CN114299078A - Medical image preprocessing method and device and computer readable storage medium - Google Patents
Medical image preprocessing method and device and computer readable storage medium Download PDFInfo
- Publication number
- CN114299078A CN114299078A CN202111471045.3A CN202111471045A CN114299078A CN 114299078 A CN114299078 A CN 114299078A CN 202111471045 A CN202111471045 A CN 202111471045A CN 114299078 A CN114299078 A CN 114299078A
- Authority
- CN
- China
- Prior art keywords
- dimensional image
- historical
- target
- organ
- predetermined
- 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.)
- Withdrawn
Links
Images
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a medical image preprocessing method and device and a computer readable storage medium. Wherein, the method comprises the following steps: acquiring a two-dimensional image sequence of a predetermined organ of a target object; dividing each two-dimensional image slice in the two-dimensional image sequence according to each part of a predetermined organ through a predetermined division model to obtain a divided two-dimensional image sequence; and splicing the divided two-dimensional image sequences to obtain a target three-dimensional image of the predetermined organ. The invention solves the technical problems that in the prior art, the traditional blood vessel segmentation mode based on deep learning is to segment the whole image, the segmentation effect is poor, and the segmentation result needs to manually remove the part of the non-target blood vessel in the segmentation result, so that the blood vessel segmentation efficiency is low.
Description
Technical Field
The invention relates to the technical field of medical image processing, in particular to a medical image preprocessing method and device and a computer readable storage medium.
Background
It is important to segment each part of the heart organ from medical images (CT contrast, magnetic resonance, ultrasound images, etc.). In clinic, structural analysis of each part of a heart organ is an essential link for understanding pathological changes of each part of the heart organ to describe diseases of each part of the heart organ. The segmentation of each part of the organ in the medical image has high research value for heart position image understanding, image analysis, coronary vessel central line extraction, image segmentation and model reconstruction. The traditional coronary vessel segmentation method based on deep learning is to segment the whole image, the segmentation effect is poor, the segmented result needs to manually remove the non-coronary vessel parts in the segmentation result, such as vein vessels, and the modeling difficulty is high, and the efficiency is low.
The result of the image semantic segmentation algorithm depends on the aspect of feature extraction of the image to a great extent, and the more sufficient the feature extraction of the image is, the higher the striving rate of the semantic segmentation pixels is. The mainstream sampling method of the image semantic segmentation algorithm with a significant effect at present is as follows, which is described in detail in 3.
The first is a U-Net network model, which adopts the way of splicing features together in channel dimensions to better utilize the information of shallow characteristic diagrams. The second type is a DeepLab series network model, the DeepLab series network model adopts the hole convolution to extract image characteristics, and the use of the hole convolution increases the receptive field of the filter, so that the extraction of the characteristics is more comprehensive. The third type is a feature pyramid model, which improves the extraction efficiency of image information by combining deep feature information with shallow information.
However, the above 3 sampling methods of image features cannot solve the problem of difficult classification of boundary pixels. Because the blood vessels in the medical image are branched and disordered, the accuracy of the boundary pixel classification directly influences the segmentation effect of the whole image, and for the blood vessels close to bones, tissues, lesions and the like, the boundary characteristics of the blood vessels are not particularly obvious because the lumens of the pixels are close, and the difficulty of blood vessel segmentation is also increased.
In view of the above-mentioned problems that in the related art, the conventional blood vessel segmentation method based on deep learning is to segment the whole image, which not only results in poor segmentation effect, but also results after segmentation still require manual removal of non-target blood vessels in the segmentation result, which results in low blood vessel segmentation efficiency, no effective solution has been proposed at present.
Disclosure of Invention
The embodiment of the invention provides a medical image preprocessing method and device and a computer readable storage medium thereof, which at least solve the technical problems that in the related art, the traditional blood vessel segmentation mode based on deep learning is to segment the whole image, the segmentation effect is poor, and the segmentation result also needs to manually remove the non-target blood vessel part in the segmentation result, so that the blood vessel segmentation efficiency is low.
According to an aspect of an embodiment of the present invention, there is provided a method for preprocessing a medical image, including: acquiring a two-dimensional image sequence of a predetermined organ of a target object; segmenting each two-dimensional image slice in the two-dimensional image sequence according to each part of the predetermined organ through a predetermined segmentation model to obtain a segmented two-dimensional image sequence, wherein the predetermined segmentation model is obtained by using multiple groups of training data through machine learning training, and each group of training data in the multiple groups of training data comprises: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices; and splicing the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ.
Optionally, acquiring a two-dimensional image sequence of a predetermined organ of the target object comprises: acquiring an original three-dimensional image of the predetermined organ; slicing the original three-dimensional image to obtain a plurality of two-dimensional image slices; and obtaining the two-dimensional image sequence based on the two-dimensional image slice.
Optionally, before each two-dimensional image slice in the two-dimensional image sequence is segmented according to each part of the predetermined organ by a predetermined segmentation model, the method further includes: and carrying out normalization processing on each two-dimensional image slice in the two-dimensional image sequence.
