CN110570405A - pulmonary nodule intelligent diagnosis method based on mixed features - Google Patents

pulmonary nodule intelligent diagnosis method based on mixed features Download PDF

Info

Publication number
CN110570405A
CN110570405A CN201910790034.8A CN201910790034A CN110570405A CN 110570405 A CN110570405 A CN 110570405A CN 201910790034 A CN201910790034 A CN 201910790034A CN 110570405 A CN110570405 A CN 110570405A
Authority
CN
China
Prior art keywords
features
feature
network
nodule
hog
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201910790034.8A
Other languages
Chinese (zh)
Inventor
张国彬
杨志永
姜杉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201910790034.8A priority Critical patent/CN110570405A/en
Publication of CN110570405A publication Critical patent/CN110570405A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
    • G06V10/507—Summing image-intensity values; Histogram projection analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/64—Three-dimensional objects
    • 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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/467—Encoded features or binary features, e.g. local binary patterns [LBP]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Primary Health Care (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Epidemiology (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Image Analysis (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

The invention discloses a pulmonary nodule intelligent diagnosis method based on mixed characteristics, which comprises the following steps: acquiring CT image data comprising a chest medical image file and a corresponding diagnosis result lesion label; resampling, smoothing and normalizing the acquired CT image data; learning a 3D CT image by using a 3D residual error-tight connection network to obtain a high-dimensional depth feature, and obtaining LBP-based texture features and HOG-based shape features which are used for describing the characterization features of lung nodules; and classifying the benign and malignant lung nodules based on the high-dimensional depth feature, the LBP-based texture feature and the HOG-based shape feature by using a GBM gradient elevator. The method can rapidly distinguish the benign and malignant lung nodules in the CT medical image, reduce the misjudgment rate, improve the classification precision and enhance the generalization capability of a diagnosis network model.

