CN1993710A - Watershed segmentation to improve detection of spherical and ellipsoidal objects using cutting planes - Google Patents
Watershed segmentation to improve detection of spherical and ellipsoidal objects using cutting planes Download PDFInfo
- Publication number
- CN1993710A CN1993710A CN 200580025982 CN200580025982A CN1993710A CN 1993710 A CN1993710 A CN 1993710A CN 200580025982 CN200580025982 CN 200580025982 CN 200580025982 A CN200580025982 A CN 200580025982A CN 1993710 A CN1993710 A CN 1993710A
- Authority
- CN
- China
- Prior art keywords
- image
- desired object
- watershed
- plane
- candidate pixel
- 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
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
Abstract
In one exemplary embodiment of the present invention, a method of detecting a desired object at a candidate pixel from an image is provided. The method includes the steps of (a) selecting a representative point in the desired object; (b) determining first representative cross-sections of the desired object by passing first lower dimension planes through the representative point; (c) passing at least one second lower dimension plane through the candidate pixel; (d) using region segmentation to separate the candidate pixel containing second regions from the rest of the pixels in each of the at least one second lower dimension plane; (e) matching at least one of the second regions with at least one of the first cross-sections; (f) determining a match value based on the result of step (e); and (g) using the match value to determine if the desired object is detected at the candidate pixel.
Description
The cross reference of related application
It is No.60/576 that the application requires application number, 041 U.S. Provisional Application No., and this application was submitted on June 1st, 2004, and this application is incorporated herein by reference in this integral body.
Background of invention
1. invention field
The present invention relates generally to field, and relate more particularly to from image, detect spherical objects and ellipticity object based on computer generated image.
2. association area is described
The purpose of a plurality of computer-aided diagnosises based on image (" CAD ") instrument is to help the doctor to detect spherical in large quantities of image slices and the ellipticity structure.For example in chest image, the doctor may be interested in the detection tubercle, and this tubercle is rendered as white spheroid or hemisphere in the lung region inside of dead color.For example in colon image, the doctor may be interested in the detection polyp again, and this polyp is rendered as and is attached to epicolic spherical and hemispherical raised structures.For example in vascular image, the doctor may be interested in the detection aneurysm again, and this aneurysm is rendered as the bulbous protrusion structure beyond vessel surface.The doctor may expect to detect any various other structures in the anatomical tissue.These structures can include, but are not limited to the polyp in various types of tumours, the bladder, the hemangioma in the liver or the like.
The current method that is used for detecting from 3D rendering spherical or part chondritic normally is a plurality of 2D planes with image division.Detect loop configuration or lump then in described a plurality of planes, these images are oriented at a plurality of directions across entire image.Can be combined into 3D from the collected information in described a plurality of planes reproduces.
In addition, can carry out pre-service to original volumetric data, for example, to improve the whole result of process, perhaps after some conversion, to represent to find spherical objects with the another kind of identical image.Yet this pre-service may spended time and is made instability or consistent effectively hypothesis, thereby causes carrying out failure.
Summary of the invention
In one aspect of the invention, provide a kind of in the method that detects desired object from the candidate pixel place of image.This method comprises the following steps: that (a) selects representational point in desired object; (b) pass the first representational xsect that this representational point is determined desired object by the plane that makes the first low dimension; (c) make the plane of at least one second low dimension pass this candidate pixel; (d) rest of pixels of using Region Segmentation will comprise in each plane at least one second low dimension plane of candidate pixel and this of second area is separated; (e) at least one second area and at least one first xsect are complementary; (f) result according to step (e) determines matching value; And (g) use this matching value to determine whether to detect desired object at this candidate pixel place.
In another aspect of this invention, provide a kind of machine readable media, this machine readable media has the instruction that is stored thereon, and these instructions move the method for carrying out detect desired object from the candidate pixel place of image by processor.This method comprises the following steps: that (a) selects representational point in desired object; (b) pass the first representative xsect that this representational point is determined desired object by the plane that makes the first low dimension; (c) make the plane of at least one second low dimension pass this candidate pixel; (d) use Region Segmentation that the rest of pixels in each plane that comprises at least one second low dimension plane of candidate pixel and this of second area is separated; (e) at least one second area and at least one first xsect are complementary; (f) result based on step (e) determines matching value; And (g) use this matching value to determine whether to detect desired object at this candidate pixel place.
In one side more of the present invention, provide a kind of in the system that detects desired object from the candidate pixel place of image.This system comprises: selecting arrangement is used for selecting representational point in desired object; Determine device, be used for passing the first representative xsect that this representational point is determined desired object by the plane that makes the first low dimension; Segmenting device is used for using Region Segmentation that the rest of pixels that comprises in each plane at least one second low dimension plane of candidate pixel and this of second area is separated; Coalignment is used for: (a) at least one second area and at least one first xsect are complementary, and (b) determine matching value according to this coupling; And pick-up unit, be used to use this matching value to determine whether to detect desired object at this candidate pixel place.
The accompanying drawing summary
The present invention can understand with reference to following explanation in conjunction with the drawings, wherein identical Ref. No. sign components identical, and wherein:
Fig. 1 shows section from the three dimensional CT image of the colon that comprises polyp according to an exemplary embodiment of the present invention;
Fig. 2 shows the CT image of Fig. 1 and carry out watershed segmentation image afterwards after carrying out the inhibition of object and background local maxima according to an exemplary embodiment of the present invention; And
Fig. 3 shows the image after use is cut apart the image of Fig. 2 from the simple threshold values of watershed segmentation and information.
The detailed description of preferred implementation
Illustrated embodiment of the present invention is described below.For clear, all features to actual embodiment all are not described in instructions.Will of course be appreciated that, in the research and development of any this actual embodiment, must make the decision of a plurality of embodiment special uses for the specific objective of realizing the developer, these specific objectives are such as meeting relevant with system and relevant with commerce constraint, and these specific objectives will change because of the difference of embodiment.And, should be understood that this development may be complicated and time-consuming, but for the those of ordinary skills that have benefited from present disclosure, this only is regular works.
Although the present invention is subjected to the influence of the various modifications and the form of replacement easily, its specific implementations for example is illustrated in the accompanying drawings and here describes in detail.Yet, be to be understood that, be not intended to limit the invention to particular forms disclosed in this description to specific implementations, but opposite, the present invention will cover all modifications, equivalent and the alternative that falls into by in the appended the spirit and scope of the present invention that claim limited.
Should be appreciated that system and method described herein can various forms of hardware, software, firmware, application specific processor or its make up and implement.Especially, at least a portion of the present invention preferably is implemented as the application program that comprises programmed instruction, these programmed instruction (for example are included in one or more program storage devices really, hard disk, flexible plastic disc, RAM, ROM, CD ROM, or the like) go up and can carry out by any device or the machine that comprise suitable framework, such as carrying out by universal digital computer with processor, storer and input/output interface.It is also understood that because some shown in the accompanying drawing formed system unit and process steps is preferably implemented with software, so the mode that the connection between the system module (or logic flow of method step) may be programmed according to the present invention and different.Provide instruction at this, the those of ordinary skill in related field can reckon with these and similar embodiment of the present invention.
In following disclosure, should be appreciated that term " image " as used in this can comprise that any dimension is more than or equal to 2 image (for example, two dimensional image, three-D volumes).
Although so be not limited,, the part spherical objects in the polyp of colon environment in reference computers tomoscan (" the CT ") image is described by illustrative embodiments of the present invention here for simply.Yet, should be appreciated that as those skilled in the art expect the present invention can be applicable to form widely, these forms comprise magnetic resonance (" MR "), ultrasound wave (" US ") and positron emission computerized tomography (" PET ").It is also understood that the present invention can be used to the non-medical imaging applications, such as in geological data, finding spherical salt dome.
In the ideal case, circular object (being area-of-interest) is clearly separated with background and with other objects (being referred to as " background " hereinafter).Simple intensity threshold is enough for isolating area-of-interest.Yet, as this situation typically, circular object may not by simple threshold values or the threshold value of crossing over the unique application of image easily with background separation.
For example, consider to comprise the image of following two objects: circular object (that is area-of-interest) and background object.Also consider it to be that the intensity in the zone between two objects of background intensity is similar to the intensity of these two objects and/or owing to image acquisition and/or reconstruction but level and smooth because of partial volume effect.In this case, the optimal threshold treatment technology should be able to be suitable for finding contours of objects, no matter and background intensity and variation thereof.For example, when two objects are close to each other, the partial volume effect between these two objects make the strength ratio of background in other place more near the intensity of object.Fig. 1 is at above-mentioned example shown in computed tomography (" the CT ") image of colon.
With reference now to Fig. 1,, the section from the three dimensional CT image 100 of colon is shown, thereby represents polyp 105 by dotted circle centered on.Polyp 105 is present in inner chamber 110 (also being known as " background " or " air ") and the bodily tissue on every side 115.Bodily tissue 115 on every side and the edge between the inner chamber 110 are colon walls.It should be noted that the intensity between polyp 105 and the colon wall may be different from the intensity of background and normally uncertain.
Polyp 105 invests colon wall, and is adhering to the place, not only owing to form adhering to transition but also having partial volume effect owing to cutting into slices by volume of protuberance (ridge).As shown in Figure 1, these volume effects are rendered as the minimizing of intensity level.Refer again to Fig. 1, the intensity level of visual leap center horizontal image distribution (profile) in the left side.Dotted circle area illustrates polyp 105.During horizontal line 120 in following image, can see intensity distributions, increase then from low intensity.At the center (with the cross-hair cross section of polyp 105), can see that intensity level increases on the left side of polyp 105, and cross over polyp, but reduce once again to represent ridge.This ridge may be difficult to limit.Dividing ridge method described herein makes it possible to easily detect this transition.
As used in this, term " intensity " refers to the measuring of local X-gamma ray absorption coefficient of CT image.For visual, can use the mapping from intensity to the color.For example, in an exemplary map technology, such as the low absorption coefficient of the absorption coefficient of air corresponding to " deceiving " look, and such as the high absorption coefficient of the absorption coefficient of tissue corresponding to light grey or white.
(that is, watershed method) technology is used for separating automatically area-of-interest, thereby has overcome the problem of selecting optimal threshold to propose a kind of use watershed segmentation.Watershed method is a kind of technology that is used for cutting apart by the simulation submergence basin (basins) of intensity distributions.Term " watershed divide " is meant the zone that in the image water is discharged to minimum point along the inclined-plane downwards.The output of watershed method is the ridge point (crest point) between one group of composition that marked, that connected (component) and the one group of composition that is connected.
Watershed segmentation is a kind of morphology technology, and this technology can be described as follows intuitively:
(1) image is considered as height map, and along with water penetrates minimum point and evenly rises and cross over image fully and immerse it in water gradually; And
(2) converge part at two distinct water bodys (that is, reception basin) and place " dykes and dams ", and this process of continuing up to water arrived image have a few.Dykes and dams provide finally to be cut apart.In the image field, these dykes and dams can be interpreted as the growth of seed, and these seeds are placed on the minimum value of image highly pro rata with it simultaneously, and the growth of these seeds finally converges on the crestal line of intensity map.Term " minimum value " is meant do not had point, curve or the plane that surrounds than the pixel of low value.Term " maximal value " is meant point, curve or the plane of the pixel encirclement that is not had high value.
From background, isolate object (that is area-of-interest) and usually need to determine the border of object (that is, object begin and stop place).Yet, judge that the border of the object in the background may be uncertain for the pixel (that is, the ridge point) on border, these pixels have the intensity level between object strength range and background intensity scope.These intensity levels may be by partial volume effect or blur and generation artificially.
Watershed segmentation can be with object and background separation.Especially, watershed segmentation produces the ridge point of Ambience object of interest.Watershed segmentation will find all the ridge points in the image usually.Yet, the ridge point that just has the uncertain strength range between object strength range and background intensity scope that needs.Come only in this uncertain strength range, to find the ridge point in order to retrain the watershed divide, for example remove all local maximums in object and the background by image being carried out threshold process suitably.This can be considered to object and background local maxima suppresses.
The aftertreatment of watershed method may be necessary.Especially, hint that these one-tenth that connect belong to same object if know this situation, all ridge points between the composition that may want so to suppress to be connected, the maximum intensity of these compositions are (that is, the ridge point is pure partial volume group) not in the scope of object intensity.As used in this, term " composition that is connected " is meant gathering of " all connecting " pixel.Just, any two pixels in this group pixel can arrive one other pixel from a pixel, jump to neighbor from neighbor.In interchangeable embodiment, can also merge its maximum intensity all adjacent compositions that connected in the scope of object intensity.
With reference now to Fig. 2,, the as above watershed divide output 200 of CT image 100 after carrying out the inhibition of object and background local maxima of detailed described Fig. 1 is shown.Ridge point (that is, being decorated with the border of shade) 205 clearly separates polyp 210 and on every side bodily tissue 215 and colon wall (that is the edge between polyp 210 and on every side the bodily tissue 215).Ridge point 205 is discerned the optimum Cutting between polyp 210 and the colon wall automatically.
Then, polyp 210 and separation energy between the colon wall are used to the further calculating based on the composition that is connected.For example, consider a kind of method that is used for cutting planes that is used in the CT scan of colon, detect polyp.This cutting planes method can be used to use the present invention of image segmentation, and this image segmentation is considered cutting apart in the watershed clusters.The illustrative methods that is used for cutting planes can be that the application number of the common unsettled and common transfer that on September 20th, 2004 submitted to is no.10/945,130 and name be called in the patented claim of " METHOD AND SYSTEM FOR USINGCUTTING PLANES FOR COLON POLYP DETECTION (using the method and system of cutting planes at colon polyp detection) " and find, this application all is incorporated herein by reference at this.
Fig. 3 illustrates the result that use is cut apart the image of Fig. 2 from the simple threshold values and the information (that is ridge point) of watershed segmentation.In exemplary simple threshold values, if the intensity of pixel under threshold value, pixel is set as black so, otherwise is set as white.This simple threshold values can not be separated polyp usually with colon wall.On the other hand, these two objects (that is, polyp and colon wall) can both be by forcing and will isolating by the ridge institute separate areas of watershed divide.
Should be appreciated that as those of ordinary skills anticipate, use any different dividing method.For example, realization is cut apart can be to use fixed threshold, overall adaptive threshold (promptly, the statistical figure of use such as average, minimum value, maximal value, standard deviation, quantile and weighted type thereof, alternative manner or image histogram analysis) and local auto-adaptive threshold value (that is, but the identical local sub-volumes that is used for overall adaptive threshold).
In an illustrative embodiments, watershed segmentation can be implemented by the following step:
(1) in volumetric image, selects volume of interest (" VOI ").This can be whole volume.
(2), select seed points (perhaps,, then from local minimum, selecting) the local maximum in VOI if seek hole rather than lump in this VOI inside.
(3) remove some incoherent watershed clusters according to specified criteria, these specified criteria are minimal surface, maximum surface, smallest circle, adjacent group's numbering and position for example.
Should be appreciated that as those of ordinary skill in the art reckons with, can use any different watershed segmentation method.For example, another exemplary may not need to select seed.
In addition, the present invention can be used to the appearance of its median ridge and can be only uses the watershed divide to recover and therefore as the image of the part of dynamic process by the time series of analysis image.This can be used to detect hole (depression) with symmetric mode equally, such as detecting diverticulosis (diverticulosis).
Since the present invention can for have benefited from this instruction the conspicuous difference of those of ordinary skills but the equivalence mode revise and put into practice, so above-mentioned disclosed specific implementations only is illustrative.In addition, except as following claim described those, and be not intended to be limited to the structure that goes out shown here or the details of design.Therefore, above-mentioned disclosed specific implementations obviously can be modified or be modified, and all this variations are considered in scope and spirit of the present invention.Therefore, illustrated in following claim in this protection of looking for.
Claims (20)
1. one kind in the method that detects desired object from the candidate pixel place of image, and this method comprises the following steps:
A. in desired object, select representational point;
B. pass the first representational xsect that this representational point is determined desired object by the plane that makes the first low dimension;
C. make the plane of at least one second low dimension pass this candidate pixel;
D. use Region Segmentation that the rest of pixels in each plane that comprises at least one second low dimension plane of candidate pixel and this of second area is separated;
E. at least one second area and at least one first xsect are complementary;
F. the result according to step e determines matching value; And
G. use this matching value to determine whether to detect desired object at this candidate pixel place.
2. the method for claim 1 further may further comprise the steps:
Select image, wherein this image is two-dimentional at least.
3. method as claimed in claim 2, wherein, the step that obtains image comprises:
Select medical image.
4. method as claimed in claim 3, wherein, select the step of medical image to comprise:
Select one of computed tomography (CT) image, magnetic resonance (MR), positron emission computerized tomography (PET) image, ultrasound wave (US) image and single photon emission computerized tomography,SPECT (SPECT) image.
5. the method for claim 1, wherein pass described representational point and determine that the step of the first representational xsect of desired object comprises by the plane that makes the first low dimension:
By making two-dimensional cross sectional determine the first representational xsect of desired object by described representational point, wherein this image is three-dimensional.
6. the method for claim 1, wherein use the step of Region Segmentation to comprise:
Use threshold process.
7. method as claimed in claim 6, wherein, use the step of threshold value to comprise:
Use fixed threshold processing, the processing of overall adaptive threshold and local adaptive threshold one of to handle.
8. the method for claim 1 further comprises:
Selection has the desired object of the shape of expectation.
9. method as claimed in claim 8, wherein, the step of desired object of selecting to have the shape of expectation comprises:
Selection has the desired object of one of spherical shape and ellipsoidal shape.
10. the method for claim 1, wherein use the step of Region Segmentation to comprise:
Use watershed segmentation.
11. method as claimed in claim 10 wherein, uses the step of watershed segmentation to comprise:
Growth watershed clusters in the volume of interest (VOI) of image; And
Remove incoherent watershed clusters according to the watershed divide criterion.
12. method as claimed in claim 11 further may further comprise the steps:
Select the VOI of image inside.
13. method as claimed in claim 11 further may further comprise the steps:
Within VOI, select seed points.
14. method as claimed in claim 13, wherein, the step of growth watershed clusters comprises in VOI:
From described seed points growth watershed clusters.
15. method as claimed in claim 13 wherein, selects the step of seed points to comprise within VOI:
Select seed points in the local maximum within VOI.
16. method as claimed in claim 13 wherein, selects the step of seed points to comprise within VOI:
Select seed points in the local minimum within VOI.
17. method as claimed in claim 11, wherein, the step of removing incoherent watershed clusters according to the watershed divide criterion comprises:
Remove incoherent watershed clusters according at least a surface properties.
18. method as claimed in claim 17, wherein, the step of removing incoherent watershed clusters according at least a surface properties comprises:
Remove incoherent watershed clusters according at least a in the attribute of maximal value, minimum value, smallest circle, a plurality of adjacent group, adjacent group's position, first order derivative, second derivative and watershed clusters.
19. a machine readable media, it has the instruction that is stored thereon, and described instruction moves the method for carrying out detect desired object from the candidate pixel place of image by processor, and this method comprises the following steps:
A. in desired object, select representational point;
B. pass the first representative xsect that this representational point is determined desired object by the plane that makes the first low dimension;
C. make the plane of at least one second low dimension pass this candidate pixel;
D. use Region Segmentation that the rest of pixels in each plane that comprises at least one second low dimension plane of candidate pixel and this of second area is separated;
E. at least one second area and at least one first xsect are complementary;
F. the result according to step e determines matching value; And
G. use this matching value to determine, whether detect desired object at this candidate pixel place.
20. one kind in the system that detects desired object from the candidate pixel place of image, this system comprises:
Selecting arrangement is used for selecting representational point in desired object;
Determine device, be used for passing the first representative xsect that this representational point is determined desired object by the plane that makes the first low dimension;
Segmenting device is used for using Region Segmentation that the rest of pixels that comprises in each plane at least one second low dimension plane of candidate pixel and this of second area is separated;
Coalignment is used for: (a) at least one second area and at least one first xsect are complementary, and (b) mate to determine matching value according to this; And
Pick-up unit is used to use this matching value to determine whether to detect desired object at this candidate pixel place.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US57604104P | 2004-06-01 | 2004-06-01 | |
| US60/576,041 | 2004-06-01 | ||
| US11/065,727 | 2005-02-25 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1993710A true CN1993710A (en) | 2007-07-04 |
Family
ID=38214954
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200580025982 Pending CN1993710A (en) | 2004-06-01 | 2005-03-04 | Watershed segmentation to improve detection of spherical and ellipsoidal objects using cutting planes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1993710A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014101387A1 (en) * | 2012-12-27 | 2014-07-03 | 同方威视技术股份有限公司 | Method for processing and identifying three-dimensional data |
| CN108000731A (en) * | 2017-11-16 | 2018-05-08 | 华侨大学 | A kind of circular arc of circular saw cutting stone material and the method for elliptic contour |
| CN111127646A (en) * | 2019-12-26 | 2020-05-08 | 西南林业大学 | Construction method and system of rasterized elevation curved surface for measuring height difference of landform |
-
2005
- 2005-03-04 CN CN 200580025982 patent/CN1993710A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014101387A1 (en) * | 2012-12-27 | 2014-07-03 | 同方威视技术股份有限公司 | Method for processing and identifying three-dimensional data |
| CN108000731A (en) * | 2017-11-16 | 2018-05-08 | 华侨大学 | A kind of circular arc of circular saw cutting stone material and the method for elliptic contour |
| CN108000731B (en) * | 2017-11-16 | 2019-07-05 | 华侨大学 | A kind of method of the circular arc and elliptic contour of circular saw cutting stone material |
| CN111127646A (en) * | 2019-12-26 | 2020-05-08 | 西南林业大学 | Construction method and system of rasterized elevation curved surface for measuring height difference of landform |
| CN111127646B (en) * | 2019-12-26 | 2023-03-14 | 西南林业大学 | Construction method and system of rasterized elevation curved surface for measuring height difference of landform |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20190355117A1 (en) | Techniques for Segmentation of Lymph Nodes, Lung Lesions and Other Solid or Part-Solid Objects | |
| JP6877868B2 (en) | Image processing equipment, image processing method and image processing program | |
| US7333646B2 (en) | Watershed segmentation to improve detection of spherical and ellipsoidal objects using cutting planes | |
| Qin et al. | Learning tubule-sensitive CNNs for pulmonary airway and artery-vein segmentation in CT | |
| US9147239B2 (en) | Computing the mass of an object | |
| EP1851722B1 (en) | Image processing device and method | |
| US7583831B2 (en) | System and method for using learned discriminative models to segment three dimensional colon image data | |
| Pu et al. | Shape “break-and-repair” strategy and its application to automated medical image segmentation | |
| AU2012220301B2 (en) | Method and apparatus for isolating a potential anomaly in imaging data and its application to medical imagery | |
| US20080118136A1 (en) | Propagating Shell for Segmenting Objects with Fuzzy Boundaries, Automatic Volume Determination and Tumor Detection Using Computer Tomography | |
| CN110910405B (en) | Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network | |
| CN1670768A (en) | Local watershed operators for image segmentation | |
| WO2003090173A2 (en) | Segmentation of 3d medical structures using robust ray propagation | |
| WO2005112769A1 (en) | Nodule detection | |
| CN1682657A (en) | System and method for a semi-automatic quantification of delayed enchancement images | |
| Chen et al. | Image segmentation method using thresholds automatically determined from picture contents | |
| US7480401B2 (en) | Method for local surface smoothing with application to chest wall nodule segmentation in lung CT data | |
| AU2005280763A1 (en) | Candidate generation for lung nodule detection | |
| Maitra et al. | Accurate breast contour detection algorithms in digital mammogram | |
| CN1910590A (en) | Automatic contrast medium control in images | |
| CN111798424A (en) | Medical image-based nodule detection method and device and electronic equipment | |
| CN1993710A (en) | Watershed segmentation to improve detection of spherical and ellipsoidal objects using cutting planes | |
| Zhang et al. | Comparative analysis of U-Net and TLMDB GAN for the cardiovascular segmentation of the ventricles in the heart | |
| CN1823337A (en) | System and method for detecting a protrusion in a medical image | |
| Jeong et al. | Segmentation methodology for automated classification and differentiation of soft tissues in multiband images of high-resolution ultrasonic transmission tomography |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |