CN110070612B - CT image interlayer interpolation method based on generation countermeasure network - Google Patents
CT image interlayer interpolation method based on generation countermeasure network Download PDFInfo
- Publication number
- CN110070612B CN110070612B CN201910340691.2A CN201910340691A CN110070612B CN 110070612 B CN110070612 B CN 110070612B CN 201910340691 A CN201910340691 A CN 201910340691A CN 110070612 B CN110070612 B CN 110070612B
- Authority
- CN
- China
- Prior art keywords
- image
- layer
- countermeasure network
- generator
- interpolation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 239000011229 interlayer Substances 0.000 title claims abstract description 25
- 239000010410 layer Substances 0.000 claims abstract description 83
- 238000010606 normalization Methods 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 14
- 230000004913 activation Effects 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 7
- 238000012804 iterative process Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000013170 computed tomography imaging Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000008929 regeneration Effects 0.000 description 1
- 238000011069 regeneration method Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- 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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Graphics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a CT image interlayer interpolation method based on a generated countermeasure network; comprising the following steps: s1, acquiring a thick-layer CT image aiming at a CT image to be processed, and carrying out linear normalization processing on the thick-layer CT image; s2, combining and inputting adjacent two layers of the thick-layer CT image after normalization processing to a pre-trained generator for generating an countermeasure network; s3, taking the output of the pre-trained generator for generating the countermeasure network as a CT interlayer interpolation image; the method can automatically acquire CT interlayer interpolation images by utilizing the generated countermeasure network, and the model has the advantages of simple structure, high convergence speed, high precision and small calculated amount, and the image is accurate and convenient for the establishment of subsequent three-dimensional images.
Description
Technical Field
The invention belongs to the technical field of medical image processing, and particularly relates to a CT image interlayer interpolation method based on a generated countermeasure network.
Background
With the continued development of medical imaging technology, medical images have evolved from early X-ray films to today's sequential two-dimensional digital tomographic images. Image interpolation is particularly well-suited for medical image data processing, and generally requires interpolation processing from the generation of a medical image to post-processing. Lehmann et al have recognized that since the advent of computer graphics and digital image processing, image interpolation has been a process of image data regeneration. Tomographic imaging such as MRI and CT imaging is to perform coplanar sampling by adjusting the distance and thickness to obtain a cross-sectional image of an object under examination. In digital medical image studies, it is often desirable to recover its three-dimensional shape from two-dimensional serial slices, reconstructing a virtual organ or tissue to aid in clinical analysis and diagnosis. In some clinically complex procedures, a physician needs to know three-dimensional spatial information of tissue before the procedure, and thus, three-dimensional reconstruction of medical images has been developed. The three-dimensional reconstruction of the images requires stacking together the clinically acquired serial slices.
Based on the time and hardware cost, the current spiral CT for clinic in China is mostly 64 rows, the layer spacing is generally 3mm-7mm, and the acquired sequence slice layer spacing is far greater than the distance between adjacent pixels in the slice layer. Therefore, when a three-dimensional visualization model is built, the layer spacing of the tomographic image is too large, the layer resolution is low, and a new image layer needs to be generated by an interpolation method. The object of the tomographic image interpolation is to generate a new interpolation image between two adjacent tomographic images by an interpolation method, so that the number of the sequential tomographic images is increased, the distance between the images is reduced, and the subsequent three-dimensional reconstruction of the images is facilitated. The image data is added in the fault image interpolation, and the interpolation itself also consumes a certain time, so that a new algorithm which is relatively applicable must be researched in the fault image interpolation technology, so that the accuracy and the efficiency of three-dimensional reconstruction can be truly improved. The interpolation method of the tomographic image can be classified into a gray-scale-based interpolation method, a shape-based interpolation method, and a wavelet-based interpolation method. Although these methods improve interpolation accuracy to some extent, the algorithm is complex, so it is necessary to further improve the algorithm accuracy and reduce the computation complexity.
Disclosure of Invention
First, the technical problem to be solved
In order to solve the technical problems of slow convergence speed and poor effect of the traditional CT image interlayer interpolation method, the invention provides a CT image interlayer interpolation method based on a generation countermeasure network.
(II) technical scheme
In order to achieve the above purpose, the main technical scheme adopted by the invention comprises the following steps:
s1, acquiring a thick-layer CT image aiming at a CT image to be processed, and carrying out linear normalization processing on the thick-layer CT image;
s2, combining and inputting adjacent two layers of the thick-layer CT image after normalization processing to a pre-trained generator for generating an countermeasure network;
and S3, taking the output of the pre-trained generator for generating the countermeasure network as a CT interlayer interpolation image.
