CN115482190A - Fetal heart structure segmentation measurement method and device and computer storage medium - Google Patents
Fetal heart structure segmentation measurement method and device and computer storage medium Download PDFInfo
- Publication number
- CN115482190A CN115482190A CN202111326747.2A CN202111326747A CN115482190A CN 115482190 A CN115482190 A CN 115482190A CN 202111326747 A CN202111326747 A CN 202111326747A CN 115482190 A CN115482190 A CN 115482190A
- Authority
- CN
- China
- Prior art keywords
- fetal heart
- segmentation
- image
- segmentation model
- structural feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000011218 segmentation Effects 0.000 title claims abstract description 231
- 210000002458 fetal heart Anatomy 0.000 title claims abstract description 223
- 238000000691 measurement method Methods 0.000 title claims abstract description 33
- 238000013135 deep learning Methods 0.000 claims abstract description 72
- 238000005259 measurement Methods 0.000 claims abstract description 52
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000004458 analytical method Methods 0.000 claims abstract description 21
- 238000002604 ultrasonography Methods 0.000 claims description 75
- 238000012549 training Methods 0.000 claims description 72
- 238000007781 pre-processing Methods 0.000 claims description 32
- 230000006870 function Effects 0.000 claims description 30
- 230000002861 ventricular Effects 0.000 claims description 22
- 239000008280 blood Substances 0.000 claims description 20
- 210000004369 blood Anatomy 0.000 claims description 20
- 210000003754 fetus Anatomy 0.000 claims description 13
- 230000000747 cardiac effect Effects 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 12
- 210000004204 blood vessel Anatomy 0.000 claims description 9
- 230000001605 fetal effect Effects 0.000 claims description 9
- 238000011478 gradient descent method Methods 0.000 claims description 7
- 210000000709 aorta Anatomy 0.000 claims description 5
- 210000001147 pulmonary artery Anatomy 0.000 claims description 5
- 210000005246 left atrium Anatomy 0.000 claims description 3
- 210000005240 left ventricle Anatomy 0.000 claims description 3
- 210000005245 right atrium Anatomy 0.000 claims description 3
- 210000005241 right ventricle Anatomy 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 abstract description 5
- 239000000523 sample Substances 0.000 description 37
- 238000004590 computer program Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 230000001746 atrial effect Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000000926 separation method Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 210000000748 cardiovascular system Anatomy 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000003708 edge detection Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 210000004072 lung Anatomy 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 1
- 210000002837 heart atrium Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 210000001562 sternum Anatomy 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 210000001835 viscera Anatomy 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
The invention discloses a fetal heart structure segmentation measurement method, a device and a computer storage medium, wherein the method comprises the steps of inputting an obtained fetal heart ultrasonic image into a trained image deep learning segmentation model for analysis, and obtaining a segmentation result output by the image deep learning segmentation model as characteristic information of the fetal heart ultrasonic image, wherein the characteristic information of the fetal heart ultrasonic image comprises a section type of the fetal heart ultrasonic image and a form of at least one structural characteristic; and acquiring the segmentation edge of each structural feature according to the form of each structural feature, and measuring corresponding geometric parameters from the segmentation edge of each structural feature according to the target information of the section class. Therefore, the method can quickly identify the fetal heart section and effectively segment the fetal heart structure for measurement, improves the examination efficiency of doctors and reduces the misdiagnosis and missed diagnosis rate.
Description
Technical Field
The present invention relates to the field of image technologies, and in particular, to a fetal heart structure segmentation measurement method, a fetal heart structure segmentation measurement device, and a computer storage medium.
Background
With the enhancement of the consciousness of prenatal and postnatal care, more and more pregnant women can carry out fetal heart color Doppler ultrasound detection when the fetus develops for 26-30 weeks so as to judge the health condition of the cardiovascular system of the fetus.
During the fetal heart color ultrasound detection, a detector performs film making display by acquiring an ultrasonic image of a fetal heart structure through a heart probe, and then medical workers manually measure the sizes of ventricles, atria, blood vessels and the like in a fetal heart section through the display of the ultrasonic image so as to judge the health degree of a fetal cardiovascular system. However, practice finds that, due to the influence of personal experience of medical care workers, different medical care workers have different judgment results on the fetal heart, and the judgment efficiencies of different medical care workers are different, so that the medical care system as a whole cannot control the uniformity and efficiency of the output quality of the fetal heart color ultrasound detection, and therefore, a fetal heart structure segmentation measurement method and device need to be researched to help doctors to quickly identify and measure the fetal heart structure.
Disclosure of Invention
The invention aims to provide a fetal heart structure segmentation measurement method, a fetal heart structure segmentation measurement device and a computer storage medium, which can quickly identify a fetal heart section and effectively segment a fetal heart structure for measurement, improve the examination efficiency of doctors and reduce the misdiagnosis and missed diagnosis rate.
In order to solve the technical problem, a first aspect of the present invention discloses a fetal heart structure segmentation measurement method, which includes:
inputting the obtained fetal heart ultrasonic image into a trained image deep learning segmentation model for analysis, and obtaining a segmentation result output by the image deep learning segmentation model as characteristic information of the fetal heart ultrasonic image, wherein the characteristic information of the fetal heart ultrasonic image comprises a section type of the fetal heart ultrasonic image and a form of at least one structural feature;
and acquiring the segmentation edge of each structural feature according to the form of the structural feature, and measuring corresponding geometric parameters from the segmentation edge of each structural feature according to the target information of the section class.
As an optional implementation manner, in the first aspect of the present invention, before the acquired ultrasound image of the fetal heart is input into the trained image deep learning segmentation model for analysis, the method further includes:
and executing preprocessing operation on the fetal heart ultrasonic image based on the determined preprocessing mode to obtain the fetal heart ultrasonic image without the target type information, triggering and executing the image deep learning segmentation model input into the trained image deep learning segmentation model, wherein the fetal heart ultrasonic image input into the image deep learning segmentation model is the fetal heart ultrasonic image without the target type information, and the preprocessing operation comprises cutting and/or hiding.
In an alternative embodiment, in the first aspect of the present invention, the image deep learning segmentation model includes a U-Net segmentation model,
and the training method of the U-Net segmentation model comprises the following steps:
acquiring a training sample set of fetal heart ultrasound images, the training sample set of fetal heart ultrasound images including original ultrasound images of a fetal heart and labeled ultrasound images of the fetal heart;
under the determined deep learning framework, a U-Net segmentation model based on image deep learning is built;
and adjusting model parameters of a U-Net segmentation model according to the picture size and the sample size number of the key section of the fetal heart ultrasonic image in the training sample set to enhance the influence factor of the key section, and training the U-Net segmentation model to obtain the trained image deep learning segmentation model, wherein the model parameters comprise at least one of the number of pictures in single training and a data enhancement mode.
