CN105427325A - Automatic lung tumour segmentation method based on random forest and monotonically decreasing function - Google Patents
Automatic lung tumour segmentation method based on random forest and monotonically decreasing function Download PDFInfo
- Publication number
- CN105427325A CN105427325A CN201510890132.0A CN201510890132A CN105427325A CN 105427325 A CN105427325 A CN 105427325A CN 201510890132 A CN201510890132 A CN 201510890132A CN 105427325 A CN105427325 A CN 105427325A
- Authority
- CN
- China
- Prior art keywords
- tumour
- image
- decreasing function
- monotonically decreasing
- random forest
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10104—Positron emission tomography [PET]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30061—Lung
Abstract
The invention discloses an automatic lung tumour segmentation method based on a random forest and a monotonically decreasing function. The automatic lung tumour segmentation method comprises the following steps: obtaining the initial position of a lung tumour by utilizing the brightness variation characteristic of the tumour on a PET image, then, sufficiently utilizing metabolic information of the tumour on the PET image and anatomic information of the tumour on a CT image through characteristic extraction, and finally, realizing precision segmentation of the tumour through a random forests algorithm. The automatic lung tumour segmentation method based on the random forest and the monotonically decreasing function provided by the invention is capable of determining the position and the size of the lung tumour automatically and precisely; therefore, a doctor can be clinically assisted to treat the lung tumour; and more rapid and more precise full-automatic lung tumour segmentation can be realized without artificial intervention.
Description
Technical field
The present invention relates to a kind of automatic division method of the lung neoplasm based on random forest and monotonically decreasing function, particularly relate to a kind of random forest and monotonically decreasing function method of utilizing and the method for full-automatic dividing is carried out to lung tumors, belong to Biologic Medical Image processing technology field.
Background technology
Tumour refers to that body is under the effect of the various tumorigenesis factor, the neoformation that local organization hyperplasia is formed.According to neoplastic cell characteristics and the harmfulness degree to body, again tumour is divided into benign tumour and the large class of malignant tumour two, and cancer is the general name of malignant tumour.
Lung neoplasm is one of common malignant tumour, and nearly decades, the M & M of lung cancer had the trend obviously increased.The early diagnosis of lung cancer is the effective way improving result for the treatment of, and in the middle of Medical Imaging, the development that image is split accurately and located, for the early diagnosis of lung cancer provides advantage with accurately treating.
PET and CT, as quantitative molecule and structure imaging technology, has been widely used in the therapeutic scheme of lung neoplasm.At present, many experts both domestic and external and scholar have proposed and have employed a variety of method to split lung neoplasm, such as Threshold segmentation, region growing and the algorithm cut based on figure.But these algorithms are not have employed single mode, cannot provide with accurate segmentation result, need man-machine interaction exactly, cannot realize full automatic dividing method, the Seed Points that the algorithm such as cut based on figure needs artificial calibration maps to cut.
The present invention uses monotonically decreasing function (Downhill) to determine the rough location of lung neoplasm, then uses random forest (RandomForest) algorithm to complete accurate segmentation.By extracting different eigenwerts, taking full advantage of the metabolic information of lung neoplasm in PET image and the texture information on CT image, can reach more fast and more accurate full-automatic dividing without the need to artificially getting involved.
Summary of the invention
Technical matters to be solved by this invention is, provides a kind of position and the size of automatically accurately determining lung neoplasm, to reach the automatic division method assisted a physician clinically to the lung neoplasm based on random forest and monotonically decreasing function of the treatment of lung neoplasm.
For solving the problems of the technologies described above, the technical solution used in the present invention is:
Based on the automatic division method of the lung neoplasm of random forest and monotonically decreasing function, it is characterized in that, comprise the following steps:
S01, to lung neoplasm, patient is scanned, and obtains PET image and CT image;
S02, carries out up-sampling to PET image, and carries out affine registration to PET image and CT image, makes the position one_to_one corresponding of the pixel on PET image and CT image;
S03, to PET image and CT image filter filtering, smoothed image;
S04, splits by the method for Threshold segmentation, tentatively obtains connected region in PET image; With the opening operation of mathematical mor-phology, the assorted point of described connected region periphery is removed;
S05, according to the characteristic of monotonically decreasing function, gets rid of the interference of other organs, tentatively obtains the position of lung neoplasm in described connected region;
S06, PET image and CT image carry out feature extraction;
S07, splits lung neoplasm with the algorithm of random forest.
The standard of smoothed image described in S03 is the information keeping tumor focus border in PET image and CT image.
Feature extraction described in S06 comprises the metabolic information of tumour in PET image and the anatomic information of CT image.
Described PET image comprises the monotonic decreasing feature of tumour brightness, knub position information, the gradient information of tumor boundaries and the distributed intelligence of standardized uptake value.
Described CT image comprises gradient information and the organization of human body information of tumor boundaries.
Up-sampling described in S02 is linear up-sampling.
In PET image, first obtain maximum described standardized uptake value, then with 45% of maximum described standardized uptake value as threshold value, obtain a lot of described connected region and some highlighted assorted points, described connected region comprises tumour highlight regions, tumour adjacent organs highlight regions and vertebra highlight regions.
The formula of described monotonically decreasing function is:
In formula (1), v
iand v
jrepresent the voxel that in connected region two are different; v
maxrepresent the gray-scale value that in this connected region, voxel is maximum; X
vrepresent the position of voxel v; || || represent Euclidean distance; SUV represents standardized uptake value; SUV (v
i) represent v
ithe standardized uptake value of this voxel; SUV (v
j) represent v
jthe standardized uptake value of this voxel; R
drepresentative meets the connected region that monotonically decreasing function obtains;
By the v in connected region
maxas the Seed Points of region growing, the region meeting monotonically decreasing function is carried out the region growing of 26 neighborhoods; After growth, if v
iand v
maxbetween Euclidean distance>=v
jand v
maxbetween Euclidean distance, simultaneously v
istandardized uptake value≤v
jstandardized uptake value, so this connected region is tumour highlight regions, i.e. tumor area; If do not meet above-mentioned condition, then this connected region is tumour adjacent organs highlight regions and/or vertebra highlight regions, i.e. non-tumor area.
If described tumor area comprises some Field necrosis tumours or Cystic changes tumour, the tumour that then there will be high gray-scale value surrounds the low region of pixel value, therefore, after region growing, the tumor growth areas of described Field necrosis tumour or Cystic changes tumour is uneven state, and middle part, tumor area there will be some hollow bore, now, then use the closed operation in mathematical mor-phology, go to fill up these hollow bore.
Technical scheme of the present invention is to provide a kind of method for the automatic detection of lung neoplasm in human body PET, CT image, the method utilizes monotonically decreasing function, first determining the rough position of tumour, then by extracting the feature on PET and CT, splitting accurately in random forest sorter.The method is that the segmentation of lung neoplasm provides a kind of speed sooner, the more method of robust.The method mainly comprises the following steps: 1, carry out up-sampling to PET image, and carries out affine registration (AffineRegistration) to PET and CT image, makes the position one_to_one corresponding of the point on PET and CT image; 2, to the medical image extracted, goldstandard is demarcated: under the supervising and guiding of veteran oncologist, utilize ITK-SNAP software, tumor region is demarcated; 3, to PET and CT image filter filtering, smoothed image, but the information keeping border; 4, split in PET image by the method for Threshold segmentation, obtain initial connected region; With the opening operation of mathematical mor-phology, the assorted point of the connected region periphery obtained is removed; 5, according to the characteristic of monotonically decreasing function, get rid of the interference of the organ such as heart and liver, obtain the position of initial tumour; 6, on PET and CT image, feature extraction is carried out; 7, the algorithm being used in random forest is split tumour, the result of segmentation and goldstandard is compared, and uses weighing criteria to quantize testing result.
The advantage of this method is in conjunction with the metabolic information of lung neoplasm in PET image and the anatomic information on CT image, utilize the monotonic decreasing feature of tumour brightness in PET image, positional information, the gradient information on border and the standardized uptake value (SUVs of PET image, Standarduptakevalues) the organization of human body information that distributed intelligence and CT provide, by random forest (RandomForest) algorithm, auto Segmentation tumour, make the tumour split more accurate, for the treatment of tumour from now on lays the first stone.
The automatic division method of a kind of lung neoplasm based on random forest and monotonically decreasing function provided by the invention, the brightness variation characteristic (monotonic decreasing) of lung neoplasm in PET image is first utilized to obtain the initial position of tumour, then make full use of the metabolic information of tumour in PET image and the anatomic information of CT image by feature extraction, realize lung neoplasm finally by random forests algorithm and accurately split.
Figure of description
Fig. 1 is PET image of the present invention and CT image; A is PET image, and b is the CT image corresponding with a;
Fig. 2 is process flow diagram of the present invention;
Fig. 3 is the DSC index contrast figure of the present invention and existing IGC, RG40 and RG50 partitioning algorithm.
Embodiment
Below in conjunction with accompanying drawing, the present invention is further described.
As shown in Figure 1, first lung neoplasm dividing method of the present invention carries out the collection of PET, CT view data, by carrying out up-sampling to PET image, and carries out affine registration to PET and CT image, makes the pixel one_to_one corresponding on PET and CT image.The goldstandard of tumour is obtained under the help and supervision of clinical tumor scholar.The position of tumour is first tentatively determined by the method for Threshold segmentation and monotonic decreasing (Downhill) function, then on PET and CT image, feature extraction is carried out, information on PET and the CT image finally utilizing the Algorithms Integration analysis of random forest (Randomforest) to extract, lung neoplasm region tested and splits, drawing last testing result.
The present invention is under the patronage of First Affiliated Hospital of Soochow University,Suzhou, obtain the patient data suffering from non-small cell lung cancer, the segmentation principle of lung neoplasm mainly make use of extraction and the fusion of information on PET, CT image, utilizes random forest (Randomforest) this algorithm to carry out calculating, analyzing.
Below in conjunction with the algorithm principle figure shown in accompanying drawing 2, the embodiment of process in detail.
1, this method carries out linear up-sampling to the PET image got, and carries out affine registration to PET, CT image; Under the guidance of clinical tumor expert, determine the goldstandard of tumour.
2, the acquisition of connected region: in PET image, first obtains maximum SUV value, then with 45% of maximum SUV value as threshold value, obtain a lot of connected region, comprise tumour, heart, the highlight regions such as liver and vertebra.Owing to also there is some noises or small region on original image, so the result of Threshold segmentation has some assorted points, these assorted points are removed with the opening operation (Openingoperation) inside mathematical mor-phology by we.
3, region growing, the determination of knub position: uniform tumour meets the character of monotonically decreasing function.And the histoorgan such as heart and liver does not meet the character of monotonically decreasing function.
V in formula
itwo that represent in connected region with vj different voxels, v
maxrepresent the gray-scale value that in this connected region, voxel is maximum, X
vrepresent the position of voxel v.|| || be Euclidean distance.
The region growing of 26 neighborhoods, using the Seed Points of the pixel value of the maximum gray scale in connected region as region growing, is carried out in the region meeting monotonically decreasing function by us.The region not meeting monotonically decreasing function character as heart and liver and vertebra can not grow out, or the region grown out only has sub-fraction.The general location of last tumour will be determined.
4, in some cases, tumour is not the strict character meeting monotonically decreasing function.Because wherein some Field necrosis or Cystic changes may be comprised.Like this surround the low region of pixel value with regard to the tumour that there will be high gray-scale value.Therefore, after region growing, in the tumor growth areas that those are uneven, middle part there will be some hollow bore, and now, we use the closed operation in mathematical mor-phology, goes to fill up these holes occurred.
5, the Accurate Segmentation of tumour: after the position tentatively determining lung neoplasm, the result of the boundary segmentation of tumour is still accurate not.At this moment we use random forest this fast again accurately algorithm to carry out last segmentation.Start, 19 features are extracted from PET and CT image respectively, so not only make use of the information of high-contrast in PET image but also make use of the anatomic information of tumour on CT image.Finally use random forests algorithm to split, obtain end product and carry out data analysis.
6, this experiment adopts the tumour data that First Affiliated Hospital of Soochow University,Suzhou provides, and we have chosen 24 available patient datas, and adopt DSC coefficient to weigh the result of lesion segmentation.DSC coefficient is used for reflecting the tumour result of segmentation and the registration in goldstandard region.The quantitative test of these 24 data and other dividing method, the figure such as improved cuts (IGC, ImproveGraphcut), the region growing (RG40) of threshold value of 40% and region growing (RG50) method of the threshold value of 50% compare, and result as shown in Figure 3.
DSC coefficient (DiceSimilarityCoefficient):
U1, U2 are respectively result and the goldstandard of segmentation.
The above is only the preferred embodiment of the present invention; be noted that for those skilled in the art; under the premise without departing from the principles of the invention, can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (9)
1., based on the automatic division method of the lung neoplasm of random forest and monotonically decreasing function, it is characterized in that, comprise the following steps:
S01, to lung neoplasm, patient is scanned, and obtains PET image and CT image;
S02, carries out up-sampling to PET image, and carries out affine registration to PET image and CT image, makes the position one_to_one corresponding of the pixel on PET image and CT image;
S03, to PET image and CT image filter filtering, smoothed image;
S04, splits by the method for Threshold segmentation, tentatively obtains connected region in PET image; With the opening operation of mathematical mor-phology, the assorted point of described connected region periphery is removed;
S05, according to the characteristic of monotonically decreasing function, gets rid of the interference of other organs, tentatively obtains the position of lung neoplasm in described connected region;
S06, PET image and CT image carry out feature extraction;
S07, splits lung neoplasm with the algorithm of random forest.
2. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 1, is characterized in that: the standard of smoothed image described in S03 is the information keeping tumor focus border in PET image and CT image.
3. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 1, is characterized in that: feature extraction described in S06 comprises the metabolic information of tumour in PET image and the anatomic information of CT image.
4. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 1, is characterized in that: described PET image comprises the distributed intelligence of the monotonic decreasing feature of tumour brightness, knub position information and standardized uptake value.
5. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 1, is characterized in that: described CT image comprises gradient information and the organization of human body information of tumor boundaries.
6. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 1, is characterized in that: up-sampling described in S02 is linear up-sampling.
7. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 4, it is characterized in that: in PET image, first obtain maximum described standardized uptake value, then with 45% of maximum described standardized uptake value as threshold value, obtain a lot of described connected region and some highlighted assorted points, described connected region comprises tumour highlight regions, tumour adjacent organs highlight regions and vertebra highlight regions.
8. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 7, is characterized in that: the formula of described monotonically decreasing function is:
In formula (1), v
iand v
jrepresent the voxel that in connected region two are different; v
maxrepresent the gray-scale value that in this connected region, voxel is maximum; X
vrepresent the position of voxel v; || || represent Euclidean distance; SUV represents standardized uptake value; SUV (v
i) represent v
ithe standardized uptake value of this voxel; SUV (v
j) represent v
jthe standardized uptake value of this voxel; RD representative meets the connected region that monotonically decreasing function obtains;
By the v in connected region
maxas the Seed Points of region growing, the region meeting monotonically decreasing function is carried out the region growing of 26 neighborhoods; After growth, if v
iand v
maxbetween Euclidean distance>=v
jand v
maxbetween Euclidean distance, simultaneously v
istandardized uptake value≤v
jstandardized uptake value, so this connected region is tumour highlight regions, i.e. tumor area; If do not meet above-mentioned condition, then this connected region is tumour adjacent organs highlight regions and/or vertebra highlight regions, i.e. non-tumor area.
9. the automatic division method of the lung neoplasm based on random forest and monotonically decreasing function according to claim 8, it is characterized in that: if described tumor area comprises some Field necrosis tumours or Cystic changes tumour, the tumour that then there will be high gray-scale value surrounds the low region of pixel value, therefore, after region growing, the tumor growth areas of described Field necrosis tumour or Cystic changes tumour is uneven state, middle part, tumor area there will be some hollow bore, now, then use the closed operation in mathematical mor-phology, go to fill up these hollow bore.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510890132.0A CN105427325A (en) | 2015-12-07 | 2015-12-07 | Automatic lung tumour segmentation method based on random forest and monotonically decreasing function |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510890132.0A CN105427325A (en) | 2015-12-07 | 2015-12-07 | Automatic lung tumour segmentation method based on random forest and monotonically decreasing function |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105427325A true CN105427325A (en) | 2016-03-23 |
Family
ID=55505505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510890132.0A Pending CN105427325A (en) | 2015-12-07 | 2015-12-07 | Automatic lung tumour segmentation method based on random forest and monotonically decreasing function |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105427325A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105931224A (en) * | 2016-04-14 | 2016-09-07 | 浙江大学 | Pathology identification method for routine scan CT image of liver based on random forests |
| CN105957094A (en) * | 2016-06-15 | 2016-09-21 | 深圳维影医疗科技有限公司 | contour detection based rapid automatic segmentation method and system for lung field of chest radiograph |
| CN107833229A (en) * | 2017-11-02 | 2018-03-23 | 上海联影医疗科技有限公司 | Information processing method, apparatus and system |
| CN107845098A (en) * | 2017-11-14 | 2018-03-27 | 南京理工大学 | Liver cancer image full-automatic partition method based on random forest and fuzzy clustering |
| CN108205805A (en) * | 2016-12-20 | 2018-06-26 | 北京大学 | The dense corresponding auto-creating method of voxel between pyramidal CT image |
| CN109478321A (en) * | 2016-07-25 | 2019-03-15 | 索尼公司 | Automatic 3D brain tumor segmentation and classification |
| CN110084800A (en) * | 2019-04-28 | 2019-08-02 | 上海海事大学 | A kind of Lung metastases prediction technique for four limbs soft tissue sarcoma patient |
| CN110349662A (en) * | 2019-05-23 | 2019-10-18 | 复旦大学 | The outliers across image collection that result is accidentally surveyed for filtering pulmonary masses find method and system |
| CN112734790A (en) * | 2020-12-30 | 2021-04-30 | 武汉联影生命科学仪器有限公司 | Tumor region labeling method, system, device and readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5857030A (en) * | 1995-08-18 | 1999-01-05 | Eastman Kodak Company | Automated method and system for digital image processing of radiologic images utilizing artificial neural networks |
| CN103942785A (en) * | 2014-04-09 | 2014-07-23 | 苏州大学 | PET and CT image lung tumor segmenting method based on graph cut |
| CN104414636A (en) * | 2013-08-23 | 2015-03-18 | 北京大学 | Magnetic resonance image based cerebral micro-bleeding computer auxiliary detection system |
-
2015
- 2015-12-07 CN CN201510890132.0A patent/CN105427325A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5857030A (en) * | 1995-08-18 | 1999-01-05 | Eastman Kodak Company | Automated method and system for digital image processing of radiologic images utilizing artificial neural networks |
| CN104414636A (en) * | 2013-08-23 | 2015-03-18 | 北京大学 | Magnetic resonance image based cerebral micro-bleeding computer auxiliary detection system |
| CN103942785A (en) * | 2014-04-09 | 2014-07-23 | 苏州大学 | PET and CT image lung tumor segmenting method based on graph cut |
Non-Patent Citations (4)
| Title |
|---|
| CHERRY BALLANGAN等: "Automated lung tumor segmentation for whole body PET volume based on novel downhill region growing", 《SPIE PROCEEDING, MEDICAL IMAGING 2010》 * |
| CHERRY BALLANGAN等: "Lung tumor segmentation in PET images using graph cuts", 《COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE》 * |
| 赵涓涓: "《基于PET-CT的肺肿瘤早期计算机辅助诊断技术》", 31 May 2015, 北京：国防工业出版社 * |
| 韩玉等: "随机森林方法在医学中的应用", 《中国预防医学杂志》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105931224A (en) * | 2016-04-14 | 2016-09-07 | 浙江大学 | Pathology identification method for routine scan CT image of liver based on random forests |
| CN105957094B (en) * | 2016-06-15 | 2018-09-25 | 深圳维影医疗科技有限公司 | A kind of fast automatic dividing method of rabat lung field and its system based on contour detecting |
| CN105957094A (en) * | 2016-06-15 | 2016-09-21 | 深圳维影医疗科技有限公司 | contour detection based rapid automatic segmentation method and system for lung field of chest radiograph |
| CN109478321B (en) * | 2016-07-25 | 2022-04-26 | 索尼公司 | Automated 3D brain tumor segmentation and classification |
| CN109478321A (en) * | 2016-07-25 | 2019-03-15 | 索尼公司 | Automatic 3D brain tumor segmentation and classification |
| CN108205805B (en) * | 2016-12-20 | 2020-06-02 | 北京大学 | Automatic establishment method for dense correspondence of cone beam CT image intermediate pixels |
| CN108205805A (en) * | 2016-12-20 | 2018-06-26 | 北京大学 | The dense corresponding auto-creating method of voxel between pyramidal CT image |
| CN107833229A (en) * | 2017-11-02 | 2018-03-23 | 上海联影医疗科技有限公司 | Information processing method, apparatus and system |
| CN107845098A (en) * | 2017-11-14 | 2018-03-27 | 南京理工大学 | Liver cancer image full-automatic partition method based on random forest and fuzzy clustering |
| CN110084800A (en) * | 2019-04-28 | 2019-08-02 | 上海海事大学 | A kind of Lung metastases prediction technique for four limbs soft tissue sarcoma patient |
| CN110349662A (en) * | 2019-05-23 | 2019-10-18 | 复旦大学 | The outliers across image collection that result is accidentally surveyed for filtering pulmonary masses find method and system |
| CN110349662B (en) * | 2019-05-23 | 2023-01-13 | 复旦大学 | Cross-image set outlier sample discovery method and system for filtering lung mass misdetection results |
| CN112734790A (en) * | 2020-12-30 | 2021-04-30 | 武汉联影生命科学仪器有限公司 | Tumor region labeling method, system, device and readable storage medium |
| CN112734790B (en) * | 2020-12-30 | 2023-07-11 | 武汉联影生命科学仪器有限公司 | Tumor region labeling method, system, device and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105427325A (en) | Automatic lung tumour segmentation method based on random forest and monotonically decreasing function | |
| EP3432784B1 (en) | Deep-learning-based cancer classification using a hierarchical classification framework | |
| CN103714536B (en) | The dividing method and device of the multi-modal MRI based on rarefaction representation | |
| CN109493325B (en) | Tumor heterogeneity analysis system based on CT images | |
| CN108492300B (en) | Lung blood vessel tree segmentation method combining tubular structure enhancement and energy function | |
| CN105719278A (en) | Organ auxiliary positioning segmentation method based on statistical deformation model | |
| CN107292312A (en) | Tumour recognition methods | |
| Jony et al. | Detection of lung cancer from CT scan images using GLCM and SVM | |
| CN111179237A (en) | Image segmentation method and device for liver and liver tumor | |
| CN103942785A (en) | PET and CT image lung tumor segmenting method based on graph cut | |
| US11568538B2 (en) | Medical image-based tumor detection and diagnostic device | |
| CN105654490A (en) | Lesion region extraction method and device based on ultrasonic elastic image | |
| CN105139377A (en) | Rapid robustness auto-partitioning method for abdomen computed tomography (CT) sequence image of liver | |
| CN106780497A (en) | A kind of organ vascular tree extraction method based on statistical information | |
| CN113139981A (en) | DCE-MRI (direct current imaging-magnetic resonance imaging) breast tumor image segmentation method based on deep neural network | |
| CN108986114A (en) | A kind of abdominal CT sequence image liver automatic division method based on level set and shape description symbols | |
| CN106570868A (en) | Three-dimensional liver tumor semi-automatic segmentation method based on level set | |
| CN104915989B (en) | Blood vessel three-dimensional dividing method based on CT images | |
| CN116309551B (en) | Method, device, electronic equipment and readable medium for determining focus sampling area | |
| Bnouni et al. | Boosting CNN learning by ensemble image preprocessing methods for cervical cancer segmentation | |
| CN110049715A (en) | For showing the equipment and associated display methods of the internal organ of patient | |
| Cairone et al. | Robustness of radiomics features to varying segmentation algorithms in magnetic resonance images | |
| CN111986216B (en) | RSG liver CT image interactive segmentation algorithm based on neural network improvement | |
| Lee et al. | Breast ultrasound automated ROI segmentation with region growing | |
| Fu et al. | Segmentation of spinal canal region in CT images using 3D region growing technique |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160323 |