Optionally, the method further comprises:
s4, performing three-dimensional reconstruction by using the thick-layer CT image and the CT interlayer interpolation image to obtain a three-dimensional image.
Optionally, before step S2, it further comprises
A1, constructing a generation countermeasure network comprising a generator and a discriminator;
a2, performing linear normalization operation on the plurality of thin CT sample images, and combining the K-1 layer image and the K+1 layer image of each thin CT sample image after normalization to obtain a training sample;
a3, inputting all training samples into a generator for generating an countermeasure network by using the generated countermeasure network model, and taking an image output by the generator for generating the countermeasure network by using the generated countermeasure network model as a new K-layer interpolation image;
a4, taking the K-th layer image of the new K-th layer interpolation image and the K-th layer image of the corresponding thin layer CT sample image as the input of the discriminator for generating the countermeasure network by the generated countermeasure network model;
and (3) optimizing a pre-constructed loss function by using a back propagation method, updating the network weights of a generator and a discriminator according to the result of the loss function optimization, carrying out structural similarity evaluation on the new K-layer interpolation image, repeating the steps A1 to A5, and saving the weight used as a generated countermeasure network when the structural similarity evaluation is optimal in the iterative process to acquire the pre-trained generated countermeasure network.
Optionally, in step A1, the generator for generating the countermeasure network includes a first convolution structure, and the arbiter for generating the countermeasure network includes a second convolution structure;
the first convolution structure is used for downsampling an input image of the generator, and upsampling the downsampled feature map by adopting a nearest neighbor interpolation algorithm to restore the size of the image;
the second convolution structure is used for extracting features of the image input by the discriminator, and the convolution product with the size of the feature map and the like is adopted to enable the output of the discriminator to be a scalar.
Optionally, the first convolution structure includes: a first convolution layer, a first activation layer with a Relu as an activation function, and a first regularization layer Instance Normalization;
the second convolution structure includes: a second convolution layer, a second activation layer with a activation function of leak Relu, and a second regularization layer Instance Normalization.
Optionally, the three-dimensional image includes: three-dimensionally reconstructing a sagittal plane result image, and/or three-dimensionally reconstructing a coronal plane result image.
Optionally, the thick layer CT image has a thickness of 1mm.
Optionally, the thin layer CT image has a thickness of 5mm.
(III) beneficial effects
The beneficial effects of the invention are as follows: on one hand, the method of the invention utilizes the CT interlayer interpolation image which can be automatically generated, and the generated countermeasure network has simple structure, convenient application, high convergence speed and small calculated amount; on the other hand, the CT interlayer interpolation image generated by the method has a good accurate effect, and provides a basis for the subsequent three-dimensional reconstruction of the image.
Drawings
FIG. 1 is a flow chart of an inter-layer interpolation method for CT images based on a generated countermeasure network according to an embodiment of the present invention;
FIG. 2 is a flow chart of generating an countermeasure network using training set and test set acquisition in accordance with an embodiment of the present invention;
FIG. 3 is a diagram illustrating a network structure of a generator for generating an countermeasure network according to an embodiment of the present invention;
FIG. 4 is a diagram of a network architecture of a arbiter for generating an countermeasure network according to an embodiment of the present invention;
FIG. 5a is a new K-th layer interpolated image generated using sample set training according to an embodiment of the present invention;
FIG. 5b is a schematic illustration of an exemplary CT inter-layer interpolation image generated from a CT image to be processed;
FIG. 6a is a diagram of a three-dimensional reconstructed sagittal plane result provided by an embodiment of the present invention;
FIG. 6b is a graph of three-dimensional reconstructed coronal plane results according to an embodiment of the present invention.
Detailed Description
The invention will be better explained by the following detailed description of the embodiments with reference to the drawings.
Example 1
The embodiment provides a CT image interlayer interpolation method based on a generated countermeasure network, which specifically comprises the following steps:
as shown in fig. 2, the present invention obtains a CT inter-layer interpolation image by using a generated countermeasure network, the generated countermeasure network has the characteristics of simple structure and convenient application, and the generated countermeasure network needs to train the countermeasure network by using a large amount of data in the using process, for example, a sample set and a test set are established to verify a network model, wherein the obtaining the pre-trained generated countermeasure network includes the following steps:
a1, constructing a generation countermeasure network comprising a generator and a discriminator; for example, in step A1 shown in fig. 3, the generator that generates the countermeasure network includes a first convolution structure, and as shown in fig. 4, the arbiter that generates the countermeasure network includes a second convolution structure; the first convolution structure is used for downsampling the input image of the generator, and upsampling the downsampled feature map by adopting a nearest neighbor interpolation algorithm to restore the size of the image;
the second convolution structure is used for extracting features of the image input by the discriminator, and the convolution product with the size of the feature map and the like is adopted to enable the output of the discriminator to be a scalar.