Description

Pulmonary nodule intelligent diagnosis method based on mixed features
Technical Field
The invention relates to the technical field of medical image processing, in particular to a lung nodule intelligent diagnosis method based on mixed characteristics.
Background
The traditional automatic lung nodule diagnosis algorithm includes five stages: the method comprises the steps of CT image acquisition, lung nodule segmentation, prior feature extraction, feature screening and good and malignant nodule classification. Extraction of pulmonary nodule prior features is a crucial step. The most common features currently used include texture features, shape (geometric) features, grayscale features, size features, contour features, and morphological features. And after the effective features are selected, classifying the effective features by using a classifier based on machine learning. However, due to the heterogeneity of the lung nodule characteristics, the complex lung nodule shape and the various types, the extracted prior characteristics cannot accurately describe the nodule, and further, the phenomena of misjudgment, missed judgment and the like occur in the diagnosis process.
With the emergence and rapid development of artificial intelligence, the lung nodule diagnosis work is optimized by utilizing a deep learning technology, and destructive progress is achieved. The convolutional neural network algorithm based on deep learning can autonomously learn the depth features with high lung nodule identification degree in the CT image, span the manual feature definition link, greatly save the labor cost and improve the diagnosis efficiency. In addition, the intelligent diagnosis network of the pulmonary nodules under deep learning has high generalization capability and strong adaptability, and can efficiently diagnose different types of nodules, improve the diagnosis accuracy, reduce the probability of misjudgment and missed judgment, and improve the classification sensitivity.
However, the conventional convolutional neural classification network under deep learning is mainly based on 2DCT image to process lung nodule slices one by one, so that three-dimensional spatial information of CT images is largely ignored, and the learned feature dimension is single, and the identification capability is still limited.
Disclosure of Invention
The invention aims to provide a pulmonary nodule intelligent diagnosis method based on mixed characteristics aiming at the technical defects in the prior art, which realizes the intelligent diagnosis of pulmonary nodules by analyzing and learning training of 3D CT medical images of a chest and fusing depth characteristics and traditional characteristics, can realize the rapid, efficient and accurate classification of benign and malignant pulmonary nodules and solves the problem of intelligent diagnosis of pulmonary nodules.
the technical scheme adopted for realizing the purpose of the invention is as follows:
A lung nodule intelligent diagnosis method based on mixed characteristics comprises the following steps:
acquiring CT image data comprising a chest medical image file and a corresponding diagnosis result lesion label; resampling, smoothing and normalizing the acquired CT image data;
Learning a 3D CT image by using a 3D residual error-tight connection network to obtain a high-dimensional depth feature, and obtaining LBP-based texture features and HOG-based shape features which are used for describing the characterization features of lung nodules;
And classifying the benign and malignant lung nodules based on the high-dimensional depth feature, the LBP-based texture feature and the HOG-based shape feature by using a GBM gradient elevator.
Wherein the labeling comprises the size, contour information and grade of benign and malignant of the lung nodule.
Wherein the residual learning of the 3D residual-tightly connected network is denoted as pl＝Rl(xl-1)+xl-1Wherein p islNetwork output, x, representing layer ll-1Representing the network input of layer l, RlRepresenting a feature learning function, R, in a residual networkl(xl-1)+xl-1The operation is a jump connection.
Wherein the LBP-based texture feature and the HOG-based shape feature extraction are both based on a nodule region in the 2D CT image slice.
According to the intelligent pulmonary nodule diagnosis method provided by the invention, an advanced depth feature extraction network in a natural image is adopted, network parameters are optimized through a deepening network structure, and the prior traditional features and the depth high-dimensional features are fused, so that the method can rapidly distinguish the benign and malignant pulmonary nodules in a CT medical image, the misjudgment rate is reduced, the classification precision is improved, and the generalization capability of a diagnosis network model is enhanced.
In particular, the 3D residual error-tight connection network is used for extracting the pulmonary nodule depth features, the design structure is simple, the calculation complexity and the model complexity are low, and the generalization capability is strong. The depth features and the traditional features are fused to describe the pulmonary nodules, the defects of missing judgment, erroneous judgment and the like can be avoided to the maximum extent, the GBM classifier is used for performing final benign and malignant identification on the pulmonary nodules, and the obtained classification effect is obvious.
The method is integrated into a CAD system and can be used as a clinical auxiliary tool for radiologists to accurately and efficiently judge the benign and malignant lung nodules.
Drawings
FIG. 1 is a flow chart of a method for intelligent diagnosis of pulmonary nodules based on mixed features according to the present invention;
FIG. 2 is a schematic diagram of the basic elements of a 3D residual-tight connection network;
Fig. 3 is a schematic structural diagram of a 3D residual-tight connection network;
FIG. 4 shows a ten-fold cross-validation resulting classification accuracy distribution;
FIG. 5 shows a ROC curve;
FIG. 6 shows a portion of an example lung nodule diagnostic result.
Detailed Description
the invention is described in further detail below with reference to the figures and specific examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The invention relates to a lung nodule intelligent diagnosis method based on mixed characteristics, which comprises the following steps:
First, data acquisition and preprocessing
To ensure the repeatability of the comparative experiments and the authority of the experimental data, both training and testing datasets were derived from LUNA16, LUNA16 was a special dataset screened from the LIDC-IDRI database, containing a total of 888 cases, 1186 lung nodules, 450 benign nodules and 554 malignant nodules, each of which had a diameter greater than 3mm, and the dataset consisted of chest medical image files and corresponding diagnostic lesion labels, wherein the data was collected from the american national cancer institute for the primary purpose of studying early cancer detection in high risk groups. The voxel spacing and scanning layer thickness of CT scan images of different patients in the LUNA16 dataset are greatly different, and in order to avoid performance errors caused by objective factors, each set of CT images needs to be resampled to 1 × 1 × 1 before being input into a classification network, and the images are smoothed by gaussian filtering. The voxels in each CT slice with gray values at-1200,600 are normalized to [0,1 ]. The input of the depth feature training network is a 32 × 32 × 32 image matrix centered on the nodule centroid, and the texture feature based on LBP and the shape feature extraction based on HOG are both based on the nodule region in the 2D CT image slice.
Second, lung nodule mixed feature extraction
the features are mainly divided into two forms, wherein lung nodule data for high-dimensional depth feature extraction is in a 3D mode, and the main purpose is to fully utilize three-dimensional space information of a CT image and learn depth features with higher discrimination degree. The conventional feature extraction is mainly based on 2D CT image slices. The extraction process, both 2D and 3D lung nodules, is based on labeling information already in the LIDC-IDRI.
