CN109934831A - A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging - Google Patents
A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging Download PDFInfo
- Publication number
- CN109934831A CN109934831A CN201910203727.2A CN201910203727A CN109934831A CN 109934831 A CN109934831 A CN 109934831A CN 201910203727 A CN201910203727 A CN 201910203727A CN 109934831 A CN109934831 A CN 109934831A
- Authority
- CN
- China
- Prior art keywords
- image
- tumor
- icg
- real
- fluorescent image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The invention discloses one kind to be based on indocyanine green (Indocyanine-green, English abbreviation ICG) surgical tumor of fluorescent imaging performs the operation real-time navigation method, comprising the following steps: and 1. obtain each frame image in real-time ICG fluorescence operation video frequencies and save as ICG fluorescent image；2. being come out the tumor boundary segmentation in ICG fluorescent image using the tumor boundary segmentation model based on deep learning, and it is added in former ICG fluorescent image in a manner of colored lines and is shown；3. assessing using edge valuation functions the partitioning boundary of ICG fluorescent image, its corresponding sharpness of border coefficient is obtained；4. finding out the maximum image of sharpness of border coefficient, then its corresponding partitioning boundary is added on the tumor image under conventional illumination, and by it is superimposed as the result is shown on the screen, guidance doctor carries out tumor resection.
Description
Technical field
The present invention relates to technical field of medical image processing, especially a kind of surgery based on indocyanine green fluorescent imaging is swollen
Tumor operation real-time navigation method.
Background technique
China is the first big country of liver cancer, has a large amount of patients to die of liver cancer every year.It is completely that liver cancer is thin by surgical operation
Born of the same parents' excision is treatment liver cancer most efficient method.In art, surgeon generally according to imaging datas such as preoperative CT, MRI,
And examined by vision, palpation and intraoperative ultrasound judge borderline tumor, determine excision extension.Since that there are equipment is huge by CT, MRI
Greatly, ray radiation, check the problems such as time-consuming, imaging evaluation is only limitted to ultrasonic examination in real time in art, but ultrasonic examination
It is higher to the professional skill requirement of image interpretation, it is difficult to universal in surgery of liver.
Indocyanine green (Indocyanine-green, English abbreviation ICG), alias indocyanine green, in conjunction with haemocyanin
Macromolecular compound for entity tumor have high susceptibility, after being absorbed by lesion, according to the optical characteristics of itself combine
Optical molecule image technology, can fluorescing entities tumour.Currently, ICG is widely used in endoscopic surgery, but it is existing in ICG
The lower operation guiding system of guidance, only simple displaying fluorescent image, then doctor rely on the clinical experience of itself to image into
Row is observed and delineates boundary, and the standard of scientific quantification is lacked.And ICG dyeing program process when being one, i.e., with the absorption of cell
And metabolism, fluorescence are one by secretly gradually increasing, the process gradually to decay after peaking, while ICG fluorescent image edge compared with
It is that fuzzy, traditional image processing method can not effectively obtain boundary, and only visually observing by doctor can not obtain most clearly
Clear boundary.
Summary of the invention
In view of the above-mentioned problems, the present invention provides a kind of surgical tumor operation real-time navigation based on indocyanine green fluorescent imaging
Method.
The present invention protects a kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging, including following
Step: 1. obtain each frame image in real-time ICG fluorescence operation video frequency and save as ICG fluorescent image；2. using based on deep
The tumor boundary segmentation model of degree study comes out the tumor boundary segmentation in ICG fluorescent image, and in a manner of colored lines
It is added in former ICG fluorescent image and is shown；3. being commented using edge valuation functions the partitioning boundary of ICG fluorescent image
Estimate, obtains its corresponding sharpness of border coefficient；4. the maximum image of sharpness of border coefficient is found out, then by its corresponding segmentation side
Boundary is added on the tumor image under conventional illumination, and by it is superimposed as the result is shown on the screen, guidance doctor carries out tumour
Resection operation.