In an alternative embodiment, in the first aspect of the present invention, the U-Net segmentation model uses a cross-entropy cost function as a loss function,
and training the U-Net segmentation model to obtain the trained image deep learning segmentation model, wherein the training comprises the following steps:
in training the U-Net segmentation model, continuously adjusting the cross-entropy cost function until convergence according to the difference between the predicted value and the actual value of the U-Net segmentation model, wherein the U-Net segmentation model is input by pairs of an original ultrasound image of the fetal heart and an ultrasound image of the marked fetal heart, and the cross-entropy of the loss function of each pixel pair in the ultrasound image of the fetal heart is configured to be equal in weight.
As an optional implementation manner, in the first aspect of the present invention, the training the U-Net segmentation model to obtain the trained image deep learning segmentation model includes:
in training the U-Net segmentation model, the model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, and the parameters of the training process are controlled to be configured to comprise 1000 total iterations, the initial learning rate is 0.002, and the learning rate is descended once every 400 rounds.
As an optional implementation manner, in the first aspect of the present invention, the obtaining the segmentation edge of each structural feature according to the morphology of the structural feature, and measuring the corresponding geometric parameter from the segmentation edge of each structural feature according to the target information of the section class includes:
and extracting the boundary of each structural feature by using a Canny algorithm for each obtained structural feature, and performing measurement operation as a segmentation edge of the structural feature to obtain a geometric parameter of the structural feature.
As an alternative embodiment, in the first aspect of the present invention, all the section categories include at least one of a four-lumen cardiac section, a left ventricular outflow tract section, a right ventricular outflow tract section, and a three-vessel section;
all the structural characteristics comprise cavity structural characteristics and blood vessel structural characteristics, wherein the cavity structural characteristics comprise at least one of left ventricle blood cavity characteristics, left atrium blood cavity characteristics, right ventricle blood cavity characteristics and right atrium blood cavity characteristics, and the blood vessel structural characteristics comprise at least one of ascending aorta characteristics and pulmonary artery characteristics.
In a second aspect, the present invention discloses a fetal heart structure segmentation measuring device, which includes:
the analysis module is used for acquiring a fetal heart ultrasonic image and inputting the fetal heart ultrasonic image into the trained image deep learning segmentation model for analysis;
an obtaining module, configured to obtain a segmentation result output by the image deep learning segmentation model, as feature information of the fetal heart ultrasound image, where the feature information of the fetal heart ultrasound image includes a section category of the fetal heart ultrasound image and a form of at least one structural feature;
and the measuring module is used for acquiring the segmentation edge of each structural feature according to the form of the structural feature and measuring the corresponding geometric parameters from the segmentation edge of each structural feature according to the target information of the section class.
As an optional embodiment, in the second aspect of the present invention, the apparatus further comprises:
the preprocessing module is used for performing preprocessing operation on the acquired fetal heart ultrasonic image based on the determined preprocessing mode before the acquired fetal heart ultrasonic image is input into the trained image deep learning segmentation model for analysis to obtain the fetal heart ultrasonic image without the target type information, and triggering and executing the processing operation input into the trained image deep learning segmentation model, wherein the fetal heart ultrasonic image input into the image deep learning segmentation model is the fetal heart ultrasonic image without the target type information, and the preprocessing operation comprises cutting and/or hiding.
As an optional implementation manner, in the second aspect of the present invention, the image deep learning segmentation model includes a U-Net segmentation model, and the apparatus further includes a learning module for training the U-Net segmentation model, where the learning module includes:
a sample set acquisition module for acquiring a training sample set of fetal heart ultrasound images, the training sample set of fetal heart ultrasound images including an original ultrasound image of a fetal heart and a marked ultrasound image of the fetal heart;
the model building module is used for building a U-Net segmentation model based on image deep learning under the determined deep learning framework;
and the training module is used for adjusting model parameters of the U-Net segmentation model according to the picture size and the sample size number of the key section of the fetal heart ultrasonic image in the training sample set so as to enhance the influence factor of the key section, and training the U-Net segmentation model to obtain the trained image deep learning segmentation model, wherein the model parameters comprise at least one of the number of pictures in single training and the data enhancement mode.
As an optional implementation manner, in the second aspect of the present invention, the U-Net segmentation model uses a cross-entropy cost function as a loss function;
and the mode that the training module trains the U-Net segmentation model further comprises the following steps:
in training the U-Net segmentation model, continuously adjusting the cross-entropy cost function until convergence according to the difference between the predicted value and the actual value of the U-Net segmentation model, wherein the U-Net segmentation model is input by pairing an original ultrasound image of the fetal heart and an ultrasound image of the marked fetal heart, and the cross-entropy of the loss function of each pixel pair in the ultrasound image of the fetal heart is configured to be an equivalent weight.
As an optional implementation manner, in the second aspect of the present invention, the manner in which the training module trains the U-Net segmentation model further includes:
when the U-Net segmentation model is trained, the model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, and the parameters of the training process are controlled to be configured to comprise 1000 total iterations, the initial learning rate is 0.002, and the learning rate is descended once every 400 rounds.
As an optional implementation manner, in the second aspect of the present invention, a manner that the measurement module obtains the segmentation edge of each structural feature according to the form of the structural feature, and obtains the corresponding geometric parameter from the segmentation edge of each structural feature according to the target information of the section class includes:
and extracting the boundary of each structural feature by using a Canny algorithm for each obtained structural feature, and performing measurement operation as a segmentation edge of the structural feature to obtain a geometric parameter of the structural feature.
As an alternative embodiment, in the second aspect of the present invention, all the section categories include at least one of a four-lumen cardiac section, a left ventricular outflow tract section, a right ventricular outflow tract section, and a three-vessel section;
all the structural characteristics comprise cavity structural characteristics and vessel structural characteristics, wherein the cavity structural characteristics comprise at least one of left ventricle blood cavity characteristics, left atrium blood cavity characteristics, right ventricle blood cavity characteristics and right atrium blood cavity characteristics, and the vessel structural characteristics comprise at least one of ascending aorta characteristics and pulmonary artery characteristics.
In a third aspect, the present invention discloses another fetal heart structure segmentation measuring device, which includes:
a memory storing executable program code;
a processor coupled with the memory;
the processor calls the executable program code stored in the memory to execute the fetal heart structure segmentation measurement method disclosed by the first aspect of the invention.
In a fourth aspect, the present invention discloses a computer storage medium storing computer instructions for executing the fetal heart structure segmentation measurement method disclosed in the first aspect of the present invention when the computer instructions are called.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
in the embodiment of the invention, the obtained fetal heart ultrasonic image is input into a trained image deep learning segmentation model for analysis, and a segmentation result output by the image deep learning segmentation model is obtained and used as the characteristic information of the fetal heart ultrasonic image, wherein the characteristic information of the fetal heart ultrasonic image comprises the section type of the fetal heart ultrasonic image and the form of at least one structural characteristic; and acquiring the segmentation edge of each structural feature according to the form of the structural feature, and measuring corresponding geometric parameters from the segmentation edge of each structural feature according to the target information of the section class. Therefore, the method can realize automatic identification and measurement of the key tangent planes in the fetal heart image and effective separation of the fetal heart structure, further measure the fetal heart structure, has higher accuracy, is suitable for prenatal ultrasonic examination, is favorable for helping doctors to quickly classify and judge the key tangent planes, improves the examination efficiency of the doctors, reduces the misdiagnosis and missed diagnosis rate, can also improve the effective utilization of medical resources, has higher practical value, and is expected to be popularized and used in primary hospitals.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flow chart of a fetal heart structure segmentation measurement method according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of another fetal heart structure segmentation measurement method disclosed in the embodiment of the invention;
fig. 3 is a schematic structural diagram of a fetal heart structure segmentation measuring device according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of another fetal heart structure segmentation measuring device disclosed in the embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
The terms "first," "second," and the like in the description and in the claims, and in the drawings described above, are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements recited, but may alternatively include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.