Preferably, the first convolution structure comprises: a first convolution layer, a first activation layer with a Relu as an activation function, and a first regularization layer Instance Normalization;
the second convolution structure includes: a second convolution layer, a second activation layer with a activation function of leak Relu, and a second regularization layer Instance Normalization.
For example, the depth, the number of convolution kernels, the convolution sum and the size of the generator and the arbiter network in this embodiment may be modified according to the size of the actual input image, which is not limited in this embodiment, and the whole network architecture does not use a full connection layer;
a2, performing linear normalization operation on the plurality of thin CT sample images, and combining the K-1 layer image and the K+1 layer image of each thin CT sample image after normalization to obtain a training sample; wherein K is a positive integer greater than or equal to 1; for example, the CT images comprise a thin layer CT image and a thick layer CT image, the training of the invention by using the thin layer CT image is that the network is hoped to learn the information of the thin layer, and finally the whole network is applied to the thick layer CT image, so that the three-dimensional reconstruction effect is better.
In this embodiment, the thickness of the thick layer CT image is 5mm, the thickness of the thin layer CT image used for training is 1mm, and images with other thicknesses can be selected in the application process, and the specific thickness is not limited only for illustration.
A3, inputting the training sample into a generator for generating an countermeasure network, and taking an image output by the generator for generating the countermeasure network as a new K-th layer interpolation image; specifically, as shown in fig. 5a, in this embodiment, a training sample set is input to a generator for generating an countermeasure network, and a new K-th layer interpolation image is obtained;
a4, taking the K-th layer image of the new K-th layer interpolation image and the K-th layer image of the corresponding thin layer CT sample image as the input of the discriminator for generating the countermeasure network by the generated countermeasure network model;
and (3) optimizing a pre-constructed loss function by using a back propagation method, updating the network weights of a generator and a discriminator according to the result of the loss function optimization, carrying out structural similarity evaluation on the new K-th interpolation image in the training process, repeating the steps A1 to A5, and saving the weight used as the generated countermeasure network when the structural similarity evaluation is optimal in the iterative process to acquire the pre-trained generated countermeasure network.
Specifically, for example, firstly, in the process of generating an countermeasure network by using sample image training, a loss function is constructed in advance, and a back propagation method optimizes the loss function and updates the weights of a generator and a discriminator for generating the countermeasure network; secondly, the method also carries out structural similarity evaluation on the new K-layer interpolation image, and takes the weights of the corresponding generator and the corresponding discriminator when the structural similarity evaluation of the new K-layer interpolation image in the iteration process is optimal as the weights of the pre-trained generator and the pre-trained discriminator for generating the countermeasure network.
As shown in fig. 1, after acquiring a pre-trained generated countermeasure network, acquiring a CT inter-layer interpolation image for an arbitrary CT image to be processed includes the steps of:
s1, acquiring a thick-layer CT image aiming at the CT image to be processed, and carrying out linear normalization processing on the thick-layer CT image;
s2, combining and inputting adjacent two layers of the thick-layer CT image after normalization processing to a pre-trained generator for generating an countermeasure network;
s3, taking the output of a pre-trained generator for generating an countermeasure network as a CT interlayer interpolation image; specifically, the pre-trained CT inter-layer interpolation image output by the generator for generating the countermeasure network in this embodiment is shown in fig. 5 b;
preferably, in the actual application process, after the CT interlayer interpolation image is obtained, the method further comprises the step S4 of carrying out three-dimensional reconstruction by using the thick-layer CT image and the CT interlayer interpolation image to obtain a three-dimensional image;
for example, three-dimensional reconstruction is performed by using the thick-layer CT image of the CT image to be processed and the CT inter-layer interpolation image obtained in step S3, the obtained three-dimensional reconstruction sagittal plane result is shown in fig. 6a, and the obtained three-dimensional reconstruction coronal plane result is shown in fig. 6 b.