Further, the tumor boundary segmentation model based on deep learning is established by following steps: 1. obtain
Hundred endoscopic surgery videos using OpenCV by each frame in image data with the preservation of jpg format, and are labeled, obtain
ICG fluorescent image after mark is divided into training sample and test sample by the original I CG fluoroscopy image sequence after to mark；2.
Training sample is pre-processed；3. the boundary segmentation deep learning network based on Encoder-Decoder model is built, wherein
The Encoder stage separates convolution using the depth with residual error structure and carries out feature extraction, Decoder rank to ICG fluorescent image
Characteristic pattern is restored to original image size by up-sampling gradually and classified to each pixel in image by section；4.
Model training is carried out to pretreated training sample using the learning network for building completion, obtains tumor boundary segmentation model；
5. a pair test sample pre-processes, and is input to tumor boundary segmentation model as input parameter, just segmented image is obtained, so
Connected area disposal$ is carried out to every set image afterwards, the segmentation result after Connected area disposal$ is fed back in original I CG fluoroscopy image sequence.
Further, the Encoder stage first uses step-length to carry out ICG fluorescent image feature for 27 × 7 convolution
Extraction, be followed by BN layers and ReLU layers；Then the depth with residual error structure is recycled to separate convolution module to ICG fluorogram
As carrying out feature extraction；During the Decoder stage up-samples, first bottom-up information is merged with high layer information,
Then it is up-sampled using transposition convolution.
Further, model training uses Adam optimizer, loss function of the cross entropy loss function as network.
Beneficial effects of the present invention: clear by edge by each frame image border in deep learning Real-time segmentation video
Clear degree algorithm carries out quantitative evaluation to marginal definition, selects clearest ICG fluorescent image, demarcates borderline tumor, and by side
Edge is added in the real time video image under natural lighting, and doctor is instructed to carry out tumor resection, provides navigation for operation.
Detailed description of the invention
Fig. 1 is the flow chart of surgical tumor operation real-time navigation method；
Fig. 2 is the boundary segmentation schematic diagram under fluorescence；
Fig. 3 is that the boundary under conventional illumination is superimposed schematic diagram；
Fig. 4 is the tumor boundary segmentation model flow figure established；
Fig. 5 is that depth separates convolution schematic diagram；
Fig. 6 is that the depth with residual error structure separates convolution schematic diagram；
Fig. 7 is liver segmentation deep learning network structure.
Specific embodiment
The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.The embodiment of the present invention is
It is provided for the sake of example and description, and is not exhaustively or to limit the invention to disclosed form.Very much
Modifications and variations are obvious for the ordinary skill in the art.Selection and description embodiment are in order to more preferable
Illustrate the principle of the present invention and practical application, and makes those skilled in the art it will be appreciated that the present invention is suitable to design
In the various embodiments with various modifications of special-purpose.
Embodiment 1
A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging, such as Fig. 1, including following step
Rapid: 1. obtain each frame image in real-time ICG fluorescence operation video frequency and save as ICG fluorescent image；2. using depth is based on
The tumor boundary segmentation model of study comes out the tumor boundary segmentation in ICG fluorescent image, and is added in a manner of colored lines
It is added in former ICG fluorescent image and is shown, as shown in Figure 2；3. using edge valuation functions to the segmentation side of ICG fluorescent image
Boundary is assessed, its corresponding sharpness of border coefficient is obtained；4. finding out the maximum image of sharpness of border coefficient, then corresponded to
Partitioning boundary be added on the tumor image under conventional illumination, and by it is superimposed as the result is shown on the screen, such as Fig. 3 institute
Show, guidance doctor carries out tumor resection.
The foundation of tumor boundary segmentation model based on deep learning, such as Fig. 4, especially by following steps:
1. 100 endoscopic surgery videos are obtained, using OpenCV by each frame in image data with jpg format guarantor
It deposits, and is labeled, the ICG fluorescent image after mark is divided into trained sample by the original I CG fluoroscopy image sequence after being marked
Sheet and test sample, wherein being used for training sample 80, test sample 20 for model testing of model training.
2. pair training sample pre-processes: by the jpg Image Adjusting of preservation to 512*512 size, then using up and down
To treated, training sample carries out amplification processing for the operations such as translation, left and right translation, mirror image, rotation.
3. building the liver segmentation deep learning network based on Encoder-Decoder model, wherein Encoder stage benefit
Convolution being separated with the depth with residual error structure, feature extraction being carried out to ICG fluorescent image, the Decoder stage passes through gradually upper
Characteristic pattern is restored to original image size and classified to each pixel in image by sampling.
4. carrying out model training to pretreated training sample using the learning network for building completion, tumor boundaries are obtained
Parted pattern, concrete configuration is as follows in the present embodiment: 1. hardware configuration, 2 GeForceGTX1080Ti video cards, video memory is in total
22GB, i7-8700CPU, 64GB memory；2. software configuration: 64 versions of operating system Ubuntu16.04 LTS, deep learning library
For Tensorflow1.9.0 version, acceleration tool has used the tall and handsome cuda9.0 and cuDNN7.1.2 reached.In the training of model
Adam optimizer, loss function of the cross entropy loss function as network are used in the process.Initial learning rate is set as 1 ×
10-4, it is gradually reduced according to the increase of the number of iterations, the number of iterations of tuning is set as 100000 times, when the number of iterations reaches
When, Network termination training obtains tumor boundary segmentation model.
5. a pair test sample pre-processes, and is input to tumor boundary segmentation model as input parameter, obtain just dividing
Image is cut, Connected area disposal$ then is carried out to remove noise that may be present in segmentation result to every group of image, improves segmentation
Accuracy, the segmentation result after Connected area disposal$ are fed back in original sequence.
Encoder-Decoder (coding-decoding) is one of model framework common in deep learning, it is not one
Specific model, but a class framework.List entries is exactly converted to the vector of a regular length by so-called coding；Decoding,
The fixed vector generated before is exactly then converted into output sequence.Based on Encoder-Decoder, we be can be designed that respectively
Kind various kinds applies algorithm.
In the present embodiment, the Encoder stage first uses step-length to carry out ICG fluorescent image feature for 27 × 7 convolution
Extraction, be followed by BN layers and ReLU layers；Then the depth with residual error structure is recycled to separate convolution module to ICG fluorogram
As carrying out feature extraction；During the Decoder stage up-samples, first bottom-up information is merged with high layer information,
Then it is up-sampled using transposition convolution, characteristic pattern is gradually restored to by original image size by up-sampling gradually, and
Classify to each of these pixel；The last layer of network is softmax layers.
Depth separates convolution and refers to one 1 × 1 Standard convolution of convolution sum that Standard convolution is decomposed into one by depth
(point-by-point convolution), as shown in Figure 5.The channel of each input feature vector figure is corresponded to by depth convolution, 1 × 1 point-by-point convolution is responsible for
The feature extracted by depth convolution is merged.Depth separates convolution can reduce model in the case where not influencing result
Parameter, be conducive to carry out deeper model in the limited situation of GPU resource to build；The application of residual error structure can be effective
Solve the gradient dispersion in deep layer network.
Depth, which separates convolution, reduces required parameter than common convolution, is exemplified below.Such as a certain layer is defeated
Entering channel is 64, and output channel is 128, and the size of convolution kernel is 3 × 3, and using normal convolution, the parameter of this layer is calculated as (64
× 3 × 3) × 128=73728, and separating in convolution in depth is 64 × 3 × 3+128 × (64 × 1 × 1)=8768, i.e., first
Data with the convolution sum of 64 3 × 3 sizes respectively at 64 channels of input carry out convolution operation, obtain 64 characteristic spectrums,
Then convolution operation is carried out on 64 characteristic spectrums using 128 1 × 1 convolution kernels, the information in 64 channels is melted
It closes.Convolution operation parameter is separated by depth and is reduced to 8768 from 73728, and model parameter greatly reduces, it can be one
Determine the over-fitting that model is prevented in degree.
The core concept of residual error is to bypass parameter layer using identical mapping, and input client information is directly passed through simple phase
It attaches and is added to output end, as shown in Figure 6.Experiment shows that the use of residual error can effectively solve increase depth bring degeneration and ask
Topic makes it possible to improve accuracy rate by increasing the depth of network, while there is the network of residual error structure to be easier to optimize.
Residual error connects expression formula: Xt+1=Xt+F(Xt,Wt), wherein XtIndicate that the feature of t layers of residual block inputs, and Xt+1Indicate feature
Output, F (Xt,Wt) indicate the non-linear unit that residual error feature learning is carried out in residual block, it include batch normalization layer
(BatchNormalization, BN), amendment linear unit (Rectifiedlinearunit, ReLU).
The liver segmentation deep learning network designed in the present embodiment shares 81 layers, and first 76 layers for carrying out ICG fluorogram
As the extraction of characteristic information, concrete structure diagram is as shown in Figure 7.
Obviously, described embodiment is only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, this field and those of ordinary skill in the related art institute without creative labor
The every other embodiment obtained, all should belong to the scope of protection of the invention.
Claims (5)
- The real-time navigation method 1. a kind of surgical tumor based on indocyanine green fluorescent imaging is performed the operation, which is characterized in that including following Step,Step 1: obtaining each frame image in real-time ICG fluorescence operation video frequency and save as ICG fluorescent image；Step 2: being gone out the tumor boundary segmentation in ICG fluorescent image using the tumor boundary segmentation model based on deep learning Come, and is added in former ICG fluorescent image in a manner of colored lines and is shown；Step 3: being assessed using partitioning boundary of the edge valuation functions to ICG fluorescent image, it is clear to obtain its corresponding boundary Clear coefficient；Step 4: finding out the maximum image of sharpness of border coefficient, then its corresponding partitioning boundary is added under conventional illumination On tumor image, and by it is superimposed as the result is shown on the screen, guidance doctor carries out tumor resection.
- The real-time navigation method 2. surgical tumor according to claim 1 is performed the operation, which is characterized in that described to be based on deep learning Tumor boundary segmentation model by following steps establish,Step a: obtaining up to a hundred endoscopic surgery videos, using OpenCV by each frame in image data with jpg format guarantor It deposits, and is labeled, the ICG fluorescent image after mark is divided into trained sample by the original I CG fluoroscopy image sequence after being marked Sheet and test sample；Step b: training sample is pre-processed；Step c: building the boundary segmentation deep learning network based on Encoder-Decoder model, wherein Encoder stage benefit Convolution being separated with the depth with residual error structure, feature extraction being carried out to ICG fluorescent image, the Decoder stage passes through gradually upper Characteristic pattern is restored to original image size and classified to each pixel in image by sampling；Step d: model training is carried out to pretreated training sample using the learning network for building completion, obtains tumor boundaries Parted pattern；Step e: pre-processing test sample, and is input to tumor boundary segmentation model as input parameter, obtains just dividing Image is cut, Connected area disposal$ then is carried out to every set image, the segmentation result after Connected area disposal$ feeds back to original I CG fluorogram As in sequence.
- The real-time navigation method 3. surgical tumor according to claim 2 is performed the operation, which is characterized in that the Encoder stage It first uses step-length to carry out the extraction of ICG fluorescent image feature for 27 × 7 convolution, is followed by BN layers and ReLU layers；Then sharp again Convolution module is separated with the depth with residual error structure, and feature extraction is carried out to ICG fluorescent image.
- The real-time navigation method 4. surgical tumor according to claim 2 is performed the operation, which is characterized in that the Decoder stage During up-sampling, first bottom-up information is merged with high layer information, is then up-sampled using transposition convolution.
