CN115311258A - Method and system for automatically segmenting organs in SPECT (single photon emission computed tomography) plane image - Google Patents
Method and system for automatically segmenting organs in SPECT (single photon emission computed tomography) plane image Download PDFInfo
- Publication number
- CN115311258A CN115311258A CN202211123639.XA CN202211123639A CN115311258A CN 115311258 A CN115311258 A CN 115311258A CN 202211123639 A CN202211123639 A CN 202211123639A CN 115311258 A CN115311258 A CN 115311258A
- Authority
- CN
- China
- Prior art keywords
- organ
- spect
- dimensional
- image
- template
- 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.)
- Granted
Links
- 210000000056 organ Anatomy 0.000 title claims abstract description 246
- 238000002603 single-photon emission computed tomography Methods 0.000 title claims abstract description 126
- 238000000034 method Methods 0.000 title claims abstract description 46
- 230000011218 segmentation Effects 0.000 claims abstract description 45
- 238000001514 detection method Methods 0.000 claims abstract description 44
- 230000004927 fusion Effects 0.000 claims abstract description 18
- 239000013598 vector Substances 0.000 claims description 32
- 230000006870 function Effects 0.000 claims description 18
- 230000009466 transformation Effects 0.000 claims description 14
- 238000013135 deep learning Methods 0.000 claims description 9
- 238000013519 translation Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 229940121896 radiopharmaceutical Drugs 0.000 abstract description 6
- 239000012217 radiopharmaceutical Substances 0.000 abstract description 6
- 230000002799 radiopharmaceutical effect Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000011478 gradient descent method Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000024924 glomerular filtration Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
- 238000003325 tomography Methods 0.000 description 1
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
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/44—Analysis of texture based on statistical description of texture using image operators, e.g. filters, edge density metrics or local histograms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10108—Single photon emission computed tomography [SPECT]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Nuclear Medicine (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method and a system for automatically segmenting organs in a SPECT plane image, wherein the method comprises the following steps: detecting an organ with high degree of distinction from the background in a SPECT plane image of a detected person, determining a detection frame corresponding to the organ, and obtaining the central coordinate of the detection frame of the organ; the method comprises the steps of segmenting a target organ from three-dimensional SPECT/CT fusion tomographic data of the same examinee to obtain a three-dimensional organ template picture, and projecting the three-dimensional organ template picture to a two-dimensional plane in the front-back direction of a human body to generate a two-dimensional organ template picture; translating the two-dimensional organ template picture to align the geometric centers of a plurality of organs in the two-dimensional organ template picture with the center of the detection frame obtained in the step A; and registering the two-dimensional organ template image and the SPECT plane image, so that all organs in the two-dimensional organ template image correspond to organ areas in the SPECT plane image, and completing segmentation. The invention realizes full-automatic accurate segmentation of organs in the SPECT plane image, does not need manual participation, and can be suitable for the SPECT plane images of different radiopharmaceuticals.
Description
Technical Field
The invention relates to the technical field of medical imaging, in particular to a method and a system for automatically segmenting organs in a SPECT (single photon emission computed tomography) plane image.
Background
Acquiring planar images (anterior, posterior, or both) of a subject using dual-probe SPECT is the primary imaging modality of current clinical mainstream SPECT for diagnosis of various types of diseases and assessment of organ function. In some applications, such as pharmacokinetic studies of new radiopharmaceuticals, glomerular filtration rate measurements, etc., it is necessary to acquire planar images at multiple time points and analyze the time-dependent changes in drug uptake of certain target organs, and therefore it is necessary to segment the corresponding region of the relevant organ (two-dimensional projection) in the SPECT planar image.
At present, the common method in clinical practice is to manually delineate organ outlines to realize segmentation or semi-automatic segmentation combining manual operation and an algorithm, and the main limitations are workload of manual delineation and delineation errors caused by personnel fatigue and subjective difference. The main difficulty of the full-automatic delineation algorithm is that the SPECT plane image is the superposition of the three-dimensional space distribution of the radiopharmaceutical in two dimensions, so that the contrast of certain organs and peripheral tissues in the image is low and the edges are blurred, and great difficulty is brought to a plurality of segmentation algorithms based on rules. And the adoption of data-based methods such as deep learning and the like requires a large amount of accurate sketching data for training, and the corresponding data is difficult to obtain due to the difficulty of manual sketching and the characteristic of strong specialty in the field.