In the present invention, the hybrid feature includes a high-dimensional depth feature based on a residual-tight connection network, and a texture feature based on LBP and a shape feature based on HOG. The extraction of the depth features depends on a 3D residual error-tight connection network, which is mainly composed of two parts, the first part is a residual error network, and the second part is a tight connection network. The basic units of a 3D residual-tight connection network are shown in fig. 2. The richness of the high-dimensional depth features obtained by the convolutional neural network learning mainly depends on the depth of the network, and generally, the deeper the network is, the richer the features are, and the higher the classification degree is. However, as the depth of the network increases, there are many risks associated with them, such as gradient disappearance, gradient explosion, and network degradation. The 3D residual error-tight connection network can overcome the potential defects, the jump connection in the residual error network can maximize the feature reuse, and even under the condition that the network depth is increased, the jump connection can also effectively solve the problems of gradient loss, gradient explosion and the like in the network training process. The main advantage of full connectivity in the tightly-connected network is to explore deep new features, so the 3D residual-tightly-connected network makes the learning training model more fully utilize the features. In terms of conventional features, texture features based on Local Binary Patterns (LBP) and shape features based on Histogram of Oriented Gradient (HOG) are used to describe the characterization features of lung nodules.
The structure of the 3D residual error-tight connection network is shown in fig. 3, where input data is a 32 × 32 × 32 3DCT image matrix, a feature map is obtained after a first 3 × 3 × 3 convolution, and then 2560-dimensional high-dimensional depth features are obtained through 30 residual error-tight connection modules and finally through an average pooling layer.
The structure of each residual error-tight connection module is shown in fig. 2, and after data is input, the residual error-tight connection module is divided into two parts: "Data _ 1" and "Data _ 2" are then subjected to two convolution operations, 1 × 1 × 1 and 3 × 3 × 3, respectively, and finally the total Data is separated into "M _ 1" and "M _ 2" by 1 × 1 × 1 convolution, wherein "Data _ 1" and "M _ 2" are combined into "Data _1 '" by "Concat" connection, "Data _ 2" and "M _ 1" are combined into "Data _ 2'" by "Element-derived" connection, and "Data _1 '" and "Data _ 2'" are merged and then enter the next residual-tight connection module cycle as new input Data.
in this scheme, residual learning is denoted as pl＝Rl(xl-1)+xl-1wherein p islNetwork output, x, representing layer ll-1Representing the network input of layer l, RlRepresenting a feature learning function, R, in a residual networkl(xl-1)+xl-1The operation is an "Element-wise" connection (jump connection).
The residual error network has the greatest advantage of feature reuse to relieve the gradient disappearance phenomenon and solve the feature redundancy problem. The drawback of residual networks is the poor mining ability of new features, which is being solved by compact networks, enhancing feature propagation by receiving feature maps of all layers, enhancing the mining of new features.
Wherein the operating principle of the compact network is denoted ql＝Dl([x0,x1,...,xl-1])。qlRepresenting features extracted by the l-th layer of the tightly-connected network, DlRepresenting a feature learning function, Dl([x0,x1,...,xl-1]) Operation is achieved by means of a "Concat" connection.
By fusing a feature reuse mechanism in a residual error network and a new feature mining mechanism in a tightly connected network, the network can mine the high-discriminative depth features of the lung nodules in the CT image to the maximum extent. In the training process of the network, the training iteration number is set to 1500, the initial learning rate is set to 0.01, the initial learning rate is set to 0.001 after 750 times of iterative training, and the initial learning rate is set to 0.0001 after 1200 times of iterative training. Conventional features include LBP-based texture features and HOG-based shape features. LBP has strong grayscale and rotation invariance and is therefore the first operator to compute texture features. HOG is an efficient and robust feature descriptor that characterizes the shape of lung nodules by the distribution of pixel gradient directions and intensities.
Third, classification of benign and malignant pulmonary nodules
And (3) judging whether the lung nodule is benign or malignant by using a GBM classifier, wherein the GBM is an algorithm for gradually enhancing or improving errors under machine learning. An excellent classifier would cause the penalty function to fall along the gradient direction, so gradient descent and linear search are the main steps of the GBM algorithm. The core idea of the GBM algorithm is to construct a potential basis learning model that is closely related to the negative gradient of the penalty function. The loss function has a significant impact on the robustness of the classification model. For a binary problem, the invention adopts a cross entropy loss function to complete a classification task. The construction of the GBM algorithm is to continually perform iterative updates to the learner predictions. In each iteration, the estimated output is updated as a weighted sum of the previous base learner and the current weak learner output (the weighting is based on the optimal gradient descent step size). And finally, well training the GBM grader and obtaining a satisfactory effect.
The GBM algorithm is one of Boosting algorithms (Boosting methods). The main idea of the GBM algorithm is to build a new base learner based on the gradient descending direction of the loss function of the base learner built before, and the purpose is to hope to make the loss function of the model overall continuously descend and the learning model continuously optimize by integrating the base learners. Through experimental tests, the GBM is a preferred classifier superior to traditional algorithms such as a Support Vector Machine (SVM), a Random Forest (RF) and the like.
ten-fold cross experimental verification
The training set, validation set and test set used in the present invention are all taken from the LUNA16 database. In order to enhance the authority and the persuasiveness of experimental data, the LUNA16 data set is divided into 10 subsets randomly and evenly, a verification test experiment is carried out on the intelligent pulmonary nodule diagnosis method by adopting ten-fold cross validation, the classification accuracy, the Receiver Operating Characterization (ROC) and the Area Under Curve (AUC) are used as measuring standards to carry out analysis and comparison, the classification accuracy obtained by the ten-fold cross validation is shown in figure 4, the average classification accuracy is 93.78%, the ROC Curve is shown in figure 5, and the AUC value obtained by the experiment is 0.9687.
Example diagnostic test
Clinical cases are first obtained and pre-processed. Secondly, pulmonary nodule extraction is performed for high-dimensional depth features and conventional features. And thirdly, extracting lung nodule mixed features through a residual error-tight connection network, LBP and HOG operator, and finally, performing benign and malignant classification identification on the lung nodules through GBM to obtain a final diagnosis result. Some example lung nodule diagnoses are shown in fig. 6, where the more likely the predicted malignancy value is to be "1", the more malignant it is.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims (4)