Finally, it should be noted that: the embodiments described above are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced with equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (4)
1. An inter-layer interpolation method of CT images based on a generated countermeasure network, wherein the generated countermeasure network comprises a generator and a discriminator, and is characterized by comprising the following steps:
s1, acquiring a thick-layer CT image aiming at a CT image to be processed, and carrying out linear normalization processing on the thick-layer CT image;
s2, combining and inputting adjacent two layers of the thick-layer CT image after normalization processing to a pre-trained generator for generating an countermeasure network; the whole architecture of the generated countermeasure network does not use a full connection layer;
s3, taking the output of the pre-trained generator for generating the countermeasure network as a CT interlayer interpolation image;
s4, performing three-dimensional reconstruction by using the thick-layer CT image and the CT interlayer interpolation image to obtain a three-dimensional image;
the method further comprises the following steps before the step S2:
a1, constructing a generation countermeasure network comprising a generator and a discriminator;
a2, performing linear normalization operation on the plurality of thin CT sample images, and combining the K-1 layer image and the K+1 layer image of each thin CT sample image after normalization to obtain a training sample; k is a positive integer greater than or equal to 1;
a3, inputting all training samples to a generator for generating an countermeasure network by the generated countermeasure network model, and taking an image output by the generator for generating the countermeasure network by the generated countermeasure network model as a new K-layer interpolation image;
a4, taking the K-th layer image of the new K-th layer interpolation image and the K-th layer image of the corresponding thin layer CT sample image as the input of the discriminator for generating the countermeasure network by the generated countermeasure network model;
a5, optimizing a pre-constructed loss function by using a back propagation method, updating the network weights of a generator and a discriminator according to the optimized result of the loss function, and evaluating the structural similarity of the new K-th interpolation image;
repeating the steps A1 to A5, and saving the weight of the generated countermeasure network when the structural similarity evaluation is optimal in the iterative process, so as to acquire the pre-trained generated countermeasure network;
wherein, the thickness of the thick layer CT image is 5mm, and the thickness of the thin layer CT image used for training is 1mm.
2. The method of claim 1, wherein in step A1, the generator that generates the countermeasure network comprises a first convolution structure, and the arbiter that generates the countermeasure network comprises a second convolution structure;
the first convolution structure is used for downsampling an input image of the generator, and upsampling the downsampled feature map by adopting a nearest neighbor interpolation algorithm to restore the size of the image;
the second convolution structure is used for extracting features of the image input by the discriminator, and the convolution product with the size of the feature map and the like is adopted to enable the output of the discriminator to be a scalar.
3. The method of claim 2, wherein,
the first convolution structure includes: a first convolution layer, a first activation layer with a Relu as an activation function, and a first regularization layer Instance Normalization;
the second convolution structure includes: a second convolution layer, a second activation layer with a activation function of leak Relu, and a second regularization layer Instance Normalization.
4. The method of claim 1, wherein the three-dimensional image comprises: three-dimensionally reconstructing a sagittal plane result image, and/or three-dimensionally reconstructing a coronal plane result image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910340691.2A CN110070612B (en) | 2019-04-25 | 2019-04-25 | CT image interlayer interpolation method based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910340691.2A CN110070612B (en) | 2019-04-25 | 2019-04-25 | CT image interlayer interpolation method based on generation countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110070612A CN110070612A (en) | 2019-07-30 |
| CN110070612B true CN110070612B (en) | 2023-09-22 |
Family
ID=67369003
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910340691.2A Active CN110070612B (en) | 2019-04-25 | 2019-04-25 | CT image interlayer interpolation method based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110070612B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220114259A1 (en) * | 2020-10-13 | 2022-04-14 | International Business Machines Corporation | Adversarial interpolation backdoor detection |
| US12019747B2 (en) * | 2020-10-13 | 2024-06-25 | International Business Machines Corporation | Adversarial interpolation backdoor detection |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112633306B (en) * | 2019-09-24 | 2023-09-22 | 杭州海康威视数字技术股份有限公司 | Method and device for generating countermeasure image |
| CN110796613B (en) * | 2019-10-10 | 2023-09-26 | 东软医疗系统股份有限公司 | Automatic identification method and device for image artifacts |
| CN112435341B (en) * | 2020-11-23 | 2022-08-19 | 推想医疗科技股份有限公司 | Training method and device for three-dimensional reconstruction network, and three-dimensional reconstruction method and device |
| CN112509091B (en) * | 2020-12-10 | 2023-11-14 | 上海联影医疗科技股份有限公司 | Medical image reconstruction method, device, equipment and medium |
| CN113706358A (en) * | 2021-07-09 | 2021-11-26 | 清华大学 | Method and device for encrypting tomographic image interlamellar spacing |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107968962A (en) * | 2017-12-12 | 2018-04-27 | 华中科技大学 | A kind of video generation method of the non-conterminous image of two frames based on deep learning |
| GB201809604D0 (en) * | 2018-06-12 | 2018-07-25 | Tom Tom Global Content B V | Generative adversarial networks for image segmentation |
| CN108629816A (en) * | 2018-05-09 | 2018-10-09 | 复旦大学 | The method for carrying out thin layer MR image reconstruction based on deep learning |
| GB201902459D0 (en) * | 2019-02-22 | 2019-04-10 | Facesoft Ltd | Facial shape representation and generation system and method |
-
2019
- 2019-04-25 CN CN201910340691.2A patent/CN110070612B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107968962A (en) * | 2017-12-12 | 2018-04-27 | 华中科技大学 | A kind of video generation method of the non-conterminous image of two frames based on deep learning |
| CN108629816A (en) * | 2018-05-09 | 2018-10-09 | 复旦大学 | The method for carrying out thin layer MR image reconstruction based on deep learning |
| GB201809604D0 (en) * | 2018-06-12 | 2018-07-25 | Tom Tom Global Content B V | Generative adversarial networks for image segmentation |
| GB201902459D0 (en) * | 2019-02-22 | 2019-04-10 | Facesoft Ltd | Facial shape representation and generation system and method |
Non-Patent Citations (4)
| Title |
|---|
| Adversarial Image Alignment and Interpolation;Viren Jain;《https://arxiv.org/abs/1707.00067》;20170630;1-10 * |
| Frame Interpolation Using Generative Adversarial Networks;Mark Koren ET AL;《http://cs231n.stanford.edu/reports/2017/pdfs/317.pdf》;20170702;1-8 * |
| Reconstruction of Thin-Slice Medical Image Using Generative Adversarial Network;Zeju Li ET AL;《Machine Learning in Medical Imaging》;20170907;325-333 * |
| 基于低剂量CT的层间插值技术研究;孟博;中国优秀硕士学位论文全文数据库 信息科技辑(第11期);I138-865 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220114259A1 (en) * | 2020-10-13 | 2022-04-14 | International Business Machines Corporation | Adversarial interpolation backdoor detection |
| US12019747B2 (en) * | 2020-10-13 | 2024-06-25 | International Business Machines Corporation | Adversarial interpolation backdoor detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110070612A (en) | 2019-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070612B (en) | CT image interlayer interpolation method based on generation countermeasure network | |
| CN107610194B (en) | Magnetic resonance image super-resolution reconstruction method based on multi-scale fusion CNN | |
| CN104933683B (en) | A kind of non-convex low-rank method for reconstructing for magnetic resonance fast imaging | |
| CN111951344B (en) | Magnetic resonance image reconstruction method based on cascade parallel convolution network | |
| CN110517238B (en) | AI three-dimensional reconstruction and human-computer interaction visualization network system for CT medical image | |
| CN112258415B (en) | Chest X-ray film super-resolution and denoising method based on generation countermeasure network | |
| CN111870245B (en) | Cross-contrast-guided ultra-fast nuclear magnetic resonance imaging deep learning method | |
| Banerjee et al. | A completely automated pipeline for 3D reconstruction of human heart from 2D cine magnetic resonance slices | |
| US8873819B2 (en) | Method for sorting CT image slices and method for constructing 3D CT image | |
| CN113298710B (en) | Optical coherence tomography super-resolution imaging method based on external attention mechanism | |
| CN111091575B (en) | Medical image segmentation method based on reinforcement learning method | |
| CN111047512B (en) | Image enhancement method and device and terminal equipment | |
| CN111340903A (en) | Method and system for generating synthetic PET-CT image based on non-attenuation correction PET image | |
| CN109741439B (en) | Three-dimensional reconstruction method of two-dimensional MRI fetal image | |
| CN116823625A (en) | Cross-contrast magnetic resonance super-resolution method and system based on variational self-encoder | |
| CN114913262A (en) | Nuclear magnetic resonance imaging method and system based on joint optimization of sampling mode and reconstruction algorithm | |
| CN110942496A (en) | Propeller sampling and neural network-based magnetic resonance image reconstruction method and system | |
| CN112213673B (en) | Dynamic magnetic resonance imaging method, device, reconstruction computer and magnetic resonance system | |
| CN111161370B (en) | Human body multi-core DWI joint reconstruction method based on AI | |
| Qi et al. | Multi-task MR imaging with iterative teacher forcing and re-weighted deep learning | |
| CN114926559A (en) | PET reconstruction method based on dictionary learning thought attenuation-free correction | |
| JP2021099794A (en) | System and method for reconstructing medical images using deep neural network and recursive decimation of measurement data | |
| CN113052840A (en) | Processing method based on low signal-to-noise ratio PET image | |
| CN112967295A (en) | Image processing method and system based on residual error network and attention mechanism | |
| Islam et al. | A critical survey on developed reconstruction algorithms for computed tomography imaging from a limited number of projections |
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 |