CN111798455B - Thyroid nodule real-time segmentation method based on full convolution dense cavity network - Google Patents
Thyroid nodule real-time segmentation method based on full convolution dense cavity network Download PDFInfo
- Publication number
- CN111798455B CN111798455B CN201910914157.8A CN201910914157A CN111798455B CN 111798455 B CN111798455 B CN 111798455B CN 201910914157 A CN201910914157 A CN 201910914157A CN 111798455 B CN111798455 B CN 111798455B
- Authority
- CN
- China
- Prior art keywords
- convolution
- layer
- dense
- full
- layers
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/10132—Ultrasound image
-
- 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
- G06T2207/30096—Tumor; Lesion
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses a thyroid nodule real-time segmentation method based on a full convolution dense cavity network, which comprises the following steps of: step one, thyroid data are obtained and preprocessed; marking the obtained data as a data set for training a full convolution dense cavity network model; constructing a full convolution dense cavity network model based on dense connection, and performing parameter training; step four, replacing the convolution kernel in the convolution layer with cavity convolution and decomposing by using convolution kernel decomposition; fifthly, carrying out data standardization and nonlinear activation processing on the input of the convolution layer; and step six, analyzing and comparing the segmentation effect and efficiency of the full convolution dense cavity network model.
Description
Technical Field
The invention belongs to the field of deep learning and image processing, relates to a convolutional neural network and an image semantic segmentation technology, and particularly relates to a thyroid nodule real-time segmentation method based on a full convolutional dense hole network.
Background
Thyroid nodules are the most common abnormality in the endocrine system, and their potential malignancy makes them clinically important. Ultrasound examination is the preferred imaging method for diagnosing thyroid nodules. In clinical practice, a radiologist diagnoses benign and malignant thyroid according to macroscopic evaluation criteria such as the aspect ratio of a nodule in an ultrasound image, whether calcification exists, structure (diffuse, single shot or multiple), boundary, echo characteristics (hyperecho, equal echo and hypoecho), etc., but different diagnosis results may appear on one thyroid ultrasound image due to the influence of cognitive ability, subjective experience, fatigue degree, etc. In addition, low contrast and speckle noise in ultrasound images can also have an impact on the physician's diagnosis.
In recent years, computer-aided diagnosis based on deep learning has been paid attention to by researchers, and research on thyroid ultrasound image-aided diagnosis has been developed. The existing deep learning party [1] [2] achieves high precision, however, the models trained by the deep learning party [1] [2] have a large number of weight parameters, a large amount of calculation resources are needed, and the storage space and the processor performance of medical equipment are limited, so that the practical application of the deep learning method is limited to a certain extent. On the other hand, the purpose of computer-aided diagnosis is to shorten the diagnosis time and improve the diagnosis efficiency and accuracy, which requires a depth model with high accuracy and high real-time.
In recent years, the design of deeper neural networks achieves higher precision than many traditional computer vision algorithms in the tasks of image classification, semantic segmentation, object detection, etc., however, this also requires a large amount of computing resources and long reasoning time, which makes the depth model unable to run on some resource-constrained platforms. In order to solve the balance problem between high precision and computing resources, most of the existing methods focus on network pruning [3], low-bit quantization [4] and design of efficient network architecture. However, whether the network pruning or the low-bit quantization is performed on a trained model, the accuracy of the model is inevitably affected. In contrast, designing an efficient network architecture can reduce the computational resources required by the model without losing accuracy. The existing deep network has a very large number of parameters, however, some parameters have very little or no effect when the network is running, compared with the previous network, the network designed in the document [5] has the advantages of improved performance and unchanged parameter number, so that the parameters of the network are reduced, and the effect of improving the residual parameters is very useful for balancing the precision and operation resources of the network.
Disclosure of Invention
The invention aims to solve the problems that the existing semantic segmentation model has too many parameters and cannot be efficiently operated on computer-aided diagnosis equipment with limited computing resources.
The invention aims at realizing the following technical scheme:
a thyroid nodule real-time segmentation method based on a full convolution dense cavity network comprises the following steps:
step one: thyroid data are obtained and preprocessed;
step two: labeling the obtained data to be used as a data set for training a full convolution dense cavity network model;
step three: constructing a full convolution dense network model based on dense connection, and performing parameter training;
step four: replacing the convolution kernel in the convolution layer with cavity convolution and decomposing by using convolution kernel decomposition;
step five: performing data standardization and nonlinear activation processing on the input of the convolution layer;
step six: and analyzing and comparing the precision index and the efficiency of the segmentation effect and the efficiency of the full convolution dense cavity network model.
