the accuracy;
a hierarchical diagnostic module: diagnosing the detection sample, layering the tumor into extremely low risk lesion, extremely high risk lesion and uncertain risk lesion, and feeding back to the terminal equipment and the database platform of the doctor;
a database platform: and the module is connected with the modules and used for establishing a model, storing and calling data, continuously collecting newly added data and providing an online updating function of the model.
Preferably, the sources of the aforementioned existing data include: the data set and the case data of hospitals at all levels are disclosed. For example, for Lung cancer, the public data sets include, but are not limited to, data sets such as nlst (national Lung Screening trial) and lid-idri (the Lung Image Database patient Image collection), all levels of hospitals include community hospitals, third-generation hospitals, and the like, data sources cover all types of medical institutions and people, high-quality images and common clinical disease spectra as much as possible, Image inspection equipment includes all current home and abroad common manufacturers, such as siemens, philips, toshiba, GE, eastern softy, united shadow, and the like, and influence of other factors on the diagnosis result is reduced as much as possible.
Preferably, in the system for diagnosing malignancy risk stratification of a tumor based on artificial intelligence medical images, when acquiring lung nodule data, the existing data acquisition criteria are: all the incorporated nodules are primary nodules with the size smaller than 3cm, the benign labels are obtained from operation pathology, the malignant labels are obtained by means of operation, puncture or bronchoscopy and the like, and the image is a chest CT flat scan without obvious artifacts.
More preferably, the desensitization process includes the privacy information of the patient and the medical institution information being hidden, so that it is ensured that the privacy of the patient is not compromised.
More preferably, in the data preprocessing module, a physician searches for a lesion corresponding to a pathological result, and then performs contour segmentation and local block segmentation on the lesion by using an automatic segmentation method confirmed by physician modification or a manual segmentation method by the physician.
Further preferably, the reward and punishment mechanism is as follows: in the training set, firstly constructing a classification model and marking benign and malignant labels; identifying error images by artificial intelligence, and applying penalty factors to previous models for images which should not be subjected to error in definite diagnosis by doctors under the guidance of comprehensive opinions of three or more medical experts; identifying accurate images for artificial intelligence, and applying reward factors in model improvement; and modifying to obtain a new model by combining the reward and punishment factors and the original model, circularly iterating and optimizing the model to finally obtain an ideal benign and malignant layered model, and continuously perfecting and optimizing the model by virtue of a reward and punishment mechanism to improve the accuracy and reliability of a layered diagnosis result.
In addition, the invention also discloses an auxiliary diagnosis method for the tumor malignancy risk stratification based on the artificial intelligent medical image, which is based on the auxiliary diagnosis system and comprises the following aspects:
(1) and collecting data: acquiring the existing medical data in multiple ways, and screening and desensitizing the medical data;
(2) and image segmentation: carrying out contour segmentation and local square segmentation on the acquired image;
(3) and (3) feature extraction: extracting the image characteristics of the image in the region of interest, wherein the image characteristics comprise histogram characteristics, texture characteristics, shape characteristics, filtering transformation characteristics and convolution neural network characteristics;
(4) and selecting characteristics: selecting the stable-performance characteristics by using a characteristic stability test data set repeatedly scanned for a short time, screening highly-relevant characteristics in the test data set by adopting a univariate analysis method, and finally screening the low-redundancy and highly-relevant characteristics;
(5) and establishing a model: performing model building by adopting various machine learning and deep learning algorithms, and building a model by adopting cross validation;
(6) and optimizing the model: the integrated learning method comprises an integrated learning and reward and punishment mechanism, wherein the integrated learning is to combine a plurality of classifiers, and the generalization performance of the integrated learning is often better than that of a single classifier; the reward and punishment mechanism is that in the training set, firstly, a classification model is constructed, and benign and malignant labels are marked; identifying error images by artificial intelligence, and applying penalty factors to previous models for images which should not be subjected to error in definite diagnosis by doctors under the guidance of comprehensive opinions of three or more medical experts; identifying accurate images for artificial intelligence, and applying reward factors in model improvement; modifying to obtain a new model by combining the reward and punishment factors and the original model, and circularly iterating and optimizing the model to finally obtain an ideal benign and malignant layered model;
(7) and verifying the model: performing model verification on a verification data set, drawing an ROC curve, searching a specificity index as high as possible in a high-sensitivity model, wherein the corresponding specificity index is more than or equal to 20% when the sensitivity is 99%; searching a sensitivity index as high as possible in a high specificity model, wherein the sensitivity index corresponding to 99% of specificity is more than or equal to 20%; calculating a positive predicted value (true malignancy rate), a negative predicted value (true malignancy rate) and an accuracy index;
(8) and clinical diagnosis: collecting test data, carrying out tumor layered diagnosis and outputting a diagnosis result.
Preferably, the aforementioned filtering transformation comprises: Gaussian-Laplace filtering and/or wavelet transformation, and the univariate analysis method is a t-test method and/or an analysis of variance method.