1. A pulmonary nodule intelligent diagnosis method based on mixed characteristics is characterized by comprising the following steps:
Acquiring CT image data comprising a chest medical image file and a corresponding diagnosis result lesion label; resampling, smoothing and normalizing the acquired CT image data;
Learning a 3D CT image by using a 3D residual error-tight connection network to obtain a high-dimensional depth feature, and obtaining LBP-based texture features and HOG-based shape features which are used for describing the characterization features of lung nodules;
And classifying the benign and malignant lung nodules based on the high-dimensional depth feature, the LBP-based texture feature and the HOG-based shape feature by using a GBM gradient elevator.
2. The intelligent lung nodule diagnosis method based on mixed features as claimed in claim 1, wherein the labels include lung nodule size, contour information, benign and malignant grade.
3. The lung nodule intelligent diagnosis method based on mixed features as claimed in claim 1, wherein the residual learning of the 3D residual-tight connection network is represented as pl＝Rl(xl-1)+xl-1Wherein p islNetwork output, x, representing layer ll-1Representing the network input of layer l, Rlrepresenting a feature learning function, R, in a residual networkl(xl-1)+xl-1The operation is a jump connection.
4. The intelligent lung nodule diagnosis method based on mixed features as claimed in claim 1, wherein the texture feature based on LBP and the shape feature based on HOG are extracted based on nodule region in 2D CT image slice.
CN201910790034.8A 2019-08-26 2019-08-26 pulmonary nodule intelligent diagnosis method based on mixed features Pending CN110570405A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910790034.8A CN110570405A (en) 2019-08-26 2019-08-26 pulmonary nodule intelligent diagnosis method based on mixed features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910790034.8A CN110570405A (en) 2019-08-26 2019-08-26 pulmonary nodule intelligent diagnosis method based on mixed features