Further, marking the edge of the nodule on the obtained thyroid ultrasonic image in the second step, and normalizing the size of the thyroid ultrasonic image; the resulting data set includes a training set and a test set.
Further, training the full convolution dense cavity network model parameters by using the data set obtained in the second step; the full convolution dense hole network model is an end-to-end model for thyroid nodule segmentation, follows the architecture of an automatic encoder-decoder, and uses dense connections to cross-layer transport features extracted by the convolution layers.
In the fourth step, the full convolution dense cavity network model is composed of convolution layers, the convolution kernels in different convolution layers are replaced by cavity convolutions with different cavity rates, and the two-dimensional convolution kernels are decomposed into one-dimensional convolution kernels by convolution kernel decomposition.
Furthermore, step five uses Batch Normalization (BN) and linear correction unit activation function (ReLU) to normalize and activate the input of the convolution layer on the basis of step four, and uses random inactivation (dorpout) to process the full convolution dense cavity network model parameters; specifically, the method comprises the steps of carrying out data normalization processing on input of a convolution layer by using Batch Normalization (BN), and increasing nonlinearity of data by using a linear correction unit activation function (ReLU).
Compared with the prior art, the technical scheme of the invention has the following beneficial effects:
1. the combination of dense connection, hole convolution and convolution kernel decomposition is used in the model constructed by the invention, so that the model is more efficient while similar learning performance is maintained. The dense connection can greatly improve the feature reuse, so that a model can generate a large number of features by using a small number of convolution kernels, and the size of the final model is small. Convolution kernel decomposition decomposes one convolution kernel of 3 x 3 size into two convolution kernels of 3 x 1 and 1 x 3 sizes, the model parameters decrease, but the nonlinearity of the model increases, so the accuracy does not drop significantly. The cavity convolution can acquire more context information while maintaining the same parameter quantity, so that the precision of the model is improved.
2. The model constructed by the invention achieves similar segmentation precision on a thyroid dataset as a high-precision model, the IOU is 0.57 percent higher than the FCDenseNet model with the best overall effect, the IOU is 70.72 percent, the running time on a single NVIDATITAN Xp GPU is less than 1/6 of the high-precision model, the IOU is only 7.74ms, the IOU is as competitive as the high-efficiency model, and the IOU is only 2.24ms more than the fastest ENT.
3. The model trained by the method of the invention has good balance between the segmentation precision and the segmentation speed, the model size is only 3.9Mb, and the method is suitable for equipment which needs robustness and efficiency.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Fig. 2a is a screened thyroid ultrasound image and fig. 2b shows the label (group trunk) of the image.
Fig. 3 is a schematic diagram of the structure of a single convolutional layer. In fig. 3, c represents the number of channels of the input feature, 3×1, 1×3 represents the size of the convolution kernel, k represents the number of convolution kernels, L represents the void fraction of the convolution kernel,
representing the connection of the input to the output.
Fig. 4a shows an image input to the full convolution dense hole network model, and fig. 4b shows the segmentation result of the FCDDN model. FIG. 4c shows the segmentation result of the FC DenseNet model; FIG. 4d shows the segmentation result of the U-Net model.
Detailed Description
The invention is described in further detail below with reference to the drawings and the specific examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The invention provides a thyroid nodule segmentation method based on a full convolution dense cavity network, as shown in fig. 1, which is an overall flow schematic diagram of a specific embodiment of the thyroid nodule segmentation method of the invention, comprising the following steps:
step one: thyroid data are obtained and preprocessed;
step 101: and acquiring thyroid ultrasonic image data subjected to pathological verification from a hospital, taking out the image from a folder with medical records, modifying the name of the image and making a backup, and then screening out the data with clear image and clear physiological structure of the nodule.
Step two: labeling the obtained data to be used as a data set for training a full convolution dense cavity network model;
step 201: the node edges of the screened ultrasound images are marked with the help of the ultrasound department and radiologist, and the group trunk is obtained according to the marks, so that the data set of the invention is formed, the images are randomly divided into a training set and a test set, and then the sizes of the images and the group trunk are uniformly scaled to 256×256 (the images and the group trunk are shown in fig. 2a and 2 b).
