CN107016665B - CT pulmonary nodule detection method based on deep convolutional neural network - Google Patents
CT pulmonary nodule detection method based on deep convolutional neural network Download PDFInfo
- Publication number
- CN107016665B CN107016665B CN201710083606.XA CN201710083606A CN107016665B CN 107016665 B CN107016665 B CN 107016665B CN 201710083606 A CN201710083606 A CN 201710083606A CN 107016665 B CN107016665 B CN 107016665B
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- nodule
- dimensional
- nodules
- 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
- 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)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a CT pulmonary nodule detection method based on a deep convolutional neural network, which comprises the following steps: 1) preprocessing a CT image to enable pixel intervals to be uniform and image contrast to be uniform; 2) training a two-dimensional convolutional neural network U-net, predicting a lung nodule segmentation image, and recommending candidate nodules based on the lung nodule segmentation image; 3) and (3) training a three-dimensional depth residual neural network Resnet3D, predicting the true and false positive probability of lung nodules, and screening out false positive nodules. The CT pulmonary nodule detection method provided by the invention fully exerts the advantage of deep learning, can automatically detect pulmonary nodules in a CT image more efficiently and more accurately, and has stronger adaptability to medical big data.
Description
Technical Field
The invention relates to a pulmonary nodule detection method aiming at a CT image, in particular to a CT pulmonary nodule detection method based on a deep convolutional neural network.
Background
The lung cancer is a main cause of cancer-related death worldwide, and the examination of high-risk people by using CT scanning is an effective means for finding early-stage lung cancer, and the number of the people is huge, and the workload of imaging physicians is increased rapidly, so that computer-aided diagnosis plays an important role.
Currently, in the field of detecting lung nodules with the aid of a computer in a CT image, a great deal of research work has been carried out based on a conventional statistical machine learning method, and certain results have been achieved. The detection steps are generally divided into two steps, wherein the first step is to recommend candidate nodules and detect the regions with the nodules possibly existing in the lung CT image, and the second step is to screen out false positive nodules, identify the regions detected in the first step, judge whether suspected targets in the regions are the nodules, reduce the false positive nodules as much as possible, and judge whether the nodules are cancerated or not. However, based on the conventional statistical machine learning method, predefined morphological features of texture, such as area, effective diameter, gradient, etc., are extracted from the image, which are not sufficient to accurately represent the nodules, resulting in a large number of detected false positive lung nodules.
In recent years, deep learning attracts a great deal of research interest, and results that cannot be achieved by traditional methods are obtained in many fields. Similarly, deep learning has proved to be the most effective means in the field of medical image analysis, and currently, the mainstream computer aided detection systems all use deep learning. Olaf Ronneberger et al proposed U-Net for biological Image Segmentation in 2015, which proposed a medical Image Segmentation method based on Convolutional neural network, and the method was applied to various medical Image Segmentation tasks, such as blood vessel Segmentation and cell Segmentation, and achieved good results. In the aspect of a Pulmonary Nodule Detection method, in 2016, Setio et al put forward a ' pulse non-product Detection in CT Images ', namely, False Positive Reduction Using Multi-View connected Networks ', the method combines a traditional method with a deep learning method, adopts the traditional method still when recommending candidate nodules, designs a Multi-view two-dimensional convolutional neural network when screening out False Positive nodules, and obtains good Detection effect. Qi et al in 2016 proposed "Multi-level Contextual 3D CNNs for False Positive Reduction in Pulmonary node Detection", which proposed a three-dimensional convolutional neural network to screen out False Positive nodules and reduce the Detection rate of False Positive Pulmonary nodules. Compared with a two-dimensional convolutional neural network, the three-dimensional convolutional neural network can capture more spatial information, extract more abundant image characteristics and reduce the detection rate of false positive pulmonary nodules to a certain extent. However, the design network structure of the method is shallow, only three convolutional layers are needed, and network models with multiple scales need to be trained for fusion. And the method still adopts the traditional method to recommend the candidate nodules, and cannot fully utilize the advantage of deep learning. The invention provides a method based on a deep convolutional neural network to detect lung nodules in a CT image, and aims to further improve the accuracy and robustness of a computer-aided CT lung nodule detection method.
Disclosure of Invention
The technical problem concerned by the invention is as follows: how to automatically, efficiently and accurately detect lung nodules in a CT image by using a computer.
In order to solve the technical problems, the invention provides a method for detecting lung nodules in CT by using a deep convolutional neural network, which adopts a deep learning method when candidate nodules are recommended and false positive nodules are screened out, fully exerts the advantages of the deep learning method, and can ensure that the nodules have higher recall ratio while ensuring that the detection rate of the false positive nodules is lower. The method specifically comprises the following steps:
(1) preprocessing the CT image to ensure that the pixel intervals of all the CT images are uniform and the image contrast is uniform;
(2) training a two-dimensional convolutional neural network U-net, predicting a lung nodule segmentation image, and recommending candidate nodules based on the lung nodule segmentation image;
(3) and (3) training a three-dimensional depth residual neural network Resnet3D, predicting the true and false positive probability of lung nodules, and screening out false positive nodules.
Wherein, the step (1) is used for preprocessing the CT image
The CT images acquired by different instruments in different environments have great difference in pixel spacing (i.e. the actual distance between two adjacent pixels) and image contrast. In the step, CT images are preprocessed by means of three-dimensional linear interpolation and numerical value normalization, so that CT volume data consistent in all aspects are obtained. The spatial information and the intensity information of the CT image in the subsequent detection step are kept consistent, and the useful characteristics can be extracted in the subsequent machine learning step, so that a better effect is obtained;
the specific method comprises the following steps:
1) uniform pixel spacing
Firstly, counting pixel interval information of all CT images and diameter information of nodules, formulating uniform pixel intervals, and then carrying out scaling operation on the original CT images in a three-dimensional linear interpolation mode to enable the pixel intervals of all the CT images to be uniform;
2) unified CT image contrast
Calculating the Mean value of the ith CT image as MeaniStandard deviation of Stdi(ii) a By comparing the pixel value I of each pixel in the CT imagexyzNormalizing according to the following formula (1) so as to unify the contrast of the CT image;
Ixyz=(Ixyz-Meani)/Stdi (1)。
the step (2) is specifically as follows: and firstly, generating a standard segmentation image of a training sample in the nodule candidate recommending step as a label, and training the two-dimensional convolutional neural network U-net to obtain a network model of the two-dimensional convolutional neural network U-net. And predicting the test sample according to a network model of a two-dimensional convolutional neural network U-net to obtain a lung nodule segmentation image, performing binarization processing on the lung nodule segmentation image, distinguishing whether a pixel is a background or a nodule, reducing noise by adopting morphological corrosion operation on the binarization segmentation image, calculating barycentric coordinates of a three-dimensional connected region of the nodule pixel, namely the center of a recommended candidate lung nodule, combining lung nodule centers with a three-dimensional space Euclidean distance smaller than 3cm, and avoiding repeated detection of the same lung nodule.
The step (3) specifically adopts the following method:
1) data preparation and preprocessing
After the training samples in the step of screening out the false positive nodules are obtained by the method in the step (2), copying the training samples and all the test samples into three-dimensional image blocks of a CT image according to the recommended nodule center, labeling each three-dimensional image block of the training samples according to the positions of the lung nodules labeled in the original images of the training samples by a professional doctor, and distinguishing the true positive nodules from the false positive lung nodules; expanding the true positive nodule sample;
2) network architecture construction
Constructing a three-dimensional depth residual error neural network structure Resnet3D, wherein a three-dimensional convolution layer, a three-dimensional residual error block and a three-dimensional pooling layer are adopted in the structure;
3) network model training
Training a classifier of true and false positive nodules based on the constructed three-dimensional depth residual error neural network structure Resnet3D to obtain a network model of the three-dimensional depth residual error neural network Resnet 3D;
4) pulmonary nodule prediction
And predicting the probability of true and false positive nodules of the test sample by using the trained network model, distinguishing the true and false positive nodules according to a preset probability threshold, and recording the position coordinates of the true and false positive nodules so as to obtain the lung nodule detection result in each CT image.
In the specific scheme of step (3), further, the expansion of the true positive nodule sample in step 1) may adopt several times of operations of translation, scaling and horizontal rotation on the three-dimensional image block of the original sample randomly to obtain several new three-dimensional image blocks, that is, new true positive nodule image block extended training data.
Further, a P & C (Pooling + Cropping) Pooling layer may be used in the three-dimensional Pooling layer in step 2), where the P & C Pooling layer is to perform Pooling (Pooling) and Cropping (Cropping) operations on the output of the previous layer at the same time, and connect the results of the Pooling and Cropping operations in a channel dimension as the input of the next layer; the pooling operation is the maximum pooling that down-samples the input to the original 1/8 size, and the cropping operation is to cut the 1/8 portion in the middle of the input feature map.
Further, the network model training in step 3) may be based on a stochastic gradient descent algorithm, and may be performed by pre-training using the stochastic gradient descent algorithm, and then an optimization strategy for training using online dynamically selected difficult samples, that is, N samples are available for each iteration, loss of N samples is first tested using the current network before iteration, the N samples are sorted, K samples with the largest loss are taken out, and the K sample loss functions are propagated backwards to update the weight of the network.
Further, in the step 4), in the sample prediction, the test sample may be subjected to a plurality of expansion operations of translation, scaling and horizontal rotation, samples obtained by each expansion are predicted, and an arithmetic average of prediction results of all the expanded samples is used as a prediction result of the final sample. The invention has the following beneficial effects:
(1) the method for detecting the lung nodules in the CT by using the deep convolutional neural network is provided, so that a computer can be used for more efficiently and accurately assisting in detecting the lung nodules in the CT image;
(2) a novel Pooling layer structure for P & C (Pooling + Cropping) is provided, where Pooling operations obtain global information in the profile and Cropping operations obtain central nodule information in the profile. The generated feature maps with different scales can help to capture the features of the multi-scale lung nodules, so that a more accurate detection effect is obtained. Meanwhile, the structure enables all previous convolution calculations to be shared among nodules with different scales, and compared with the method proposed by Qi and the like in the background art, the method separately processes multi-scale data, trains a plurality of models and has higher efficiency.
(3) The optimization strategy for training the samples with difficulty in online dynamic selection is provided, most of the samples which are easy to distinguish are dynamically excluded, and the efficiency of network training is greatly improved. Meanwhile, samples difficult to classify are fully trained, and the prediction capability of the model is improved.
(4) In the method, because the invention completely adopts the deep learning method, the related deep learning network structure can automatically learn the image characteristics from a large number of CT images, has stronger adaptability to medical big data and can obtain more accurate detection effect.
Drawings
FIG. 1 is a flow chart of a method provided by the present invention;
FIG. 2 is a schematic diagram of a network structure of a deep two-dimensional convolutional neural network U-Net in a recommended nodule candidate according to the present invention;
FIG. 3 is a schematic diagram of the network structure of the deep three-dimensional convolutional neural network Resnet3D in the present invention for screening out false positive nodules;
FIG. 4 is a graphical representation of lung nodule detection results (FROC curves) for an embodiment of the present invention;
FIG. 5 is a graph showing the lung nodule detection results (FROC curves) disclosed by the method of Setio et al;
FIG. 6 is a schematic representation of the lung nodule detection results (FROC curves) disclosed by the Qi et al method.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention, but are not intended to limit the scope thereof:
the data used in the specific experiments of the present invention were from the LIDC/IDRI database, containing CT images of 888 patients, and a total of 1086 nodules. The data were obtained by two image annotations by 4 experienced chest radiologists on CT images of the patient, the first being blind and the second being corrected with reference to other physicians. However, the 888 CT images are acquired by different instruments, the pixel spacing of the CT images is different and has a large variation range, the pixel spacing of the Z-axis dimension ranges from 0.45 mm to 2.50mm, and the pixel spacing of the X, Y-axis dimension ranges from 0.46 mm to 0.97 mm. Also, these CT images differ in HU range and contrast.
Fig. 1 is a processing flow chart of a CT lung nodule detection method based on a deep convolutional neural network according to the present invention. The method comprises the following steps:
CT image preprocessing
In the step, CT volume data with all aspects consistent are obtained by preprocessing CT images acquired by different instruments in different environments, so that a machine learning method in the subsequent detection step can obtain a better effect, and the method specifically comprises the following substeps:
1.1, uniform pixel spacing
Firstly, the pixel interval information of all CT images and the diameter information of nodules are counted, a uniform pixel interval is formulated, so that the difference between the uniform pixel interval and an original image is not too far, and meanwhile, the nodules are guaranteed to have a proper pixel length range in any dimension. And then, the original CT image is subjected to scaling operation in a three-dimensional linear interpolation mode, so that the pixel intervals of all the CT images are uniform. Therefore, inconsistency of spatial information of the nodes can be avoided, and the characteristic information of the nodes is prevented from generating fuzziness in the forward propagation process of the network, so that the accuracy of subsequent steps is prevented from being influenced. The pixel spacing used in the experiments detailed in the present invention: x, Y dimension is 0.75mm and the Z dimension is 1.25 mm.