Optionally, before each two-dimensional image slice in the two-dimensional image sequence is segmented according to each part of the predetermined organ by a predetermined segmentation model, the method further includes: obtaining the preset segmentation model through machine learning training by utilizing the multiple groups of training data; wherein obtaining the predetermined segmentation model through machine learning training using the plurality of sets of training data comprises: acquiring a plurality of historical original three-dimensional images; slicing the plurality of historical original three-dimensional images to obtain a plurality of groups of historical two-dimensional image slice sets; normalizing each historical two-dimensional image slice in each historical two-dimensional image set in the multiple sets of historical two-dimensional image sets to obtain multiple sets of historical two-dimensional image slice sets after normalization; on a Graphic Processing Unit (GPU), training a preset network model by using the multiple groups of historical two-dimensional image slice sets after the normalization processing and historical segmentation results corresponding to the multiple groups of historical two-dimensional image slice sets to obtain the preset segmentation model.
Optionally, the predetermined organ is a heart organ, and after the segmented two-dimensional image sequence is spliced to obtain a target three-dimensional image of the predetermined organ, the method further includes: carrying out internal contraction processing and external expansion processing on the target three-dimensional image; acquiring a first region outside the boundary of the target three-dimensional image after the target three-dimensional image is subjected to retraction processing and a second region outside the boundary of the target three-dimensional image after the target three-dimensional image is subjected to outward expansion processing; and deleting the first region and the second region in the target three-dimensional image to obtain images of the aorta and coronary vessel regions in the heart organ.
Optionally, performing an internal contraction process and an external expansion process on the target three-dimensional image, including: determining ventricles and atria in the target three-dimensional image; determining the geometric centers of the ventricle and the atrium; and respectively connecting the points on the boundary of the ventricle and the atrium with the corresponding geometric centers, controlling the points on the boundary of the ventricle and the atrium to move along the connecting line to the direction of the corresponding geometric center for retraction processing, and controlling the points on the boundary of the ventricle and the atrium to move along the connecting line to the opposite direction of the corresponding geometric center for expansion processing.
According to another aspect of the embodiments of the present invention, there is also provided a medical image preprocessing apparatus including: the first acquisition module is used for acquiring a two-dimensional image sequence of a predetermined organ of a target object; a segmentation module, configured to segment each two-dimensional image slice in the two-dimensional image sequence according to each part of the predetermined organ through a predetermined segmentation model to obtain a segmented two-dimensional image sequence, where the predetermined segmentation model is obtained through machine learning training using multiple sets of training data, and each set of training data in the multiple sets of training data includes: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices; and the splicing module is used for splicing the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ.
Optionally, the obtaining module includes: a first acquisition unit for acquiring an original three-dimensional image of the predetermined organ; the first processing unit is used for carrying out slicing processing on the original three-dimensional image to obtain a plurality of two-dimensional image slices; and the second acquisition unit is used for obtaining the two-dimensional image sequence based on the two-dimensional image slice.
Optionally, the apparatus further comprises: and the first processing module is used for carrying out normalization processing on each two-dimensional image slice in the two-dimensional image sequence before each two-dimensional image slice in the two-dimensional image sequence is segmented according to each part of the predetermined organ through a preset segmentation model.
Optionally, the apparatus further comprises: the second acquisition module is used for obtaining a preset segmentation model through machine learning training by utilizing the multiple groups of training data before segmenting each two-dimensional image slice in the two-dimensional image sequence according to each part of the preset organ through the preset segmentation model; wherein the second obtaining module includes: the third acquisition unit is used for acquiring a plurality of historical original three-dimensional images; the second processing unit is used for carrying out slicing processing on the plurality of historical original three-dimensional images to obtain a plurality of groups of historical two-dimensional image slice sets; the third processing unit is used for carrying out normalization processing on each historical two-dimensional image slice in each historical two-dimensional image set in the multiple sets of historical two-dimensional image sets to obtain multiple sets of historical two-dimensional image slice sets after normalization processing; and the training unit is used for training a preset network model by utilizing the multiple groups of historical two-dimensional image slice sets after the normalization processing and historical segmentation results corresponding to the multiple groups of historical two-dimensional image slice sets on a Graphics Processing Unit (GPU) to obtain the preset segmentation model.
Optionally, the apparatus further comprises: the second processing module is used for splicing the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ, and then performing retraction processing and external expansion processing on the target three-dimensional image; the third acquisition module is used for acquiring a first area outside the target three-dimensional image after the target three-dimensional image is subjected to retraction processing and a second area outside the target three-dimensional image after the target three-dimensional image is subjected to expansion processing; and the fourth acquisition module is used for deleting the first area and the second area in the target three-dimensional image to obtain images of the aorta and coronary vessel areas in the heart organ.
Optionally, the second processing module includes: a first determination unit for determining a ventricle and an atrium in the target three-dimensional image; a second determination unit for determining the geometric centers of the ventricle and the atrium; and the connecting unit is used for respectively connecting the points on the boundary of the ventricle and the atrium with the corresponding geometric centers, controlling the points on the boundary of the ventricle and the atrium to move towards the corresponding geometric center direction along the connecting line for retraction processing, and controlling the points on the boundary of the ventricle and the atrium to move towards the opposite direction of the corresponding geometric center direction along the connecting line for expansion processing.