Step three: constructing a full convolution dense network model based on dense connection, and performing parameter training;
step 301: the full convolution dense hole network model is built according to the architecture of an automatic encoder-decoder, and can be called as a Fully Convolutional Dense DilatedNet model; abbreviated FCDDN model; wherein layers 1 to 8 constitute an encoder and layers 9 to 15 constitute a decoder. Layer 1 is a convolution layer with a convolution kernel size of 3 x 3, a number of 48; the 2 nd to 8 th layers are sequentially 1 convolution layer, pooling layer, 4 convolution layers and pooling layer, wherein the convolution kernel size of the 1 convolution layer is 3 multiplied by 3, the number of the convolution layers is 16, the pooling layer size is 2 multiplied by 2, and the step length is 2; the 9 th to 14 th layers are deconvolution layers, 1 convolution layer, deconvolution layer and 1 convolution layer in sequence, wherein the deconvolution size is 3 multiplied by 3, the step length is 2, the convolution kernel size of the convolution layers is 3 multiplied by 3, and the number is 16; layer 15 is a convolution layer with a convolution kernel size of 1 x 1 and a number of convolution kernels of 2, and is used to classify features.
Step 302: and training parameters of the full convolution dense cavity network model by using a training set in the thyroid dataset. The initial learning rate was 0.0001, the number of rounds per training divided by 2, the total training was 240 rounds, the final learning rate was 0.00005, and dropout was 0.8, and training was performed using a random gradient descent optimizer theAdam optimization.
Step four: replacing the convolution kernel in the convolution layer with cavity convolution and decomposing by using convolution kernel decomposition;
step 401: the full-convolution dense hole network model comprises a plurality of convolution layers, the convolution kernels in different convolution layers are replaced by using the hole convolutions with different hole ratios, the hole ratio is set according to the number of the convolution layers contained in the layers of the full-convolution dense hole network model, the layers containing one convolution layer are 2, the hole ratio of the hole convolutions is 2, the layers containing 4 convolution layers are 2, 4, 8 and 16.
Step 402: it is proposed to decompose the 2-dimensional convolution kernels in layers 2 to 14 into one-dimensional convolution kernels, which are all 3 x 3 in size, using a convolution kernel decomposition, which can be decomposed into two consecutive convolution kernels of 3 x 1 and 1 x 3 in size. For a convolution kernel of size 3 x 3, the convolution kernel decomposition can reduce the model parameters to the original 2/3.
Step five: performing data batch standardization and nonlinear activation processing on the input of the convolution layer;
step 501: the data input into the convolution layer is subjected to batch standardization processing, specifically, the data mean value is calculated, the data variance is calculated, the data is subjected to the data standardization processing by using a standard deviation formula, and finally two parameters of gamma and beta are introduced, and the data is subjected to translation and scaling processing.
Step 502: and carrying out nonlinear activation processing on the data after batch normalization processing by using a Relu function. The structure of the convolution layer is built, the complete schematic diagram is shown in fig. 3, the data of the input convolution layer is subjected to batch standardization processing, nonlinear activation processing and convolution kernels with the sizes of 3×1 and 1×3, finally the data of the input convolution layer and the output data are connected together, dropout only randomly deactivates part of neurons when the network is trained, and the trained network does not deactivate the neurons any more.
Step six: and analyzing and comparing the segmentation effect and efficiency of the full convolution dense cavity network model.
The trained full convolution dense cavity network model is used for nodule segmentation of the test set, the accuracy is an important aspect for measuring the segmentation effect, the main indexes for evaluating the accuracy are IOU, TPF and FPF, the segmentation efficiency is an important aspect for measuring the quality of the model, and the only indexes for evaluating the efficiency are the running time and the storage size. Table (1) quantitatively compares the segmentation effect and efficiency of the FCDDN model with those of other models, and from the perspective of IOU, TPF and FPF of each model, the proposed FCDDN models IOU, TPF and FPF are 72.72%, 96.10% and 0.677%, respectively, and the overall gap is not large compared with the best U-Net and FC DenseNet models in accuracy, although the IOU of U-Net is 3.76% higher than that of the model of the invention, the running time is nearly doubled, the parameters are two orders of magnitude more, and from the perspective of fig. 4a to 4d, the segmentation result of U-Net is always smaller than that of groundtruth, which can cause a certain influence on the doctor diagnosing the thyroid benign malignancy. From the viewpoint of running time and Model size (Model size), the full convolution dense network Model provided by the invention is as high as an efficient Model (ENT, ERFNet), the Model size is even much smaller, and the segmentation accuracy is higher than that of the efficient Model. The full convolution dense network model provided by the invention is optimal in all comparison models by integrating a plurality of evaluation indexes.