- The real-time navigation method 5. surgical tumor according to claim 2 is performed the operation, which is characterized in that the mould in the step d Type training uses Adam optimizer, loss function of the cross entropy loss function as network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910203727.2A CN109934831A (en) | 2019-03-18 | 2019-03-18 | A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910203727.2A CN109934831A (en) | 2019-03-18 | 2019-03-18 | A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109934831A true CN109934831A (en) | 2019-06-25 |
Family
ID=66987331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910203727.2A Pending CN109934831A (en) | 2019-03-18 | 2019-03-18 | A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109934831A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112396601A (en) * | 2020-12-07 | 2021-02-23 | 中山大学 | Real-time neurosurgical instrument segmentation method and device based on endoscope image and storage medium |
| WO2021044590A1 (en) * | 2019-09-05 | 2021-03-11 | オリンパス株式会社 | Endoscope system, treatment system, endoscope system operation method and image processing program |
| CN112741596A (en) * | 2019-10-31 | 2021-05-04 | 南京大学 | Mobile terminal heart sound classification method based on deep separable convolutional neural network |
| CN113693739A (en) * | 2021-08-27 | 2021-11-26 | 南京诺源医疗器械有限公司 | Tumor navigation correction method and device and portable fluorescent image navigation equipment |
| CN116563284A (en) * | 2023-07-10 | 2023-08-08 | 南京信息工程大学 | System for quantitatively describing brain glioma characteristic boundary change evaluation index in MRI |
| CN117045178A (en) * | 2023-08-28 | 2023-11-14 | 南京诺源医疗器械有限公司 | Image quality intelligent monitoring method and system for fluorescent endoscope imaging |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110280810A1 (en) * | 2010-03-12 | 2011-11-17 | Carl Zeiss Meditec, Inc. | Surgical optical systems for detecting brain tumors |
| CN106683064A (en) * | 2016-12-13 | 2017-05-17 | 西北工业大学 | Multi-focusing image fusion method based on two-dimensional coupling convolution |
| CN107610109A (en) * | 2017-09-06 | 2018-01-19 | 艾瑞迈迪医疗科技（北京）有限公司 | Method for displaying image, the apparatus and system of endoscope micro-wound navigation |
| CN107977969A (en) * | 2017-12-11 | 2018-05-01 | 北京数字精准医疗科技有限公司 | A kind of dividing method, device and the storage medium of endoscope fluorescence image |
| CN108230335A (en) * | 2017-12-29 | 2018-06-29 | 北京数字精准医疗科技有限公司 | Multispectral fluorescence dividing method and device |
| CN108510446A (en) * | 2018-04-10 | 2018-09-07 | 四川和生视界医药技术开发有限公司 | The stacking method of retinal images and the stacking apparatus of retinal images |
| CN109242864A (en) * | 2018-09-18 | 2019-01-18 | 电子科技大学 | Image segmentation result quality evaluating method based on multiple-limb network |
| CN109345538A (en) * | 2018-08-30 | 2019-02-15 | 华南理工大学 | A kind of Segmentation Method of Retinal Blood Vessels based on convolutional neural networks |
-
2019
- 2019-03-18 CN CN201910203727.2A patent/CN109934831A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110280810A1 (en) * | 2010-03-12 | 2011-11-17 | Carl Zeiss Meditec, Inc. | Surgical optical systems for detecting brain tumors |
| CN106683064A (en) * | 2016-12-13 | 2017-05-17 | 西北工业大学 | Multi-focusing image fusion method based on two-dimensional coupling convolution |
| CN107610109A (en) * | 2017-09-06 | 2018-01-19 | 艾瑞迈迪医疗科技（北京）有限公司 | Method for displaying image, the apparatus and system of endoscope micro-wound navigation |
| CN107977969A (en) * | 2017-12-11 | 2018-05-01 | 北京数字精准医疗科技有限公司 | A kind of dividing method, device and the storage medium of endoscope fluorescence image |
| CN108230335A (en) * | 2017-12-29 | 2018-06-29 | 北京数字精准医疗科技有限公司 | Multispectral fluorescence dividing method and device |
| CN108510446A (en) * | 2018-04-10 | 2018-09-07 | 四川和生视界医药技术开发有限公司 | The stacking method of retinal images and the stacking apparatus of retinal images |
| CN109345538A (en) * | 2018-08-30 | 2019-02-15 | 华南理工大学 | A kind of Segmentation Method of Retinal Blood Vessels based on convolutional neural networks |