According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium including a stored computer program, wherein when the computer program is executed by a processor, the computer-readable storage medium is controlled by an apparatus to execute any one of the above methods for preprocessing a medical image.
According to another aspect of the embodiments of the present invention, there is further provided a processor, configured to execute a computer program, wherein the computer program executes to execute any one of the methods for preprocessing a medical image.
In the embodiment of the invention, a two-dimensional image sequence of a predetermined organ of a target object is obtained; segmenting each two-dimensional image slice in the two-dimensional image sequence according to each part of a predetermined organ through a predetermined segmentation model to obtain a two-dimensional image sequence after segmentation, wherein the predetermined segmentation model is obtained by using a plurality of groups of training data through machine learning training, and each group of training data in the plurality of groups of training data comprises: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices; and splicing the divided two-dimensional image sequences to obtain a target three-dimensional image of the predetermined organ. By the medical image preprocessing method, the purpose of splicing a target three-dimensional image of the predetermined organ based on a new two-dimensional image sequence obtained by segmenting a two-dimensional image slice in a two-dimensional image sequence of the predetermined organ is achieved, so that the technical effect of improving the segmentation efficiency of the predetermined organ is achieved, and the technical problem that in the related technology, the traditional blood vessel segmentation mode based on deep learning is to segment the whole image, the segmentation effect is poor, and the segmented result also needs to manually remove the non-target blood vessel part in the segmentation result, so that the blood vessel segmentation efficiency is low is solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
fig. 1 is a flowchart of a method of preprocessing a medical image according to an embodiment of the present invention;
FIG. 2 is a diagram of a comparison of a predicted cross-section of a segmentation result with a truly labeled cross-section, in accordance with an embodiment of the present invention;
FIG. 3 is a process diagram of a CTA image segmenting various portions of a heart organ, according to an embodiment of the invention;
FIG. 4(a) is a schematic cross-sectional view of a segmentation result according to an embodiment of the present invention;
FIG. 4(b) is a schematic diagram of a coronal plane of a segmentation result according to an embodiment of the present invention;
FIG. 4(c) is a sagittal sectional view of a segmentation result according to an embodiment of the present invention;
FIG. 5 is a flow chart of segmentation of various portions of a heart organ according to an embodiment of the present invention;
FIG. 6 is a graph of the effect of an organ after retraction and expansion according to an embodiment of the present invention;
FIG. 7 is a schematic representation of a coronary vessel containing a region of space formed between an internal and an external dilation in accordance with an embodiment of the present invention;
FIG. 8 is a schematic comparison of an original image of a cardiac organ after pretreatment in accordance with an embodiment of the present invention;
FIG. 9 is a schematic illustration of an overall 3D view of a segmentation result according to an embodiment of the invention;
FIG. 10 is a graph of the result of abdominal organ segmentation according to an embodiment of the present invention;
fig. 11 is a schematic diagram of a medical image preprocessing apparatus according to an 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 derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
In accordance with an embodiment of the present invention, there is provided a method embodiment of a method for pre-processing medical images, it is noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than that presented herein.
Fig. 1 is a flowchart of a medical image preprocessing method according to an embodiment of the present invention, as shown in fig. 1, the medical image preprocessing method includes the following steps:
step S102, a two-dimensional image sequence of a predetermined organ of a target object is acquired.
In the above steps, the predetermined organs include, but are not limited to: heart, brain organs, and abdominal organs.
Step S104, each two-dimensional image slice in the two-dimensional image sequence is segmented according to each part of a predetermined organ through a predetermined segmentation model to obtain a segmented two-dimensional image sequence, wherein the predetermined segmentation model is obtained by using a plurality of groups of training data through machine learning training, and each group of training data in the plurality of groups of training data comprises: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices.
And step S106, splicing the divided two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ.
As can be seen from the above, in the embodiment of the present invention, first, a two-dimensional image sequence of a predetermined organ of a target object may be acquired; then, each two-dimensional image slice in the two-dimensional image sequence can be segmented according to each part of a predetermined organ through a predetermined segmentation model to obtain a segmented two-dimensional image sequence, wherein the predetermined segmentation model is obtained by using a plurality of groups of training data through machine learning training, and each group of training data in the plurality of groups of training data comprises: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices; and finally, splicing the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ. By the medical image preprocessing method, the purpose of splicing a target three-dimensional image of the predetermined organ based on a new two-dimensional image sequence obtained by segmenting a two-dimensional image slice in a two-dimensional image sequence of the predetermined organ is achieved, so that the technical effect of improving the segmentation efficiency of the predetermined organ is achieved, and the technical problem that in the related technology, the traditional blood vessel segmentation mode based on deep learning is to segment the whole image, the segmentation effect is poor, and the segmented result also needs to manually remove the non-target blood vessel part in the segmentation result, so that the blood vessel segmentation efficiency is low is solved.