Publications (1)

Publication Number Publication Date
CN110570405A true CN110570405A (en) 2019-12-13

Family

ID=68775940

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910790034.8A Pending CN110570405A (en) 2019-08-26 2019-08-26 pulmonary nodule intelligent diagnosis method based on mixed features

Country Status (1)

Country Link
CN (1) CN110570405A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN112530578A (en) * 2020-12-02 2021-03-19 中国科学院大学宁波华美医院 Viral pneumonia intelligent diagnosis system based on multi-mode information fusion
CN113706517A (en) * 2021-09-01 2021-11-26 什维新智医疗科技(上海)有限公司 Device is judged to good or malignant node based on GULBP operator
CN116883768A (en) * 2023-07-26 2023-10-13 郑州大学 Lung nodule intelligent grading method and system based on multi-modal feature fusion

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106650830A (en) * 2017-01-06 2017-05-10 西北工业大学 Deep model and shallow model decision fusion-based pulmonary nodule CT image automatic classification method
CN107945179A (en) * 2017-12-21 2018-04-20 王华锋 A kind of good pernicious detection method of Lung neoplasm of the convolutional neural networks of feature based fusion
CN108288271A (en) * 2018-02-06 2018-07-17 上海交通大学 Image detecting system and method based on three-dimensional residual error network
CN109598709A (en) * 2018-11-29 2019-04-09 东北大学 Mammary gland assistant diagnosis system and method based on fusion depth characteristic

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106650830A (en) * 2017-01-06 2017-05-10 西北工业大学 Deep model and shallow model decision fusion-based pulmonary nodule CT image automatic classification method
CN107945179A (en) * 2017-12-21 2018-04-20 王华锋 A kind of good pernicious detection method of Lung neoplasm of the convolutional neural networks of feature based fusion
CN108288271A (en) * 2018-02-06 2018-07-17 上海交通大学 Image detecting system and method based on three-dimensional residual error network
CN109598709A (en) * 2018-11-29 2019-04-09 东北大学 Mammary gland assistant diagnosis system and method based on fusion depth characteristic

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
GUOBIN ZHANG ET AL.: "Classification of benign and malignant lung nodules from CT images based on hybrid features", 《PHYSICS IN MEDICINE & BIOLOGY》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN112530578A (en) * 2020-12-02 2021-03-19 中国科学院大学宁波华美医院 Viral pneumonia intelligent diagnosis system based on multi-mode information fusion
CN113706517A (en) * 2021-09-01 2021-11-26 什维新智医疗科技(上海)有限公司 Device is judged to good or malignant node based on GULBP operator
CN113706517B (en) * 2021-09-01 2024-05-24 什维新智医疗科技(上海)有限公司 Device is judged to benign malignancy of tuberosity based on GULBP operator
CN116883768A (en) * 2023-07-26 2023-10-13 郑州大学 Lung nodule intelligent grading method and system based on multi-modal feature fusion

Similar Documents

Publication Publication Date Title
Xu et al. Stacked sparse autoencoder (SSAE) for nuclei detection on breast cancer histopathology images
Wang et al. A multi-view deep convolutional neural networks for lung nodule segmentation
CN109685776B (en) Pulmonary nodule detection method and system based on CT image
CN108492297B (en) MRI brain tumor positioning and intratumoral segmentation method based on deep cascade convolution network
CN110570405A (en) pulmonary nodule intelligent diagnosis method based on mixed features
CN104751178B (en) Lung neoplasm detection means and method based on shape template matching combining classification device
CN106599854B (en) Automatic facial expression recognition method based on multi-feature fusion
Alilou et al. A comprehensive framework for automatic detection of pulmonary nodules in lung CT images
CN110929789A (en) Liver tumor automatic classification method and device based on multi-stage CT image analysis
Wang et al. Weakly supervised learning for whole slide lung cancer image classification
CN111144474B (en) Multi-view, multi-scale and multi-task lung nodule classification method
CN108010013A (en) A kind of lung CT image pulmonary nodule detection methods
Kuo et al. Automatic lung nodule detection system using image processing techniques in computed tomography
CN109214397A (en) The dividing method of Lung neoplasm in a kind of lung CT image
CN113344933B (en) Glandular cell segmentation method based on multi-level feature fusion network
CN111767952A (en) Interpretable classification method for benign and malignant pulmonary nodules
CN112700461A (en) System for pulmonary nodule detection and characterization class identification
Fu et al. Automatic detection of lung nodules using 3D deep convolutional neural networks
CN115393293A (en) Electron microscope red blood cell segmentation and positioning method based on UNet network and watershed algorithm
Yuan et al. Pulmonary nodule detection using 3-d residual u-net oriented context-guided attention and multi-branch classification network
D’Arnese et al. On the automation of radiomics-based identification and characterization of nsclc
Gopikha et al. Regularised Layerwise Weight Norm Based Skin Lesion Features Extraction and Classification.
CN112508860B (en) Artificial intelligence interpretation method and system for positive check of immunohistochemical image
CN113935961A (en) Robust breast molybdenum target MLO (Multi-level object) visual angle image pectoral muscle segmentation method
CN116777893B (en) Segmentation and identification method based on characteristic nodules of breast ultrasound transverse and longitudinal sections

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
CB03 Change of inventor or designer information
CB03 Change of inventor or designer information

Inventor after: Yang Zhiyong

Inventor after: Zhang Guobin

Inventor after: Jiang Shan

Inventor before: Zhang Guobin

Inventor before: Yang Zhiyong

Inventor before: Jiang Shan

WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20191213