Disclosure of Invention
The invention aims to provide a method and a system for automatically segmenting organs in a SPECT (single photon emission computed tomography) plane image, which are used for segmenting the organs in the SPECT plane image based on the organ segmentation result of three-dimensional SPECT/CT fusion tomography data of the same examinee, so that the organs in the SPECT plane image can be segmented automatically and accurately without manual participation, and the method and the system can be suitable for SPECT plane images of different radiopharmaceuticals.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method of automatically segmenting an organ in a SPECT planar image comprising the steps of:
A. detecting an organ with high degree of difference from the background in a SPECT plane image of a detected person, determining a detection frame corresponding to the organ, and obtaining the central coordinate of the detection frame of the organ;
B. dividing a target organ from three-dimensional SPECT/CT fusion fault data of the same examinee to obtain a three-dimensional organ template map, and projecting the three-dimensional organ template map to a two-dimensional plane in the anteroposterior direction of a human body to generate a two-dimensional organ template map;
C. translating the two-dimensional organ template picture to align the geometric centers of a plurality of organs in the two-dimensional organ template picture with the center of the detection frame obtained in the step A;
D. and registering the two-dimensional organ template image and the SPECT plane image, so that all organs in the two-dimensional organ template image correspond to organ areas in the SPECT plane image to finish segmentation.
Further, in the step a, a deep learning algorithm is used for detecting the SPECT planar image, and a detection frame center coordinate [ xcenter, ycenter ] of an organ which is greatly different from the background is output;
the step of detecting the SPECT plane image by the deep learning algorithm comprises the following steps:
s101: taking ResNet50 as a basic feature extraction network of the detection network, and extracting features to obtain a feature map;
s102: the features extracted in the step S101 are up-sampled, and the size of the features is returned to 1/N of the resolution of the original image;
s103: acquiring a heatmap of the data sampled in the S102 through the sub-network 1, extracting a central position coordinate of the target to be detected through a heatmap extreme point position, predicting the sub-network by utilizing position offset, and acquiring a central point offset;
s104: and correcting the coordinates of the center point of the target through the center point offset acquired in the step S103, so as to acquire the position of the center point of the target to be detected finally.
Further, in the step B, in the three-dimensional SPECT/CT fusion tomographic data of the same subject, pixels belonging to the same organ are set to the same encoding value, different organs correspond to different encoding values, and pixels not belonging to any organ are encoded to be 0;
and projecting the three-dimensional template picture to a two-dimensional plane in the front-back direction of the human body, and enabling pixel values on the two-dimensional plane to correspond to the coding values or 0 of each organ, so that the two-dimensional organ template picture can be generated.
Further, in the step B, a unet + + segmentation network is used to segment the main organs in the three-dimensional SPECT/CT fusion tomographic data, including the following steps:
s201: taking each layer of the CT tomogram as data, and obtaining a segmentation result of the layer through a unet + + segmentation network;
s202: putting all the segmentation results of each layer of data of the CT tomographic image together, and integrating the three-dimensional tomographic segmentation results of each organ through connected domain analysis;
s203: and optimizing the S202 three-dimensional fault segmentation result by using the contrast information of the SPECT fault image to obtain a three-dimensional organ template map.
Further, in the step B, the step of obtaining the three-dimensional organ template map is as follows:
s301: extracting features of the SPECT tomographic image and the CT tomographic image respectively to obtain feature maps;
s302: enabling the feature maps of the two modes to enter an SPP network, pooling the feature maps by the SPP network to generate fixed-length output, namely featuremap, and linking the featuremaps of the two modes to enable the featuremaps with the same size of the two modes to be spliced according to channel dimensions to obtain a spliced multi-modal result;
s303: and decoding the spliced multi-modal result into a segmentation result through sampling on N layers to obtain the three-dimensional organ template picture.
Further, the step C includes the following sub-steps:
defining the coordinates of the center of the detection frame of the organ in the SPECT plane image as
Wherein i represents the number of the organ;
defining the geometric center coordinate of the organ in the two-dimensional organ template picture as (x) i ，y i )，i＝1，2，..，I；
Finding the optimal translation vector (x) of the two-dimensional organ template map by the following formula 0 ，y 0 ) And minimizing the sum of squared distances between the geometric center of the organ in the two-dimensional organ template map and the center of the detection frame of the organ in the SPECT plane image:
solving the above formula yields:
further, in the step D, the two-dimensional organ template image and the SPECT planar image are integrally registered based on an image registration method of mutual information cost function or B-spline non-rigid transformation.