Table 1 model segmentation results and comparison
Specifically, the experiment verifies that three evaluation indexes of the interface-over-unit (IOU), true Positive Fraction (TPF) and False Positive Fraction (FPF) are adopted to evaluate the experimental result. The three evaluation indexes are calculated by the following formulas (1), (2) and (3).
IOU＝area(A∩B)/area(A∪B) (1)
TPF＝area(A∩B)/area(A) (2)
FPF＝(area(A)-area(A∩B))/(area(C)-area(A)) (3)
Where A is the nodule region in the ground trunk, B is the nodule region in the model segmentation result, and C is the ground trunk. The larger the IOU and TPF, the smaller the FPF, indicating a better segmentation effect. In addition, the running time of the model is tested and compared, and the shorter the running time is, the higher the model segmentation efficiency is.
Experimental results show that compared with a high-precision network, the full convolution dense network model provided by the invention achieves similar precision on the IOU, but the number of parameters is obviously smaller, and the running time is obviously shorter. U-Net [6] is less than 4% higher on IOU but 11% lower on TPF than our method, as can be seen from the effect graph of segmentation, since the method of the present invention can contain group trunk entirely, while U-Net does not. The parameters of the high-precision network are more than those of the network provided by the invention, so that a great deal of redundancy of the parameters of the depth model can be illustrated. As can be seen from the comparison of FCDDN and FC DenseNet [7], the design of a high-efficiency network architecture can improve efficiency without losing accuracy. Compared to an efficient network, it is only slightly longer in run time than the fastest ENT [8], but 2.49% more accurate than the highest ERFNet [9 ]. The final experimental result shows that the model in the invention has high efficiency while maintaining high precision.
The invention provides a high-precision and high-efficiency semantic segmentation network structure, which solves the problem that a large amount of computing resources are required to run on medical equipment in real time. The combination of dense connections, hole convolutions and convolution kernels is used in the model, making it more efficient while maintaining similar learning performance. The FCDDN model achieves similar segmentation accuracy on the thyroid dataset as the high accuracy model, and runs less than 1/6 of the high accuracy model on a single NVIDATITAN Xp GPU, with the same competitiveness as the high efficiency model. Finally, the model trained by the method of the invention achieves a good balance between segmentation accuracy and speed, making it suitable for use on devices that require both robustness and efficiency.
The invention is not limited to the embodiments described above. The above description of specific embodiments is intended to describe and illustrate the technical aspects of the present invention, and is intended to be illustrative only and not limiting. Numerous specific modifications can be made by those skilled in the art without departing from the spirit of the invention and scope of the claims, which are within the scope of the invention.
Reference is made to:
[1]Liu Shu,Qi Lu,Haifang Qin,Jianping Shi,and Jiaya Jia.Path aggregation network for instance segmentation.2018.
[2]Liang Chieh Chen,Yukun Zhu,George Papandreou,Florian Schroff,and Hartwig Adam.Encoder-decoder with atrous separable convolution for semantic image segmentation.2018
[3]Yihui He,Xiangyu Zhang,and Jian Sun.Channel pruning for accelerating very deep neural networks.2017
[4]Bohan Zhuang,Chunhua Shen,Mingkui Tan,Lingqiao Liu,and Ian Reid.Towards effective low-bitwidth convolutional neural networks.2017.
[5]Francois Chollet.Xception:Deep learning with depthwise separable convolutions.In IEEE Conference on Computer Vision&Pattern Recognition,2016.
[6]Olaf Ronneberger,Philipp Fischer,and Thomas Brox.U-net:Convolutional networks for biomedical image segmentation.In International Conference on Medical Image Computing&Computer-assisted Intervention,2015.
[7]Simon Jegou,Michal Drozdzal,David′Vazquez,Adriana Romero,and Yoshua Bengio.The one hundred layers tiramisu:Fully convolutional densenets for semantic segmentation.2016.
[8]Adam Paszke,Abhishek Chaurasia,
Sangpil Kim,and Eugenio Culurciello.Enet:A deep neural network architecture for real-time semantic segmentation.2016.
[9]Eduardo Romera,Jose M.Alvarez,Luis M.Bergasa,Roberto Arroyo,Eduardo Romera,Jose M.Alvarez,Luis M.Bergasa,Roberto Arroyo,Eduardo Romera,and Jose M.Alvarez.Erfnet:Efficient residual factorized convnet for real-time semantic segmentation.IEEE Transactions on Intelligent Transportation Systems,PP(99):1–10,2017.
Claims (4)
1. a thyroid nodule real-time segmentation method based on a full convolution dense cavity network is characterized by comprising the following steps:
step one: thyroid data are obtained and preprocessed;
step two: labeling the obtained data to be used as a data set for training a full convolution dense cavity network model;
the full convolution dense hole network model is built according to the architecture of an automatic encoder-decoder, and is called FCDDN model for short; wherein layers 1 to 8 constitute an encoder and layers 9 to 15 constitute a decoder; layer 1 is a convolution layer with a convolution kernel size of 3 x 3, a number of 48; the 2 nd to 8 th layers are sequentially 1 convolution layer, pooling layer, 4 convolution layers and pooling layer, wherein the convolution kernel size of the 1 convolution layer is 3 multiplied by 3, the number of the convolution layers is 16, the pooling layer size is 2 multiplied by 2, and the step length is 2; the 9 th to 14 th layers are deconvolution layers, 1 convolution layer, deconvolution layer and 1 convolution layer in sequence, wherein the deconvolution size is 3 multiplied by 3, the step length is 2, the convolution kernel size of the convolution layers is 3 multiplied by 3, and the number is 16; layer 15 is a convolution layer with convolution kernel size of 1×1 and number of convolution kernels of 2, and is used for classifying features;
step three: constructing a full convolution dense cavity network model based on dense connection, and performing parameter training;
step four: replacing the convolution kernel in the convolution layer with cavity convolution and decomposing by using convolution kernel decomposition;
step five: performing data standardization and nonlinear activation processing on the input of the convolution layer;
step six: and analyzing and comparing the precision index and the efficiency of the segmentation effect and the efficiency of the full convolution dense cavity network model.
2. The real-time thyroid nodule segmentation method based on the full convolution dense cavity network, which is characterized by comprising the following steps of marking the nodule edges of an obtained thyroid ultrasound image in the second step, and normalizing the size of the thyroid ultrasound image; the resulting data set includes a training set and a test set.
3. The thyroid nodule real-time segmentation method based on the full convolution dense hole network according to claim 1, wherein in the third step, training is performed on the full convolution dense hole network model parameters by using the data set obtained in the second step; the full convolution dense hole network model is an end-to-end model for thyroid nodule segmentation, follows the architecture of an automatic encoder-decoder, and uses dense connections to cross-layer transport features extracted by the convolution layers.
4. The thyroid nodule real-time segmentation method based on the full-convolution dense hole network according to claim 1, wherein in the fourth step, the full-convolution dense hole network model is composed of convolution layers, convolution kernels in different convolution layers are replaced by hole convolutions with different void ratios, and two-dimensional convolution kernels are decomposed into one-dimensional convolution kernels by convolution kernel decomposition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910914157.8A CN111798455B (en) | 2019-09-25 | 2019-09-25 | Thyroid nodule real-time segmentation method based on full convolution dense cavity network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910914157.8A CN111798455B (en) | 2019-09-25 | 2019-09-25 | Thyroid nodule real-time segmentation method based on full convolution dense cavity network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111798455A CN111798455A (en) | 2020-10-20 |
| CN111798455B true CN111798455B (en) | 2023-07-04 |
Family
ID=72805521
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910914157.8A Active CN111798455B (en) | 2019-09-25 | 2019-09-25 | Thyroid nodule real-time segmentation method based on full convolution dense cavity network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111798455B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112381846B (en) * | 2020-12-11 | 2024-03-08 | 江南大学 | Ultrasonic thyroid nodule segmentation method based on asymmetric network |