| CN109242864A (en) * | 2018-09-18 | 2019-01-18 | 电子科技大学 | Image segmentation result quality evaluating method based on multiple-limb network |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021044590A1 (en) * | 2019-09-05 | 2021-03-11 | オリンパス株式会社 | Endoscope system, treatment system, endoscope system operation method and image processing program |
| CN112741596A (en) * | 2019-10-31 | 2021-05-04 | 南京大学 | Mobile terminal heart sound classification method based on deep separable convolutional neural network |
| CN112396601A (en) * | 2020-12-07 | 2021-02-23 | 中山大学 | Real-time neurosurgical instrument segmentation method and device based on endoscope image and storage medium |
| CN113693739A (en) * | 2021-08-27 | 2021-11-26 | 南京诺源医疗器械有限公司 | Tumor navigation correction method and device and portable fluorescent image navigation equipment |
| CN116563284A (en) * | 2023-07-10 | 2023-08-08 | 南京信息工程大学 | System for quantitatively describing brain glioma characteristic boundary change evaluation index in MRI |
| CN116563284B (en) * | 2023-07-10 | 2023-11-07 | 南京信息工程大学 | System for quantitatively describing brain glioma characteristic boundary change evaluation index in MRI |
| CN117045178A (en) * | 2023-08-28 | 2023-11-14 | 南京诺源医疗器械有限公司 | Image quality intelligent monitoring method and system for fluorescent endoscope imaging |
| CN117045178B (en) * | 2023-08-28 | 2024-02-02 | 南京诺源医疗器械有限公司 | Image quality intelligent monitoring method and system for fluorescent endoscope imaging |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109934831A (en) | A kind of surgical tumor operation real-time navigation method based on indocyanine green fluorescent imaging | |
| CN109949309A (en) | A kind of CT image for liver dividing method based on deep learning | |
| CN112086197B (en) | Breast nodule detection method and system based on ultrasonic medicine | |
| Bai et al. | Automatic segmentation of cervical region in colposcopic images using K-means | |
| CN111179237A (en) | Image segmentation method and device for liver and liver tumor | |
| CN112200763A (en) | Liver function grading method based on liver CT image | |
| CN110189293A (en) | Cell image processing method, device, storage medium and computer equipment | |
| CN112071418B (en) | Gastric cancer peritoneal metastasis prediction system and method based on enhanced CT image histology | |
| CN113744271A (en) | Neural network-based automatic optic nerve segmentation and compression degree measurement and calculation method | |
| Liu et al. | Cascaded atrous dual attention U-Net for tumor segmentation | |
| Jha et al. | Instance segmentation for whole slide imaging: end-to-end or detect-then-segment | |
| Wu et al. | Fatty liver diagnosis using deep learning in ultrasound image | |
| CN113539476A (en) | Stomach endoscopic biopsy Raman image auxiliary diagnosis method and system based on artificial intelligence | |
| CN111402207B (en) | Ultrasonic contrast video data analysis method based on composite neural network | |
| WO2021072873A1 (en) | Grayscale image processing method, system, device, and computer-readable storage medium | |
| Xue et al. | A unified set of analysis tools for uterine cervix image segmentation | |
| CN109509204A (en) | The extraction method of CT image silicotic nodule | |
| US20070286525A1 (en) | Generation of imaging filters based on image analysis | |
| CN114898862A (en) | Cervical cancer computer-aided diagnosis method based on convolutional neural network and pathological section image | |
| Vijayalakshmi et al. | Liver tumor detection using CNN | |
| Plancoulaine et al. | Building of a composite virtual slide from contiguous tissue samples | |
| CN117153343B (en) | Placenta multiscale analysis system | |
| CN114360718B (en) | Feature fitting-based PET/CT automatic lung cancer diagnosis and classification system and construction method | |
| Brehler et al. | An automated approach for annotation Gleason patterns in whole-mount prostate cancer histology using deep learning | |
| Jeronimo et al. | A tool for collection of region based data from uterine cervix images for correlation of visual and clinical variables related to cervical neoplasia |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190625 |
|
| RJ01 | Rejection of invention patent application after publication |