The medical image preprocessing method provided by the embodiment of the invention can be applied to data preprocessing of blood vessel extraction at brain, lung and other parts.
As an alternative embodiment, in step S102, acquiring a two-dimensional image sequence of a predetermined organ of the target object includes: acquiring an original three-dimensional image of a predetermined organ; slicing an original three-dimensional image to obtain a plurality of two-dimensional image slices; and obtaining a two-dimensional image sequence based on the two-dimensional image slice.
In the above optional embodiment, an original three-dimensional image of a predetermined organ is first acquired, where the acquisition is performed in a manner including, but not limited to, acquisition by mathematical modeling software such as 3DMAX, MATLAB, and the like, then the acquired three-dimensional image is sliced, the three-dimensional image is cut into two-dimensional image slices, and finally a sequence of two-dimensional images is acquired based on the two-dimensional image slices.
As an alternative embodiment, before each two-dimensional image slice in the two-dimensional image sequence is segmented according to the parts of the predetermined organ by the predetermined segmentation model, the method for preprocessing the medical image further includes: and carrying out normalization processing on each two-dimensional image slice in the two-dimensional image sequence.
In the above alternative embodiment, the image segmentation algorithm includes, but is not limited to, the PointRend image segmentation algorithm. The PointRend image segmentation model is an image semantic segmentation algorithm, and the algorithm introduces supervision on each point in an image to improve the boundary segmentation condition of a target object, so that the overall segmentation result of the image is greatly improved. Compared with the traditional image segmentation algorithm model, the PointRend image segmentation model achieves remarkable improvement on the segmentation result of the images of the natural scene and the road scene.
As an alternative embodiment, before each two-dimensional image slice in the two-dimensional image sequence is segmented according to the parts of the predetermined organ by the predetermined segmentation model, the method for preprocessing the medical image further includes: obtaining a preset segmentation model through machine learning training by utilizing a plurality of groups of training data; wherein, utilize the predetermined segmentation model of multiunit training data through machine learning training, include: acquiring a plurality of historical original three-dimensional images; slicing a plurality of historical original three-dimensional images to obtain a plurality of groups of historical two-dimensional image slice sets; normalizing each historical two-dimensional image slice in each historical two-dimensional image set in the multiple sets of historical two-dimensional image sets to obtain multiple sets of historical two-dimensional image slice sets after normalization; on a Graphic Processing Unit (GPU), training a preset network model by utilizing the multiple groups of historical two-dimensional image slice sets after normalization processing and historical segmentation results corresponding to the multiple groups of historical two-dimensional image slice sets to obtain a preset segmentation model.
Fig. 2 is a comparison diagram of a cross section predicted by a segmentation result and a cross section actually labeled according to an embodiment of the present invention, as shown in fig. 2, a column a of pictures is a cross section of a CTA image, that is, an organ image actually captured by a CTA image technology, a column B of pictures is a predicted segmentation diagram of a cross section of a CTA image, that is, a picture predicted by organ segmentation through machine learning, and a diagram C is an actual labeling diagram of a cross section of a CTA image, that is, a segmentation labeling diagram before deep learning training of a cross section of a CTA image. It should be noted that CTA imaging technology, also called Computer Tomography (CTA) is a technology for examining blood vessels (or other organs) by taking a photograph using computer tomography, and can detect all arteries and veins of the whole body, including blood vessels of the heart, brain, lung, kidney, limbs, and the like, and most commonly, the coronary artery system of the heart, the arterial artery system of the brain, and the aortic artery system of the brain, and the like, as provided in this embodiment. The principle of the computerized tomography is to generate a cross-sectional image of an object in a specific scanning area by using a computer processing combination of a plurality of X-ray measurement values at different angles, such that the internal influence of an organ can be obtained without cutting in this embodiment.
As an alternative embodiment, the predetermined organ is a heart organ, and after the segmented two-dimensional image sequences are spliced to obtain the target three-dimensional image of the predetermined organ, the method further includes: carrying out internal contraction processing and external expansion processing on the target three-dimensional image; acquiring a first region outside the boundary after the target three-dimensional image is subjected to retraction processing and a second region outside the boundary after the target three-dimensional image is subjected to outward expansion processing; and deleting the first region and the second region in the target three-dimensional image to obtain images of the aorta and coronary vessel regions in the heart organ.
Fig. 3 is a process diagram of segmenting parts of a heart organ by a CTA image according to an embodiment of the present invention, as shown in fig. 3, first, as shown in part a in fig. 3, an overall CTA image is obtained, then, as shown in part B in fig. 3, a slice of the CTA image, that is, a two-dimensional image sequence is obtained, as shown in part C in fig. 3, the CTA two-dimensional image slice is normalized, next, as shown in part D in fig. 3, data is input into a machine learning model to obtain segmentation results of parts of a central organ of the slice, and finally, as shown in part E in fig. 3, segmentation results of parts of the central organ of the CTA are merged together.
It should be noted that the normalization process is to change a dimensional expression into a dimensionless expression, so that the data can be processed conveniently.