More preferably, the low Redundancy, highly correlated features are screened using a minimum Redundancy Maximum correlation (mrmr), minimum absolute value shrinkage and selector (Least absolute value shrinkage and selection operator).
Further preferably, the aforementioned machine learning and deep learning algorithm includes: linear and non-linear Support Vector Machines (SVMs), random forests, decision trees, k-nearest neighbors, logistic regression, convolutional neural networks, challenge generation networks, and migratory learning.
The invention has the advantages that:
(1) the auxiliary diagnosis system for the malignant tumor risk layering can realize successive layering of benign and malignant tumor risks by means of an artificial intelligence technology, simulate a clinical diagnosis idea, and automatically diagnose occupied lesion with clear image characteristics by means of high-precision benign lesion detection and malignant tumor detection capabilities of an artificial intelligence model, so that clinical management decisions of the occupied lesion can be substantially assisted, the existing working process of clinical diagnosis is improved, diagnosis confidence of doctors is increased, working pressure is relieved, and anxiety of low-malignant-risk lesion patients is reduced;
(2) the system can continuously verify and optimize the model, solves the problems that computer auxiliary software cannot feed back and self-learn and self-update in the prior art, and improves the accuracy and reliability of tumor stratification, through the diagnosis system, the sensitivity of the high-sensitivity model reaches over 99 percent, the specificity reaches over 20 percent, the specificity of the high-specificity model reaches over 99 percent, the sensitivity reaches over 20 percent, and the diagnosis rate of benign lesions and malignant tumors is greatly improved;
(3) the system has a reward and punishment mechanism, provides a more effective mode and higher efficiency for the improvement of the artificial intelligent learning ability by relying on a high-quality doctor knowledge base, and is expected to realize the landing implementation of artificial intelligent clinical auxiliary diagnosis by relying on a clinical diagnosis flow.
Drawings
FIG. 1 is a schematic structural diagram of a system for auxiliary diagnosis of tumor malignancy risk stratification according to the present invention;
FIG. 2 is a schematic diagram of the hierarchical management decision of the system for auxiliary diagnosis of malignancy risk hierarchy of tumor of the present invention;
FIG. 3 is a schematic flow chart of the tumor malignancy risk stratification aided diagnosis method of the present invention.
Detailed Description
The present invention will be described in detail with reference to the following embodiments.
The medical image of the present invention includes, but is not limited to, X-ray images, CT images, MRI images, and ultrasonic detection, and the system of the present invention may be integrated with existing medical imaging devices, or may be installed independently or on a terminal such as a server.
Referring to fig. 1, the system for diagnosing malignancy risk stratification of tumor according to the present invention is implemented by artificial intelligence medical imaging, and comprises:
(1) a data acquisition module: the data desensitization processing system is used for collecting the existing data, including image images, clinical information, gene information and pathological benign and malignant labels, carrying out data desensitization processing and sending the data desensitization processing to the data preprocessing module after verifying and evaluating the data integrity and the data quality, wherein the desensitization processing includes hiding privacy information of patients and information of medical institutions, and therefore the privacy of the patients can be guaranteed not to be leaked.
Wherein, the sources of the existing data comprise: the data set and the case data of hospitals at all levels are disclosed. For example, the public data sets include, but are not limited to, data sets such as nlst (national Lung Screening trial) and lid-idri (the Lung Image Database patient Image collection), hospitals of all levels include hospitals of community, third-generation, and the like, data sources include various medical institutions and people, high-quality images and common clinical disease spectra as much as possible, Image inspection equipment includes various common manufacturers at home and abroad at present, such as siemens, philips, toshiba, GE, eastern softy, united shadows, and the like, and influence of other factors on the hierarchical diagnosis result is reduced as much as possible.
Taking a CT image of a lung as an example, the existing data acquisition standard is: all the incorporated pulmonary nodules are primary nodules with the size smaller than 3cm, the benign labels are obtained from operation pathology, the malignant labels are obtained by means of operations, puncture or bronchoscopy biopsy and the like, the image images are chest CT flat scans without obvious artifacts, and data are effectively screened according to the standards, so that the data validity can be better guaranteed. Of course, the collection method is also suitable for organs such as stomach, thyroid gland, mammary gland and the like, and corresponding collection standards are provided according to clinical guidance.
(2) A data preprocessing module: and carrying out noise reduction and data normalization processing on the acquired existing data, searching for a pathological change corresponding to a pathological result, and carrying out pathological change labeling and segmentation on the image. In this module, a physician finds a lesion corresponding to a pathological result, and then performs contour segmentation and local block segmentation on the lesion by using an automatic segmentation method confirmed by the physician or a manual segmentation method by the physician. At present, segmentation can be realized by an automatic segmentation algorithm for lung segmentation, and other parts except for lung have no automatic segmentation algorithm and mostly depend on manual segmentation of doctors.
(3) A model building module: based on the lesions obtained by the data preprocessing module, carrying out feature extraction and feature selection on the region of interest, screening out low-redundancy high-correlation features, establishing a tumor malignant risk hierarchical model based on medical image images, and selecting model parameters and threshold values of a high-sensitivity benign lesion detection model and a high-specificity malignant lesion detection model according to different application scenes: for a clinical screening application scene, selecting a high-sensitivity benign lesion detection threshold value, enabling a negative prediction value/true benign rate to be more than 99%, and layering extremely-low-risk lesions; aiming at a malignant lesion diagnosis scene, a high-specificity malignant lesion detection threshold is selected, so that the positive predictive value/true malignancy rate is more than 99%, and extremely high-risk lesions are layered.
(4) A model verification and optimization module: the method is used for evaluating the effectiveness, the generalization and the robustness of the model under different parameters and different scenes, iterating the performance of the model by adopting a reward and punishment mechanism, gradually modifying the parameters of the model on the premise of ensuring high sensitivity and high specificity, and improving the diagnosis capability and the accuracy.
The reward and punishment mechanism in the step plays a crucial role in model optimization, the problems that computer auxiliary software cannot feed back and self-learn and self-update in the prior art are solved, and the accuracy and the reliability of tumor layering are improved. The specific implementation of the mechanism is as follows: in the training set, firstly constructing a classification model and marking benign and malignant labels; identifying error images by artificial intelligence, and applying penalty factors to previous models for images which should not be subjected to error in definite diagnosis by doctors under the guidance of comprehensive opinions of three or more medical experts; identifying accurate images for artificial intelligence, and applying reward factors in model improvement; and modifying to obtain a new model by combining the reward and punishment factors and the original model, circularly iterating and optimizing the model to finally obtain an ideal benign and malignant layered model, and continuously perfecting and optimizing the model by virtue of a reward and punishment mechanism to improve the accuracy and reliability of a layered diagnosis result.
(5) A hierarchical diagnostic module: and diagnosing the detection sample, layering the tumor into extremely low risk lesion, extremely high risk lesion and uncertain risk lesion, and feeding back to the terminal equipment and the database platform of the doctor.
(6) A database platform: the module is connected with the modules and used for establishing data storage and calling of the model, continuously collecting new data and providing the online updating function of the model, thereby continuously enriching and optimizing the database and improving the performance of the system.
The system for auxiliary diagnosis of the malignant tumor risk stratification is composed of five modules and a platform, and when the actual system is applied, the functions of the modules can be split or combined according to requirements and actual conditions, so long as the functions can be realized.
A schematic diagram of a tumor malignant risk hierarchical management decision-making based on the system is shown in fig. 2, and the system firstly collects patient data (medical images and the like) for preliminary screening, and hierarchically classifies patients with extremely low risk lesions according to a high-sensitivity benign lesion detection model to give reasonable clinical opinions, such as periodic review; then, further diagnosing the possible tumors with malignant risk lesions, and providing a treatment scheme for patients with extremely high risk lesions in a layered manner according to a high-specificity malignant lesion detection model to perform active treatment in time so as to avoid delay of disease conditions; for the pathological changes with uncertain risks, a professional doctor needs to perform analysis and judgment by combining other clinical indexes, and follow-up visit or treatment is performed according to clinical guidelines. The score in the figure is a score of each nodule calculated by a model, and the benign line and the malignant line are grade scores obtained by a high-sensitivity benign lesion detection model and a high-specificity malignant lesion detection model, respectively.
In addition, in order to better implement and understand the present invention, a brief description is provided below of a tumor malignancy risk stratification auxiliary diagnosis method based on the system, which mainly includes the following aspects:
(a) collecting data: acquiring the existing medical data in multiple ways, and screening and desensitizing the medical data;
(b) image segmentation: carrying out contour segmentation and local square segmentation on the acquired image;
(c) feature extraction: the method comprises the following steps of extracting the imaging characteristics of an image in an interested area, wherein the imaging characteristics comprise histogram characteristics, texture characteristics, shape characteristics, filtering transformation characteristics and convolution neural network characteristics, and the filtering transformation comprises the following steps: a gaussian-laplacian filter and/or a wavelet transform;
(d) selecting characteristics: selecting the characteristic with stable performance by using a characteristic stability test data set repeatedly scanned for a short time, then screening the highly relevant characteristic in the test data set by adopting a univariate analysis method, and finally screening the low Redundancy and highly relevant characteristic by using a minimum Redundancy Maximum correlation method (mrmr), a minimum absolute value shrinkage and selector method (Least absolute shrinkage and selection operator); wherein the univariate analysis method is a t-test method and/or an analysis of variance method;
(e) establishing a model: performing model building by adopting various machine learning and deep learning algorithms, and building a model by adopting cross validation; wherein, the machine learning and deep learning algorithm comprises: one or more of a linear and non-linear Support Vector Machine (SVM), a random forest, a decision tree, k-nearest neighbor, logistic regression, convolutional neural networks, a challenge generation network, and transfer learning;
(f) optimizing the model: the integrated learning method comprises an integrated learning and reward and punishment mechanism, wherein the integrated learning is to combine a plurality of classifiers to obtain more excellent generalization performance than a single classifier; the reward and punishment mechanism is that in the training set, firstly, a classification model is constructed, and benign and malignant labels are marked; identifying error images by artificial intelligence, and applying penalty factors to previous models for images which should not be subjected to error in definite diagnosis by doctors under the guidance of comprehensive opinions of three or more medical experts; identifying accurate images for artificial intelligence, and applying reward factors in model improvement; modifying to obtain a new model by combining the reward and punishment factors and the original model, and circularly iterating and optimizing the model to finally obtain an ideal benign and malignant layered model;
(g) and (3) verifying the model: performing model verification on a verification data set, drawing an ROC curve, searching a specificity index as high as possible in a high-sensitivity model, wherein the corresponding specificity index is more than or equal to 20% when the sensitivity is 99%; searching a sensitivity index as high as possible in a high specificity model, wherein the sensitivity index corresponding to 99% of specificity is more than or equal to 20%; calculating a positive predicted value (true malignancy rate), a negative predicted value (true malignancy rate) and an accuracy index;
(h) and (3) clinical diagnosis: collecting test data, carrying out tumor layered diagnosis and outputting a diagnosis result.