Reference herein to "an embodiment" means that a subject feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
The invention discloses a fetal heart structure segmentation measurement method and a fetal heart structure segmentation measurement device, which can be used for automatically acquiring the category and the contour of structural features of a fetus, measuring the contour of the structural features according to a corresponding measurement mode, and quickly acquiring a high-precision measurement result of the structural features of the fetus without manually acquiring geometric parameters of the structural features of the fetus, thereby accurately determining the growth and development conditions of the fetus; and the acquisition efficiency of the measurement result of the structural characteristics of the fetus can be improved by inputting the ultrasonic image of the heart of the fetus into the image deep learning segmentation model. The following are detailed below.
Example one
Referring to fig. 1, fig. 1 is a schematic flow chart of a fetal heart structure segmentation measurement method according to an embodiment of the present invention. The fetal heart structure segmentation measurement method described in fig. 1 may be applied to a parameter measurement server, where the parameter measurement server may include a local parameter measurement server or a cloud parameter measurement server, and the embodiment of the present invention is not limited thereto. As shown in fig. 1, the fetal heart structure segmentation measurement method may include the following operations:
101. and inputting the acquired fetal heart ultrasonic image into a trained image deep learning segmentation model for analysis, and acquiring a segmentation result output by the image deep learning segmentation model as feature information of the fetal heart ultrasonic image, wherein the feature information of the fetal heart ultrasonic image comprises a section category of the fetal heart ultrasonic image and a form of at least one structural feature.
In the embodiment of the invention, the fetal heart ultrasonic image is configured to be an ultrasonic image of the cardiac structure of a 18-35-week fetus, and is used for displaying the development condition of the fetal viscera in positive and negative positions and in non-positioning. The fetal heart ultrasonic image acquisition mode can adopt at least one of two-dimensional measurement, three-dimensional measurement, M-type measurement and Doppler spectrum measurement, and is implemented by the wavelength being less than or equal to 15mm and the frequency being more than or equal to 20 KHz.
Further, the fetal heart ultrasound image may be one of a single frame still picture, a plurality of consecutive still pictures, or a dynamic video. When the fetal heart ultrasound image is configured as a single frame of still image, the fetal heart ultrasound image in step 101 is limited to a slice image of a designated fetal heart, so that the single frame of still image contains specific information that can be identified and analyzed by the image deep learning segmentation model. The fetal heart ultrasonic image can also be configured into a plurality of frames of continuous static images, at this time, the static images can be image sets acquired at the same position at equal time intervals by using ultrasonic waves, or image sets acquired at equal time intervals in the process of moving the probe along a specific track, when the image sets are input, the fetal heart ultrasonic images can be continuously input into the image deep learning segmentation model for analysis according to a predetermined frame rate, so that the fetal heart ultrasonic images of continuous frames are input into the image deep learning segmentation model for analysis, for the image sets at the same position at equal time intervals, the image sets at the same position at equal time intervals are beneficial to multiple analysis on the same section, and therefore, the profile of the structural features in the section can be favorably measured for multiple times, the measurement accuracy of the structural features of the fetal heart ultrasonic images is improved, or for the image sets acquired at equal time intervals in the process of moving the probe along the specific track, a feature association relationship between adjacent images can be established, the profile of the structural features in the section can be favorably verified through the front-back association, and the measurement accuracy of the structural features of the fetal heart ultrasonic images is improved. Still further, for the image sets at the same position and at equal time intervals, the configuration of the time intervals may be related to the section type of the fetus during measurement, that is, the size of the acquired time intervals is selected according to the section type of the fetus, so that the corresponding acquisition time intervals are selected according to the section of the fetal heart ultrasound image to be measured, which is beneficial to improving the measurement efficiency and accuracy of the geometric parameters of the structural features of the fetal heart ultrasound image. In addition, when the fetal heart ultrasound image is configured as a dynamic video, a frequency dividing device needs to be arranged and used for intercepting and forming multiple frames of continuous static pictures at a medium time interval in the dynamic video, and then the fetal heart ultrasound image is continuously input into the image deep learning segmentation model for analysis according to a predetermined frame rate, wherein the frequency dividing device can be used for manually dividing frequency by using computer video processing software, and has the advantages that frequency dividing parameters can be flexibly set, the frequency dividing device can also be configured as a program module, and at the moment, the frequency dividing device automatically divides the frequency of the dynamic video according to the preset parameters, so that the manual operation process is simplified. Still further, a measurement operation may be performed on a structural feature of the fetal heart ultrasound image other than the heart (such as a sternum, a right lung, a left lung, a spine, etc.), so as to facilitate determining the health condition of the fetus corresponding to the fetal heart ultrasound image according to the position parameter between the structural feature of the heart and the structural feature of the non-heart.
In an embodiment of the present invention, the section type of the fetal heart ultrasound image includes at least one of a four-chamber heart section, a left ventricular outflow tract section, a right ventricular outflow tract section, and a three-blood-vessel section, which is not limited in the embodiment of the present invention. Furthermore, four sections, such as a four-cavity heart section, a left-cavity outflow tract section, a right-cavity outflow tract section, a three-blood-vessel section and the like, are defined as key sections, when a training sample set is obtained, the proportion of the sample volume of the key sections in all the sample volumes is controlled to be large, such as at least more than 80%, and the proportions of the four-cavity heart section, the left-cavity outflow tract section, the right-cavity outflow tract section and the three-blood-vessel section are different, so that a training side weight is formed on an image deep learning segmentation model, and the recognition efficiency of the key sections and the recognition accuracy of structural characteristics are favorably provided.
102. And acquiring the segmentation edge of each structural feature according to the morphology of the structural feature.
In an embodiment of the present invention, the structural features are specifically a chamber structural feature and a blood vessel structural feature, where the chamber structural feature includes at least one of a left ventricular blood cavity feature, a left atrial blood cavity feature, a right ventricular blood cavity feature, and a right atrial blood cavity feature, and the blood vessel structural feature includes at least one of an ascending aorta feature and a pulmonary artery feature, which is not limited in the embodiment of the present invention.
In the embodiment of the invention, each structural feature has a fixed three-dimensional form, but the two-dimensional forms shown by the same structural feature are different under different sections, so that the form of the structural feature needs to be subjected to edge analysis to obtain the segmentation edge of the structural feature.
103. And measuring corresponding geometric parameters from the segmentation edges of each structural feature according to the target information of the section class.
Because the same structural feature shows different forms in different section categories, in the embodiment of the invention, different target information is configured for different sections, so that the measurement of the geometric parameters of the same structural feature in different dimensions is improved, and the measurement integrity of the structural feature of the fetal heart ultrasonic image is improved.
In this embodiment of the present invention, optionally, the target information of the section category may further include at least one of a graphic coordinate, a size, a shape, a distance between adjacent structural features, and an angle of the structural feature of the fetal heart ultrasound image, which is not limited in this embodiment of the present invention.
Therefore, the fetal heart structure segmentation measurement method described in the embodiment of fig. 1 can realize automatic identification and measurement of key tangent planes in the fetal heart image and effective separation of the fetal heart structure, and further measure the fetal heart structure, has higher accuracy, is suitable for prenatal ultrasonic examination, is beneficial to helping doctors to quickly classify and judge the key tangent planes, improves the examination efficiency of the doctors, reduces misdiagnosis and missed diagnosis rate, can also improve the effective utilization of medical resources, has higher practical value, and is expected to be popularized and used to primary hospitals.
In an optional embodiment, the image deep learning segmentation model is specifically configured as a U-Net segmentation model based on image deep learning, wherein the training method of the U-Net segmentation model includes:
acquiring a training sample set of fetal heart ultrasound images, wherein the training sample set of fetal heart ultrasound images is configured to contain an original ultrasound image of a fetal heart and a labeled ultrasound image of the fetal heart;
under the determined deep learning frame Pythrch, building a U-Net segmentation model based on image deep learning;
according to the picture size and the sample size number of the key section of the fetal heart ultrasonic image in the training sample set, adjusting model parameters of a U-Net segmentation model to enhance influence factors of the key section, and training the U-Net segmentation model to obtain a trained image deep learning segmentation model, wherein the model parameters comprise at least one of the number of pictures in single training and a data enhancement mode.
In the optional embodiment, the U-Net segmentation model is developed and tested under a deep learning frame Pythrch, so that the building integrity and systematicness of the U-Net segmentation model are ensured, in addition, according to the picture size and the sample quantity number of the key section of the fetal heart ultrasonic image in the training sample set, the model parameters of the U-Net segmentation model are adjusted to enhance the influence factor of the key section, the training of the U-Net segmentation model is subjected to side redistribution, and the accuracy of the output result of the U-Net segmentation model is improved.
In this alternative embodiment, the source of the training sample set may be from a hospital inspector, or may be a deformed sample image obtained by a computer tool through rotation, scaling or pixel change, so that the data amount of the sample library may be increased, and the training effect may be improved.
As an optional implementation, further, the U-Net segmentation model uses a cross entropy cost function as a loss function, in this case, in the case of this optional implementation, the training is performed on the U-Net segmentation model, and the obtained trained image deep learning segmentation model includes:
in training the U-Net segmentation model, continuously adjusting the cross-entropy cost function until convergence according to the difference between the predicted value and the actual value of the U-Net segmentation model, wherein the U-Net segmentation model is input by pairs of an original ultrasound image of the fetal heart and an ultrasound image of the marked fetal heart, and the cross-entropy of the loss function of each pixel pair in the ultrasound image of the fetal heart is configured to be equal in weight.
Therefore, in the alternative embodiment, the cross entropy cost function is continuously adjusted through the difference between the predicted value and the actual value until convergence, and the cross entropy of the loss function of each pixel pair in the fetal heart ultrasound image is configured to be the equivalent weight, so that the function can be quickly lost, the parameters of the U-Net model can be optimized, and the accuracy and the efficiency of model analysis can be further improved.
In this optional embodiment, further, in a case that the U-Net segmentation model uses a cross entropy cost function as a loss function, training the U-Net segmentation model, and obtaining a trained image deep learning segmentation model, includes:
in training the U-Net segmentation model, model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, and parameters of a control training process are configured to comprise 1000 total iterations, an initial learning rate of 0.002, and a descending learning rate every 400 rounds.
In the optional embodiment, model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, so that a global optimal solution can be easily obtained, and in consideration of the real-time performance and high-speed rate requirements of system operation, parameters of a control training process are configured to be 1000 total iterations, the initial learning rate is 0.002, and the learning rate is reduced once every 400 iterations, so that the overall iteration time can be quickly shortened on the premise of meeting the requirement of fetal heart ultrasonic image identification, thereby seeking the optimal balance between the accuracy and the high speed rate of system identification in a scene of detecting fetal heart ultrasonic images, and improving the competitiveness of the fetal heart structure segmentation measurement method.
In another optional embodiment, the obtaining the segmentation edge of each structural feature according to the morphology of the structural feature, and measuring the corresponding geometric parameter from the segmentation edge of each structural feature according to the target information of the section class includes:
and extracting the boundary of each structural feature by using a Canny algorithm on each obtained structural feature, and performing measurement operation as a segmentation edge of the structural feature to obtain a geometric parameter of the structural feature.
As an optional implementation manner, further, extracting the boundary of each structural feature by using a Canny algorithm on each obtained structural feature, including:
graying the fetal heart ultrasonic image, wherein graying is carried out after pixel colors are converted according to the section type of the fetal heart ultrasonic image, namely weighted average is carried out according to sampling values of all channels of the image;
performing Gaussian filtering on the grayed fetal heart ultrasonic image, wherein the Gaussian filtering can be realized by using two one-dimensional Gaussian kernels to respectively perform weighting twice, and can also be realized by using one two-dimensional Gaussian kernel to perform convolution once;
calculating the amplitude and the direction of the gradient by using the finite difference of the first-order partial derivatives, wherein the gradient of the grayed fetal heart ultrasonic image is approximated by using the first-order finite difference to obtain two matrixes of partial derivatives of the image in the x and y directions, and the gradient amplitude and the gradient direction are solved according to the two matrixes;
carrying out non-maximum suppression on the gradient amplitude, namely searching a local maximum of a pixel point, setting a gray value corresponding to a non-maximum point as 0, and determining a real and potential edge according to double-threshold detection;
edge detection is accomplished by suppressing isolated weak edges.
Therefore, the optional embodiment extracts the boundary of each structural feature as the segmentation edge of the structural feature through the Canny algorithm, and can identify the actual edge in the image as much as possible and as close as possible, so that the true weak edge is detected, the edge information is more restored, and the measurement efficiency and accuracy of the geometric parameters of the structural feature of the required fetal heart ultrasonic image are improved.
Example two
Referring to fig. 2, fig. 2 is a schematic flow chart of another fetal heart structure segmentation measurement method according to an embodiment of the present invention. The fetal heart structure segmentation measurement method described in fig. 2 may be applied to a parameter measurement server, where the parameter measurement server may include a local parameter measurement server or a cloud parameter measurement server, and the embodiment of the present invention is not limited thereto. As shown in fig. 2, the fetal heart structure segmentation measurement method may include the following operations:
201. and inputting the acquired fetal heart ultrasonic image into a trained image deep learning segmentation model for analysis, and acquiring a segmentation result output by the image deep learning segmentation model as feature information of the fetal heart ultrasonic image, wherein the feature information of the fetal heart ultrasonic image comprises a section category of the fetal heart ultrasonic image and a form of at least one structural feature.
202. And performing preprocessing operation on the fetal heart ultrasonic image based on the determined preprocessing mode to obtain the fetal heart ultrasonic image without the target type information, and triggering and executing the image deep learning segmentation model input into the trained image.
In embodiments of the present invention, the pre-processing operations include clipping and/or hiding.
In the embodiment of the invention, the target type information may be normal noise in the fetal heart ultrasound image, or may be privacy information of the fetal heart ultrasound image about the detecting person.
As can be seen, after acquiring the ultrasound image of the fetal heart, the embodiment of the present invention further performs a preprocessing operation on the ultrasound image of the fetal heart, such as: and cutting and/or hiding can filter noise/privacy information in the fetal heart ultrasonic image, and the acquisition efficiency and accuracy of the geometric parameters of the structural features can be improved.
203. And acquiring the segmentation edge of each structural feature according to the morphology of the structural feature.
204. And measuring corresponding geometric parameters from the segmentation edges of each structural feature according to the target information of the section class.
In the embodiment of the present invention, please refer to the detailed description in the first embodiment for the other descriptions of step 201, step 203, and step 204, which is not repeated herein.
In an alternative embodiment, the preprocessing operation is performed on the fetal heart ultrasound image based on the determined preprocessing mode, and includes:
acquiring at least one preset fixed region position parameter, and performing preprocessing operation on a corresponding region of the fetal heart ultrasonic image according to each fixed region position parameter to obtain a fetal heart ultrasonic image with target type information removed; and/or
And acquiring position parameters of selected areas defined by a user in a sliding manner on a human-computer interaction interface, and executing preprocessing operation on corresponding areas of the fetal heart ultrasonic image according to the position parameters of each selected area to obtain the fetal heart ultrasonic image with the target type information removed.
Therefore, in an optional embodiment, by flexibly configuring multiple optional preprocessing modes, the adaptation degree of the system can be improved, the user side can conveniently and rapidly operate, and the identification efficiency of the system on the geometric parameters of the structural features is improved in a variable phase manner.
In this alternative embodiment, the ultrasound image of the fetal heart requiring the preprocessing operation may include at least one of a four-chamber cardiac section, a left ventricular outflow tract section, a right ventricular outflow tract section, and a three-vessel section, which is not limited by the embodiment of the present invention.
EXAMPLE III
Referring to fig. 3, fig. 3 is a schematic structural diagram of a fetal heart structure segmentation measuring device according to an embodiment of the present invention. The fetal heart structure segmentation measurement apparatus described in fig. 3 may be applied to a parameter measurement server, where the parameter measurement server may include a local parameter measurement server or a cloud parameter measurement server, and the embodiment of the present invention is not limited thereto. As shown in fig. 3, the fetal heart structure segmentation measuring device may include an analyzing module 301, an acquiring module 302, and a measuring module 303, wherein:
the analysis module 301 is configured to acquire an ultrasound image of a fetal heart, input the ultrasound image into a trained image deep learning segmentation model, and analyze the ultrasound image;
an obtaining module 302, configured to obtain a segmentation result output by the image deep learning segmentation model as feature information of a fetal heart ultrasound image, where the feature information of the fetal heart ultrasound image includes a section category of the fetal heart ultrasound image and a form of at least one structural feature;
in an embodiment of the present invention, the section type of the fetal heart ultrasound image includes at least one of a four-chamber heart section, a left ventricular outflow tract section, a right ventricular outflow tract section, and a three-blood-vessel section, which is not limited in the embodiment of the present invention. Furthermore, four sections, such as a four-lumen heart section, a left ventricular outflow tract section, a right ventricular outflow tract section, a three-blood-vessel section and the like, are defined as key sections, when a training sample set is obtained, the proportion of the sample volume of the key sections in all sample volumes is controlled to be larger, such as at least more than 80%, and the proportions of the four-lumen heart section, the left ventricular outflow tract section, the right ventricular outflow tract section and the three-blood-vessel section are different, so that a training weight is formed on an image deep learning segmentation model, and the recognition efficiency of the key sections and the recognition accuracy of structural features are favorably provided.
In an embodiment of the present invention, the structural features are specifically a chamber structural feature and a blood vessel structural feature, where the chamber structural feature includes at least one of a left ventricular blood cavity feature, a left atrial blood cavity feature, a right ventricular blood cavity feature, and a right atrial blood cavity feature, and the blood vessel structural feature includes at least one of an ascending aorta feature and a pulmonary artery feature, which is not limited in the embodiment of the present invention.
The measuring module 303 is configured to obtain a segmentation edge of each structural feature according to a form of the structural feature, and measure a corresponding geometric parameter from the segmentation edge of each structural feature according to the target information of the section class.
It can be seen that, implementing the fetal heart structure segmentation measuring device described in fig. 3 can realize automatic identification and measurement of key tangent planes in the fetal heart image, and effective separation of the fetal heart structure, and then measure the fetal heart structure, has higher accuracy, is suitable for prenatal ultrasonic examination, is favorable to helping doctors to classify and judge the key tangent planes quickly, improves doctor's efficiency of examination, reduces misdiagnosis and missed diagnosis rate, and simultaneously can also improve effective utilization of medical resources, has great practical value, and is expected to be popularized and used to primary hospitals.
In an optional embodiment, the apparatus further comprises:
and the preprocessing module is used for performing preprocessing operation on the acquired fetal heart ultrasonic image based on the determined preprocessing mode before the acquired fetal heart ultrasonic image is input into the trained image deep learning segmentation model for analysis to obtain the fetal heart ultrasonic image without the target type information, triggering the executed fetal heart ultrasonic image input into the trained image deep learning segmentation model, wherein the fetal heart ultrasonic image input into the image deep learning segmentation model is the fetal heart ultrasonic image without the target type information.
In this alternative embodiment, the preprocessing operation includes cropping and/or hiding, and the target type information may be normal noise in the fetal heart ultrasound image, or may be privacy information of the fetal heart ultrasound image about the detecting person.
As can be seen, after the fetal heart ultrasound image is acquired, the embodiment of the present invention further performs preprocessing operations on the fetal heart ultrasound image, such as: and cutting and/or hiding can filter noise/privacy information in the fetal heart ultrasonic image, and the acquisition efficiency and accuracy of the geometric parameters of the structural features can be improved.
As an alternative embodiment, the way in which the preprocessing module performs the preprocessing operation on the ultrasound image of the fetal heart based on the determined preprocessing way includes:
acquiring at least one preset fixed region position parameter, and performing preprocessing operation on a corresponding region of the fetal heart ultrasonic image according to each fixed region position parameter to obtain a fetal heart ultrasonic image with target type information removed; and/or
And acquiring the position parameters of the selected area defined by the user in the human-computer interaction interface in a sliding way, and executing preprocessing operation on the corresponding area of the fetal heart ultrasonic image according to the position parameters of each selected area to obtain the fetal heart ultrasonic image with the target type information removed.
Therefore, in an optional embodiment, by flexibly configuring multiple optional preprocessing modes, the adaptation degree of the system can be improved, the user side can conveniently and rapidly operate, and the identification efficiency of the system on the geometric parameters of the structural features is improved in a variable phase manner.
In this alternative embodiment, the ultrasound image of the fetal heart requiring the preprocessing operation may include at least one of a four-chamber cardiac section, a left ventricular outflow tract section, a right ventricular outflow tract section, and a three-vessel section, which is not limited by the embodiment of the present invention.
As another optional embodiment, the image deep learning segmentation model is specifically configured as a U-Net segmentation model based on image deep learning, and the apparatus further includes a learning module for training the U-Net segmentation model, where the learning module includes:
the system comprises a sample set acquisition module, a comparison module and a comparison module, wherein the sample set acquisition module is used for acquiring a training sample set of fetal heart ultrasonic images, and the training sample set of the fetal heart ultrasonic images comprises an original ultrasonic image of a fetal heart and an ultrasonic image of a marked fetal heart;
the model building module is used for building a U-Net segmentation model based on image deep learning under the determined deep learning framework;
and the training module is used for adjusting model parameters of the U-Net segmentation model according to the picture size and the sample size number of the key section of the fetal heart ultrasonic image in the training sample set so as to enhance the influence factors of the key section, training the U-Net segmentation model and obtaining the trained image deep learning segmentation model, wherein the model parameters comprise at least one of the number of pictures in single training and the data enhancement mode.
In the optional embodiment, the U-Net segmentation model is developed and tested under a deep learning frame Pythrch, so that the integrity and systematicness of the U-Net segmentation model building are ensured, in addition, the model parameters of the U-Net segmentation model are adjusted according to the picture size and the number of samples of the key section of the fetal heart ultrasonic image in the training sample set to enhance the influence factors of the key section, the training of the U-Net segmentation model is subjected to side redistribution, and the accuracy of the output result of the U-Net segmentation model is improved.
In this alternative embodiment, the source of the training sample set may be from a hospital inspector, or may be a deformed sample image obtained by a computer tool through rotation, scaling or pixel change, so that the data amount of the sample library may be increased, and the training effect may be improved.
As an optional implementation manner, further, the U-Net segmentation model uses a cross entropy cost function as a loss function, in this case, in the case of this optional implementation manner, the manner in which the training module trains the U-Net segmentation model to obtain the trained image deep learning segmentation model includes:
in training the U-Net segmentation model, continuously adjusting the cross-entropy cost function until convergence according to the difference between the predicted value and the actual value of the U-Net segmentation model, wherein the U-Net segmentation model is input by pairing an original ultrasound image of the fetal heart and an ultrasound image of the marked fetal heart, and the cross-entropy of the loss function of each pixel pair in the ultrasound image of the fetal heart is configured to be an equivalent weight.
Therefore, in the alternative embodiment, the cross entropy cost function is continuously adjusted through the difference between the predicted value and the actual value until convergence, and the cross entropy of the loss function of each pixel pair in the fetal heart ultrasound image is configured to be the equivalent weight, so that the function can be quickly lost, the parameters of the U-Net model can be optimized, and the accuracy and the efficiency of model analysis can be further improved.
In this optional implementation manner, still further, in a case that the U-Net segmentation model uses a cross entropy cost function as a loss function, the training module trains the U-Net segmentation model, and a manner of obtaining a trained image deep learning segmentation model includes:
in training the U-Net segmentation model, model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, and parameters of a control training process are configured to comprise 1000 total iterations, an initial learning rate of 0.002, and a descending learning rate every 400 rounds.
In the optional embodiment, model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, so that a global optimal solution can be easily obtained, and in consideration of the real-time performance and high-speed rate requirements of system operation, parameters of a control training process are configured to be 1000 total iterations, the initial learning rate is 0.002, and the learning rate is reduced once every 400 iterations, so that the overall iteration time can be quickly shortened on the premise of meeting the requirement of fetal heart ultrasonic image identification, thereby seeking the optimal balance between the accuracy and the high speed rate of system identification in a scene of detecting fetal heart ultrasonic images, and improving the competitiveness of the fetal heart structure segmentation measurement method.
In another optional embodiment, the manner in which the measurement module obtains the segmentation edge of each structural feature according to the form of the structural feature, and measures the corresponding geometric parameter from the segmentation edge of each structural feature according to the target information of the tangent plane category includes:
and extracting the boundary of each structural feature by using a Canny algorithm on each obtained structural feature, and performing measurement operation as a segmentation edge of the structural feature to obtain a geometric parameter of the structural feature.
As an optional implementation manner, further, the manner in which the measurement module extracts the boundary of each structural feature by using a Canny algorithm to each obtained structural feature includes:
graying the fetal heart ultrasonic image, wherein graying is carried out after pixel colors are converted according to the section type of the fetal heart ultrasonic image, namely weighted average is carried out according to sampling values of all channels of the image;
performing Gaussian filtering on the grayed fetal heart ultrasonic image, wherein the Gaussian filtering can be realized by using two one-dimensional Gaussian kernels to respectively perform weighting twice, and can also be realized by using one two-dimensional Gaussian kernel to perform convolution once;
calculating the amplitude and the direction of the gradient by using the finite difference of the first-order partial derivatives, wherein the gradient of the grayed fetal heart ultrasonic image is approximated by using the first-order finite difference to obtain two matrixes of partial derivatives of the image in the x and y directions, and the gradient amplitude and the gradient direction are solved according to the two matrixes;
carrying out non-maximum suppression on the gradient amplitude, namely searching a local maximum of a pixel point, setting a gray value corresponding to a non-maximum point as 0, and determining a real and potential edge according to double-threshold detection;
edge detection is accomplished by suppressing isolated weak edges.
Therefore, in the optional embodiment, the boundary of each structural feature is extracted by the Canny algorithm to serve as the segmentation edge of the structural feature, so that the actual edge in the image can be identified as much as possible and as close as possible, thereby detecting the true weak edge, more reducing the edge information, and improving the measurement efficiency and accuracy of the geometric parameters of the structural feature of the fetal heart ultrasound image.
Example four
Referring to fig. 4, fig. 4 is a view of another fetal heart structure segmentation measuring device according to an embodiment of the present invention. The fetal heart structure segmentation measurement apparatus described in fig. 4 may be applied to a parameter measurement server, where the parameter measurement server may include a local parameter measurement server or a cloud parameter measurement server, and the embodiment of the present invention is not limited thereto. As shown in fig. 4, the fetal heart structure segmentation measuring device may include:
a memory 401 storing executable program code;
a processor 402 coupled with the memory 401;
further, an input interface 403 and an output interface 404 coupled to the processor 402 may be included;
the processor 402 calls the executable program code stored in the memory 401 to perform part or all of the steps of the fetal heart structure segmentation measurement method described in the first embodiment or the second embodiment.