Further, the target organ can be segmented by a model obtained through machine learning from multiple angles, so that the target organ can be observed better, and a larger research value is provided. The results of the segmentation of the target organ at each angle will be described in detail below.
Fig. 4(a) is a schematic cross-sectional view of a cross-section of a segmentation result according to an embodiment of the present invention, as shown in fig. 4(a), different blood vessels of a target organ are segmented and represented by different colors, and it should be noted that the cross-section of the target organ refers to a cross-section perpendicular to a midline direction through a midline post.
Fig. 4(b) is a schematic sectional view of the coronal plane of the segmentation result according to the embodiment of the present invention, and as shown in fig. 4(b), different blood vessels of the target organ are segmented and represented by different colors, and it should be noted that the coronal plane of the target organ refers to a plane passing through the vertical plane and the horizontal axis and all planes parallel thereto.
Fig. 4(c) is a schematic view of a sagittal section of a segmentation result according to an embodiment of the present invention, and as shown in fig. 4(c), different blood vessels of a target organ are segmented and represented by different colors, it should be noted that the sagittal section of the target organ refers to a left and right section that divides the target organ into a left and right part, and it should be noted that the sagittal section and the coronal section are perpendicular to each other.
As an alternative embodiment, the target is subjected to three-dimensional imaging
Performing retraction processing and external expansion processing, comprising: determining ventricles and atria in the target three-dimensional image; determining the geometric centers of the ventricles and the atria; the points on the boundary of the ventricle and the atrium are respectively connected with the corresponding geometric centers, the points on the boundary of the ventricle and the atrium are controlled to move along the connecting line to the direction of the corresponding geometric center for retraction processing, and the points on the boundary of the ventricle and the atrium are controlled to move along the connecting line to the opposite direction of the corresponding geometric center for expansion processing.
Fig. 5 is a flow chart of the segmentation of various parts of the heart organ according to the embodiment of the invention, as shown in fig. 5, taking the coronary vessel segmentation of the heart as an example.
S1, the CTA image is converted into a 2D image slice (i.e., a two-dimensional image slice).
S2, normalization processing is performed on the obtained 2D image slice.
S3, training a heart organ segmentation model on the GPU using the 2D image dataset.
And S4, performing normalization processing on the 3D image needing prediction according to the steps 1 and 2.
And S5, obtaining the segmentation result of each part of the heart organ by the slice generated in the step 4 through a deep learning model.
And S6, orderly splicing the segmentation results of each slice to obtain the segmentation results of each part of the heart organ of the whole 3D image.
S7, since the coronary blood vessels are all attached to the atria of the ventricle, the geometric center of each atria of the ventricle is found, the point on the boundary of the atria of the ventricle is connected with the geometric center point, the point on the boundary is moved along the connection line toward the geometric center direction to be called retraction, the point on the boundary is moved along the connection line toward the opposite direction of the geometric center direction to be called expansion, and the effect graph after retraction and expansion is shown in fig. 6 (fig. 6 is the effect graph after organ retraction and expansion according to the embodiment of the present invention). Through experiments, when the ventricular atrium is retracted and extended by 10mm, the space region formed between the retracted and extended regions can include all coronary vessels, specifically as shown in fig. 7 (fig. 7 is a schematic diagram of the coronary vessels including the space region formed between the retracted and extended regions according to the embodiment of the present invention), so that the image information within the boundary after the ventricular atrium is retracted is deleted, and the image information (except for the aorta) outside the boundary after the ventricular atrium is extended is deleted, so that the image information of the lower aorta and the coronary vessel region can be retained, as shown in fig. 8.
It should be noted that the contraction and the expansion are common image preprocessing means, the main algorithm of the contraction is a corrosion algorithm, and the common algorithm of the expansion is an expansion algorithm, so that the generated image is clearer. The comparison of the effects of the image preprocessing will be described in detail below.
Fig. 8 is a schematic diagram comparing an original image of a heart organ with a pre-processed image, as shown in fig. 8, the left side is the original image of the heart, the middle is a result graph of heart segmentation obtained by inputting a machine learning model, and the right side is an effect graph after the original image is pre-processed, as can be seen from the above, in the middle graph, only all different blood vessels of the heart organ are segmented respectively, and after the pre-processing in the right graph, coronary blood vessels can be segmented separately, which is helpful for further research of researchers, and has great clinical significance.
Further, by the processing steps provided in the above alternative embodiment, a coronary vessel segmentation map of the heart may be obtained. The whole of the heart segmentation result will be described in detail below.
Fig. 9 is a schematic diagram of an overall 3D view of a segmentation result according to an embodiment of the present invention, and as shown in fig. 9, after a heart organ is segmented by a 2D image, the 2D image segmentation map is merged to obtain an overall effect map of the 3D image, which has a great clinical significance.
Further, the technical solution provided by the embodiment of the present invention can be applied not only to coronary vessel segmentation of heart organs, but also to other organs, including but not limited to: brain organs, limbs, and abdominal organs, etc. The following is a detailed description of an embodiment of the present invention applied to the abdominal cavity organ.
Fig. 10 is a diagram of a segmentation result of an abdominal organ according to an embodiment of the present invention, as shown in fig. 10, in which a column in fig. 10 is a cross-sectional image of an initial CTA image, and a predicted segmentation map of a cross-section of the CTA image in a column B in fig. 10 can be obtained by inputting a machine learning model, then a real labeled map of the cross-section of the CTA image in a training process in a column C in fig. 10, and finally a corresponding position map of the predicted segmentation map in the cross-section of the CTA image in a column D in fig. 10 is obtained through several iterative training of machine learning.
It can be known from the above that, in the embodiment of the present invention, first, a 3D image is converted into a 2D image slice, then, the 2D image is normalized, then, a segmentation model is trained on a GPU using a 2D image data set, then, the 3D image to be predicted is continuously converted into the 2D image slice and normalized, the obtained slice data is sequentially subjected to segmentation prediction through the model, then, the predicted segmentation result is sequentially spliced to generate a predicted 3D segmentation result graph, finally, the heart organ segmentation result in the predicted 3D segmentation result graph generated in the previous step is deleted from the original image, the preprocessed coronary artery blood vessel extraction image is obtained after the internal contraction and the external expansion, the coronary artery can be preprocessed quickly, accurately and efficiently, and also, the non-coronary artery part organs in the CTA image can be accurately segmented, then, the non-coronary blood vessel part is automatically deleted, so that the pretreatment of the coronary blood vessel in the CTA image can be accurately and efficiently completed.
Example 2
According to another aspect of the embodiments of the present invention, there is also provided a medical image preprocessing apparatus, fig. 11 is a schematic diagram of a medical image preprocessing apparatus according to an embodiment of the present invention, as shown in fig. 11, the medical image preprocessing apparatus includes: a first acquisition module 1101, a segmentation module 1103, and a stitching module 1105. The following describes the apparatus for preprocessing the medical image.
A first acquisition module 1101 is configured to acquire a two-dimensional image sequence of a predetermined organ of a target object.
A segmentation module 1103, configured to segment each two-dimensional image slice in the two-dimensional image sequence according to each part of a predetermined organ through a predetermined segmentation model, to obtain a segmented two-dimensional image sequence, where the predetermined segmentation model is obtained by using multiple sets of training data through machine learning training, and each set of training data in the multiple sets of training data includes: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices.
A stitching module 1105, configured to stitch the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ.
It should be noted here that the first obtaining module 1101, the dividing module 1103, and the splicing module 1105 correspond to steps S102 to S106 in embodiment 1, and the modules are the same as the corresponding steps in the implementation example and application scenario, but are not limited to the disclosure in embodiment 1. It should be noted that the modules described above as part of an apparatus may be implemented in a computer system such as a set of computer-executable instructions.
As can be seen from the above, in the embodiment of the present invention, first, a two-dimensional image sequence of a predetermined organ of a target object may be acquired by the first acquisition module 1101; next, the segmentation module 1103 is used to segment each two-dimensional image slice in the two-dimensional image sequence according to each part of a predetermined organ through a predetermined segmentation model, so as to obtain a segmented two-dimensional image sequence, where the predetermined segmentation model is obtained by using multiple sets of training data through machine learning training, and each set of training data in the multiple sets of training data includes: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices; finally, the split two-dimensional image sequence is spliced by using the splicing module 1105 to obtain a target three-dimensional image of the predetermined organ. By the medical image preprocessing device, the purpose of splicing a target three-dimensional image of a predetermined organ based on a new two-dimensional image sequence obtained by slicing and dividing a two-dimensional image in the two-dimensional image sequence of the predetermined organ is achieved, so that the technical effect of improving the segmentation efficiency of the predetermined organ is achieved, and the technical problem that in the related art, the traditional blood vessel segmentation mode based on deep learning is to segment the whole image, the segmentation effect is poor, and the segmented result needs to manually remove the non-target blood vessel part in the segmentation result, so that the blood vessel segmentation efficiency is low is solved.
Optionally, the obtaining module includes: a first acquisition unit for acquiring an original three-dimensional image of a predetermined organ; the first processing unit is used for carrying out slicing processing on the original three-dimensional image to obtain a plurality of two-dimensional image slices; and the second acquisition unit is used for obtaining a two-dimensional image sequence based on the two-dimensional image slice.
Optionally, the apparatus for preprocessing medical images further comprises: the first processing module is used for carrying out normalization processing on each two-dimensional image slice in the two-dimensional image sequence before each two-dimensional image slice in the two-dimensional image sequence is divided according to each part of a preset organ through a preset division model.
Optionally, the apparatus for preprocessing medical images further comprises: the second acquisition module is used for obtaining a preset segmentation model through machine learning training by utilizing a plurality of groups of training data before segmenting each two-dimensional image slice in the two-dimensional image sequence according to each part of a preset organ through the preset segmentation model; wherein, the second acquisition module includes: the third acquisition unit is used for acquiring a plurality of historical original three-dimensional images; the second processing unit is used for carrying out slicing processing on the plurality of historical original three-dimensional images to obtain a plurality of groups of historical two-dimensional image slice sets; the third processing unit is used for carrying out normalization processing on each historical two-dimensional image slice in each historical two-dimensional image set in the multiple sets of historical two-dimensional image sets to obtain multiple sets of historical two-dimensional image slice sets after normalization processing; and the training unit is used for training the preset network model by utilizing the multiple groups of historical two-dimensional image slice sets after normalization processing and the historical segmentation results corresponding to the multiple groups of historical two-dimensional image slice sets on the GPU to obtain the preset segmentation model.
Optionally, the apparatus for preprocessing medical images further comprises: the second processing module is used for splicing the two-dimensional image sequences after the preset organ is a heart organ to obtain a target three-dimensional image of the preset organ, and then carrying out internal contraction processing and external expansion processing on the target three-dimensional image; the third acquisition module is used for acquiring a first area outside the target three-dimensional image after the target three-dimensional image is subjected to internal contraction processing and a second area outside the target three-dimensional image after the target three-dimensional image is subjected to external expansion processing; and the fourth acquisition module is used for deleting the first area and the second area in the target three-dimensional image to obtain images of the aorta and coronary vessel areas in the heart organ.
Optionally, the second processing module includes: a first determination unit for determining a ventricle and an atrium in the target three-dimensional image; a second determination unit for determining the geometric centers of the ventricles and the atria; and the connecting unit is used for respectively connecting the points on the boundary of the ventricle and the atrium with the corresponding geometric centers, controlling the points on the boundary of the ventricle and the atrium to move towards the direction of the corresponding geometric centers along the connecting line for retraction processing, and controlling the points on the boundary of the ventricle and the atrium to move towards the opposite direction of the corresponding geometric centers along the connecting line for expansion processing.
Example 3
According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium including a stored computer program, wherein when the computer program is executed by a processor, the apparatus where the computer-readable storage medium is located is controlled to execute the method for preprocessing a medical image according to any one of the above.
Example 4
According to another aspect of the embodiments of the present invention, there is further provided a processor for executing a computer program, wherein the computer program executes to execute any one of the methods for preprocessing a medical image.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (10)
1. A method for preprocessing a medical image, comprising:
acquiring a two-dimensional image sequence of a predetermined organ of a target object;
segmenting each two-dimensional image slice in the two-dimensional image sequence according to each part of the predetermined organ through a predetermined segmentation model to obtain a segmented two-dimensional image sequence, wherein the predetermined segmentation model is obtained by using multiple groups of training data through machine learning training, and each group of training data in the multiple groups of training data comprises: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices;
and splicing the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ.
2. The method of claim 1, wherein acquiring a two-dimensional image sequence of a predetermined organ of a target object comprises:
acquiring an original three-dimensional image of the predetermined organ;
slicing the original three-dimensional image to obtain a plurality of two-dimensional image slices;
and obtaining the two-dimensional image sequence based on the two-dimensional image slice.
3. The method of claim 1, wherein before segmenting each two-dimensional image slice in the two-dimensional image sequence according to the respective portion of the predetermined organ by a predetermined segmentation model, the method further comprises:
and carrying out normalization processing on each two-dimensional image slice in the two-dimensional image sequence.
4. The method of claim 1, wherein before segmenting each two-dimensional image slice in the two-dimensional image sequence according to the respective portion of the predetermined organ by a predetermined segmentation model, the method further comprises: obtaining the preset segmentation model through machine learning training by utilizing the multiple groups of training data;
wherein obtaining the predetermined segmentation model through machine learning training using the plurality of sets of training data comprises:
acquiring a plurality of historical original three-dimensional images;
slicing the plurality of historical original three-dimensional images to obtain a plurality of groups of historical two-dimensional image slice sets;
normalizing each historical two-dimensional image slice in each historical two-dimensional image set in the multiple sets of historical two-dimensional image sets to obtain multiple sets of historical two-dimensional image slice sets after normalization;
on a Graphic Processing Unit (GPU), training a preset network model by using the multiple groups of historical two-dimensional image slice sets after the normalization processing and historical segmentation results corresponding to the multiple groups of historical two-dimensional image slice sets to obtain the preset segmentation model.
5. The method according to any one of claims 1 to 4, wherein the predetermined organ is a heart organ, and after the step of stitching the segmented two-dimensional image sequences to obtain a target three-dimensional image of the predetermined organ, the method further comprises:
carrying out internal contraction processing and external expansion processing on the target three-dimensional image;
acquiring a first region outside the boundary of the target three-dimensional image after the target three-dimensional image is subjected to retraction processing and a second region outside the boundary of the target three-dimensional image after the target three-dimensional image is subjected to outward expansion processing;
and deleting the first region and the second region in the target three-dimensional image to obtain images of the aorta and coronary vessel regions in the heart organ.