In conclusion, the auxiliary diagnosis system for the malignant tumor risk stratification can realize successive stratification of benign and malignant tumor risks by means of an artificial intelligence technology, stratify the tumor into extremely low-risk lesion, extremely high-risk lesion and uncertain risk lesion, can substantially assist clinical management decision of occupying lesion, improve the existing working process of clinical diagnosis, increase diagnosis confidence of doctors, relieve working pressure and reduce anxiety of patients with low-malignant tumor risk lesion; moreover, the system can continuously verify and optimize the model, solves the problems that computer auxiliary software cannot feed back and self-learn and self-update in the prior art, and improves the accuracy and reliability of tumor stratification, and through the diagnosis system, the sensitivity of the high-sensitivity model reaches 99 percent, the specificity reaches more than 20 percent, the specificity of the high-specificity model reaches 99 percent, the sensitivity reaches more than 20 percent, and the diagnosis rate of benign nodules and malignant nodules is determined.
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It should be understood by those skilled in the art that the above embodiments do not limit the present invention in any way, and all technical solutions obtained by using equivalent alternatives or equivalent variations fall within the scope of the present invention.
Claims (9)
1. The malignant risk stratification of tumor auxiliary diagnostic system of artificial intelligence medical image, characterized by, include:
a data acquisition module: the system is used for acquiring the existing data, including image images, clinical information, gene information and pathological benign and malignant labels, verifying and evaluating the data integrity and the data quality, performing data desensitization processing and sending the data desensitization processing to a data preprocessing module;
a data preprocessing module: carrying out noise reduction and normalization processing on the acquired existing data, and marking and dividing an interested area corresponding to a pathological result on an image;
a model building module: feature extraction and feature selection are carried out on the region of interest obtained based on the data preprocessing module, low-redundancy high-correlation features are screened out, a tumor malignant risk hierarchical model is established, and model parameters and threshold values of a high-sensitivity benign lesion detection model and a high-specificity malignant lesion detection model are selected according to different application scenes: for a clinical screening application scene, a high-sensitivity benign lesion detection model is selected, and the negative predictive value is more than 99%; specifically diagnosing a clinical scene aiming at malignant lesions, selecting a high-specificity malignant lesion detection model to ensure that the positive predictive value is more than 99 percent, and layering extremely high-risk lesions;
a model verification and optimization module: the method is used for evaluating the effectiveness, the generalization and the robustness of the model under different parameters and different scenes, iterating the performance of the model by adopting a reward and punishment mechanism, gradually modifying the parameters of the model on the premise of ensuring high sensitivity and high specificity, and improving the diagnosis capability and the accuracy; the model verification and optimization mode is as follows: performing model verification on a verification data set, drawing an ROC curve, and searching a specificity index which is more than or equal to 20% of a corresponding specificity index when the sensitivity is 99% in a high-sensitivity model; searching a sensitivity index of which the corresponding sensitivity index is not less than 20% when the specificity is 99% in a high-specificity model; calculating a positive predicted value, a negative predicted value and an accuracy index; the reward and punishment mechanism is as follows: in the training set, firstly constructing a classification model and marking benign and malignant labels; the method comprises the steps of (1) applying a penalty factor to a previous model for an image which is definitely diagnosed by a doctor and should not make an error by means of comprehensive opinion guidance of more than three medical experts for artificially and intelligently identifying an error image; identifying accurate images for artificial intelligence, and applying reward factors in model improvement; modifying to obtain a new model by combining the reward and punishment factors and the original model, and circularly iterating and optimizing the model to finally obtain an ideal tumor malignant risk layered model;
a hierarchical diagnostic module: diagnosing the detection sample, layering the tumor into extremely low risk lesion, extremely high risk lesion and uncertain risk lesion, and feeding back to the terminal equipment and the database platform of the doctor;
a database platform: and the module is connected with the modules and used for establishing a model, storing and calling data, continuously collecting newly added data and providing an online updating function of the model.
2. The system of claim 1, wherein the sources of the existing data include: the data set and the case data of hospitals at all levels are disclosed.
3. The system of claim 1, wherein when acquiring lung nodule data, the existing data acquisition criteria are: all the incorporated nodules are primary nodules with the size smaller than 3cm, the benign labels are obtained from operation pathology, the malignant labels are obtained by means of operation, puncture or bronchoscope biopsy, and the image is a chest CT flat scan without obvious artifacts.
4. The system of claim 2, wherein the desensitization process includes privacy information and medical institution information.
5. The system of claim 1, wherein the data preprocessing module is configured to search for a lesion corresponding to a pathological result by a physician, and then perform contour segmentation and local block segmentation on the lesion by an automatic segmentation method confirmed by the physician or a manual segmentation method by the physician.
6. The method for layered auxiliary diagnosis of malignant tumor risk based on artificial intelligence medical image is characterized in that the auxiliary diagnosis system based on any one of claims 1 to 5 comprises the following aspects:
(1) collecting data: acquiring the existing medical data in multiple ways, and screening and desensitizing the medical data;
(2) image segmentation: carrying out contour segmentation and local square segmentation on the acquired image;
(3) feature extraction: extracting the image characteristics of the image in the region of interest, wherein the image characteristics comprise histogram characteristics, texture characteristics, shape characteristics, filtering transformation characteristics and convolution neural network characteristics;
(4) selecting characteristics: selecting the stable-performance characteristics by using a characteristic stability test data set repeatedly scanned for a short time, screening highly-relevant characteristics in the test data set by adopting a univariate analysis method, and finally screening the low-redundancy and highly-relevant characteristics;
(5) establishing a model: adopting various machine learning algorithms to establish a model, and adopting cross validation to establish the model;
(6) optimizing the model: the integrated learning method comprises an integrated learning and reward and punishment mechanism, wherein the integrated learning is to combine a plurality of classifiers to obtain more excellent generalization performance than a single classifier; the reward and punishment mechanism is that in the training set, firstly, a classification model is constructed, and benign and malignant labels are marked; the method comprises the steps of (1) applying a penalty factor to a previous model for an image which is definitely diagnosed by a doctor and should not make an error by means of comprehensive opinion guidance of more than three medical experts for artificially and intelligently identifying an error image; identifying accurate images for artificial intelligence, and applying reward factors in model improvement; modifying to obtain a new model by combining the reward and punishment factors and the original model, and circularly iterating and optimizing the model to finally obtain an ideal benign and malignant layered model;
(7) and (3) verifying the model: performing model verification on a verification data set, drawing an ROC curve, and searching a specificity index which is more than or equal to 20% of a corresponding specificity index when the sensitivity is 99% in a high-sensitivity model; searching a sensitivity index of which the corresponding sensitivity index is not less than 20% when the specificity is 99% in a high-specificity model; calculating a positive predicted value, a negative predicted value and an accuracy index;
(8) and (3) clinical diagnosis: collecting test data, carrying out tumor layered diagnosis and outputting a diagnosis result.
7. The method of claim 6, wherein the filter transformation comprises: Gaussian-Laplace filtering and/or wavelet transformation, and the univariate analysis method is a t-test method and/or an analysis of variance method.
8. The method of claim 6, wherein the low redundancy, highly correlated features are screened by using a minimum redundancy maximum correlation method, a minimum absolute value shrinkage and a selector method.
9. The method of claim 6, wherein the machine learning algorithm comprises: one or more of a Support Vector Machine (SVM), a random forest, a decision tree, k-nearest neighbors, logistic regression, convolutional neural networks, a challenge generation network, and migratory learning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810866088.3A CN109166105B (en) | 2018-08-01 | 2018-08-01 | Tumor malignancy risk layered auxiliary diagnosis system based on artificial intelligent medical image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810866088.3A CN109166105B (en) | 2018-08-01 | 2018-08-01 | Tumor malignancy risk layered auxiliary diagnosis system based on artificial intelligent medical image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109166105A CN109166105A (en) | 2019-01-08 |
| CN109166105B true CN109166105B (en) | 2021-01-26 |
Family
ID=64898661
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810866088.3A Active CN109166105B (en) | 2018-08-01 | 2018-08-01 | Tumor malignancy risk layered auxiliary diagnosis system based on artificial intelligent medical image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109166105B (en) |
Families Citing this family (55)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109949349B (en) * | 2019-01-24 | 2021-09-21 | 北京大学第三医院（北京大学第三临床医学院） | Multi-mode three-dimensional image registration and fusion display method |
| CN109522974B (en) * | 2019-01-24 | 2020-12-18 | 湖北省肿瘤医院（湖北省肿瘤研究所） | Focus level selection method for improving positive rate of puncture biopsy |
| CN109785430B (en) * | 2019-01-28 | 2023-07-18 | 上海交通大学医学院附属上海儿童医学中心 | Construction method of risk prediction model of postoperative restenosis on TAPVC (graft versus graft) core |
| US11062459B2 (en) * | 2019-02-07 | 2021-07-13 | Vysioneer INC. | Method and apparatus for automated target and tissue segmentation using multi-modal imaging and ensemble machine learning models |
| CN109858562A (en) * | 2019-02-21 | 2019-06-07 | 腾讯科技（深圳）有限公司 | A kind of classification method of medical image, device and storage medium |
| CN111613302B (en) * | 2019-02-22 | 2024-03-05 | 曹生 | Tumor Ai processing method and product based on medical image |
| CN109903839A (en) * | 2019-02-22 | 2019-06-18 | 武汉凯德维斯生物技术有限公司 | A kind of medical imaging diagosis system based on cloud platform |
| CN110033432B (en) * | 2019-03-01 | 2023-04-18 | 广州市番禺区中心医院(广州市番禺区人民医院、广州市番禺区心血管疾病研究所) | Urinary calculus component analysis method and system based on machine learning and energy spectrum CT |
| CN110458883B (en) * | 2019-03-07 | 2021-07-13 | 腾讯科技（深圳）有限公司 | Medical image processing system, method, device and equipment |
| CN114255198A (en) * | 2019-03-28 | 2022-03-29 | 芨影(厦门)科技有限公司 | Artificial intelligence modeling and identification system for medical graphic image AI |
| CN109948806A (en) * | 2019-03-28 | 2019-06-28 | 医渡云（北京）技术有限公司 | Decision model optimization method, device, storage medium and equipment |
| CN110021431A (en) * | 2019-04-11 | 2019-07-16 | 上海交通大学 | Artificial intelligence assistant diagnosis system, diagnostic method |
| CN110060246B (en) * | 2019-04-15 | 2021-08-24 | 上海商汤智能科技有限公司 | Image processing method, device and storage medium |
| CN110070125A (en) * | 2019-04-19 | 2019-07-30 | 四川大学华西医院 | A kind of liver and gall surgical department's therapeutic scheme screening technique and system based on big data analysis |
| CN110895968B (en) * | 2019-04-24 | 2023-12-15 | 苏州图灵微生物科技有限公司 | Automatic diagnostic system and method for artificial intelligent medical image |
| CN110175697B (en) * | 2019-04-25 | 2021-05-25 | 胡盛寿 | Adverse event risk prediction system and method |
| CN110136108B (en) * | 2019-05-08 | 2022-09-13 | 郑州大学 | Method for carrying out benign and malignant characteristic statistics on breast cancer by adopting machine learning algorithm |
| CN110265144A (en) * | 2019-07-01 | 2019-09-20 | 边源医疗科技(杭州)有限公司 | Medical artificial intelligence's applied valuation system based on physician feedback |
| CN110310741A (en) * | 2019-07-01 | 2019-10-08 | 边源医疗科技(杭州)有限公司 | A kind of electric market system of medical artificial intelligence's application |
| US20210090718A1 (en) * | 2019-09-25 | 2021-03-25 | Vingroup Joint Stock Company | Labeling apparatus and method, and machine learning system using the labeling apparatus |
| CN110796656A (en) * | 2019-11-01 | 2020-02-14 | 上海联影智能医疗科技有限公司 | Image detection method, image detection device, computer equipment and storage medium |
| CN110837572B (en) * | 2019-11-15 | 2020-10-13 | 北京推想科技有限公司 | Image retrieval method and device, readable storage medium and electronic equipment |
| CN112927179A (en) * | 2019-11-21 | 2021-06-08 | 粘晓菁 | Intelligent analysis method for liver tumor |
| CN111476754B (en) * | 2020-02-28 | 2022-12-09 | 中国人民解放军陆军军医大学第二附属医院 | Bone marrow cell image artificial intelligence auxiliary grading diagnosis system and method |
| CN111310851B (en) * | 2020-03-03 | 2023-04-28 | 四川大学华西第二医院 | Artificial intelligence ultrasonic auxiliary system and application thereof |
| CN111583208B (en) * | 2020-04-29 | 2023-08-11 | 重庆工程职业技术学院 | System for assisting medical focus image recognition according to multiple characteristic points and working method |
| CN111899867B (en) * | 2020-07-28 | 2022-08-05 | 四川大学华西医院 | Operation complication prediction and avoidance aid decision-making system based on deep learning |
| CN111915596A (en) * | 2020-08-07 | 2020-11-10 | 杭州深睿博联科技有限公司 | Method and device for predicting benign and malignant pulmonary nodules |
| CN112102343B (en) * | 2020-08-12 | 2024-03-29 | 海南大学 | PTC diagnostic system based on ultrasonic image |
| CN113450910A (en) * | 2020-09-27 | 2021-09-28 | 四川大学华西医院 | Isolated lung nodule malignancy risk prediction system based on logistic regression model |
| CN112201336B (en) * | 2020-09-28 | 2023-08-11 | 暨南大学 | Remote consultation system and method based on AI diagnosis |
| CN112330644A (en) * | 2020-11-11 | 2021-02-05 | 复旦大学 | Medical image diagnosis system based on deep learning |
| CN112562816A (en) * | 2020-11-13 | 2021-03-26 | 陈卫霞 | System and method for correspondence and evaluation of diagnosis result and pathological result of tumor image report |
| CN112562849B (en) * | 2020-12-08 | 2023-11-17 | 中国科学技术大学 | Clinical automatic diagnosis method and system based on hierarchical structure and co-occurrence structure |
| CN112735569B (en) * | 2020-12-31 | 2022-09-02 | 四川大学华西医院 | System and method for outputting glioma operation area result before multi-modal MRI of brain tumor |
| CN112785256A (en) * | 2021-01-14 | 2021-05-11 | 田进伟 | Real-time assessment method and system for clinical endpoint events in clinical trials |
| CN112801958A (en) * | 2021-01-18 | 2021-05-14 | 青岛大学附属医院 | Ultrasonic endoscope, artificial intelligence auxiliary identification method, system, terminal and medium |
| CN113113130A (en) * | 2021-03-15 | 2021-07-13 | 湖南医云智享医疗科技有限公司 | Tumor individualized diagnosis and treatment scheme recommendation method |
| CN112967778B (en) * | 2021-03-17 | 2024-02-02 | 复旦大学附属华山医院 | Accurate medicine application method and system for inflammatory bowel disease based on machine learning |
| CN113096796B (en) * | 2021-04-01 | 2022-09-02 | 四川大学华西医院 | Intelligent prediction system and method for cerebral hemorrhage hematoma expansion risk |
| CN113239972A (en) * | 2021-04-19 | 2021-08-10 | 温州医科大学 | Artificial intelligence auxiliary diagnosis model construction system for medical images |
| CN113707298A (en) * | 2021-08-25 | 2021-11-26 | 景元明 | Prediction method for tumor diagnosis based on medical big data |
| CN113643776A (en) * | 2021-08-31 | 2021-11-12 | 平安医疗健康管理股份有限公司 | Patient marking method, device, equipment and storage medium based on hierarchical calculation |
| CN113823413A (en) * | 2021-10-22 | 2021-12-21 | 上海长征医院 | Lung function small airway disease prediction system, method, medium and electronic device |
| CN114068013B (en) * | 2021-11-16 | 2022-09-23 | 高峰 | Cerebral artery occlusion artificial intelligence assistant decision system |
| CN114331972B (en) * | 2021-12-08 | 2024-05-17 | 复旦大学附属中山医院 | Artificial intelligence assisted medical biological information establishing accurate evaluation system for arteritis imaging |
| CN114219807B (en) * | 2022-02-22 | 2022-07-12 | 成都爱迦飞诗特科技有限公司 | Mammary gland ultrasonic examination image grading method, device, equipment and storage medium |
| CN114648509B (en) * | 2022-03-25 | 2023-05-12 | 中国医学科学院肿瘤医院 | Thyroid cancer detection system based on multi-classification task |
| CN114431836A (en) * | 2022-04-11 | 2022-05-06 | 中南大学湘雅医院 | Artificial intelligence-based benign and malignant onychomycosis line prediction system |
| CN114998644B (en) * | 2022-05-10 | 2023-02-07 | 中国医学科学院肿瘤医院 | Tumor diagnosis system, construction method thereof, terminal device and storage medium |
| CN114782397B (en) * | 2022-05-12 | 2022-12-23 | 中晗控股集团有限公司 | Artificial intelligence tumor diagnosis system based on medical image and machine learning |
| CN115346658A (en) * | 2022-07-15 | 2022-11-15 | 一选(浙江)医疗科技有限公司 | Early intelligent auxiliary lung cancer detection system and method based on big data technology |
| CN115458160B (en) * | 2022-10-12 | 2023-06-20 | 浙江大学 | Whole-course intelligent management system, method, equipment and medium for breast tumor patients |
| CN115588504A (en) * | 2022-10-28 | 2023-01-10 | 大连大学附属中山医院 | Monitoring management system based on molecular image imaging technology |
| CN117711578B (en) * | 2024-02-06 | 2024-04-30 | 重庆医科大学绍兴柯桥医学检验技术研究中心 | Medical image data analysis management system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226155A (en) * | 2007-12-21 | 2008-07-23 | 中国人民解放军第八一医院 | Intelligentize lung cancer early cell pathological picture recognition processing method |
| CN106530298A (en) * | 2016-11-14 | 2017-03-22 | 同济大学 | Three-way-decision-based liver tumor CT image classification method |