1.2, unify CT image contrast
Calculating the Mean value of the ith CT image as MeaniStandard deviation of Stdi. By comparing the pixel value I of each pixel in the CT imagexyzThe contrast of the CT image is unified by performing normalization as shown in the following equation (1). Therefore, the problem that the extraction of the feature information of the nodule is influenced due to inconsistent signal intensity of the nodule can be avoided.
Ixyz=(Ixyz-Meani)/Stdi (1)
2. Recommending nodule candidates
The method comprises the following steps of training a U-net two-dimensional convolutional neural network based on the U-net two-dimensional convolutional neural network shown in the figure 2, predicting a lung nodule segmentation image, and recommending candidate nodules based on the lung nodule segmentation image; the method specifically comprises the following substeps:
2.1 Generation of Standard segmented images of training samples
For the training sample of the recommended nodule candidate step, dividing the training sample into a cross-sectional two-dimensional image according to the Z axis, and obtaining the pixel value I of each pixel of the Z-th layerxy,zCalculated according to formula (2), wherein PxyzThe coordinates of the pixel are then calculated,is the pixel coordinate, r, of the n-th nodule marked in the CT imagenIs the nodule radius.
2.2 training U-net network
Using a two-dimensional convolutional neural network architecture U-net proposed by 01af Ronneberger et al, which is shown in fig. 2, a Dropout layer was added after each convolutional layer to mitigate overfitting in a specific experiment of the present invention. And training by directly adopting an optimization strategy of random gradient descent and a cross validation training mode to obtain a network model of the CT volume data, and predicting a lung nodule segmentation image of each layer in the CT volume data.
2.3, extracting candidate nodule centers
Firstly, predicting a test sample according to a network model of a two-dimensional convolutional neural network U-net to obtain a lung nodule segmentation image, and performing binarization processing on the lung nodule segmentation image, wherein 0 represents that the pixel is a background, and 1 represents that the pixel belongs to a nodule. And then, reducing noise by adopting a morphological erosion operation on the binary segmentation image, and calculating the barycentric coordinate of each three-dimensional connected region with the pixel value of 1, namely the center of the recommended lung nodule candidate. Finally, the centers of the lung nodules with too close three-dimensional space distance (generally, the Euclidean distance in three-dimensional space can be set to be less than 3cm) are merged, and repeated detection of the same lung nodule is avoided.
3. Screening out false positive nodules
And (3) more false positive nodules in the recommended candidate nodules, and the characteristics of the lung nodules are obtained by adopting a three-dimensional depth residual neural network Resnet3D mode in the step, so that the false positive nodules are screened out. The method specifically comprises the following substeps:
3.1 data preparation and preprocessing
And (3) adopting the node center recommended in the step (2) for the training sample of the step of screening out the false positive nodes, copying the training sample and all the test samples into a three-dimensional image block (image patch) of the CT image according to the recommended node center, wherein the size of the image block adopted in the specific experiment of the invention is 44x44x28 pixels. For the training sample, labeling each three-dimensional image block according to the position of the lung nodule labeled in the original image by a professional physician, wherein 1 represents a true positive nodule, and 0 represents a false positive lung nodule. In order to balance positive and negative samples in the training data and obtain a better classification effect, the true positive nodule sample data is expanded, namely, the three-dimensional image blocks of the original samples are subjected to a plurality of times of micro translation, scaling and horizontal rotation operations randomly to obtain a plurality of new three-dimensional image blocks, namely, the new true positive nodule image blocks are used for expanding the training data.
3.2 network architecture construction
A three-dimensional depth residual error neural network structure Resnet3D shown in fig. 3 is constructed to screen false positive nodules, three-dimensional convolution layers are adopted in the structure, richer space information can be obtained compared with common two-dimensional convolution, and more representative features can be extracted by the network. The structure uses the thought of the existing two-dimensional residual error network for reference, the three-dimensional residual error block is designed, and the deep residual error network often has better data expression capability compared with the common convolutional neural network. A unique P & C (Pooling + Cropping) Pooling layer is designed in the structure to fuse multi-scale information, namely Pooling (Pooling) and Cropping (Cropping) operations are simultaneously carried out on the output of the previous layer, and the result of the Pooling operation and the result of the Cropping operation are connected in channel dimension to serve as the input of the next layer. The pooling operation is the maximum pooling that down-samples the input to the original 1/8 size, and the cropping operation is the cutting of the middle 1/8 portion of the input profile. Because the three-dimensional convolutional neural network has higher requirement on computing resources, in comprehensive consideration of experimental data and computing resources, the number N of three-dimensional residual blocks between two pooling layers is set to be 3 in the example, and 27 layers of three-dimensional depth residual neural networks are adopted.
3.3 network model training
Training a classifier of true and false positive nodules based on the three-dimensional depth residual error neural network structure Resnet3D constructed in 3.2 to obtain a network model of the three-dimensional depth residual error neural network Resnet 3D. The network training is based on a random gradient descent algorithm, and on the basis of the random gradient descent algorithm, an optimization strategy of training by taking online dynamically selected difficult samples is adopted, namely N samples exist in each iteration, the loss of the N samples is tested by using the current network before the iteration, the N samples are sequenced, K samples with the largest loss are taken out, and the loss function of the K samples is transmitted backwards to update the weight of the network.
The optimization strategy can increase the training difficulty of the network to a certain extent, and in order to ensure normal network convergence, a pre-training technique can be further adopted, namely, the network is directly trained by adopting a random gradient descent algorithm, the convergence speed of the network is higher at the moment, after a certain number of iterations, the optimization strategy is continuously adopted to train the network, the network is easier to train at the moment, and the convergence can be ensured.
3.4 pulmonary nodule prediction
And (3) predicting the probability of the true and false positive nodules of the test sample by using the network model obtained by training in the step 3.3, distinguishing the true and false positive nodules according to a preset probability threshold value, and recording the position coordinates of the true and false positive nodules, so that the lung nodule detection result in each CT is obtained. In order to obtain a more robust prediction result, an optimization method is further adopted in sample prediction, namely, a test sample is subjected to a plurality of times of micro translation, scaling and horizontal rotation expansion operations, samples obtained by each expansion are predicted, and finally the prediction result of the sample is the arithmetic average of the prediction results of all the expanded samples. The method can effectively reduce the influence of network overfitting on the result, is similar to the result of a model fusing a plurality of same networks, can improve the accuracy of the final prediction result, and has higher practicability because the time spent on expanding test data and testing for multiple times is far shorter than the time spent on training a plurality of models.
Fig. 4 shows a schematic diagram of an FROC (free recipient operating characteristic curve) curve of an embodiment of the present invention, and it can be seen that, compared to a method of using a multi-view two-dimensional convolutional neural network by the person of part et al and a method of using a shallow three-dimensional convolutional neural network by the person of Qi et al to improve the results obtained in the step of screening out false positive nodules (see fig. 5 and 6, both of which are performed by using the same data set as the present invention, specifically, see the corresponding documents in the background section), a more accurate detection effect can be obtained by using the method of the present invention.
Claims (8)
1. A CT pulmonary nodule detection method based on a deep convolutional neural network is characterized by comprising the following steps:
(1) preprocessing the CT image to ensure that the pixel intervals of all the CT images are uniform and the image contrast is uniform;
(2) training a two-dimensional convolutional neural network U-net, predicting a lung nodule segmentation image, and recommending candidate nodules based on the lung nodule segmentation image;
(3) training a three-dimensional depth residual error neural network Resnet3D, predicting the true and false positive probability of lung nodules, and screening out false positive nodules;
the specific method of the step (3) is as follows:
1) data preparation and preprocessing
After the training samples in the step of screening out the false positive nodules are obtained by the method in the step (2), copying the training samples and all the test samples into three-dimensional image blocks of a CT image according to the recommended nodule center, labeling each three-dimensional image block of the training samples according to the positions of the lung nodules labeled in the original images of the training samples by a professional doctor, and distinguishing the true positive nodules from the false positive lung nodules; expanding the true positive nodule sample;
2) network architecture construction
Constructing a three-dimensional depth residual error neural network structure Resnet3D, wherein a three-dimensional convolution layer, a three-dimensional residual error block and a three-dimensional pooling layer are adopted in the structure;
3) network model training
Training a classifier of true and false positive nodules based on the constructed three-dimensional depth residual error neural network structure Resnet3D to obtain a network model of the three-dimensional depth residual error neural network Resnet 3D;
4) pulmonary nodule prediction
And predicting the probability of true and false positive nodules of the test sample by using the trained network model, distinguishing the true and false positive nodules according to a preset probability threshold, and recording the position coordinates of the true and false positive nodules so as to obtain the lung nodule detection result in each CT image.
2. The CT pulmonary nodule detection method based on the deep convolutional neural network as claimed in claim 1, wherein the step (1) comprises the following steps:
1) uniform pixel spacing
Firstly, counting pixel interval information of all CT images and diameter information of nodules, formulating uniform pixel intervals, and then carrying out scaling operation on the original CT images in a three-dimensional linear interpolation mode to enable the pixel intervals of all the CT images to be uniform;
2) unified CT image contrast
Calculating the Mean value of the ith CT image as MeaniStandard deviation of Stdi(ii) a By comparing the pixel value I of each pixel in the CT imagexyzNormalizing according to the following formula (1) so as to unify the contrast of the CT image;
Ixyz=(Ixyz-Meani)/Stdi (1)。
3. the CT pulmonary nodule detection method based on the deep convolutional neural network as claimed in claim 1, wherein in step (2), a standard segmentation image of a training sample of the nodule candidate recommending step is generated as a label, and a two-dimensional convolutional neural network U-net is trained to obtain a network model of the two-dimensional convolutional neural network U-net.
4. The CT pulmonary nodule detection method based on the deep convolutional neural network as claimed in claim 3, wherein the candidate nodule is recommended based on the lung nodule segmentation image in the step (2), and the specific method is as follows: firstly, predicting a test sample according to a network model of a two-dimensional convolutional neural network U-net to obtain a lung nodule segmentation image, carrying out binarization processing on the lung nodule segmentation image, distinguishing whether a pixel is a background or a nodule, reducing noise by adopting morphological corrosion operation on the binarization segmentation image, calculating barycentric coordinates of a three-dimensional connected region of the nodule pixel, namely the center of a recommended candidate lung nodule, combining lung nodule centers with a three-dimensional space Euclidean distance smaller than 3cm, and avoiding repeated detection of the same lung nodule.
5. The CT pulmonary nodule detection method based on the deep convolutional neural network as claimed in claim 1, wherein the true positive nodule sample is expanded in step 1) and the three-dimensional image blocks of the original sample are randomly translated, scaled and horizontally rotated for several times to obtain several new three-dimensional image blocks, i.e. new true positive nodule image block extended training data.
6. The CT pulmonary nodule detection method based on the deep convolutional neural network of claim 1, wherein a P & C pooling layer is adopted in the three-dimensional pooling layer in step 2), and the P & C pooling layer is used for simultaneously performing pooling operation and cropping operation on the output of the previous layer and connecting the results of the pooling operation and the cropping operation in a channel dimension to serve as the input of the next layer; the pooling operation is the maximum pooling that down-samples the input to the original 1/8 size, and the cropping operation is to cut the 1/8 portion in the middle of the input feature map.
7. The CT pulmonary nodule detection method based on the deep convolutional neural network as claimed in claim 1, wherein the network model training in step 3) is based on a stochastic gradient descent algorithm, the stochastic gradient descent algorithm is firstly adopted for pre-training, then an optimization strategy of online dynamic selection of difficult sample training is adopted, namely N samples exist in each iteration, the loss of the N samples is firstly tested by using the current network before iteration, the N samples are sequenced, K samples with the largest loss are taken out, and the loss function of the K samples is transmitted backwards to update the weight of the network.