Further, the image registration method of the B-spline non-rigid transformation is as follows:
defining SPECT plane image vector as P, organ template image vector as T, organ template image vector after B spline conversion as B (T, s) 1 ，s 2 ，...，s n), wherein s1 ，s 2 ，...，s n Representing a set of parameters of the B-spline transformation, the minimization problem can be expressed as the following equation:
(s 1 ，s 2 ，...，s n ) opt ＝argmin{-I(P，B(T，s 1 ，s 2 ，...，s n ))} (3)
i.e. find an optimized set of B-spline transformation parameters such that P and B (T, s) 1 ，s 2 ，...，s n ) Mutual information of I (P, B (T, s) 1 ，s 2 ，...，s n ) Maximum, define p i And b i Respectively represent P and B (T, s) 1 ，s 2 ，...，s n ) The value of the ith pixel of (2) is defined as follows:
wherein Prob (p) i ，b i ) Representing a joint probability function, prob (p) i ) The probability function of the distribution of pixel values, prob (b), representing the image P i ) Representative image B (T, s) 1 ，s 2 ，...，s n ) The probability function is distributed over the pixel values.
Further, the image registration method based on the mutual information cost function is as follows:
defining the SPECT plane image vector as P and the corresponding template image vector with organ number as i as T i ，T i The pixel value of the region corresponding to the organ number i in the image is i, the pixel values of other regions are 0, and the rigid-transformed organ i template image vector is
wherein 
Relative to the geometric center (x) of the organ i ，y i ) A two-bit vector of the translation is,
for the angle at which the organ is rotationally transformed around the translated geometric center, the above registration problem is equivalent to the following optimization problem:
A system for automatically segmenting an organ in a SPECT planar image, the system employing the method for automatically segmenting an organ in a SPECT planar image described above, the system comprising:
the detection frame acquisition module is used for detecting an organ with high degree of distinction from the background in a SPECT plane image of a detected person, determining a detection frame corresponding to the organ and obtaining the center coordinates of the detection frame of the organ;
the segmentation module is used for segmenting a target organ from three-dimensional SPECT/CT fusion tomographic data of the same examinee to obtain a three-dimensional organ template map, and projecting the three-dimensional organ template map to a two-dimensional plane in the anteroposterior direction of a human body to generate a two-dimensional organ template map;
an alignment module, which is used for translating the two-dimensional organ template drawing to align the geometric centers of a plurality of organs in the two-dimensional organ template drawing with the center of the detection frame obtained in the step A;
a registration module to register the two-dimensional organ template map with the SPECT planar image such that all organs in the two-dimensional organ template map correspond to organ regions in the SPECT planar image.
The technical scheme provided by the invention can have the following beneficial effects:
1. the method of the invention is based on the organ segmentation result of the three-dimensional SPECT/CT fusion fault data of the same examinee, segments the SPECT plane image, realizes full-automatic accurate segmentation of the organ in the SPECT plane image, does not need manual participation, and improves the accuracy of organ segmentation;
2. aiming at the condition that the organ overlap and shelter are serious in the SPECT plane image, the method does not need a deep learning method for training by adopting a large amount of accurate delineation data, is low in difficulty, and can also be suitable for the SPECT plane images of different radiopharmaceuticals.
Drawings
FIG. 1 is a flow chart of a method of automatically segmenting an organ in a SPECT planar image in accordance with one embodiment of the present invention;
FIG. 2 is a flow chart for deriving coordinates of a center of a detection box corresponding to an organ in a SPECT plane image according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an organ correspondence detection box in a SPECT planar image according to an embodiment of the present invention;
FIG. 4 is a flow chart of the segmentation of a target organ from three-dimensional SPECT/CT fusion tomographic data in accordance with an embodiment of the present invention;
FIG. 5 is an image of the same portion of a CT tomogram and a three-dimensional template according to an embodiment of the present invention, FIG. 5A is a CT tomogram, and FIG. 5B is a three-dimensional template;
fig. 6 is a schematic diagram of registering a two-dimensional organ template map with a SPECT plane image, fig. 6A is a two-dimensional organ template map, fig. 6B is a SPECT plane image (geometric mean map), and fig. 6C is a schematic diagram of a two-dimensional organ template map after registration with a SPECT plane image, in one embodiment of the invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
A method and system for automatically segmenting an organ in a SPECT planar image according to embodiments of the present invention are described below with reference to fig. 1 to 6.
As shown in fig. 1, an embodiment of the present invention provides a method for automatically segmenting an organ in a SPECT planar image, including the following steps:
A. detecting an organ with high degree of distinction from the background in a SPECT plane image of a detected person, determining a detection frame corresponding to the organ, and obtaining the central coordinate of the detection frame of the organ;
B. the method comprises the steps of segmenting a target organ from three-dimensional SPECT/CT fusion tomographic data of the same examinee to obtain a three-dimensional organ template picture, and projecting the three-dimensional organ template picture to a two-dimensional plane in the front-back direction of a human body to generate a two-dimensional organ template picture;
C. translating the two-dimensional organ template picture to align the geometric centers of a plurality of organs in the two-dimensional organ template picture with the center of the detection frame obtained in the step A;
D. and registering the two-dimensional organ template image and the SPECT plane image, so that all organs in the two-dimensional organ template image correspond to organ areas in the SPECT plane image, and completing segmentation.
According to the method, an organ with high degree of distinction from the background in the SPECT plane image is detected, then a detection frame of the organ is obtained, the central coordinate of the detection frame is calculated, then a three-dimensional template image is obtained from three-dimensional SPECT/CT fusion fault data of the same examinee, then a two-dimensional organ template image is generated, the organ in the two-dimensional organ template image is aligned with the center of the detection frame in the step A, then image registration is carried out, and organ segmentation in the SPECT plane image can be completed. Therefore, the method provided by the invention is used for segmenting the SPECT plane image based on the organ segmentation result of the three-dimensional SPECT/CT fusion fault data of the same examinee, so that the organ in the SPECT plane image is segmented automatically and accurately without manual participation, and the organ segmentation accuracy is improved; in addition, aiming at the condition that the organ overlap and shelter are serious in the SPECT plane image, the method does not need to adopt a large amount of deep learning method for training by accurately delineating data, is low in difficulty, and can be suitable for the SPECT plane images of different radiopharmaceuticals.
In step a, a SPECT plane image is detected by a deep learning algorithm, and a central coordinate [ xcenter, ycenter ] of a detection frame of an organ greatly different from a background is output;
as shown in fig. 2, the step of detecting the SPECT planar image with the deep learning algorithm includes:
s101: taking ResNet50 as a basic Feature extraction network of the detection network, and performing Feature extraction to obtain a Feature Map (Feature Map);
s102: the features extracted in step S101 are up-sampled (Conv 2D _ transit deconvolution) and the size thereof is returned to 1/N of the original image resolution, preferably, 1/N is set to 1/4 or 1/2.
S103: for the data sampled in S102, a heatmap is obtained through the subnetwork 1 (Conv combination network), the coordinates of the Center position of the target to be detected are extracted through the extreme point position of the heatmap, and the Center offset (Center offset) is obtained by using the position offset prediction subnetwork;
s104: the center point coordinates of the target are corrected by the center point offset obtained in S103, and the center point position (peak as object center) of the target to be detected can be finally obtained.
The detection frame generation result chart in the step a is shown in fig. 3, and the detection frame generation algorithm is simple. The SPECT planar image in step a is a geometric mean image of the anterior image, a geometric mean image of the posterior image, or a weighted arithmetic mean and geometric mean image of the anterior and posterior bits.
In step B, in the three-dimensional SPECT/CT fusion tomographic data of the same subject, pixels belonging to the same organ are set to the same encoding value, different organs correspond to different encoding values, and pixels not belonging to any organ are encoded to be 0; and projecting the three-dimensional template picture to a two-dimensional plane in the front-back direction of the human body, and enabling pixel values on the two-dimensional plane to correspond to the coding values or 0 of each organ, so that the two-dimensional organ template picture can be generated.
Specifically, in the step B, a unet + + segmentation network is used to segment the main organs in the three-dimensional SPECT/CT fusion tomographic data, as shown in fig. 4, including the following steps:
s201: firstly, preprocessing a CT tomogram, namely windowing transformation and normalization of pixel values, and then taking each layer of the CT tomogram as data to obtain a layer segmentation result through a unet + + segmentation network;
s202: putting all the segmentation results of each layer of data of the CT tomographic image together, and integrating the three-dimensional tomographic segmentation results of each organ through connected domain analysis;
s203: optimizing the S202 three-dimensional fault segmentation result by using the contrast information of the SPECT fault image to obtain a three-dimensional organ template map, wherein the optimization method specifically comprises the following steps: and in the segmentation region corresponding to the SPECT image organ, calculating a SPECT image pixel value statistical histogram, performing secondary confirmation on the pixel values at the edge of the organ segmentation region based on the SPECT image pixel value statistical histogram, and giving an organ code value to the final segmentation result to obtain a three-dimensional organ template image.
In another embodiment, in step B, the step of obtaining the three-dimensional organ template map is as follows:
s301: extracting features of the SPECT tomographic image and the CT tomographic image respectively to obtain feature maps;
s302: enabling the feature maps of the two modes to enter an SPP network, pooling the feature maps by the SPP network to generate fixed-length output, namely, featuremaps, and linking the featuremaps of the two modes to enable the featuremaps with the same size of the two modes to be spliced according to channel dimensions to obtain a spliced multi-mode result;
s303: and decoding the spliced multi-modal result into a segmentation result through sampling on the N layers to obtain the three-dimensional organ template picture.
It should be noted that, in the algorithm for projecting the three-dimensional template onto the two-dimensional plane in step B, for a plurality of organs that overlap and are shielded in the three-dimensional space, the organ number with a higher pixel value in the SPECT tomographic image is selected as the number of the pixel corresponding to the final two-dimensional projection template.
More specifically, the step C of aligning the geometric center of the organ of the two-dimensional organ template map and the center of the SPECT plane atlas organ detection box is implemented by an algorithm for seeking the square sum of the distances of the corresponding organ centers to be minimized, and specifically, the step C comprises the following sub-steps:
defining the coordinates of the center of the detection frame of the organ in the SPECT plane image as
Wherein i represents the number of the organ;
defining the geometric center coordinate of the organ in the two-dimensional organ template picture as (x) i ，y i )，i＝1，2，..，I；
Finding the optimal translation vector (x) of the two-dimensional organ template map by the following formula 0 ，y 0 ) And minimizing the sum of squared distances between the geometric center of the organ in the two-dimensional organ template map and the center of the detection frame of the organ in the SPECT plane image:
solving the above formula yields:
in step D, the two-dimensional organ template map and the SPECT planar image are integrally registered by using an image registration method based on mutual information cost function or B-spline non-rigid transformation.
Specifically, the image registration method of B-spline non-rigid transformation is as follows:
defining SPECT plane image vector as P, organ template image vector as T, organ template image vector after B spline conversion as B (T, s) 1 ，s 2 ，...，s n), wherein s1 ，s 2 ，...，s n Representing a set of parameters of the B-spline transformation, the minimization problem can be expressed as the following equation:
(s 1 ，s 2 ，...，s n ) opt ＝argmin{-I(P，B(T，s 1 ，s 2 ，...，s n ))} (3)
i.e. find an optimized set of B-spline transformation parameters such that P and B (T, s) 1 ，s 2 ，...，s n ) Mutual information of I (P, B (T, s) 1 ，s 2 ，...，s n ) Maximum, define p i And b i Respectively represent P and B (T, s) 1 ，s 2 ，...，s n ) The value of the ith pixel of (2) is defined as follows:
wherein Prob (p) i ，b i ) Representing a joint probability function, prob (p) i ) The probability function of the distribution of pixel values, prob (b), representing the image P i ) Representative image B (T, s) 1 ，s 2 ，...，s n ) The probability function is distributed over the pixel values.
The solution of the formula (3) is a typical nonlinear function minimization problem, and can be solved by a steepest gradient descent method.
Further, the image registration method based on the mutual information cost function is as follows:
defining the SPECT plane image vector as P and the corresponding template image vector with organ number as i as T i ，T i The pixel value of the region corresponding to the organ number i in the image is i, the pixel values of other regions are 0, and the rigid-transformed organ i template image vector is
wherein 
Relative to the geometric center (x) of the organ i ，y i ) A two-bit vector of the translation is,
for the angle at which the organ is rotationally transformed around the translated geometric center, the above registration problem is equivalent to the following optimization problem:
Wherein, by applying the steepest gradient descent method to solve equation (5), the optimized transformation parameters for three separate rigid registrations for each organ i can be obtained, thereby performing independent registration for each organ.
The embodiment of the invention also provides a system for automatically segmenting organs in the SPECT plane image, which adopts the method for automatically segmenting the organs in the SPECT plane image and comprises a detection frame acquisition module, a segmentation module, an alignment module and a registration module, wherein the modules cooperate with each other to complete the automatic segmentation of the organs in the SPECT plane image.
Specifically, the detection frame acquisition module is used for detecting an organ with a high degree of distinction from a background in a SPECT plane image of a subject, determining a detection frame corresponding to the organ, and obtaining a detection frame center coordinate of the organ;
the segmentation module is used for segmenting a target organ from three-dimensional SPECT/CT fusion fault data of the same examinee to obtain a three-dimensional organ template picture, and projecting the three-dimensional organ template picture to a two-dimensional plane in the anteroposterior direction of a human body to generate a two-dimensional organ template picture;
the alignment module is used for translating the two-dimensional organ template drawing to align the geometric centers of a plurality of organs in the two-dimensional organ template drawing with the center of the detection frame obtained in the step A;
the registration module is used for registering the two-dimensional organ template image with the SPECT plane image, so that all organs in the two-dimensional organ template image correspond to organ areas in the SPECT plane image.
Other components and operations of a method and system for automatically segmenting an organ in a SPECT planar image according to embodiments of the present invention are known to those of ordinary skill in the art and will not be described in detail herein.
In the description herein, references to the description of the terms "embodiment," "example," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
Claims (10)
1. A method for automatically segmenting an organ in a SPECT planar image, comprising the steps of:
A. detecting an organ with high degree of distinction from the background in a SPECT plane image of a detected person, determining a detection frame corresponding to the organ, and obtaining the central coordinate of the detection frame of the organ;
B. dividing a target organ from three-dimensional SPECT/CT fusion fault data of the same examinee to obtain a three-dimensional organ template map, and projecting the three-dimensional organ template map to a two-dimensional plane in the anteroposterior direction of a human body to generate a two-dimensional organ template map;
C. translating the two-dimensional organ template picture to align the geometric centers of a plurality of organs in the two-dimensional organ template picture with the center of the detection frame obtained in the step A;
D. and registering the two-dimensional organ template image and the SPECT plane image, so that all organs in the two-dimensional organ template image correspond to organ areas in the SPECT plane image, and completing segmentation.
2. The method for automatically segmenting an organ in a SPECT planar image according to claim 1, wherein in step a, the SPECT planar image is detected by a deep learning algorithm, and the center coordinates [ xcenter, ycenter ] of the organ greatly different from the background is output;
the step of detecting the SPECT plane image by the deep learning algorithm comprises the following steps:
s101: taking ResNet50 as a basic feature extraction network of the detection network, and extracting features to obtain a feature map;
s102: the features extracted in the step S101 are up-sampled, and the size of the features is returned to 1/N of the resolution of the original image;
s103: acquiring a heatmap of the data sampled in the S102 through the sub-network 1, extracting a central position coordinate of the target to be detected through a heatmap extreme point position, predicting the sub-network by utilizing position offset, and acquiring a central point offset;
s104: and correcting the coordinates of the center point of the target through the center point offset acquired in the step S103, so as to acquire the position of the center point of the target to be detected finally.
3. The method for automatically segmenting organs in the SPECT plane image according to claim 1, wherein in the step B, pixels belonging to the same organ are set to the same encoding value in the three-dimensional SPECT/CT fusion tomographic data of the same subject, different organs correspond to different encoding values, and pixels not belonging to any organ are encoded to be 0;
and projecting the three-dimensional template picture to a two-dimensional plane in the front-back direction of the human body, and enabling pixel values on the two-dimensional plane to correspond to the coding values or 0 of each organ, so that the two-dimensional organ template picture can be generated.
4. The method for automatically segmenting organs in the SPECT plane image according to claim 3, wherein in the step B, the main organs in the three-dimensional SPECT/CT fusion fault data are segmented by using a unet + + segmentation network, and the method comprises the following steps:
s201: taking each layer of the CT tomogram as data, and obtaining a segmentation result of the layer through a unet + + segmentation network;
s202: putting all the segmentation results of each layer of data of the CT tomographic image together, and integrating the three-dimensional tomographic segmentation results of each organ through connected domain analysis;
s203: and optimizing the S202 three-dimensional fault segmentation result by using the contrast information of the SPECT fault image to obtain a three-dimensional organ template map.
5. The method for automatically segmenting an organ in a SPECT plane image according to claim 3, wherein in the step B, the step of obtaining a three-dimensional organ template map is as follows:
s301: extracting features of the SPECT tomographic image and the CT tomographic image respectively to obtain feature maps;
s302: enabling the feature maps of the two modes to enter an SPP network, pooling the feature maps by the SPP network to generate fixed-length output, namely, featuremaps, and linking the featuremaps of the two modes to enable the featuremaps with the same size of the two modes to be spliced according to channel dimensions to obtain a spliced multi-mode result;
s303: and decoding the spliced multi-modal result into a segmentation result through sampling on the N layers to obtain the three-dimensional organ template picture.
6. The method of automatically segmenting an organ in a SPECT plane image of claim 1 wherein step C includes the sub-steps of:
defining the coordinates of the center of the detection frame of the organ in the SPECT plane image as
Wherein i represents the number of the organ;
defining the geometric center coordinate of the organ in the two-dimensional organ template picture as (x) i ，y i )，i＝1，2，..，I；
Finding the optimal translation vector (x) of the two-dimensional organ template map by the following formula 0 ，y 0 ) Minimizing the sum of squares of the distances corresponding to the geometric center of the organ of the two-dimensional organ template map and the center of the detection box of the organ in the SPECT plane image:
solving the above formula yields:
7. the method for automatically segmenting an organ in the SPECT plane image according to claim 1, wherein in the step D, the two-dimensional organ template map is integrally registered with the SPECT plane image based on an image registration method of mutual information cost function or B-spline non-rigid transformation.
8. The method for automatically segmenting an organ in a SPECT plane image of claim 7 wherein the image registration method of the B-spline non-rigid transformation is as follows:
defining SPECT plane image vector as P, organ template image vector as T, organ template image vector after B spline conversion as B (T, s) 1 ,s 2 ,...,s n), wherein s1 ，s 2 ，...，s n Representing a set of parameters of the B-spline transformation, the minimization problem can be expressed as the following equation:
(s 1 ，s 2 ，...，s n ) opt ＝argmin{-I(P，B(T，s 1 ，s 2 ，...，s n ))} (3)
i.e. finding an optimized set of B-spline transformation parameters such that P and B (T, s) 1 ，s 2 ，...，s n ) Mutual information of I (P, B (T, s) 1 ，s 2 ，...，s n ) Maximum, define p i And b i Respectively represent P and B (T, s) 1 ，s 2 ，...，s n ) The value of the ith pixel of (2) is defined as follows:
wherein Prob (p) i ，b i ) Representing a joint probability function, prob (p) i ) The probability function of the distribution of pixel values, prob (b), representing the image P i ) Representative image B (T, s) 1 ，s 2 ，...，s n ) The probability function is distributed over the pixel values.
9. The method for automatically segmenting organs in the SPECT plane image of claim 7 wherein the image registration method based on the mutual information cost function is as follows:
defining the SPECT plane image vector as P and the corresponding template image vector with organ number as i as T i ，T i The pixel value of the region corresponding to the organ number i in the image is i, the pixel values of other regions are 0, and the rigid-transformed organ i template image vector is
wherein 
Relative to the geometric center (x) of the organ i ，y i ) A two-bit vector of the translation is,
for the angle at which the organ is rotationally transformed around the translated geometric center, the above registration problem is equivalent to the following optimization problem:
10. A system for automatically segmenting an organ in a SPECT planar image, the system employing the method for automatically segmenting an organ in a SPECT planar image of any one of claims 1-9, the system comprising:
the detection frame acquisition module is used for detecting an organ with high degree of distinction from the background in a SPECT plane image of a detected person, determining a detection frame corresponding to the organ and obtaining the center coordinates of the detection frame of the organ;
the segmentation module is used for segmenting a target organ from three-dimensional SPECT/CT fusion tomographic data of the same examinee to obtain a three-dimensional organ template map, and projecting the three-dimensional organ template map to a two-dimensional plane in the anteroposterior direction of a human body to generate a two-dimensional organ template map;
an alignment module, which is used for translating the two-dimensional organ template drawing to align the geometric centers of a plurality of organs in the two-dimensional organ template drawing with the center of the detection frame obtained in the step A;
a registration module to register the two-dimensional organ template map with the SPECT planar image such that all organs in the two-dimensional organ template map correspond to organ regions in the SPECT planar image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211123639.XA CN115311258B (en) | 2022-09-15 | 2022-09-15 | Method and system for automatically segmenting organs in SPECT planar image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211123639.XA CN115311258B (en) | 2022-09-15 | 2022-09-15 | Method and system for automatically segmenting organs in SPECT planar image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115311258A true CN115311258A (en) | 2022-11-08 |
| CN115311258B CN115311258B (en) | 2023-05-02 |
Family
ID=83865861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211123639.XA Active CN115311258B (en) | 2022-09-15 | 2022-09-15 | Method and system for automatically segmenting organs in SPECT planar image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115311258B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188617A (en) * | 2023-04-21 | 2023-05-30 | 有方（合肥）医疗科技有限公司 | CT image data processing method, device and CT system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009097612A1 (en) * | 2008-01-31 | 2009-08-06 | The Johns Hopkins University | Automated image analysis for magnetic resonance imaging |
| CN114359642A (en) * | 2022-01-12 | 2022-04-15 | 大连理工大学 | Multi-modal medical image multi-organ positioning method based on one-to-one target query Transformer |
| CN115035089A (en) * | 2022-06-28 | 2022-09-09 | 华中科技大学苏州脑空间信息研究院 | Brain anatomy structure positioning method suitable for two-dimensional brain image data |
-
2022
- 2022-09-15 CN CN202211123639.XA patent/CN115311258B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009097612A1 (en) * | 2008-01-31 | 2009-08-06 | The Johns Hopkins University | Automated image analysis for magnetic resonance imaging |
| CN114359642A (en) * | 2022-01-12 | 2022-04-15 | 大连理工大学 | Multi-modal medical image multi-organ positioning method based on one-to-one target query Transformer |
| CN115035089A (en) * | 2022-06-28 | 2022-09-09 | 华中科技大学苏州脑空间信息研究院 | Brain anatomy structure positioning method suitable for two-dimensional brain image data |
Non-Patent Citations (1)
| Title |
|---|
| 高瑞婷 等 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116188617A (en) * | 2023-04-21 | 2023-05-30 | 有方（合肥）医疗科技有限公司 | CT image data processing method, device and CT system |
| CN116188617B (en) * | 2023-04-21 | 2023-08-08 | 有方（合肥）医疗科技有限公司 | CT image data processing method, device and CT system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115311258B (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7117026B2 (en) | Physiological model based non-rigid image registration | |
| US10332267B2 (en) | Registration of fluoroscopic images of the chest and corresponding 3D image data based on the ribs and spine | |
| Hill et al. | A strategy for automated multimodality image registration incorporating anatomical knowledge and imager characteristics | |
| US9361686B2 (en) | Method and apparatus for the assessment of medical images | |
| US20030216631A1 (en) | Registration of thoracic and abdominal imaging modalities | |
| CN108520542B (en) | Reconstruction method for time phase matching of PET/CT data | |
| Klinder et al. | Automated model-based rib cage segmentation and labeling in CT images | |
| EP1652122B1 (en) | Automatic registration of intra-modality medical volume images using affine transformation | |
| CN114066953A (en) | Three-dimensional multi-modal image deformable registration method for rigid target | |
| CN115311258B (en) | Method and system for automatically segmenting organs in SPECT planar image | |
| CN116650115A (en) | Orthopedic surgery navigation registration method based on UWB mark points | |
| CN107292351A (en) | The matching process and device of a kind of tubercle | |
| CN109872353B (en) | White light data and CT data registration method based on improved iterative closest point algorithm | |
| Li et al. | 3D intersubject warping and registration of pulmonary CT images for a human lung model | |
| Huesman et al. | Deformable registration of multimodal data including rigid structures | |
| CN111166373B (en) | Positioning registration method, device and system | |
| Zhang et al. | Performance analysis of active shape reconstruction of fractured, incomplete skulls | |
| Sargent et al. | Cross modality registration of video and magnetic tracker data for 3D appearance and structure modeling | |
| Shimizu et al. | Simultaneous extraction of multiple organs from abdominal CT | |
| Byrnes et al. | Multi-Frame CT-Video Registration for 3D Airway-Wall Analysis | |
| De Moor et al. | Non-rigid registration with position dependent rigidity for whole body PET follow-up studies | |
| Wang et al. | Elastic medical image registration based on image intensity | |
| CN114159085B (en) | PET image attenuation correction method and device, electronic equipment and storage medium | |
| Tang et al. | Validation of mutual information-based registration of CT and bone SPECT images in dual-isotope studies | |
| Balasubramanian et al. | Registration of PET and MR images of human brain using normalized cross correlation algorithm and spatial transformation techniques |
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 |