| CN113610859B (en) * | 2021-06-07 | 2023-10-31 | 东北大学 | Automatic thyroid nodule segmentation method based on ultrasonic image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056595A (en) * | 2015-11-30 | 2016-10-26 | 浙江德尚韵兴图像科技有限公司 | Method for automatically identifying whether thyroid nodule is benign or malignant based on deep convolutional neural network |
| CN108776969A (en) * | 2018-05-24 | 2018-11-09 | 复旦大学 | Breast ultrasound image lesion segmentation approach based on full convolutional network |
| CN109087327A (en) * | 2018-07-13 | 2018-12-25 | 天津大学 | A kind of thyroid nodule ultrasonic image division method cascading full convolutional neural networks |
| CN109344883A (en) * | 2018-09-13 | 2019-02-15 | 西京学院 | Fruit tree diseases and pests recognition methods under a kind of complex background based on empty convolution |
-
2019
- 2019-09-25 CN CN201910914157.8A patent/CN111798455B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056595A (en) * | 2015-11-30 | 2016-10-26 | 浙江德尚韵兴图像科技有限公司 | Method for automatically identifying whether thyroid nodule is benign or malignant based on deep convolutional neural network |
| CN108776969A (en) * | 2018-05-24 | 2018-11-09 | 复旦大学 | Breast ultrasound image lesion segmentation approach based on full convolutional network |
| CN109087327A (en) * | 2018-07-13 | 2018-12-25 | 天津大学 | A kind of thyroid nodule ultrasonic image division method cascading full convolutional neural networks |
| CN109344883A (en) * | 2018-09-13 | 2019-02-15 | 西京学院 | Fruit tree diseases and pests recognition methods under a kind of complex background based on empty convolution |
Non-Patent Citations (1)
| Title |
|---|
| Semantic Segmentation of Pathological Lung Tissue with Dilated Fully Convolutional Networks;Marios Anthimopoulos 等;arXiv;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111798455A (en) | 2020-10-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | A survey on deep learning for neuroimaging-based brain disorder analysis | |
| Hussein et al. | Lung and pancreatic tumor characterization in the deep learning era: novel supervised and unsupervised learning approaches | |
| Chhikara et al. | Deep convolutional neural network with transfer learning for detecting pneumonia on chest X-rays | |
| Jiang et al. | Learning efficient, explainable and discriminative representations for pulmonary nodules classification | |
| El-Shafai et al. | Efficient Deep-Learning-Based Autoencoder Denoising Approach for Medical Image Diagnosis. | |
| Hao et al. | A transfer learning–based active learning framework for brain tumor classification | |
| Naheed et al. | Importance of features selection, attributes selection, challenges and future directions for medical imaging data: a review | |
| Tataru et al. | Deep Learning for abnormality detection in Chest X-Ray images | |
| Karim et al. | DeepCOVIDExplainer: Explainable COVID-19 diagnosis based on chest X-ray images | |
| CN111798455B (en) | Thyroid nodule real-time segmentation method based on full convolution dense cavity network | |
| Aslam et al. | Multiple sclerosis diagnosis using machine learning and deep learning: challenges and opportunities | |
| Khanna et al. | Radiologist-level two novel and robust automated computer-aided prediction models for early detection of COVID-19 infection from chest X-ray images | |
| Pitroda et al. | An explainable AI model for interpretable lung disease classification | |
| CN113421228A (en) | Thyroid nodule identification model training method and system based on parameter migration | |
| Jin et al. | Machine learning techniques for pulmonary nodule computer-aided diagnosis using CT images: A systematic review | |
| Liu et al. | Segmenting lung parenchyma from CT images with gray correlation‐based clustering | |
| Beaulieu-Jones et al. | Learning contextual hierarchical structure of medical concepts with poincairé embeddings to clarify phenotypes | |
| Goyal et al. | Musculoskeletal abnormality detection in medical imaging using GnCNNr (group normalized convolutional neural networks with regularization) | |
| Boman et al. | Evaluating a deep convolutional neural network for classification of skin cancer | |
| Ghosh et al. | Ensemble based neural network for the classification of mura dataset | |
| Jacaruso | Accuracy improvement for Fully Convolutional Networks via selective augmentation with applications to electrocardiogram data | |
| Rodrigues-Jr et al. | Patient trajectory prediction in the Mimic-III dataset, challenges and pitfalls | |
| Pradeep et al. | A Certain Investigation on Intravascular Optical Coherence Tomography for Clustering Based Image Segmentation | |
| Ganesan et al. | Internet of medical things and cloud enabled brain tumour diagnosis model using deep learning with kernel extreme learning machine | |
| Zang et al. | AmtNet: Attentional multi-scale temporal network for phonocardiogram signal classification |
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 |