EXAMPLE five
The embodiment of the invention discloses a computer-readable storage medium for storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps of the fetal heart structure segmentation measurement method described in the first embodiment or the second embodiment.
EXAMPLE six
The embodiment of the invention discloses a computer program product, which comprises a non-transitory computer readable storage medium storing a computer program, wherein the computer program is operable to make a computer execute part or all of the steps of the fetal heart structure segmentation measurement method described in the first embodiment or the second embodiment.
The above-described embodiments of the apparatus are merely illustrative, and the modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above detailed description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this understanding in mind, the above technical solutions may essentially or in part contribute to the prior art, be embodied in the form of a software product, which may be stored in a computer-readable storage medium, including a Read-Only Memory (ROM), a Random Access Memory (RAM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an electronically Erasable Programmable Read-Only Memory (EEPROM), an optical Disc-Read (CD-ROM) or other storage medium capable of storing data, a magnetic tape, or any other computer-readable medium capable of storing data.
Finally, it should be noted that: the method and apparatus for segmenting and measuring fetal heart structures disclosed in the embodiments of the present invention are only preferred embodiments of the present invention, and are only used for illustrating the technical solutions of the present invention, not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art; the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A fetal cardiac structure segmentation measurement method, the method comprising:
inputting the obtained fetal heart ultrasonic image into a trained image deep learning segmentation model for analysis, and obtaining a segmentation result output by the image deep learning segmentation model as characteristic information of the fetal heart ultrasonic image, wherein the characteristic information of the fetal heart ultrasonic image comprises a section type of the fetal heart ultrasonic image and a form of at least one structural feature;
and acquiring the segmentation edge of each structural feature according to the form of each structural feature, and measuring corresponding geometric parameters from the segmentation edge of each structural feature according to the target information of the section class.
2. The fetal heart structure segmentation measurement method of claim 1, wherein the acquired fetal heart ultrasound image is input into a trained image deep learning segmentation model for analysis, and the method further comprises:
and executing preprocessing operation on the fetal heart ultrasonic image based on the determined preprocessing mode to obtain the fetal heart ultrasonic image without the target type information, triggering and executing the image deep learning segmentation model input into the trained image deep learning segmentation model, wherein the fetal heart ultrasonic image input into the image deep learning segmentation model is the fetal heart ultrasonic image without the target type information, and the preprocessing operation comprises cutting and/or hiding.
3. The fetal heart structure segmentation measurement method of claim 1 wherein the image deep learning segmentation model comprises a U-Net segmentation model,
and the training method of the U-Net segmentation model comprises the following steps:
acquiring a training sample set of fetal heart ultrasound images, the training sample set of fetal heart ultrasound images including an original ultrasound image of a fetal heart and a labeled ultrasound image of the fetal heart;
under the determined deep learning framework, a U-Net segmentation model based on image deep learning is built;
and adjusting model parameters of a U-Net segmentation model according to the picture size and the sample size number of the key section of the fetal heart ultrasonic image in the training sample set to enhance the influence factor of the key section, and training the U-Net segmentation model to obtain the trained image deep learning segmentation model, wherein the model parameters comprise at least one of the number of pictures in single training and a data enhancement mode.
4. The fetal cardiac structure segmentation measurement method of claim 3, wherein the U-Net segmentation model uses a cross-entropy cost function as a loss function,
and training the U-Net segmentation model to obtain the trained image deep learning segmentation model, wherein the training comprises the following steps:
in training the U-Net segmentation model, continuously adjusting the cross-entropy cost function until convergence according to the difference between the predicted value and the actual value of the U-Net segmentation model, wherein the U-Net segmentation model is input by pairs of an original ultrasound image of the fetal heart and an ultrasound image of the marked fetal heart, and the cross-entropy of the loss function of each pixel pair in the ultrasound image of the fetal heart is configured to be equal in weight.
5. The fetal heart structure segmentation measurement method of claim 4, wherein the training of the U-Net segmentation model to obtain the trained image deep learning segmentation model comprises:
in training the U-Net segmentation model, the model parameters of the U-Net segmentation model are optimized by using a random gradient descent method, and the parameters of the training process are controlled to be configured to comprise 1000 total iterations, the initial learning rate is 0.002, and the learning rate is descended once every 400 rounds.
6. The fetal heart structure segmentation measurement method of claim 1, wherein the obtaining the segmentation edge of each structural feature according to the morphology of the structural feature and measuring the corresponding geometric parameter from the segmentation edge of each structural feature according to the target information of the section class comprises:
and extracting the boundary of each structural feature by using a Canny algorithm for each obtained structural feature, and performing measurement operation as a segmentation edge of the structural feature to obtain a geometric parameter of the structural feature.
7. The fetal cardiac structure segmentation measurement method of any one of claims 1-6,
all the section categories comprise at least one of a four-cavity heart section, a left ventricular outflow tract section, a right ventricular outflow tract section and a three-blood-vessel section;
all the structural characteristics comprise cavity structural characteristics and blood vessel structural characteristics, wherein the cavity structural characteristics comprise at least one of left ventricle blood cavity characteristics, left atrium blood cavity characteristics, right ventricle blood cavity characteristics and right atrium blood cavity characteristics, and the blood vessel structural characteristics comprise at least one of ascending aorta characteristics and pulmonary artery characteristics.
8. A fetal cardiac structure segmentation measurement apparatus, the apparatus comprising:
the analysis module is used for acquiring an ultrasonic image of the heart of the fetus, inputting the ultrasonic image into the trained image deep learning segmentation model and analyzing the ultrasonic image;
an obtaining module, configured to obtain a segmentation result output by the image deep learning segmentation model, as feature information of the fetal heart ultrasound image, where the feature information of the fetal heart ultrasound image includes a section category of the fetal heart ultrasound image and a form of at least one structural feature;
and the measuring module is used for acquiring the segmentation edge of each structural feature according to the form of the structural feature and measuring the corresponding geometric parameters from the segmentation edge of each structural feature according to the target information of the section class.
9. A fetal cardiac structure segmentation measurement apparatus, the apparatus comprising:
a memory storing executable program code;
a processor coupled with the memory;
the processor invokes the executable program code stored in the memory to perform the fetal cardiac structure segmentation measurement method of any one of claims 1-7.
10. A computer storage medium storing computer instructions which, when invoked, perform a fetal cardiac structure segmentation measurement method according to any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111326747.2A CN115482190A (en) | 2021-11-10 | 2021-11-10 | Fetal heart structure segmentation measurement method and device and computer storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111326747.2A CN115482190A (en) | 2021-11-10 | 2021-11-10 | Fetal heart structure segmentation measurement method and device and computer storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115482190A true CN115482190A (en) | 2022-12-16 |
Family
ID=84420656
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111326747.2A Pending CN115482190A (en) | 2021-11-10 | 2021-11-10 | Fetal heart structure segmentation measurement method and device and computer storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115482190A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115760851A (en) * | 2023-01-06 | 2023-03-07 | 首都儿科研究所附属儿童医院 | Ultrasonic image data processing method and system based on machine learning |
CN116681958A (en) * | 2023-08-04 | 2023-09-01 | 首都医科大学附属北京妇产医院 | Fetal lung ultrasonic image maturity prediction method based on machine learning |
CN116912236A (en) * | 2023-09-08 | 2023-10-20 | 首都医科大学附属北京妇产医院 | Method, system and storable medium for predicting fetal congenital heart disease risk based on artificial intelligence |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110279433A (en) * | 2018-09-21 | 2019-09-27 | 四川大学华西第二医院 | A kind of fetus head circumference automatic and accurate measurement method based on convolutional neural networks |
CN110555836A (en) * | 2019-09-05 | 2019-12-10 | 李肯立 | Automatic identification method and system for standard fetal section in ultrasonic image |
CN111012377A (en) * | 2019-12-06 | 2020-04-17 | 北京安德医智科技有限公司 | Echocardiogram heart parameter calculation and myocardial strain measurement method and device |
CN111481233A (en) * | 2019-01-28 | 2020-08-04 | 四川大学华西第二医院 | Thickness measuring method for transparent layer of fetal cervical item |
CN111695644A (en) * | 2020-08-10 | 2020-09-22 | 华侨大学 | Tumor ultrasonic image classification method, device and medium based on optical density transformation |
CN111904469A (en) * | 2020-08-07 | 2020-11-10 | 长沙大端信息科技有限公司 | Heart section detection method and system capable of realizing parallel processing |
CN112184656A (en) * | 2020-09-24 | 2021-01-05 | 广州爱孕记信息科技有限公司 | Method and device for determining fetus section based on ultrasonic dynamic image |
CN112233167A (en) * | 2020-09-24 | 2021-01-15 | 广州爱孕记信息科技有限公司 | Automatic measurement method and device for structural characteristics of fetus |
CN112348780A (en) * | 2020-10-26 | 2021-02-09 | 首都医科大学附属北京安贞医院 | Fetal heart measuring method and device |
-
2021
- 2021-11-10 CN CN202111326747.2A patent/CN115482190A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110279433A (en) * | 2018-09-21 | 2019-09-27 | 四川大学华西第二医院 | A kind of fetus head circumference automatic and accurate measurement method based on convolutional neural networks |
CN111481233A (en) * | 2019-01-28 | 2020-08-04 | 四川大学华西第二医院 | Thickness measuring method for transparent layer of fetal cervical item |
CN110555836A (en) * | 2019-09-05 | 2019-12-10 | 李肯立 | Automatic identification method and system for standard fetal section in ultrasonic image |
CN111012377A (en) * | 2019-12-06 | 2020-04-17 | 北京安德医智科技有限公司 | Echocardiogram heart parameter calculation and myocardial strain measurement method and device |
CN111904469A (en) * | 2020-08-07 | 2020-11-10 | 长沙大端信息科技有限公司 | Heart section detection method and system capable of realizing parallel processing |
CN111695644A (en) * | 2020-08-10 | 2020-09-22 | 华侨大学 | Tumor ultrasonic image classification method, device and medium based on optical density transformation |
CN112184656A (en) * | 2020-09-24 | 2021-01-05 | 广州爱孕记信息科技有限公司 | Method and device for determining fetus section based on ultrasonic dynamic image |
CN112233167A (en) * | 2020-09-24 | 2021-01-15 | 广州爱孕记信息科技有限公司 | Automatic measurement method and device for structural characteristics of fetus |
CN112348780A (en) * | 2020-10-26 | 2021-02-09 | 首都医科大学附属北京安贞医院 | Fetal heart measuring method and device |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115760851A (en) * | 2023-01-06 | 2023-03-07 | 首都儿科研究所附属儿童医院 | Ultrasonic image data processing method and system based on machine learning |
CN115760851B (en) * | 2023-01-06 | 2023-05-09 | 首都儿科研究所附属儿童医院 | Ultrasonic image data processing equipment, system and computer readable storage medium based on machine learning |
CN116681958A (en) * | 2023-08-04 | 2023-09-01 | 首都医科大学附属北京妇产医院 | Fetal lung ultrasonic image maturity prediction method based on machine learning |
CN116681958B (en) * | 2023-08-04 | 2023-10-20 | 首都医科大学附属北京妇产医院 | Fetal lung ultrasonic image maturity prediction method based on machine learning |
CN116912236A (en) * | 2023-09-08 | 2023-10-20 | 首都医科大学附属北京妇产医院 | Method, system and storable medium for predicting fetal congenital heart disease risk based on artificial intelligence |
CN116912236B (en) * | 2023-09-08 | 2023-12-26 | 首都医科大学附属北京妇产医院 | Method, system and storable medium for predicting fetal congenital heart disease risk based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN115482190A (en) | Fetal heart structure segmentation measurement method and device and computer storage medium | |
CN113194836B (en) | Automated clinical workflow | |
KR101121396B1 (en) | System and method for providing 2-dimensional ct image corresponding to 2-dimensional ultrasound image | |
CN107292835B (en) | Method and device for automatically vectorizing retinal blood vessels of fundus image | |
CN111275755B (en) | Mitral valve orifice area detection method, system and equipment based on artificial intelligence | |
CN104182984B (en) | Method and system for rapidly and automatically collecting blood vessel edge forms in dynamic ultrasonic image | |
CN114419500A (en) | Method and device for screening diastolic and systolic images based on cardiac ultrasound video | |
CN112597982B (en) | Image classification method, device, equipment and medium based on artificial intelligence | |
CN112348780A (en) | Fetal heart measuring method and device | |
CN106910194A (en) | The dividing method of the intracardiac adventitia in cardiac function MRI | |
CN116130090A (en) | Ejection fraction measuring method and device, electronic device, and storage medium | |
CN112233167A (en) | Automatic measurement method and device for structural characteristics of fetus | |
CN111583250A (en) | Deep learning-based ultrasonic image mitral valve positioning method and system | |
EP2498222A2 (en) | Method and system for regression-based 4D mitral valve segmentation from 2D+T magnetic resonance imaging slices | |
CN109674493B (en) | Method, system and equipment for medical ultrasonic automatic tracking of carotid artery blood vessel | |
CN113469963B (en) | Pulmonary artery image segmentation method and device | |
CN117017347B (en) | Image processing method and system of ultrasonic equipment and ultrasonic equipment | |
EP3437068B1 (en) | System and methods for diagnostic image analysis and image quality assessment | |
CN112258476A (en) | Echocardiography myocardial abnormal motion mode analysis method, system and storage medium | |
CN116419716A (en) | Analysis method of periodicity parameters and ultrasonic imaging system | |
JP2801536B2 (en) | Ultrasound image processing device | |
CN113222985B (en) | Image processing method, image processing device, computer equipment and medium | |
CN114708283A (en) | Image object segmentation method and device, electronic equipment and storage medium | |
Czap | Impact of preprocessing features on the performance of ultrasound tongue contour tracking, via dynamic programming | |
CN111904469A (en) | Heart section detection method and system capable of realizing parallel processing |
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 |