6. The method of claim 5, wherein the performing the shrinking and expanding processes on the target three-dimensional image comprises:
determining ventricles and atria in the target three-dimensional image;
determining the geometric centers of the ventricle and the atrium;
and respectively connecting the points on the boundary of the ventricle and the atrium with the corresponding geometric centers, controlling the points on the boundary of the ventricle and the atrium to move along the connecting line to the direction of the corresponding geometric center for retraction processing, and controlling the points on the boundary of the ventricle and the atrium to move along the connecting line to the opposite direction of the corresponding geometric center for expansion processing.
7. A device for preprocessing a medical image, comprising:
the first acquisition module is used for acquiring a two-dimensional image sequence of a predetermined organ of a target object;
a segmentation module, configured to segment each two-dimensional image slice in the two-dimensional image sequence according to each part of the predetermined organ through a predetermined segmentation model to obtain a segmented two-dimensional image sequence, where the predetermined segmentation model is obtained through machine learning training using multiple sets of training data, and each set of training data in the multiple sets of training data includes: historical two-dimensional image slices and historical segmentation results corresponding to the historical two-dimensional image slices;
and the splicing module is used for splicing the segmented two-dimensional image sequence to obtain a target three-dimensional image of the predetermined organ.
8. The apparatus of claim 7, wherein the obtaining module comprises:
a first acquisition unit for acquiring an original three-dimensional image of the predetermined organ;
the first processing unit is used for carrying out slicing processing on the original three-dimensional image to obtain a plurality of two-dimensional image slices;
and the second acquisition unit is used for obtaining the two-dimensional image sequence based on the two-dimensional image slice.
9. A computer-readable storage medium, comprising a stored computer program, wherein when the computer program is executed by a processor, the computer-readable storage medium controls an apparatus to execute the method for preprocessing medical images according to any one of claims 1 to 6.
10. A processor, wherein the processor is configured to execute a computer program, wherein the computer program executes to execute the method for preprocessing medical images according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111471045.3A CN114299078A (en) | 2021-12-03 | 2021-12-03 | Medical image preprocessing method and device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111471045.3A CN114299078A (en) | 2021-12-03 | 2021-12-03 | Medical image preprocessing method and device and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114299078A true CN114299078A (en) | 2022-04-08 |
Family
ID=80965045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111471045.3A Withdrawn CN114299078A (en) | 2021-12-03 | 2021-12-03 | Medical image preprocessing method and device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114299078A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI838785B (en) * | 2022-07-06 | 2024-04-11 | 中國醫藥大學 | Three dimension medical image constructing method |
-
2021
- 2021-12-03 CN CN202111471045.3A patent/CN114299078A/en not_active Withdrawn
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI838785B (en) * | 2022-07-06 | 2024-04-11 | 中國醫藥大學 | Three dimension medical image constructing method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105118056B (en) | Coronary artery blood vessel extraction method based on three dimensional morphology | |
| CN109598722B (en) | Image analysis method based on recurrent neural network | |
| KR102050649B1 (en) | Method for extracting vascular structure in 2d x-ray angiogram, computer readable medium and apparatus for performing the method | |
| JP5225387B2 (en) | Volume rendering apparatus and method | |
| JP2002282235A (en) | Automatic segment dividing method for medical image | |
| Tozaki et al. | Pulmonary organs analysis for differential diagnosis based on thoracic thin-section CT images | |
| JPH11242739A (en) | Method for generating emphasized image of desired structure and device therefor | |
| Chen et al. | Coronary artery segmentation in cardiac CT angiography using 3D multi-channel U-net | |
| CN112184690B (en) | Coronary vessel trend prediction method, prediction model training method and device | |
| CN111815663A (en) | Hepatic vessel segmentation system based on Hessian matrix and gray scale method | |
| CN116503607B (en) | CT image segmentation method and system based on deep learning | |
| CN113140035B (en) | Full-automatic human cerebral vessel reconstruction method and device based on multi-modal image fusion technology | |
| Wang et al. | An automatic seeding method for coronary artery segmentation and skeletonization in CTA | |
| CN113935976A (en) | Method and system for automatically segmenting blood vessels in internal organs by enhancing CT (computed tomography) image | |
| CN114299078A (en) | Medical image preprocessing method and device and computer readable storage medium | |
| CN115705638A (en) | Medical image optimization method, system, electronic device and storage medium | |
| CN113470060A (en) | Coronary artery multi-angle curved surface reconstruction visualization method based on CT image | |
| CN113362327B (en) | Region segmentation method, device, terminal and storage medium in chest image | |
| CN112862827B (en) | Method, device, terminal and storage medium for determining opening parameters of left atrial appendage | |
| CN110378868B (en) | Pericardium segmentation method based on CT sequence and method for roughly positioning pericardium area from pericardium sequence center layer slice | |
| Seada et al. | Automatically seeded region growing approach for automatic segmentation of ascending aorta | |
| Alirr et al. | An automated liver vasculature segmentation from CT scans for hepatic surgical planning | |
| Hati et al. | Weakly Supervised Geodesic Segmentation of Egyptian Mummy CT Scans | |
| Alom et al. | Automatic slice growing method based 3D reconstruction of liver with its vessels | |
| CN113706568B (en) | Image processing method and device |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20220408 |
|
| WW01 | Invention patent application withdrawn after publication |