8. The CT pulmonary nodule detection method based on the deep convolutional neural network as claimed in claim 1, wherein in the step 4), the test sample is subjected to a plurality of expansion operations of translation, scaling and horizontal rotation in the sample prediction, the sample obtained by each expansion is predicted, and the arithmetic mean of the prediction results of all the expanded samples is used as the prediction result of the final sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710083606.XA CN107016665B (en) | 2017-02-16 | 2017-02-16 | CT pulmonary nodule detection method based on deep convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710083606.XA CN107016665B (en) | 2017-02-16 | 2017-02-16 | CT pulmonary nodule detection method based on deep convolutional neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107016665A CN107016665A (en) | 2017-08-04 |
CN107016665B true CN107016665B (en) | 2021-05-04 |
Family
ID=59440399
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710083606.XA Active CN107016665B (en) | 2017-02-16 | 2017-02-16 | CT pulmonary nodule detection method based on deep convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107016665B (en) |
Families Citing this family (129)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107247971B (en) * | 2017-06-28 | 2020-10-09 | 中国人民解放军总医院 | Intelligent analysis method and system for ultrasonic thyroid nodule risk index |
CN107274406A (en) * | 2017-08-07 | 2017-10-20 | 北京深睿博联科技有限责任公司 | A kind of method and device of detection sensitizing range |
CN107480665B (en) * | 2017-08-09 | 2020-08-11 | 北京小米移动软件有限公司 | Character detection method and device and computer readable storage medium |
CN107424152B (en) * | 2017-08-11 | 2020-12-18 | 联想(北京)有限公司 | Detection device for organ lesion, method for training neural network and electronic equipment |
CN107610129B (en) * | 2017-08-14 | 2020-04-03 | 四川大学 | CNN-based multi-modal nasopharyngeal tumor joint segmentation method |
CN107507197B (en) * | 2017-08-18 | 2020-11-24 | 东北大学 | Lung parenchyma extraction method based on clustering algorithm and convolutional neural network |
CN107590443A (en) * | 2017-08-23 | 2018-01-16 | 上海交通大学 | Limiter stage live video automatic testing method and system based on the study of depth residual error |
CN109426773A (en) * | 2017-08-24 | 2019-03-05 | 浙江宇视科技有限公司 | A kind of roads recognition method and device |
CN107680088A (en) * | 2017-09-30 | 2018-02-09 | 百度在线网络技术(北京)有限公司 | Method and apparatus for analyzing medical image |
CN107665736B (en) * | 2017-09-30 | 2021-05-25 | 百度在线网络技术(北京)有限公司 | Method and apparatus for generating information |
CN107808377B (en) * | 2017-10-31 | 2019-02-12 | 北京青燕祥云科技有限公司 | The positioning device of lesion in a kind of lobe of the lung |
CN109615636B (en) * | 2017-11-03 | 2020-06-12 | 杭州依图医疗技术有限公司 | Blood vessel tree construction method and device in lung lobe segment segmentation of CT (computed tomography) image |
CN108154164A (en) * | 2017-11-15 | 2018-06-12 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | Signal of communication modulation classification system and method based on deep learning |
US20190156204A1 (en) * | 2017-11-20 | 2019-05-23 | Koninklijke Philips N.V. | Training a neural network model |
CN107886510A (en) * | 2017-11-27 | 2018-04-06 | 杭州电子科技大学 | A kind of prostate MRI dividing methods based on three-dimensional full convolutional neural networks |
CN107895369B (en) * | 2017-11-28 | 2022-11-22 | 腾讯科技(深圳)有限公司 | Image classification method, device, storage medium and equipment |
CN108038860A (en) * | 2017-11-30 | 2018-05-15 | 杭州电子科技大学 | Spine segmentation method based on the full convolutional neural networks of 3D |
CN107977963A (en) * | 2017-11-30 | 2018-05-01 | 北京青燕祥云科技有限公司 | Decision method, device and the realization device of Lung neoplasm |
CN108078581B (en) * | 2017-12-12 | 2019-02-12 | 北京青燕祥云科技有限公司 | The good pernicious judgement system of lung cancer and realization device based on convolutional neural networks |
CN108171694B (en) * | 2017-12-28 | 2021-05-14 | 开立生物医疗科技(武汉)有限公司 | Method, system and equipment for detecting nodule based on convolutional neural network |
CN108090903A (en) * | 2017-12-29 | 2018-05-29 | 苏州体素信息科技有限公司 | Lung neoplasm detection model training method and device, pulmonary nodule detection method and device |
CN108154468B (en) * | 2018-01-12 | 2022-03-01 | 平安科技(深圳)有限公司 | Pulmonary nodule detection method, application server and computer readable storage medium |
CN108257128B (en) * | 2018-01-30 | 2020-09-18 | 浙江大学 | Establishment method of pulmonary nodule detection device based on 3D convolutional neural network |
CN108447062B (en) * | 2018-02-01 | 2021-04-20 | 浙江大学 | Pathological section unconventional cell segmentation method based on multi-scale mixed segmentation model |
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 |
CN110119736B (en) * | 2018-02-07 | 2021-12-28 | 浙江宇视科技有限公司 | License plate position identification method and device and electronic equipment |
CN108389190B (en) * | 2018-02-08 | 2022-12-16 | 贵州联科卫信科技有限公司 | Pulmonary nodule automatic detection method based on deep learning method |
CN108764263A (en) * | 2018-02-12 | 2018-11-06 | 北京佳格天地科技有限公司 | The atural object annotation equipment and method of remote sensing image |
CN108470185A (en) * | 2018-02-12 | 2018-08-31 | 北京佳格天地科技有限公司 | The atural object annotation equipment and method of satellite image |
CN108257139B (en) * | 2018-02-26 | 2020-09-08 | 中国科学院大学 | RGB-D three-dimensional object detection method based on deep learning |
CN108510502B (en) * | 2018-03-08 | 2020-09-22 | 华南理工大学 | Melanoma image tissue segmentation method and system based on deep neural network |
CN108428229B (en) | 2018-03-14 | 2020-06-16 | 大连理工大学 | Lung texture recognition method based on appearance and geometric features extracted by deep neural network |
CN108389201B (en) * | 2018-03-16 | 2020-06-30 | 北京推想科技有限公司 | Lung nodule benign and malignant classification method based on 3D convolutional neural network and deep learning |
CN108537784B (en) * | 2018-03-30 | 2021-08-24 | 四川元匠科技有限公司 | CT image pulmonary nodule detection method based on deep learning |
CN108647568B (en) * | 2018-03-30 | 2022-05-17 | 电子科技大学 | Grassland degradation automatic extraction method based on full convolution neural network |
CN108648172B (en) * | 2018-03-30 | 2021-08-03 | 四川元匠科技有限公司 | CT (computed tomography) map pulmonary nodule detection system based on 3D-Unet |
CN108648178A (en) * | 2018-04-17 | 2018-10-12 | 杭州依图医疗技术有限公司 | A kind of method and device of image nodule detection |
CN108629764A (en) * | 2018-04-17 | 2018-10-09 | 杭州依图医疗技术有限公司 | A kind of good pernicious method and device of determining Lung neoplasm |
CN108765369B (en) * | 2018-04-20 | 2023-05-02 | 平安科技(深圳)有限公司 | Method, apparatus, computer device and storage medium for detecting lung nodule |
CN108596247A (en) * | 2018-04-23 | 2018-09-28 | 南方医科大学 | A method of fusion radiation group and depth convolution feature carry out image classification |
CN108615237B (en) * | 2018-05-08 | 2021-09-07 | 上海商汤智能科技有限公司 | Lung image processing method and image processing equipment |
CN110163834B (en) * | 2018-05-14 | 2023-08-25 | 腾讯科技(深圳)有限公司 | Object recognition method and device and storage medium |
CN108648192B (en) * | 2018-05-17 | 2021-08-17 | 杭州依图医疗技术有限公司 | Method and device for detecting nodule |
CN108986067B (en) * | 2018-05-25 | 2020-08-14 | 上海交通大学 | Cross-modality-based pulmonary nodule detection method |
CN108846835B (en) * | 2018-05-31 | 2020-04-14 | 西安电子科技大学 | Image change detection method based on depth separable convolutional network |
CN108670285A (en) * | 2018-06-05 | 2018-10-19 | 胡晓云 | A kind of CT pulmonary tuberculosis detection artificial intelligence diagnosis and therapy system |
CN108876772B (en) * | 2018-06-05 | 2021-10-12 | 南华大学 | Lung cancer image diagnosis system and method based on big data |
CN108805209B (en) * | 2018-06-14 | 2021-11-19 | 清华大学深圳研究生院 | Lung nodule screening method based on deep learning |
CN108876779A (en) * | 2018-06-22 | 2018-11-23 | 中山仰视科技有限公司 | Lung cancer method for early prediction, electronic equipment based on deep learning |
CN108898595B (en) * | 2018-06-27 | 2021-02-19 | 慧影医疗科技(北京)有限公司 | Construction method and application of positioning model of focus region in chest image |
CN108986085B (en) * | 2018-06-28 | 2021-06-01 | 深圳视见医疗科技有限公司 | CT image pulmonary nodule detection method, device and equipment and readable storage medium |
CN108742679B (en) * | 2018-06-29 | 2022-07-05 | 上海联影医疗科技股份有限公司 | Nodule detection apparatus and method |
CN109215014B (en) * | 2018-07-02 | 2022-03-04 | 中国科学院深圳先进技术研究院 | Training method, device and equipment of CT image prediction model and storage medium |
CN109064477A (en) * | 2018-07-06 | 2018-12-21 | 陕西师范大学 | With the method for improved U-Net detection cell nuclear periphery |
US10878570B2 (en) * | 2018-07-17 | 2020-12-29 | International Business Machines Corporation | Knockout autoencoder for detecting anomalies in biomedical images |
CN109035234B (en) * | 2018-07-25 | 2020-12-01 | 腾讯科技(深圳)有限公司 | Nodule detection method, device and storage medium |
CN109190481B (en) * | 2018-08-06 | 2021-11-23 | 中国交通通信信息中心 | Method and system for extracting road material of remote sensing image |
CN109190682B (en) * | 2018-08-13 | 2020-12-18 | 北京安德医智科技有限公司 | Method and equipment for classifying brain abnormalities based on 3D nuclear magnetic resonance image |
CN109102506B (en) * | 2018-08-20 | 2021-08-13 | 东北大学 | Automatic segmentation method for abdominal CT liver lesion image based on three-level cascade network |
CN109086782A (en) * | 2018-08-21 | 2018-12-25 | 广东工业大学 | Feature Descriptor generation method, device, equipment and computer readable storage medium |
CN109087703B (en) * | 2018-08-24 | 2022-06-07 | 南京大学 | Peritoneal transfer marking method of abdominal cavity CT image based on deep convolutional neural network |
CN109300136B (en) * | 2018-08-28 | 2021-08-31 | 众安信息技术服务有限公司 | Automatic segmentation method for organs at risk based on convolutional neural network |
NL2021559B1 (en) * | 2018-09-04 | 2020-04-30 | Aidence B V | Determination of a growth rate of an object in 3D data sets using deep learning |
CN110766693B (en) * | 2018-09-06 | 2022-06-21 | 北京连心医疗科技有限公司 | Method for jointly predicting radiotherapy structure position based on multi-model neural network |
CN109191476B (en) * | 2018-09-10 | 2022-03-11 | 重庆邮电大学 | Novel biomedical image automatic segmentation method based on U-net network structure |
CN109308695A (en) * | 2018-09-13 | 2019-02-05 | 镇江纳兰随思信息科技有限公司 | Based on the cancer cell identification method for improving U-net convolutional neural networks model |
CN109389584A (en) * | 2018-09-17 | 2019-02-26 | 成都信息工程大学 | Multiple dimensioned rhinopharyngeal neoplasm dividing method based on CNN |
CN109242962A (en) * | 2018-09-28 | 2019-01-18 | 北京青燕祥云科技有限公司 | The 3D display method, apparatus and electronic equipment of Lung neoplasm |
CN109410185B (en) * | 2018-10-10 | 2019-10-25 | 腾讯科技(深圳)有限公司 | A kind of image partition method, device and storage medium |
CN109447960A (en) * | 2018-10-18 | 2019-03-08 | 神州数码医疗科技股份有限公司 | A kind of object identifying method and device |
CN109492547B (en) * | 2018-10-24 | 2022-03-08 | 腾讯科技(深圳)有限公司 | Nodule identification method and device and storage medium |
CN109544510B (en) * | 2018-10-24 | 2021-10-26 | 广州大学 | Three-dimensional lung nodule identification method based on convolutional neural network |
CN109544528B (en) * | 2018-11-19 | 2021-02-12 | 上海馨孚科技有限公司 | Lung nodule image identification method and device |
CN113168684B (en) * | 2018-11-26 | 2024-04-05 | Oppo广东移动通信有限公司 | Method, system and computer readable medium for improving quality of low brightness images |
CN109544534B (en) | 2018-11-26 | 2020-10-16 | 上海联影智能医疗科技有限公司 | Focal image detection device, method and computer-readable storage medium |
CN109685768B (en) * | 2018-11-28 | 2020-11-20 | 心医国际数字医疗系统(大连)有限公司 | Pulmonary nodule automatic detection method and system based on pulmonary CT sequence |
CN109544545A (en) * | 2018-11-30 | 2019-03-29 | 济南浪潮高新科技投资发展有限公司 | A kind of salt mine intelligent detecting method and system based on convolutional neural networks |
CN109829880A (en) * | 2018-12-07 | 2019-05-31 | 清影医疗科技(深圳)有限公司 | A kind of CT image detecting method based on deep learning, device and control equipment |
CN109636788A (en) * | 2018-12-11 | 2019-04-16 | 中国石油大学(华东) | A kind of CT image gall stone intelligent measurement model based on deep neural network |
CN109523547A (en) * | 2018-12-21 | 2019-03-26 | 四川大学华西医院 | A kind of method and device of image tubercle detection |
CN109886307A (en) * | 2019-01-24 | 2019-06-14 | 西安交通大学 | A kind of image detecting method and system based on convolutional neural networks |
CN109919961A (en) * | 2019-02-22 | 2019-06-21 | 北京深睿博联科技有限责任公司 | A kind of processing method and processing device for aneurysm region in encephalic CTA image |
CN110008992B (en) * | 2019-02-28 | 2023-04-11 | 合肥工业大学 | Deep learning method for prostate cancer auxiliary diagnosis |
CN110123347B (en) * | 2019-03-22 | 2023-06-16 | 杭州深睿博联科技有限公司 | Image processing method and device for breast molybdenum target |
CN110148114A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A kind of deep learning model training method based on 2D faulted scanning pattern data set |
CN110148112A (en) * | 2019-04-02 | 2019-08-20 | 成都真实维度科技有限公司 | A method of it acquires and marks the progress data set foundation of tomoscan diagram data |
CN110021430A (en) * | 2019-04-09 | 2019-07-16 | 科大讯飞股份有限公司 | A kind of the attribute information prediction technique and device of lesion |
CN110059697B (en) * | 2019-04-29 | 2023-04-28 | 上海理工大学 | Automatic lung nodule segmentation method based on deep learning |
CN110084810B (en) * | 2019-05-06 | 2021-11-05 | 成都医云科技有限公司 | Pulmonary nodule image detection method, model training method, device and storage medium |
CN110084813B (en) * | 2019-05-13 | 2022-05-13 | 山东知微智成电子科技有限公司 | Lung nodule benign and malignant prediction method based on three-dimensional deep learning network |
CN110148467B (en) * | 2019-05-16 | 2023-05-23 | 东北大学 | Pulmonary nodule computer-aided diagnosis device and method based on improved CNN |
CN110309860B (en) * | 2019-06-06 | 2022-11-08 | 昆明理工大学 | Method for classifying malignancy degree of lung nodule based on convolutional neural network |
CN110473168A (en) * | 2019-07-09 | 2019-11-19 | 天津大学 | A kind of Lung neoplasm automatic checkout system squeezing-motivate residual error network based on 3D |
CN110705583B (en) * | 2019-08-15 | 2024-03-15 | 平安科技(深圳)有限公司 | Cell detection model training method, device, computer equipment and storage medium |
CN110660045B (en) * | 2019-08-30 | 2021-12-10 | 杭州电子科技大学 | Lymph node identification semi-supervision method based on convolutional neural network |
CN110706200B (en) * | 2019-09-02 | 2022-08-05 | 杭州深睿博联科技有限公司 | Data prediction method and device |
CN110689112A (en) * | 2019-09-02 | 2020-01-14 | 杭州深睿博联科技有限公司 | Data processing method and device |
CN110705570B (en) * | 2019-09-20 | 2022-06-21 | 桂林电子科技大学 | Image feature identification method |
CN110689547B (en) * | 2019-09-25 | 2022-03-11 | 重庆大学 | Pulmonary nodule segmentation method based on three-dimensional CT image |
CN110717916B (en) * | 2019-09-29 | 2022-08-30 | 华中科技大学 | Pulmonary embolism detection system based on convolutional neural network |
CN110766713A (en) * | 2019-10-30 | 2020-02-07 | 上海微创医疗器械(集团)有限公司 | Lung image segmentation method and device and lung lesion region identification equipment |
CN112819747A (en) * | 2019-10-31 | 2021-05-18 | 深圳市华嘉生物智能科技有限公司 | Method for automatically diagnosing benign and malignant nodules based on lung tomography image |
TWI760657B (en) * | 2019-11-14 | 2022-04-11 | 和碩聯合科技股份有限公司 | Training method for anomaly detection model and electronic device using the same |
CN110910404B (en) * | 2019-11-18 | 2020-08-04 | 西南交通大学 | Anti-noise data breast ultrasonic nodule segmentation method |
CN110890143B (en) * | 2019-11-21 | 2022-03-08 | 重庆邮电大学 | 2D convolution method introducing spatial information |
CN111079821A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Derailment automatic braking pull ring falling fault image identification method |
CN111080625B (en) * | 2019-12-18 | 2020-12-29 | 推想医疗科技股份有限公司 | Training method and training device for lung image strip and rope detection model |
CN111310624B (en) * | 2020-02-05 | 2023-11-21 | 腾讯科技(深圳)有限公司 | Occlusion recognition method, occlusion recognition device, computer equipment and storage medium |
CN111340756B (en) * | 2020-02-13 | 2023-11-28 | 北京深睿博联科技有限责任公司 | Medical image lesion detection merging method, system, terminal and storage medium |
CN110969623B (en) * | 2020-02-28 | 2020-06-26 | 北京深睿博联科技有限责任公司 | Lung CT multi-symptom automatic detection method, system, terminal and storage medium |
CN111340825B (en) * | 2020-03-05 | 2023-05-09 | 上海市肺科医院(上海市职业病防治院) | Method and system for generating mediastinum lymph node segmentation model |
CN111476766B (en) * | 2020-03-31 | 2023-08-22 | 哈尔滨商业大学 | Lung nodule CT image detection system based on deep learning |
CN111583271A (en) * | 2020-04-13 | 2020-08-25 | 华东师范大学 | Method for automatically predicting gene expression categories based on cancer CT images |
CN111784593A (en) * | 2020-06-04 | 2020-10-16 | 广东省智能制造研究所 | Lung nodule CT image data enhancement method and system for deep learning |
CN111815591B (en) * | 2020-06-29 | 2024-02-27 | 郑州大学 | Lung nodule detection method based on CT image |
CN111768382B (en) * | 2020-06-30 | 2023-08-15 | 重庆大学 | Interactive segmentation method based on lung nodule growth morphology |
CN112241948A (en) * | 2020-09-23 | 2021-01-19 | 深圳视见医疗科技有限公司 | Lung nodule detection and analysis method and system capable of adapting to thickness of layer |
CN112184657A (en) * | 2020-09-24 | 2021-01-05 | 上海健康医学院 | Pulmonary nodule automatic detection method, device and computer system |
CN112529870B (en) * | 2020-12-10 | 2024-04-12 | 重庆大学 | Multi-scale CNNs lung nodule false positive eliminating method based on combination of source domain and frequency domain |
CN112581436A (en) * | 2020-12-11 | 2021-03-30 | 佛山市普世医学科技有限责任公司 | Lung nodule identification and segmentation method and system based on deep learning |
CN112541909B (en) * | 2020-12-22 | 2022-05-17 | 南开大学 | Lung nodule detection method and system based on three-dimensional neural network of slice perception |
CN112669284A (en) * | 2020-12-29 | 2021-04-16 | 天津大学 | Method for realizing pulmonary nodule detection by generating confrontation network |
CN112991269A (en) * | 2021-02-07 | 2021-06-18 | 复旦大学 | Identification and classification method for lung CT image |
CN112669319B (en) * | 2021-03-22 | 2021-11-16 | 四川大学 | Multi-view multi-scale lymph node false positive inhibition modeling method |
CN113113132A (en) * | 2021-03-25 | 2021-07-13 | 华南理工大学 | Method for diagnosing congenital megacolon and homologous diseases based on deep convolutional neural network |
CN113256634B (en) * | 2021-07-13 | 2021-11-23 | 杭州医策科技有限公司 | Cervical carcinoma TCT slice vagina arranging method and system based on deep learning |
CN113658162A (en) * | 2021-08-24 | 2021-11-16 | 中山仰视科技有限公司 | Refined lung nodule detection method, system, equipment and medium |
CN114677383B (en) * | 2022-03-03 | 2024-03-15 | 西北工业大学 | Pulmonary nodule detection and segmentation method based on multitask learning |
CN116012355B (en) * | 2023-02-07 | 2023-11-21 | 重庆大学 | Adaptive false positive lung nodule removing method based on deep learning |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1462884A (en) * | 2003-06-24 | 2003-12-24 | 南京大学 | Method of recognizing image of lung cancer cells with high accuracy and low rate of false negative |
CN102542562A (en) * | 2011-11-23 | 2012-07-04 | 首都医科大学 | Method for extracting texture based on lung nodules three orthogonality position computed tomography (CT) image and method for forecasting lung cancers |
CN102842132A (en) * | 2012-07-12 | 2012-12-26 | 上海联影医疗科技有限公司 | CT pulmonary nodule detection method |
CN104700118A (en) * | 2015-03-18 | 2015-06-10 | 中国科学院自动化研究所 | Pulmonary nodule benignity and malignancy predicting method based on convolutional neural networks |
CN105005714A (en) * | 2015-06-18 | 2015-10-28 | 中国科学院自动化研究所 | Non-small cell lung cancer prognosis method based on tumor phenotypic characteristics |
CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
CN105469419A (en) * | 2016-01-07 | 2016-04-06 | 于翠妮 | Pulmonary nodule detection method based on CT image |
CN106203327A (en) * | 2016-07-08 | 2016-12-07 | 清华大学 | Lung tumor identification system and method based on convolutional neural networks |
CN106372390A (en) * | 2016-08-25 | 2017-02-01 | 姹ゅ钩 | Deep convolutional neural network-based lung cancer preventing self-service health cloud service system |
-
2017
- 2017-02-16 CN CN201710083606.XA patent/CN107016665B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1462884A (en) * | 2003-06-24 | 2003-12-24 | 南京大学 | Method of recognizing image of lung cancer cells with high accuracy and low rate of false negative |
CN102542562A (en) * | 2011-11-23 | 2012-07-04 | 首都医科大学 | Method for extracting texture based on lung nodules three orthogonality position computed tomography (CT) image and method for forecasting lung cancers |
CN102842132A (en) * | 2012-07-12 | 2012-12-26 | 上海联影医疗科技有限公司 | CT pulmonary nodule detection method |
CN104700118A (en) * | 2015-03-18 | 2015-06-10 | 中国科学院自动化研究所 | Pulmonary nodule benignity and malignancy predicting method based on convolutional neural networks |
CN105005714A (en) * | 2015-06-18 | 2015-10-28 | 中国科学院自动化研究所 | Non-small cell lung cancer prognosis method based on tumor phenotypic characteristics |
CN105160361A (en) * | 2015-09-30 | 2015-12-16 | 东软集团股份有限公司 | Image identification method and apparatus |
CN105469419A (en) * | 2016-01-07 | 2016-04-06 | 于翠妮 | Pulmonary nodule detection method based on CT image |
CN106203327A (en) * | 2016-07-08 | 2016-12-07 | 清华大学 | Lung tumor identification system and method based on convolutional neural networks |
CN106372390A (en) * | 2016-08-25 | 2017-02-01 | 姹ゅ钩 | Deep convolutional neural network-based lung cancer preventing self-service health cloud service system |
Also Published As
Publication number | Publication date |
---|---|
CN107016665A (en) | 2017-08-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107016665B (en) | CT pulmonary nodule detection method based on deep convolutional neural network | |
CN108986067B (en) | Cross-modality-based pulmonary nodule detection method | |
CN107103187B (en) | Lung nodule detection grading and management method and system based on deep learning | |
CN110766051A (en) | Lung nodule morphological classification method based on neural network | |
Naresh et al. | Early detection of lung cancer using neural network techniques | |
CN112150442A (en) | New crown diagnosis system based on deep convolutional neural network and multi-instance learning | |
Rafael-Palou et al. | Re-identification and growth detection of pulmonary nodules without image registration using 3D Siamese neural networks | |
CN112258461B (en) | Pulmonary nodule detection method based on convolutional neural network | |
CN112819747A (en) | Method for automatically diagnosing benign and malignant nodules based on lung tomography image | |
CN116091490A (en) | Lung nodule detection method based on YOLOv4-CA-CBAM-K-means++ -SIOU | |
Farhangi et al. | Automatic lung nodule detection in thoracic CT scans using dilated slice‐wise convolutions | |
Wu et al. | 3d centroidnet: nuclei centroid detection with vector flow voting | |
Zhang et al. | Multicontext 3D residual CNN for false positive reduction of pulmonary nodule detection | |
De Nunzio et al. | Approaches to juxta-pleural nodule detection in CT images within the MAGIC-5 Collaboration | |
Khouadja et al. | Lung Cancer Detection with Machine Learning and Deep Learning: A Narrative Review | |
CN114581698A (en) | Target classification method based on space cross attention mechanism feature fusion | |
CN113889235A (en) | Unsupervised feature extraction system for three-dimensional medical image | |
Yang et al. | Leveraging auxiliary information from EMR for weakly supervised pulmonary nodule detection | |
Doğan et al. | Computer-aided detection of brain tumors using morphological reconstruction | |
Saraswathi et al. | Detection of juxtapleural nodules in lung cancer cases using an optimal critical point selection algorithm | |
Elbouknify et al. | CT-xCOV: a CT-scan based Explainable Framework for COVid-19 diagnosis | |
Jain et al. | SumNet Convolution Neural network based Automated pulmonary nodule detection system | |
CN115331220B (en) | Candida cervicales infection screening method based on target detection and heat map regression | |
Wang et al. | A Lung Nodule Detector Based on U-Net and 3D-CNN Model | |
Wu et al. | An improved attention mechanism based YOLOv4 structure for lung nodule detection |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |