CN115861600B - ROI (region of interest) area identification method and system for SPECT (Single photon emission computed tomography) image - Google Patents
ROI (region of interest) area identification method and system for SPECT (Single photon emission computed tomography) image Download PDFInfo
- Publication number
- CN115861600B CN115861600B CN202211635885.3A CN202211635885A CN115861600B CN 115861600 B CN115861600 B CN 115861600B CN 202211635885 A CN202211635885 A CN 202211635885A CN 115861600 B CN115861600 B CN 115861600B
- Authority
- CN
- China
- Prior art keywords
- hot
- areas
- contour
- area
- determining
- 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
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E30/00—Energy generation of nuclear origin
- Y02E30/30—Nuclear fission reactors
Abstract
The invention relates to a method and a system for identifying a region of interest (ROI) of a Single Photon Emission Computed Tomography (SPECT) image, which belong to the field of medical image processing, and comprise the following steps: acquiring a front bitmap and a rear bitmap of a tester from SPECT imaging equipment; constructing a parallel contour segmentation network and a hot zone segmentation network; extracting peripheral outlines of the front bitmap and the rear bitmap based on the outline segmentation network, and extracting hot areas of the front bitmap and the rear bitmap based on the hot area segmentation network; determining an axis of symmetry based on the peripheral profile; determining candidate comparison areas of the hot zone according to the symmetry axis, wherein the candidate comparison areas comprise a first comparison area and a second comparison area; performing similarity calculation on the candidate comparison areas to obtain a similarity result; and determining the ROI region according to the similarity result to obtain an ROI region identification result based on the SPECT image. The invention can improve the recognition accuracy of the ROI region in the SPECT image.
Description
Technical Field
The invention relates to the field of medical image processing, in particular to a method and a system for identifying a region of interest (ROI) of a Single Photon Emission Computed Tomography (SPECT) image.
Background
Medical imaging technology is one of the most important and fastest growing fields in the modern medical field, and has undergone the development from traditional structural imaging modes such as traditional X-ray imaging, computed tomography (ComputerizedTomography, CT), ultrasonic imaging and the like to functional imaging modes such as single photon emission computed tomography (Singlephotonemission computedtomography, SPECT) and positron emission tomography (PositronEmission Tomography, PET) and the like, and has prompted the development of various mixed imaging modes such as SPECT/CT, PET/CT and the like with higher quality.
However, SPECT has poor imaging quality, and for example, SPECT whole-body bone scan imaging requires that whole-body bone imaging data be stored in a 1024×256 data matrix, with high image resolution and low contrast. Scientists have been dedicated to study how to help improve the efficiency of medical practitioners in medical image analysis work and reduce medical costs through computer technology. Although computer-assisted medical technology has a long history, it has not progressed to the rapid development stage until convolutional neural networks have emerged. The convolution layer in the convolution neural network can automatically extract deep features from the medical image, and through further extraction of the deep features, the convolution neural network can execute a series of medical image auxiliary tasks, wherein the segmentation task can achieve outlining of a region of interest (RegionOfInterest, ROI), and is a basis for ROI extraction, clinical tests, specific tissue measurement and three-dimensional reconstruction.
However, medical science has difficulty in obtaining a large number of usable data sets, and the U-Net proposal enables a deep semantic segmentation model to obtain better results on small data sets. Segmentation of SPECT images is difficult with other imaging modalities due to the large differences between images and the high sensitivity of the model, which can easily manifest non-ROI results in the delineated hot-zone.
Symmetry is an important feature of an object and is also an important component of the human visual perception tissue system. In medical whole-body imaging, symmetry is manifested not only on the peripheral contours, but also on organ and tissue imaging, as well as on the approximate symmetry. Therefore, in the application of medical auxiliary clinic, the symmetry theory has important value, namely, by combining the symmetry of the medical image, doctors can pay more attention to important areas, and the working efficiency of medical image analysis of doctors is improved. In SPECT bone imaging of lung cancer patients, the high-temperature region symmetrical to the spine can be generally considered as normal clinical manifestation, but existing semantic segmentation networks are only sensitive to the hot region (high-absorption region), some of which will be described as ROIs, affecting the actual work. The peripheral outline is the most main symmetrical presentation mode in SPECT whole body imaging, limbs and ribs are orderly arranged on two sides of the vertebra, and the size, the position and the shape are basically the same. By combining the peripheral outline, the method can help us find symmetrical information of the tissue and provide more accurate ROI, but the current mainstream model is difficult to segment out the hot zone and the peripheral curve.