| CN107563383A (en) * | 2017-08-24 | 2018-01-09 | 杭州健培科技有限公司 | A kind of medical image auxiliary diagnosis and semi-supervised sample generation system |
| CN107680082A (en) * | 2017-09-11 | 2018-02-09 | 宁夏医科大学 | Lung tumor identification method based on depth convolutional neural networks and global characteristics |
-
2018
- 2018-08-01 CN CN201810866088.3A patent/CN109166105B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226155A (en) * | 2007-12-21 | 2008-07-23 | 中国人民解放军第八一医院 | Intelligentize lung cancer early cell pathological picture recognition processing method |
| CN106530298A (en) * | 2016-11-14 | 2017-03-22 | 同济大学 | Three-way-decision-based liver tumor CT image classification method |
| CN107563383A (en) * | 2017-08-24 | 2018-01-09 | 杭州健培科技有限公司 | A kind of medical image auxiliary diagnosis and semi-supervised sample generation system |
| CN107680082A (en) * | 2017-09-11 | 2018-02-09 | 宁夏医科大学 | Lung tumor identification method based on depth convolutional neural networks and global characteristics |
Non-Patent Citations (4)
| Title |
|---|
| "人工智能医学影像";卢光明等;《医学研究生学报》;20180731;文献第683-687页 * |
| "基于BP神经网络的乳腺肿瘤计算机辅助诊断方法研究";吴华锋等;《现代商贸工业》;20161231;文献第212-213页 * |
| "基于集成卷积神经网络的肺部肿瘤计算机辅助诊断模型";王媛媛等;《生物医学工程学杂志》;20170831;文献摘要、引言、第1-4节 * |
| "孤立性肺结节良恶性预测模型的建立";张满;《中国博士学位论文全文数据库 医药卫生科技辑》;20170215;文献第一部分、第二部分 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109166105A (en) | 2019-01-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109166105B (en) | Tumor malignancy risk layered auxiliary diagnosis system based on artificial intelligent medical image | |
| Monkam et al. | Detection and classification of pulmonary nodules using convolutional neural networks: a survey | |
| CN112292691B (en) | Methods and systems for improving cancer detection using deep learning | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| CN108389201B (en) | Lung nodule benign and malignant classification method based on 3D convolutional neural network and deep learning | |
| CN108564026B (en) | Network construction method and system for thyroid tumor cytology smear image classification | |
| Bellotti et al. | A CAD system for nodule detection in low‐dose lung CTs based on region growing and a new active contour model | |
| Xu et al. | Lung cancer diagnosis in CT images based on Alexnet optimized by modified Bowerbird optimization algorithm | |
| CN113208640A (en) | Method for predicting axillary lymph node metastasis based on PET (positron emission tomography) imaging omics special for mammary gland | |
| CN112820399A (en) | Method and device for automatically diagnosing benign and malignant thyroid nodules | |
| CN112819747A (en) | Method for automatically diagnosing benign and malignant nodules based on lung tomography image | |
| Guo et al. | DeepLN: an artificial intelligence-based automated system for lung cancer screening | |
| CN116030325A (en) | Lung nodule CT image recognition method based on deep hybrid learning framework | |
| CN114140378A (en) | Scanned image processing method, electronic device, and readable medium | |
| JP2024043567A (en) | Training method, training device, electronic device, storage medium, and pathological image classification system for pathological image feature extractor based on feature separation | |
| Tian et al. | Radiomics and its clinical application: artificial intelligence and medical big data | |
| Li et al. | Reconstruction-assisted feature encoding network for histologic subtype classification of non-small cell lung cancer | |
| Kumar et al. | A methodical exploration of imaging modalities from dataset to detection through machine learning paradigms in prominent lung disease diagnosis: a review | |
| CN117711615A (en) | Lymph node metastasis state classification prediction method and device based on image histology | |
| CN115631387B (en) | Method and device for predicting lung cancer pathology high-risk factor based on graph convolution neural network | |
| Wankhade et al. | Lung cell cancer identification mechanism using deep learning approach | |
| CN113936006A (en) | Segmentation method and device for processing high-noise low-quality medical image | |
| Pradhan | An early diagnosis of lung nodule using CT images based on hybrid machine learning techniques | |
| Bandyopadhyay et al. | Artificial-intelligence-based diagnosis of brain tumor diseases | |
| Gowri et al. | An improved classification of MR images for cervical cancer using convolutional neural networks |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210106 Address after: 305 Zhongshan East Road, Xuanwu District, Nanjing City, Jiangsu Province 210000 Applicant after: EASTERN THEATER GENERAL HOSPITAL OF PLA Address before: 210002 No. 305 Shandong Road, Xuanwu District, Nanjing City, Jiangsu Province Applicant before: NANJING GENERAL HOSPITAL OF NANJING MILITARY REGION OF PLA |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |