CN112508949B - Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image - Google Patents
Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image Download PDFInfo
- Publication number
- CN112508949B CN112508949B CN202110135783.4A CN202110135783A CN112508949B CN 112508949 B CN112508949 B CN 112508949B CN 202110135783 A CN202110135783 A CN 202110135783A CN 112508949 B CN112508949 B CN 112508949B
- Authority
- CN
- China
- Prior art keywords
- image
- left ventricle
- spect
- network
- predicted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- 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
- G06T2207/10108—Single photon emission computed tomography [SPECT]
-
- 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
- G06T2207/30048—Heart; Cardiac
Abstract
The invention discloses a method for automatically segmenting a left ventricle of a SPECT three-dimensional reconstruction image, which comprises the steps of carrying out equal-scale reduction on linear interpolation of an original SPECT chest three-dimensional image, extracting rigid registration parameter characteristics from the reduced image by using a characteristic extraction network, automatically steering the SPECT image by using a spatial transformation network and the parameter characteristics to obtain a predicted image of a standard view, cutting a central part from the predicted image to obtain a heart image, and automatically segmenting the image by using a U-NET network to obtain a left ventricle structure segmentation result under the standard view. The invention synchronously extracts the position characteristics and the semantic characteristics of the image by using the deep learning network of the multi-task learning, realizes the integrated automatic steering, the heart positioning and the left ventricle structure segmentation from different angles to the standard view by utilizing the mutual supervision of the double network characteristics to achieve the effect of network integrated training, reduces the complexity and the artificial errors of manual steering and segmentation, realizes the full automation of the image operation and improves the accuracy.
Description
Technical Field
The invention relates to the field of medical images and the field of deep learning, in particular to a method for automatically segmenting a left ventricle of a SPECT three-dimensional reconstruction image based on a deep learning network.
Background
SPECT heart imaging is the current gold standard for clinical diagnosis of cardiovascular diseases such as coronary heart disease and myocardial ischemia, curative effect evaluation and prognosis judgment, can non-invasively provide functional information of myocardial tissues to detect potential pathological changes which do not cause structural change yet, and provides more detailed functional activity information of the myocardial tissues. During clinical SPECT examination, a series of operations and analysis need to be performed on a reconstructed SPECT image, wherein the calculation of a left ventricular ejection coefficient is an important index for evaluating the cardiac function, and the left ventricular chamber and the ventricular wall need to be segmented to extract the volumes of the left ventricular chambers in different heartbeat cycles for calculation. The clinically standard SPECT cardiac view is in the short axis SA direction, and since the long axis of the left ventricle is not parallel to the long axis of the human body, it is usually necessary to manually rotate the reconstructed SPECT image to obtain the standard SA view, and perform image segmentation in this view to calculate the left ventricular ejection coefficients. Meanwhile, left ventricular images based on standard SA views can be used to prepare cardiac polar plots for activity analysis of the left ventricular myocardium.
The transfer of images from conventional RA views to standard SA views of the heart for clinical analysis often requires manual manipulation by the physician, which is subjective, introduces random errors that affect the accuracy of the analysis, and consumes a long time for manual manipulation. For left ventricle image segmentation, currently, common clinical nuclear medicine heart image analysis software mostly adopts a segmentation method of conventional image processing, for example, a segmentation method based on a left ventricle wall center line, a segmentation method based on a left ventricle model, a segmentation method based on a heart atlas, a segmentation method based on threshold or k-means clustering, and the like. Therefore, for clinical processing and analysis of SPECT cardiac images, how to achieve stable and accurate automatic image steering and positioning from a conventional reconstruction view to a standard SA view used for specific clinical analysis, and how to achieve accurate segmentation and extraction of left ventricular structures in SPECT cardiac images of lower image resolution are technical problems faced in clinical SPECT cardiac processing.
Disclosure of Invention
The invention aims to provide a method for automatically segmenting the left ventricle of a SPECT three-dimensional reconstruction image based on a deep learning network, aiming at the defects of the prior art.