Based on the problems, the existing segmentation network has poor sketching effect on the ROI in the SPECT image, and the method for directly introducing symmetry judgment in the sketching task of the ROI is less.
Disclosure of Invention
The invention aims to provide a method and a system for identifying a region of interest (ROI) of a Single Photon Emission Computed Tomography (SPECT) image, which can improve the identification accuracy of the region of interest (ROI) in the SPECT image.
In order to achieve the above object, the present invention provides the following solutions:
a method of ROI area identification for SPECT images, comprising:
acquiring a front bitmap and a rear bitmap of a tester from SPECT imaging equipment;
constructing a parallel contour segmentation network and a hot zone segmentation network;
extracting peripheral outlines of the front bitmap and the rear bitmap based on the outline segmentation network, and extracting hot areas of the front bitmap and the rear bitmap based on the hot area segmentation network;
determining an axis of symmetry based on the peripheral profile;
determining a candidate comparison area of the hot zone according to the symmetry axis, wherein the candidate comparison area comprises a first comparison area and a second comparison area;
performing similarity calculation on the candidate comparison areas to obtain a similarity result;
and determining the ROI according to the similarity result to obtain an ROI identification result of the SPECT image.
Optionally, the contour segmentation network adopts an active contour model based on a region, and a loss function of the contour segmentation network is as follows:
where ds is the Euclidean length,is the length of curve C, f is the image to be segmented, Ω C is a closed subset of f over the image domain Ω, and the mean values outside and inside curve C are denoted as C, respectively 1 And c 2 Lambda is used to control the regularization process and c 1 、c 2 Parameters of the balance between.
Optionally, the hot zone segmentation network adopts a semantic segmentation network.
Optionally, the determining the symmetry axis based on the peripheral contour specifically includes:
acquiring a contour matrix of the peripheral contour;
determining a contour coincidence region according to the contour matrix;
and determining a symmetry axis according to the contour overlapping region and the overlapping function.
Optionally, the coincidence function is:
wherein w and h respectively represent the width and the height of the contour overlapping region, I ij Representing the intensity value magnitude of position (i, j) in the matrix,for the outline matrix +.>And the outline matrix is a turnover matrix obtained after the turnover of the outline matrix.
Optionally, the determining the candidate contrast area of the hot zone according to the symmetry axis specifically includes:
extracting all the hot areas through threshold segmentation, and recording the extracted hot areas as first hot areas;
acquiring all the hot areas extracted by the hot area segmentation network, turning over the hot areas, and marking the turned hot areas as second hot areas;
judging whether the first hot zone and the second hot zone have overlapping areas or not to obtain a judging result;
when the judgment result is yes, calculating the superposition area of the first hot area and the second hot area;
acquiring the center of a minimum circumscribing circle of the overlapping region and marking the center as a first center;
acquiring the minimum radius of the circumscribed circle of the second hot zone, and marking the minimum radius as a first radius;
and determining a first comparison area according to the first circle center and the first radius.
Optionally, the determining the candidate contrast area of the hot zone according to the symmetry axis specifically includes:
and when the judging result is negative, taking the minimum circumscribed circle of the second hot zone as a second comparison area.
Optionally, the similarity calculation is performed using the following formula:
wherein h is 1 Represents the first hot zone, h 2 The candidate contrast regions are indicated, max (·) and mean (·) represent the maximum and average of the radiation intensities.
Optionally, the determining the ROI area according to the similarity result specifically includes:
and screening candidate comparison areas with the similarity result smaller than or equal to a set threshold value, namely the ROI area.
A ROI area identification system for SPECT images, comprising:
the image acquisition module is used for acquiring a front bitmap and a rear bitmap of the tester from the SPECT imaging equipment;
the network construction module is used for constructing a parallel contour segmentation network and a hot zone segmentation network;
a peripheral contour and hot zone extraction module for extracting peripheral contours of the front bitmap and the rear bitmap based on the contour segmentation network, and extracting hot zones of the front bitmap and the rear bitmap based on the hot zone segmentation network;
a symmetry axis determination module for determining a symmetry axis based on the peripheral contour;
a candidate contrast region determining module configured to determine a candidate contrast region of the hot zone according to the symmetry axis, where the candidate contrast region includes a first contrast region and a second contrast region;
the similarity calculation module is used for calculating the similarity of the candidate comparison areas to obtain a similarity result;
and the ROI region identification module is used for determining the ROI region according to the similarity result to obtain an ROI region identification result of the SPECT image.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention mainly extracts the peripheral outline and the hot zone in the SPECT image by constructing the double-path segmentation frame, and then further improves the outlining effect of the ROI region in the SPECT image by combining the similarity judgment of the bilateral symmetry region.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for identifying the ROI area of a SPECT image according to the present invention;
fig. 2 is a block diagram of an ROI area identification system for SPECT images of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a flow chart of a ROI region identification method of a SPECT image, which can improve the identification precision of the ROI region in the SPECT image.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the invention discloses a ROI area identification method of SPECT images, comprising:
step 101: the front and rear bitmaps of the tester are acquired from the SPECT imaging device.
Specifically, a DICOM format file containing bone scan data of the front and back body positions of a tester is obtained from a SPECT imaging device, and a data matrix of the front bitmap and the back bitmap is derived therefrom.
Further, the step 101 includes:
step 1.1: and (3) performing allergy experiments and other tabu examination on the bone imaging agent on the testers, injecting a certain dose of the bone imaging agent into the patients under the condition of meeting the requirements, collecting the metabolic condition of the radioactive drug in the patients by using a nuclide detector, and obtaining the whole-body bone imaging image and DICOM file of the patients by using a SPECT system.
Step 1.2: because the DICOM file has complex content, the content is analyzed by a program, and bone scan data of front and rear positions of a tester are extracted and stored.
Step 102: and constructing a parallel contour segmentation network and a hot zone segmentation network.
Further, the step 102 includes:
step 2.1: the contour sketching branch adopts an unsupervised method to divide the body contour, the part adopts an active contour model based on areas, the pixel information of the image is regarded as energy, the image dividing problem is skillfully converted into a problem for solving the minimum value of the energy functional, and the loss function is as follows:
wherein the method comprises the steps ofDs is the euclidean length, and,is the length of curve C, f is the image to be segmented, Ω c Is a closed subset of f over the image domain Ω, the mean of the outside and inside of curve C being denoted as C, respectively 1 And c 2 Lambda is used to control the regularization process and c 1 、c 2 Parameters of the balance between.
The curve C gradually approaches the peripheral contour under the drive of the loss function and finally cuts out the desired contour.
Step 2.2: the hotspot segmentation branch segments the hotspots of the image.
The hot zone segmentation branches can adopt common semantic segmentation networks such as U-Net, U-Net++, and the like, and the deep semantic segmentation networks can obtain better performance on a small amount of medical data segmentation.
Step 103: and extracting peripheral outlines of the front bitmap and the rear bitmap based on the outline segmentation network, and extracting hot areas of the front bitmap and the rear bitmap based on the hot area segmentation network.
Further, the extracting the peripheral outline in step 103 includes:
step 3.1: and extracting the outline.
In the contour segmentation result, the value of the body region in the image depends on the intensity value of each part, the background filling part is 0, and the contour is represented by 1. For this case, a closed contour is found using an algorithm. In the matrix, the contour may be represented as a set of points, and the closed contour extracted from the result may be represented as a list containing m-term (n, 1, 2) matrices, where m represents the number of closed contours present in the result and n is the number of points contained in a contour. There are also some noise profiles that may be affected in the unsupervised results, so the matrix with the largest n value is selected from among them to extract the profile point matrix.
Step 3.2: and constructing a contour matrix. And (3) creating a full zero matrix (256), traversing the matrix obtained in the step (3.1), and setting the data of the corresponding position in the full zero matrix to be 1 to obtain the contour matrix.
Step 104: an axis of symmetry is determined based on the peripheral profile.
Further, the step 104 finds the symmetry axis based on the contour matrix extracted in the step 3.2, which is essentially a process of finding the symmetry axis maximizing the degree of the left and right side overlapping. For a candidate symmetry axis, we describe its similarity by measuring the degree of coincidence of the profile matrix o with its flip matrix f, the specific coincidence function is as follows:
wherein w and h respectively represent the width and the height of the contour overlapping region, I ij Representing the intensity value magnitude of position (i, j) in the matrix,for the outline matrix +.>And the outline matrix is a turnover matrix obtained after the turnover of the outline matrix.
Since there are a large number of candidate symmetry axes, we use a three-stage algorithm to search to maximize the degree of coincidence of the two sides, where each stage depends on the optimal solution of the previous stage.
Further, the algorithm used in this process includes the steps of:
step 4.1: and determining the contour coincidence range. In the SPECT imaging process, the condition that the placement of the parts such as arms is irregular possibly affects the selection of the subsequent symmetry axis, so that the main imaging areas such as the chest, the abdomen and the like are adopted for determining the contour coincidence in the process, and the method comprises the following steps:
step 4.1.1: an upper bound and a lower bound of the image are determined. The main target area is approximately at the center of the matrix (256), and the non-zero point count is performed by starting with the rows on both sides with x=128 as an initial boundary line, and the rows with the largest non-zero point numbers on both sides are respectively used as an upper boundary and a lower boundary.
Step 4.1.2: and (3) extracting a coincident comparison region, wherein the lower bound obtained in the step 4.1.1 is taken as a starting point, and 60% of the distance between the upper bound and the lower bound is taken as a distance extraction main region.
Step 4.2: since the patient is mostly in a vertical state in the imaging result, and few deviations exist in left and right angles, the optimal points (0, x 1) and (255, x 2) can be directly found on x=0 and x=255 to determine a symmetry axis, so the following three-stage method is constructed to find the symmetry axis:
step 4.2.1: generating a set P1= { [ (0, x1), (255, x2) ]|x1, x2 ε [0,255], x1, x2=8k, k ε N }, and obtaining a combination [ (0, best_x1), (255, best_x2) ] in the set P1, which maximizes the coincidence function S.
Step 4.2.2: generating a set P2= { [ (0, x1), (255, x2) ] |x1∈ [ best_x1-4, best_x1+4], x2∈ [ best_x2-4, best_x2+4], x1, x2=k, k∈n }, and obtaining a combination [ (0, best_x1), (255, best_x2) ] in the set P2, which maximizes the coincidence function S.
Step 4.2.3: generating a set P3= { [ (0, x1), (255, x2) ] |x1 epsilon [ best_x1-2, best_x1+2], x2 epsilon [ best_x2-2, best_x2+2], x1, x2=0.1k, k epsilon N }, and obtaining a combination [ (0, best_x1), (255, best_x2) ] in the set P3, wherein the maximum coincidence function S is obtained by the combination of two points, namely, a symmetry axis obtained by final search.
Step 105: and determining candidate comparison areas of the hot zone according to the symmetry axis, wherein the candidate comparison areas comprise a first comparison area and a second comparison area.
In particular, since the hotspots are not necessarily completely symmetrical, preparation for the next symmetrical similarity calculation may be made by the method of generating candidate regions.
Observing the segmentation results of the segmentation network, wherein the non-ROI hotspots are mainly in two cases: (1) both sides are characterized as hot zones; (2) because the individual's variability predicts non-hotspots as ROIs. Respective candidate contrast regions are generated differently for the two cases.
Further, the step 105 includes:
step 5.1: for the case where the symmetrical portion contains a hot zone, in order to make the contrast area relatively accurate, respectively: (1) extracting all possible hot areas by a threshold segmentation method, and marking the extracted possible hot areas as first hot areas; (2) turning over the prediction result of the hot zone segmentation network according to the symmetry axis to obtain a second hot zone; (3) calculating the superposition area of the second hot area and the first hot area; (4) taking the circle center of the minimum circumscribing circle of each overlapping area as a first circle center, and the radius of the minimum circumscribing circle of the second hot area corresponding to the overlapping area as a first radius to generate a circular first comparison area;
step 5.2: for the ROI of the non-hot zone, it is difficult to find a suitable relative region, so by selecting the center of the second hot zone minimum circumcircle in the symmetrical region of the ROI as the second center, the radius of the second hot zone minimum circumcircle is the second radius, and the second contrast region is generated.
Step 106: and carrying out similarity calculation on the candidate comparison areas to obtain a similarity result.
Specifically, step 106 uses TBR to calculate the similarity of each group of symmetric regions using the following formula:
wherein h is 1 Represents the first hot zone, h 2 Representing candidate contrast regions, h 1 And h 2 For a set of symmetric regions, max (·) and mean (·) represent the maximum versus average function.
Step 107: and determining the ROI region according to the similarity result to obtain an ROI region identification result based on the SPECT image.
In step 107, the similarity of the two regions is determined by using the value 2.2 as a threshold, the contrast region with a similarity value greater than 2.2 is considered as a non-similar region, and the candidate contrast region with a similarity value less than or equal to 2.2 is left as the ROI region.
Based on the above method, the invention also discloses a ROI area identification system of SPECT image, as shown in fig. 2, comprising:
an image acquisition module 201 is used to acquire the front and rear bitmaps of the tester from the SPECT imaging device.
The network construction module 202 is configured to construct a parallel contour segmentation network and a hot zone segmentation network.
And the peripheral outline and hot zone extraction module 203 is configured to extract peripheral outlines of the front bitmap and the rear bitmap based on the outline segmentation network, and extract hot zones of the front bitmap and the rear bitmap based on the hot zone segmentation network.
An axis of symmetry determination module 204 for determining an axis of symmetry based on the peripheral profile.
A candidate contrast region determination module 205 is configured to determine a candidate contrast region of the hot zone according to the symmetry axis, where the candidate contrast region includes a first contrast region and a second contrast region.
And the similarity calculation module 206 is configured to perform similarity calculation on the candidate comparison area to obtain a similarity result.
And the ROI area identification module 207 is configured to determine an ROI area according to the similarity result, and obtain an ROI area identification result based on the SPECT image.
The invention also discloses the following technical effects:
the invention reduces the occurrence of non-ROI in the depth semantic segmentation network based on the symmetry information so as to improve the accuracy of the result.
The contour segmentation network and the hot zone segmentation network can simultaneously obtain the peripheral contour and the hot zone, so as to define the human body curve; the contour extraction and symmetry axis search module extracts a trunk area from a contour segmentation result to serve as a definition area for symmetric information search, so that influence caused by gestures is effectively prevented, and calculation amount can be reduced; the comparison region generation and similarity judgment module can reduce the occurrence of non-ROI regions in the result by carrying out similarity calculation on the symmetrical hot regions.
In conclusion, the method and the device solve the problem that the SPECT image is too high in sensitivity to cause excessive non-ROI in the semantic segmentation model result, and can improve the model precision.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (6)
1. A method for identifying a ROI area of a SPECT image, comprising:
acquiring a front bitmap and a rear bitmap of a tester from SPECT imaging equipment;
constructing a parallel contour segmentation network and a hot zone segmentation network;
extracting peripheral outlines of the front bitmap and the rear bitmap based on the outline segmentation network, and extracting hot areas of the front bitmap and the rear bitmap based on the hot area segmentation network;
determining an axis of symmetry based on the peripheral profile;
determining a candidate comparison area of the hot zone according to the symmetry axis, wherein the candidate comparison area comprises a first comparison area and a second comparison area;
performing similarity calculation on the candidate comparison areas to obtain a similarity result;
determining an ROI region according to the similarity result to obtain an ROI region identification result based on the SPECT image;
the determining the symmetry axis based on the peripheral outline specifically comprises the following steps:
acquiring a contour matrix of the peripheral contour;
determining a contour coincidence region according to the contour matrix;
determining a symmetry axis according to the contour overlapping region and the overlapping function; the coincidence function is:
wherein w and h respectively represent the width and the height of the contour overlapping region, I ij Representing the intensity value magnitude of position (i, j) in the matrix,for the outline matrix +.>A turnover matrix is obtained after the profile matrix is turned over;
determining candidate contrast areas of the hot zone according to the symmetry axis, wherein the candidate contrast areas comprise:
extracting all the hot areas through threshold segmentation, and recording the extracted hot areas as first hot areas;
acquiring all the hot areas extracted by the hot area segmentation network, turning over the hot areas, and marking the turned hot areas as second hot areas;
judging whether the first hot zone and the second hot zone have overlapping areas or not to obtain a judging result;
when the judgment result is yes, calculating the superposition area of the first hot area and the second hot area;
acquiring the center of a minimum circumscribing circle of the overlapping region and marking the center as a first center;
acquiring the minimum radius of the circumscribed circle of the second hot zone, and marking the minimum radius as a first radius;
determining a first comparison area according to the first circle center and the first radius;
and when the judging result is negative, taking the minimum circumscribed circle of the second hot zone as a second comparison area.
2. The SPECT image ROI region identification method of claim 1 wherein the contour segmentation network employs an active contour model based on regions, the loss function of the contour segmentation network being:
where ds is Euclidean length, +. 0 Length(C) ds is the length of curve C, f is the image to be segmented, Ω C is a closed subset of f over the image domain Ω, and the mean values outside and inside curve C are denoted as C, respectively 1 And c 2 Lambda is used to control the regularization process and c 1 、c 2 Parameters of the balance between.
3. The SPECT image ROI region identification method of claim 1 wherein the hot zone segmentation network employs a semantic segmentation network.
4. The method of ROI area identification of SPECT images of claim 1 wherein the similarity calculation is performed using the formula:
wherein h is 1 Represents the first hot zone, h 2 The candidate contrast regions are indicated, max (·) and mean (·) represent the maximum and average of the radiation intensities.
5. The method for identifying the ROI area of the SPECT image according to claim 1, wherein the determining the ROI area according to the similarity result specifically comprises:
and screening candidate comparison areas with the similarity result smaller than or equal to a set threshold value, namely the ROI area.
6. A ROI area identification system of a SPECT image, comprising:
the image acquisition module is used for acquiring a front bitmap and a rear bitmap of the tester from the SPECT imaging equipment;
the network construction module is used for constructing a parallel contour segmentation network and a hot zone segmentation network;
a peripheral contour and hot zone extraction module for extracting peripheral contours of the front bitmap and the rear bitmap based on the contour segmentation network, and extracting hot zones of the front bitmap and the rear bitmap based on the hot zone segmentation network;
a symmetry axis determination module for determining a symmetry axis based on the peripheral contour;
a candidate contrast region determining module configured to determine a candidate contrast region of the hot zone according to the symmetry axis, where the candidate contrast region includes a first contrast region and a second contrast region;
the similarity calculation module is used for calculating the similarity of the candidate comparison areas to obtain a similarity result;
the ROI region identification module is used for determining an ROI region according to the similarity result to obtain an ROI region identification result based on the SPECT image;
the determining the symmetry axis based on the peripheral outline specifically comprises the following steps:
acquiring a contour matrix of the peripheral contour;
determining a contour coincidence region according to the contour matrix;
determining a symmetry axis according to the contour overlapping region and the overlapping function; the coincidence function is:
wherein w and h respectively represent the width and height of the contour overlapping regionDegree, I ij Representing the intensity value magnitude of position (i, j) in the matrix,for the outline matrix +.>A turnover matrix is obtained after the profile matrix is turned over;
determining candidate contrast areas of the hot zone according to the symmetry axis, wherein the candidate contrast areas comprise:
extracting all the hot areas through threshold segmentation, and recording the extracted hot areas as first hot areas;
acquiring all the hot areas extracted by the hot area segmentation network, turning over the hot areas, and marking the turned hot areas as second hot areas;
judging whether the first hot zone and the second hot zone have overlapping areas or not to obtain a judging result;
when the judgment result is yes, calculating the superposition area of the first hot area and the second hot area;
acquiring the center of a minimum circumscribing circle of the overlapping region and marking the center as a first center;
acquiring the minimum radius of the circumscribed circle of the second hot zone, and marking the minimum radius as a first radius;
determining a first comparison area according to the first circle center and the first radius;
and when the judging result is negative, taking the minimum circumscribed circle of the second hot zone as a second comparison area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211635885.3A CN115861600B (en) | 2022-12-20 | 2022-12-20 | ROI (region of interest) area identification method and system for SPECT (Single photon emission computed tomography) image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211635885.3A CN115861600B (en) | 2022-12-20 | 2022-12-20 | ROI (region of interest) area identification method and system for SPECT (Single photon emission computed tomography) image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115861600A CN115861600A (en) | 2023-03-28 |
| CN115861600B true CN115861600B (en) | 2023-09-05 |
Family
ID=85674248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211635885.3A Active CN115861600B (en) | 2022-12-20 | 2022-12-20 | ROI (region of interest) area identification method and system for SPECT (Single photon emission computed tomography) image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115861600B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921860A (en) * | 2018-07-10 | 2018-11-30 | 北京大学 | A kind of prostate magnetic resonance image full-automatic partition method |
| CN109064476A (en) * | 2018-07-24 | 2018-12-21 | 西安电子科技大学 | A kind of CT rabat lung tissue image partition method based on level set |
| CN112598070A (en) * | 2020-12-25 | 2021-04-02 | 创新奇智(广州)科技有限公司 | Target detection method and device, electronic equipment and storage medium |
| CN114399493A (en) * | 2022-01-14 | 2022-04-26 | 汕头市超声仪器研究所股份有限公司 | Automatic detection and display method for ultrasonic brain abnormal area |
| CN115239716A (en) * | 2022-09-22 | 2022-10-25 | 杭州影想未来科技有限公司 | Medical image segmentation method based on shape prior U-Net |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112055555A (en) * | 2018-05-14 | 2020-12-08 | 富士胶片株式会社 | Medical image processing apparatus, medical image processing method, and endoscope system |
| KR102078876B1 (en) * | 2019-09-03 | 2020-02-20 | 주식회사 루닛 | Method and system for detecting pneumothorax |
-
2022
- 2022-12-20 CN CN202211635885.3A patent/CN115861600B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108921860A (en) * | 2018-07-10 | 2018-11-30 | 北京大学 | A kind of prostate magnetic resonance image full-automatic partition method |
| CN109064476A (en) * | 2018-07-24 | 2018-12-21 | 西安电子科技大学 | A kind of CT rabat lung tissue image partition method based on level set |
| CN112598070A (en) * | 2020-12-25 | 2021-04-02 | 创新奇智(广州)科技有限公司 | Target detection method and device, electronic equipment and storage medium |
| CN114399493A (en) * | 2022-01-14 | 2022-04-26 | 汕头市超声仪器研究所股份有限公司 | Automatic detection and display method for ultrasonic brain abnormal area |
| CN115239716A (en) * | 2022-09-22 | 2022-10-25 | 杭州影想未来科技有限公司 | Medical image segmentation method based on shape prior U-Net |
Non-Patent Citations (1)
| Title |
|---|
| 颜俊哲等.基于对称响应滤波和分水岭的脊椎图像分割方法.《湘南学院学报》.2021,第42卷(第5期),第120-125页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115861600A (en) | 2023-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223352B (en) | Medical image scanning automatic positioning method based on deep learning | |
| US20210401392A1 (en) | Deep convolutional neural networks for tumor segmentation with positron emission tomography | |
| US11869202B2 (en) | Method and system for processing multi-modality image | |
| CN107798682B (en) | Image segmentation system, method, apparatus and computer-readable storage medium | |
| CN107545584B (en) | Method, device and system for positioning region of interest in medical image | |
| CN107622492B (en) | Lung fissure segmentation method and system | |
| CN106157320B (en) | A kind of image blood vessel segmentation method and device | |
| CN112508965B (en) | Automatic outline sketching system for normal organs in medical image | |
| US9135695B2 (en) | Method for creating attenuation correction maps for PET image reconstruction | |
| CN107203741B (en) | Blood vessel extraction method, device and system | |
| US8290568B2 (en) | Method for determining a property map of an object, particularly of a living being, based on at least a first image, particularly a magnetic resonance image | |
| CN109493325A (en) | Tumor Heterogeneity analysis system based on CT images | |
| CN107451983A (en) | The three-dimensional fusion method and system of CT images | |
| WO2020001086A1 (en) | Deep learning-based cardiac ct image segmentation method and apparatus, device, and medium | |
| US11315254B2 (en) | Method and device for stratified image segmentation | |
| WO2016033458A1 (en) | Restoring image quality of reduced radiotracer dose positron emission tomography (pet) images using combined pet and magnetic resonance (mr) | |
| CN110874860A (en) | Target extraction method of symmetric supervision model based on mixed loss function | |
| Baydoun et al. | Dixon-based thorax synthetic CT generation using Generative Adversarial Network | |
| CN113989231A (en) | Method and device for determining kinetic parameters, computer equipment and storage medium | |
| CN114445429A (en) | Whole-heart ct segmentation method and device based on multiple labels and multiple decoders | |
| Zhang et al. | Automatic parotid gland segmentation in MVCT using deep convolutional neural networks | |
| CN115861600B (en) | ROI (region of interest) area identification method and system for SPECT (Single photon emission computed tomography) image | |
| CN113822298A (en) | Time activity curve extraction method and system | |
| EP3588378B1 (en) | Method for determining at least one enhanced object feature of an object of interest | |
| Tan et al. | A segmentation method of lung parenchyma from chest CT images based on dual U-Net |
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 |