The purpose of the invention is realized by the following technical scheme:
a method for automatically segmenting the left ventricle of a SPECT three-dimensional reconstruction image comprises the following steps:
the method comprises the following steps: carrying out reduction and resampling on a conventional view RA of a SPECT three-dimensional reconstruction image, and taking the reduced conventional view RA-r after resampling as an input of a feature extraction network, wherein the feature extraction network consists of a convolution module and a full connection layer, the reduced conventional view RA-r is subjected to feature extraction by utilizing the convolution module, and is fully connected and unfolded to form a 6-dimensional feature vector T-r, and the 6-dimensional feature vector T-r comprises translation parameters in 3 directions and rotation parameters at 3 angles;
step two: adjusting T-r into a characteristic vector T by utilizing the proportional relation between the conventional view RA and the reduced conventional view RA-r, wherein the rotation parameters are unchanged, and the translation parameters are amplified in equal proportion;
step three: applying the characteristic vector T-r to a reduced conventional view RA-r by utilizing a spatial transform network to obtain a predicted reduced image SA-r ', and simultaneously applying the characteristic vector T to a conventional view RA to obtain a predicted image SA';
step four: taking the center of the predicted image SA ' as the center, intercepting the extracted heart part in the image to form a predicted heart image SA-H ', and simultaneously carrying out image gradient calculation on the heart image SA-H ' to obtain a corresponding gradient image SA-G;
step five: fusing the heart image SA-H' and the gradient map SA-G into a dual-channel image, extracting image characteristics through down-sampling and up-sampling of a three-dimensional U-NET network, and performing segmentation processing through a softmax layer to obtain a predicted left ventricle structure segmentation result F;
the characteristic extraction network, the spatial transformation network and the U-NET network jointly adopt multi-task joint learning training, and a total training loss function L = delta L-par + mu L-img + lambda L-seg;
wherein δ, μ, and λ are weight coefficients; l-img is an image loss function between the predicted reduced image SA-r' and the reduced standard view SA-r, L-par is a parameter loss function between the feature vector T-r and the rigid registration parameter P-r, and L-seg is a label loss function between the predicted segmentation result F and the SA direction segmentation label G; the reduced standard view SA-r is obtained by manually steering a conventional view RA of a SPECT three-dimensional reconstruction image and reducing the conventional view RA in equal proportion; the rigid registration parameter P-r is a registration parameter between a reduced conventional view RA-r and a reduced standard view SA-r calculated by a rigid registration algorithm, and comprises translation parameters in 3 directions and rotation parameters at 3 angles; the SA direction segmentation label G comprises 3 numerical values of a left ventricle cavity, a ventricular wall and a background, and is obtained by taking a heart image SA-H captured by taking the center of a standard view SA as the center and manually drawing the ventricle cavity and the ventricular wall of the left ventricle of the heart.
Further, the size of the reduced regular view RA-r and the reduced standard view SA-r is preferably 64 × 64 voxel size; the cardiac image SA-H and the predicted cardiac image SA-H' should be optimized to cover the entire cardiac image while containing as little other high intensity organ information as possible, preferably 32 x 32 voxel size.
Further, in the third step, by constructing a transformation matrix P = [ R M =T]Obtaining a predicted image, wherein M = [ tx, ty, tz =]Representing a displacement matrix, tx, ty and tz are translation parameters in 3 corresponding directions in the feature vector respectively, R is a rotation matrix, and R is set from an Euler angle R = [ beta, alpha, gamma ]]Conversion to world coordinate system parameters:
further, the gradient map SA-G obtains a calculation formula of:
wherein the content of the first and second substances,g x , g y , g z respectively the gradient in the x, y and z directions, and the calculation formula is：
Wherein i, j and k are coordinate indexes of x, y and z directions in the image.
Because the gradient image can better provide image boundary information, the synchronous reading of the gradient image can optimize the identification of the network to the image edge, thereby further improving the effectiveness of the network to the image segmentation.
Furthermore, the number of convolution modules and full connection layers of the feature extraction network is 3, and each convolution module comprises a convolution layer and a pooling layer; the down-sampling and the up-sampling of the three-dimensional U-NET network are both 4 times, wherein the down-sampling comprises a convolutional layer and a pooling layer, and the up-sampling comprises a convolutional layer and a deconvolution layer.
Further, in the SA direction division label G, the background is set to 0, the left ventricular cavity is set to 1, and the left ventricular wall is set to 2.
Further, the values of δ, μ and λ are 1,100 and 10, respectively.
Further, the image loss function L-img adopts a mean square error function.
Further, the parameter loss function L-par employs an absolute value loss function L1 or a norm loss function L2.
Further, the label loss function L-seg adopts a Dice-loss function, wherein the label loss functions L-seg-1 and L-seg-2 are calculated for the left ventricular cavity and the left ventricular wall respectively, and finally L-seg = L-seg-2+ L-seg-1.
The invention has the beneficial effects that: the invention synchronously extracts the position characteristics and the semantic characteristics of the image by using a deep learning network of multi-task learning, achieves the effect of network integrated training by utilizing the mutual supervision of double network characteristics to realize the integrated automatic steering, heart positioning and left ventricle structure segmentation from different angles to standard views, reduces the complexity and artificial errors of manual steering and segmentation, realizes the full automation of image operation and improves the accuracy, and further adopts a gradient image to increase the accuracy of image data edge segmentation.
Drawings
FIG. 1 is a flow chart of the SPECT three-dimensional image left ventricle automatic segmentation of the invention.
Fig. 2 is a schematic diagram of an automatic steering and positioning module in a SPECT three-dimensional image left ventricle automatic segmentation model structure.
Fig. 3 is a schematic diagram of an automatic segmentation module in the structure of an automatic left ventricle segmentation model of a SPECT three-dimensional image.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings.
The invention provides a SPECT three-dimensional reconstruction image left ventricle automatic segmentation method based on a deep learning network, which comprises the following steps: the method comprises the steps of scaling an original SPECT chest three-dimensional reconstruction image to 64 x 64 voxel size through linear interpolation, extracting rigid registration parameter features from the reduced image through a feature extraction network, automatically steering the SPECT three-dimensional reconstruction image through a space transformation network and the extracted rigid registration parameter features to obtain a predicted image of a standard view, cutting a central 32 x 32 voxel part from the predicted image of the standard view to obtain a heart image, and automatically segmenting the image through a U-NET network to obtain a left ventricle structure segmentation result under the standard view, wherein the process is shown in FIG. 1. The feature extraction network, the spatial transformation network and the U-NET network jointly adopt multi-task joint learning training, and a joint loss function of a total loss function image loss function L-img, a parameter loss function L-par and a label loss function L-seg is trained; the multitask co-learning can ensure the co-learning of the front part and the back part of the network and the optimization of the target.
In the following, a SPECT three-dimensional image left ventricle automatic segmentation model (structure is shown in fig. 2 and fig. 3) is provided to realize integrated automatic steering, positioning and segmentation of the method of the present invention, and the construction and training of the model specifically includes the following steps:
the method comprises the following steps: acquiring a conventional view RA of 600 SPECT three-dimensional reconstruction images, manually turning to a standard view SA for clinical analysis, wherein the left ventricle is located at the center of the image, cutting out 32 x 32 voxel size images in the standard view SA by taking the center of the image as the center to obtain a heart image SA-H, manually delineating the ventricular cavity and the ventricular wall of the left ventricle of the heart to obtain an SA direction segmentation label G, calculating a rigid registration parameter P between the conventional view RA and the standard view SA by a rigid registration algorithm, wherein 6 parameters of the rigid registration parameter P are translation parameters in 3 directions and rotation angle parameters P = [ tx, ty, tz, beta, alpha and gamma ] in 3 angles respectively, and forming a mapping database of the conventional view RA, the standard view SA, the SA direction segmentation label G and the rigid registration parameter P of the SPECT image. And proportionally resampling the conventional view RA and the standard view SA to a reduced conventional view RA-r and a reduced standard view SA-r with the size of 64 x 64 voxel by adopting linear interpolation, and adjusting the rigid registration parameter P to the reduced rigid registration parameter P-r to form a mapping data set of RA-r, SA-r and P-r of the SPECT image. Similarly, the reduced rigid registration parameter P-r comprises translation parameters of 3 directions and rotation parameters of 3 angles, the rotation parameters of P and P-r are the same, and the translation parameters are reduced in an equal proportion according to the scaling; the SA directional segmentation label G includes 3 values of the left ventricular cavity, the ventricular wall, and the background, where the background value is 0, the left ventricular cavity value is 1, and the left ventricular wall value is 2.
Step two: inputting the reduced conventional view RA-r into a feature extraction network, performing feature extraction on the reduced conventional view RA-r by using a convolution module, fully connecting and expanding the reduced conventional view RA-r to form a 6-dimensional feature vector T-r, adjusting the T-r to a feature vector T with the same proportion as P by using the equal proportion relation between P-r and P, wherein the rotation parameter is not changed, and the translation parameter is amplified in equal proportion; the 6-dimensional vector of T-R and T can be split into a displacement matrix M = [ tx, ty, tz ] and a rotation matrix R = [ beta, alpha, gamma ], and R is converted from Euler angles to world coordinate system parameters:
reconstruction of transform matrix as T' = [ R MT]To adapt the parameter input of the subsequent spatial transformation network. FIG. 2 shows an exemplary configuration of an auto-steering module, which includes a feature extraction network and a nullAn inter-transform network. The convolution module in the feature extraction network consists of a convolution unit of 3 multiplied by 3 and a Relu activation function unit.
Step three: applying the feature vector T-r to the conventional view RA-r via a spatial transform network to obtain a predicted reduced image SA-r ', and applying the feature vector T to the SPECT reconstructed image conventional view RA to obtain a predicted image SA' (FIG. 2); the image of 32 x 32 voxel size is truncated to extract the heart portion, centered on the center of the predicted image SA ', forming a heart image SA-H'. Calculating and generating a gradient map SA-G (FIG. 3) of the SA-r' image, wherein the gradient map SA-G obtains a calculation formula as follows:
wherein the content of the first and second substances,g x , g y , g z respectively the gradient in the x, y and z directions, and the calculation formula is：
Wherein i, j and k are coordinate indexes of x, y and z directions in the image.
Step four: the cardiac image SA-H' and the gradient map SA-G are further integrated into a dual-channel image, and a three-dimensional U-NET network is utilized to obtain a predicted left ventricle structure segmentation result F. Fig. 3 is an exemplary structure of an image segmentation module, where the Convolution module is composed of a 3 × 3 × 3 Convolution unit (Convolution, Conv.) and a Relu activation function unit, and the upsampling module is composed of a 3 × 3 × 3 transposed Convolution unit (transconvolution, trans. Conv.) and a Relu activation function unit. And the last module is connected with the softmax layer through the convolution unit of 1 multiplied by 1 to realize the output of the final segmentation result. The dashed lines represent a double layer of information that is manipulated to copy and crop the data to combine the image and features.
Step five: and constructing an image loss function L-img between the predicted reduced image SA-r' and the reduced standard view RA-r, a parameter loss function L-par between the feature vector T-r and the rigid registration parameter P-r, and a label loss function L-seg between the predicted segmentation result F and the SA direction segmentation label G, and training and optimizing the network to obtain the SPECT three-dimensional image left ventricle automatic segmentation model. The specific implementation thereof is subdivided into the following sub-steps:
(5.1) the training of the automatic segmentation model is a process of multitask learning, the loss matrix of which will contain constraint information for the RA-r to SA-r steering model and the constraint information for the segmentation model from SA-H to the segmentation result F, the learning objectives of the multitask together forming the constraint for the whole network to train the automatic segmentation model. The overall loss matrix of the model is designed as a combined loss function L = δ L-par + μ L-img + λ L-seg of an image loss function L-img, a parametric loss function L-par, and a tag loss function L-seg, where δ, μ, and λ are weighting coefficients that empirically take values of 1,100, and 10, respectively.
(5.2) the image loss function L-img of the steering model part uses a mean square error function between the predicted image and the reference image, and the corresponding rigid registration parameter loss function L-par uses an absolute value loss function L1 or a norm loss function L2.
(5.3) adopting a Dice-loss function for the segmentation model part label loss function L-seg, wherein label loss functions L-seg-1 and L-seg-2 are calculated for the left ventricular cavity and the left ventricular wall respectively, and finally L-seg = L-seg-2+ L-seg-1.
The above-described embodiments are intended to illustrate rather than to limit the invention, and any modifications and variations of the present invention are within the spirit of the invention and the scope of the appended claims.
Claims (10)
1. A method for automatically segmenting a left ventricle of a SPECT three-dimensional reconstruction image is characterized by comprising the following steps:
the method comprises the following steps: carrying out reduction and resampling on a conventional view RA of a SPECT three-dimensional reconstruction image, and taking the reduced conventional view RA-r after resampling as an input of a feature extraction network, wherein the feature extraction network consists of a convolution module and a full connection layer, the reduced conventional view RA-r is subjected to feature extraction by utilizing the convolution module, and is fully connected and unfolded to form a 6-dimensional feature vector T-r, and the 6-dimensional feature vector T-r comprises translation parameters in 3 directions and rotation parameters at 3 angles;
step two: adjusting T-r into a characteristic vector T by utilizing the proportional relation between the conventional view RA and the reduced conventional view RA-r, wherein the rotation parameters are unchanged, and the translation parameters are amplified in equal proportion;
step three: applying the characteristic vector T-r to a reduced conventional view RA-r by utilizing a spatial transform network to obtain a predicted reduced image SA-r ', and simultaneously applying the characteristic vector T to a conventional view RA to obtain a predicted image SA';
step four: taking the center of the predicted image SA ' as the center, intercepting the heart part in the image to form a predicted heart image SA-H ', and simultaneously carrying out image gradient calculation on the predicted heart image SA-H ' to obtain a corresponding gradient image SA-G;
step five: fusing the predicted heart image SA-H' and the gradient map SA-G into a dual-channel image, extracting image characteristics through down sampling and up sampling of a three-dimensional U-NET network, and performing segmentation processing through a softmax layer to obtain a predicted left ventricle structure segmentation result F;
the feature extraction network, the spatial transformation network and the three-dimensional U-NET network jointly adopt multi-task joint learning training, and a total training loss function L = delta L-par + mu L-img + lambda L-seg;
wherein δ, μ, and λ are weight coefficients; l-img is an image loss function between the predicted reduced image SA-r' and the reduced standard view SA-r, L-par is a parameter loss function between the feature vector T-r and the rigid registration parameter P-r, and L-seg is a label loss function between the predicted segmentation result F and the SA direction segmentation label G; the reduced standard view SA-r is obtained by manually steering a conventional view RA of a SPECT three-dimensional reconstruction image and reducing the conventional view RA in equal proportion; the rigid registration parameter P-r is a registration parameter between a reduced conventional view RA-r and a reduced standard view SA-r calculated by a rigid registration algorithm, and comprises translation parameters in 3 directions and rotation parameters at 3 angles; the SA direction segmentation label G comprises 3 numerical values of a left ventricle cavity, a ventricular wall and a background, and is obtained by taking a heart image SA-H captured by taking the center of a standard view SA as the center and manually drawing the ventricle cavity and the ventricular wall of the left ventricle of the heart.
2. The method for automatic left ventricle segmentation of the SPECT three-dimensional reconstructed image of claim 1 wherein the reduced regular view RA-r and the reduced standard view SA-r are sized to 64 x 64 voxel size; the size of the cardiac image SA-H and the predicted cardiac image SA-H' is 32 x 32 voxel size.
3. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein in the third step, the spatial transform network is constructed by constructing a transform matrix P = [ R M ]T]Obtaining a predicted image, wherein M = [ tx, ty, tz =]Representing a displacement matrix, wherein tx, ty and tz are translation parameters in 3 corresponding directions in the feature vector respectively, R is a rotation matrix and is represented by a world coordinate system:
r = [ β, α, γ ] is the euler angle representation of the rotation matrix.
4. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein the gradient map SA-G is obtained by the following calculation formula:
wherein，g x , g y , g z Respectively the gradient in the x, y and z directions, and the calculation formula is：
Wherein i, j and k are coordinate indexes of x, y and z directions in the image.
5. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein the number of convolution modules and full connection layers of the feature extraction network is 3, and each convolution module comprises a convolution layer and a pooling layer; the down-sampling and the up-sampling of the three-dimensional U-NET network are 4, wherein the down-sampling comprises a convolutional layer and a pooling layer, and the up-sampling comprises a convolutional layer and a deconvolution layer.
6. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein in the SA direction segmentation label G, the background is set to 0, the left ventricle cavity is set to 1, and the left ventricle wall is set to 2.
7. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein the values of delta, mu and lambda are 1,100 and 10 respectively.
8. The method for automatic left ventricular segmentation of the SPECT three-dimensional reconstruction image as claimed in claim 1, wherein the image loss function L-img employs a mean square error function.
9. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein the parameter loss function L-par adopts an absolute value loss function L1 or a norm loss function L2.
10. The method for automatically segmenting the left ventricle of the SPECT three-dimensional reconstruction image according to claim 1, wherein the label loss function L-seg adopts a Dice-loss function, wherein the label loss functions L-seg-1 and L-seg-2 are respectively calculated for the left ventricle cavity and the left ventricle wall, and finally L-seg = L-seg-2+ L-seg-1.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110135783.4A CN112508949B (en) | 2021-02-01 | 2021-02-01 | Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image |
| PCT/CN2022/072199 WO2022161192A1 (en) | 2021-02-01 | 2022-01-15 | Method for automatically segmenting left ventricle of spect three-dimensional reconstruction image |
| JP2022564825A JP7270319B2 (en) | 2021-02-01 | 2022-01-15 | Left ventricle automatic segmentation method of SPECT three-dimensional reconstructed image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110135783.4A CN112508949B (en) | 2021-02-01 | 2021-02-01 | Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112508949A CN112508949A (en) | 2021-03-16 |
| CN112508949B true CN112508949B (en) | 2021-05-11 |
Family
ID=74953088
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110135783.4A Active CN112508949B (en) | 2021-02-01 | 2021-02-01 | Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image |
Country Status (3)
| Country | Link |
|---|---|
| JP (1) | JP7270319B2 (en) |
| CN (1) | CN112508949B (en) |
| WO (1) | WO2022161192A1 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508949B (en) * | 2021-02-01 | 2021-05-11 | 之江实验室 | Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image |
| CN113516658B (en) * | 2021-09-14 | 2021-12-17 | 之江实验室 | Automatic steering and segmenting method for left ventricle of PET three-dimensional image |
| CN116934768A (en) * | 2023-08-16 | 2023-10-24 | 中国人民解放军总医院 | Method and system for improving blood vessel segmentation accuracy in CTA image mode |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20150099169A (en) * | 2014-02-21 | 2015-08-31 | 전남대학교병원 | Method for measuring myocardial perfusion and method for diagnosing cardiac ischemia using the same |
| CN107545584B (en) * | 2017-04-28 | 2021-05-18 | 上海联影医疗科技股份有限公司 | Method, device and system for positioning region of interest in medical image |
| CN107220965B (en) * | 2017-05-05 | 2021-03-09 | 上海联影医疗科技股份有限公司 | Image segmentation method and system |
| JP6483875B1 (en) * | 2018-01-25 | 2019-03-13 | 日本メジフィジックス株式会社 | Myocardial image display method, myocardial image display processing program, and myocardial image processing apparatus |
| CN109285157A (en) * | 2018-07-24 | 2019-01-29 | 深圳先进技术研究院 | Myocardium of left ventricle dividing method, device and computer readable storage medium |
| CN109801294A (en) * | 2018-12-14 | 2019-05-24 | 深圳先进技术研究院 | Three-dimensional atrium sinistrum dividing method, device, terminal device and storage medium |
| CN109584254B (en) * | 2019-01-07 | 2022-12-20 | 浙江大学 | Heart left ventricle segmentation method based on deep full convolution neural network |
| JP2019128358A (en) * | 2019-02-13 | 2019-08-01 | 日本メジフィジックス株式会社 | Myocardial image display method, myocardial image display processing program, and myocardial image processor |
| CN110120051A (en) * | 2019-05-10 | 2019-08-13 | 上海理工大学 | A kind of right ventricle automatic division method based on deep learning |
| CN110163876B (en) * | 2019-05-24 | 2021-08-17 | 山东师范大学 | Left ventricle segmentation method, system, device and medium based on multi-feature fusion |
| CN111242956A (en) * | 2020-01-09 | 2020-06-05 | 西北工业大学 | U-Net-based ultrasonic fetal heart and fetal lung deep learning joint segmentation method |
| CN111340186B (en) * | 2020-02-17 | 2022-10-21 | 之江实验室 | Compressed representation learning method based on tensor decomposition |
| CN111369537A (en) * | 2020-03-05 | 2020-07-03 | 上海市肺科医院（上海市职业病防治院） | Automatic segmentation system and method for pulmonary milled glass nodules |
| CN111739161B (en) * | 2020-07-23 | 2020-11-20 | 之江实验室 | Human body three-dimensional reconstruction method and device under shielding condition and electronic equipment |
| CN112508949B (en) * | 2021-02-01 | 2021-05-11 | 之江实验室 | Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image |
-
2021
- 2021-02-01 CN CN202110135783.4A patent/CN112508949B/en active Active
-
2022
- 2022-01-15 JP JP2022564825A patent/JP7270319B2/en active Active
- 2022-01-15 WO PCT/CN2022/072199 patent/WO2022161192A1/en active Application Filing
Also Published As
| Publication number | Publication date |
|---|---|
| JP7270319B2 (en) | 2023-05-10 |
| WO2022161192A1 (en) | 2022-08-04 |
| JP2023516227A (en) | 2023-04-18 |
| CN112508949A (en) | 2021-03-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112508949B (en) | Method for automatically segmenting left ventricle of SPECT three-dimensional reconstruction image | |
| Kirisli et al. | Fully automatic cardiac segmentation from 3D CTA data: a multi-atlas based approach | |
| CN109598722B (en) | Image analysis method based on recurrent neural network | |
| WO2023040628A1 (en) | Method for automatically turning and segmenting left ventricle in pet three-dimensional image | |
| CN110232691A (en) | A kind of dividing method of multi-modal CT images | |
| CN110599499B (en) | MRI image heart structure segmentation method based on multipath convolutional neural network | |
| US20030099384A1 (en) | Detection and analysis of lesions in contact with a structural boundary | |
| CN110838140A (en) | Ultrasound and nuclear magnetic image registration fusion method and device based on hybrid supervised learning | |
| CN110751651A (en) | MRI pancreas image segmentation method based on multi-scale migration learning | |
| CN115830016B (en) | Medical image registration model training method and equipment | |
| CN107909653B (en) | Heart soft tissue three-dimensional reconstruction method based on sparse principal component analysis | |
| CN112164447B (en) | Image processing method, device, equipment and storage medium | |
| CN113689441A (en) | DeepLabV3 network-based left ventricle ultrasonic dynamic segmentation method | |
| CN111724395B (en) | Four-dimensional context segmentation method, device, storage medium and apparatus for heart image | |
| CN111862320B (en) | Automatic steering method for SPECT three-dimensional reconstruction image to standard view | |
| CN113889261B (en) | Pathological feature assistance-based PET/CT automatic lung cancer diagnosis classification model training method | |
| WO2022183851A1 (en) | Lung lobe segmentation method based on digital human technology | |
| CN115409859A (en) | Coronary artery blood vessel image segmentation method and device, storage medium and terminal | |
| Che et al. | Segmentation of bone metastases based on attention mechanism | |
| Xia et al. | A multi-scale network with the encoder-decoder structure for CMR segmentation | |
| CN111242928A (en) | Atrial full-automatic segmentation tracking and positioning method based on multi-view learning | |
| CN114820754B (en) | Automatic positioning method for four-cavity heart based on template matching | |
| Jia et al. | SCTransNet: 3D Medical Image Segmentation Model Based on the Fusion of CNN and Transformer | |
| Hu et al. | Abdominal Multi-Organ Segmentation Based on nnUNet | |
| CN114820754A (en) | Heart four-chamber heart automatic positioning method based on template matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |