WO2020164468A1 - 医学图像分割方法、图像分割方法及相关装置、系统 - Google Patents

医学图像分割方法、图像分割方法及相关装置、系统 Download PDF

Info

Publication number
WO2020164468A1
WO2020164468A1 PCT/CN2020/074712 CN2020074712W WO2020164468A1 WO 2020164468 A1 WO2020164468 A1 WO 2020164468A1 CN 2020074712 W CN2020074712 W CN 2020074712W WO 2020164468 A1 WO2020164468 A1 WO 2020164468A1
Authority
WO
WIPO (PCT)
Prior art keywords
time
processed
medical image
cross
image
Prior art date
Application number
PCT/CN2020/074712
Other languages
English (en)
French (fr)
Inventor
王亮
张军
Original Assignee
腾讯科技（深圳）有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 腾讯科技（深圳）有限公司 filed Critical 腾讯科技（深圳）有限公司
Priority to EP20755319.9A priority Critical patent/EP3926537A4/en
Publication of WO2020164468A1 publication Critical patent/WO2020164468A1/zh
Priority to US17/239,532 priority patent/US11954864B2/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/149—Segmentation; Edge detection involving deformable models, e.g. active contour models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/64—Three-dimensional objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • G06T2207/10096—Dynamic contrast-enhanced magnetic resonance imaging [DCE-MRI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10116—X-ray image
    • 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
    • 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
    • G06T2207/30068—Mammography; Breast
    • 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
    • G06T2207/30096—Tumor; Lesion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/49—Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/695—Preprocessing, e.g. image segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • This application relates to the field of computer technology, specifically, to image segmentation technology.
  • Image segmentation is the technique and process of dividing an image into several specific areas with unique properties and extracting the target of interest.
  • medical image segmentation has become the top priority of medical analysis technology.
  • Medical image segmentation is the key to determining whether medical images can provide a reliable basis for clinical diagnosis and treatment.
  • the development of medical image segmentation technology not only affects the development of other related technologies in medical image processing, such as visualization, three-dimensional reconstruction, etc., but also occupies an extremely important position in the analysis of biomedical images.
  • the deep learning network model is trained by using medical images at all time points, and then the deep learning network model is used to compare the medical images Perform segmentation on the lesion area.
  • the embodiments of the present application provide a medical image segmentation method, an image segmentation method, and related devices, thereby reducing the workload of labeling at least to a certain extent, making the training of the medical image segmentation model easier and improving the diagnosis efficiency of doctors.
  • a medical image segmentation method including: acquiring a set of medical images to be processed, the set of medical images to be processed includes a plurality of medical images to be processed corresponding to different time points; The time point corresponding to the medical image to be processed and the medical image to be processed is processed in the time dimension on the medical image set to be processed to obtain a time dynamic image; the target area is extracted from the time dynamic image through the medical image segmentation model Features to obtain the target area.
  • a medical image segmentation device including: an acquisition module for acquiring a set of medical images to be processed, the set of medical images to be processed includes a plurality of medical images corresponding to different time points Image; processing module for processing the set of medical images to be processed in time dimension according to the time points corresponding to the medical image to be processed and the medical image to be processed to obtain time dynamic images; segmentation module for The target area feature is extracted from the time dynamic image through the medical image segmentation model to obtain the target area.
  • each of the to-be-processed medical image sets corresponds to a cross section
  • each of the to-be-processed medical image sets corresponds to all the sets of images at the same time point.
  • the medical image to be processed constitutes a three-dimensional medical image; based on the foregoing solution, the medical image segmentation device includes: a cross-section determination module, configured to determine the coordinate plane corresponding to the cross-section according to the three dimensions of the three-dimensional medical image and A cross-sectional coordinate axis, the cross-sectional coordinate axis being perpendicular to the coordinate plane.
  • the processing module includes: a four-dimensional data acquisition unit, configured to determine a relationship with the cross section according to the medical image data to be processed corresponding to the cross section and the time point. Corresponding four-dimensional data; a first time component acquisition unit for analyzing the four-dimensional data corresponding to the cross section to obtain the time component corresponding to each cross section; a second time component acquisition unit for The time component corresponding to each cross section determines the target time component corresponding to the plurality of medical image sets to be processed; the post-processing unit is configured to perform post-processing on the target time component to obtain the time dynamic image.
  • the four-dimensional data acquisition unit includes: a spatial coordinate determination unit, configured to determine the first coordinate, the second coordinate, and the cross section according to the three-dimensional medical image data corresponding to the cross section.
  • the first coordinate, the second coordinate, and the cross-sectional coordinate are perpendicular to each other;
  • the time coordinate determining unit is used to determine the time coordinate according to the time point;
  • the four-dimensional data determining unit is used to determine the time coordinate according to the first A coordinate, the second coordinate, the cross-sectional coordinate, and the time coordinate construct a four-dimensional coordinate axis, and the four-dimensional data is determined according to the four-dimensional coordinate axis.
  • the first time component acquiring unit includes: a first image data acquiring unit, configured to determine a target cross-section according to the cross-sectional coordinates, and obtain a cross-section with the target Corresponding first image data, the first image data includes the first coordinates, the second coordinates, and the time coordinates; a multi-dimensional analysis unit for multi-dimensional analysis of the first image data to obtain The time component corresponding to the target cross section; repeat the above steps until the time component corresponding to each of the cross sections is obtained.
  • the second time component acquisition unit is configured to: according to the time component corresponding to each cross section, the first coordinate and the first coordinate corresponding to each cross section Two coordinates and cross-sectional coordinates determine the target time component.
  • the multi-dimensional analysis unit is configured to perform multi-dimensional analysis of the first image data through three-dimensional Clifford algebra to obtain a time component corresponding to the target cross section .
  • the post-processing unit is configured to: according to the target time component, respectively determine the sub-time component corresponding to each cross section at each time point, The number of the sub-time components is the same as the number of the time points; the sub-time components corresponding to each cross section are respectively added and averaged to obtain a target average value; and constructed according to the target average value The time dynamic image.
  • the post-processing unit is configured to: according to the target time component, respectively determine the sub-time component corresponding to each cross section at each time point, The number of the sub-time components is the same as the number of the time points; the maximum value of the sub-time components corresponding to each cross section is obtained respectively; the maximum value of the sub-time components is constructed according to the maximum value of the sub-time components. Time dynamic image.
  • the post-processing unit is configured to: according to the target time component, respectively determine the sub-time component corresponding to each cross section at each time point, The number of the sub-time components is the same as the number of the time points; the maximum value and the minimum value of the sub-time components corresponding to each cross section are obtained, and the maximum value and the minimum value are calculated Obtain the target difference; construct the time dynamic image according to the target difference.
  • the medical image segmentation device further includes: a sample acquisition module for acquiring temporal dynamic image samples and target region annotation samples corresponding to the temporal dynamic samples; training module , Used to train the medical image segmentation model to be trained according to the temporal dynamic image sample and the target region label sample to obtain the medical image segmentation model.
  • the three-dimensional medical image is a three-dimensional dynamic contrast enhanced magnetic resonance imaging image.
  • a medical image segmentation system including: a detection device for scanning and detecting a detection object to obtain a set of medical images to be processed, the set of medical images to be processed includes a plurality of Corresponding to medical images to be processed at different time points; electronic equipment, the electronic equipment is connected to the detection equipment, and the electronic equipment includes a storage device and a processor, wherein the storage device is used to store one or more programs, When the one or more programs are executed by the processor, the processor is caused to implement the aforementioned medical image segmentation method.
  • an image segmentation method including: acquiring a set of images to be processed, the set of images to be processed includes a plurality of images to be processed corresponding to different time points; At the time point corresponding to the image to be processed, the image set to be processed is processed in the time dimension to obtain a time dynamic image; the target area feature is extracted from the time dynamic image through an image segmentation model to obtain the target area.
  • the images to be processed corresponding to the same time point in each of the image sets to be processed form a three-dimensional image
  • the multiple The images to be processed corresponding to different time points form a three-dimensional image sequence.
  • the medical image set to be processed includes a plurality of medical images to be processed corresponding to different time points; and then according to the medical image to be processed and the corresponding time Perform time dimension processing on the medical image set to be processed to obtain a time dynamic image.
  • the time dynamic image is collected from the image to be processed and can reflect the changes of image data at different time points. Therefore, the time dynamic image can be processed by the medical image segmentation model to extract the target area from the time dynamic image.
  • this method extracts a time dynamic image from a set of images to be processed, and replaces multiple medical images to be processed corresponding to different time points by the time dynamic image, reducing the number of images, ensuring the accuracy of medical image segmentation, and effectively helping doctors make the most accurate Under the premise of the treatment plan, the workload of labeling is reduced, the training of the medical image segmentation model is easier, and the diagnosis efficiency is improved.
  • FIG. 1 shows a schematic diagram of an exemplary system architecture to which the technical solutions of the embodiments of the present application can be applied;
  • Figure 2 schematically shows a flowchart of a medical image segmentation method according to an embodiment of the present application
  • FIG. 3 schematically shows a structure diagram of a set of to-be-processed medical images corresponding to multiple different time points in a to-be-processed medical image according to an embodiment of the present application;
  • Figure 4 schematically shows a cross-sectional distribution diagram according to an embodiment of the present application
  • Fig. 5 schematically shows a flow chart of acquiring a time dynamic image according to an embodiment of the present application
  • FIG. 6 schematically shows a flowchart of mining topic words and opinion words contained in target text data according to an embodiment of the present application to obtain target topic words and target opinion words with a confidence higher than a predetermined value;
  • Figures 7A-7C schematically show DCE-MRI images at a certain time point after the injection of a contrast agent in the related art
  • Fig. 9 schematically shows a flowchart of training a medical image segmentation model to be trained according to an embodiment of the present application
  • 10A-10C schematically show a schematic diagram of an interface for performing tumor segmentation on a background enhancement type medical image according to an embodiment of the present application
  • Fig. 11 schematically shows a flowchart of an image segmentation method according to an embodiment of the present application
  • Fig. 12 schematically shows a block diagram of a medical image segmentation device according to an embodiment of the present application
  • Fig. 13 schematically shows a structural diagram of a medical image segmentation system according to an embodiment of the present application
  • Fig. 14 shows a schematic structural diagram of a computer system suitable for implementing an electronic device according to an embodiment of the present application.
  • FIG. 1 shows a schematic diagram of an exemplary system architecture to which the technical solutions of the embodiments of the present application can be applied.
  • the system architecture 100 may include terminal devices (such as one or more of a desktop computer 101, a tablet computer 102, and a portable computer 103 as shown in FIG. 1, and of course, it may also be other terminals with a display screen. Equipment, etc.), network 104 and server 105.
  • the network 104 is used as a medium for providing a communication link between the terminal device and the server 105.
  • the network 104 may include various connection types, such as wired communication links, wireless communication links, and so on.
  • the numbers of terminal devices, networks, and servers in FIG. 1 are merely illustrative. According to implementation needs, there can be any number of terminal devices, networks and servers.
  • the server 105 may be a server cluster composed of multiple servers.
  • a user can use a desktop computer 101 (or a tablet computer 102 or a portable computer 103) to upload a set of medical images to be processed to the server 105, and the medical images to be processed included in the set of medical images to be processed can be It is any inspection image, such as computer tomography (Computed Tomography, CT), that is, the image, it can be a Magnetic Resonance Imaging (MRI) image, or other image information that changes over time Check the inspection image.
  • CT computer tomography
  • MRI Magnetic Resonance Imaging
  • the server 105 After acquiring the medical image set to be processed, the server 105 performs time-dimensional processing on the medical image set to be processed according to the medical image to be processed and the corresponding time point, so as to convert the original medical image set to be processed into a time dynamic image.
  • the time dynamic image can effectively reflect the difference between the lesion area and the non-lesion area; then the time dynamic image is input into the trained medical image segmentation model, and the target area feature in the time dynamic image is extracted through
  • the technical solution of this embodiment obtains time dynamic images by processing multiple medical image sets to be processed in the time dimension, so that the time dynamic images can clearly present the lesion area, avoiding the need for doctors to obtain multiple medical images at different time points.
  • the image is analyzed to determine the lesion area, which further improves the diagnosis efficiency of the doctor, and segmenting the target area in the time dynamic image through the medical image segmentation model can improve the segmentation accuracy of the lesion area and provide support for the doctor's clinical diagnosis. Since the time dynamic image replaces multiple medical images acquired at different time points, the workload of labeling is reduced, and the training of the medical image segmentation model is easier.
  • the medical image segmentation method provided by the embodiment of the present application is generally executed by the server 105, and accordingly, the medical image segmentation device is generally set in the server 105.
  • the terminal device may also have a similar function to the server, so as to execute the medical image segmentation solution provided in the embodiments of the present application.
  • the amount of image information is huge, such as 5 time points.
  • the embodiment of the present application first proposes a medical image segmentation method.
  • the implementation details of the technical solution of the embodiment of the present application are described in detail below:
  • FIG. 2 schematically shows a flowchart of a medical image segmentation method according to an embodiment of the present application.
  • the medical image segmentation method may be executed by a server, and the server may be the server shown in FIG. 1.
  • the medical image segmentation method includes at least S210 to S230, and the details are as follows:
  • the server obtains a set of medical images to be processed, and the set of medical images to be processed includes a plurality of medical images to be processed corresponding to different time points.
  • the medical image set to be processed may be an image set obtained by performing sample detection on samples obtained from a patient's body in various departments of the hospital or performing a physical examination on the patient, such as the medical image set to be processed.
  • the processed medical images can be CT scan images, MRI images, X-ray images, or other images that can detect changes in the characteristics of the lesion area over time. This application does not specify the types of medical images to be processed limited.
  • the MRI image will be used as the medical image to be processed in the following, and the segmentation of the MRI image of the breast tumor will be taken as a routine description.
  • the MRI image can also be the lung tumor, stomach tumor, MRI images of liver tumors and other diseased areas obtained by MRI.
  • contrast agent is chemicals that are injected (or taken) into human tissues or organs to enhance the effect of image observation, such as iron, manganese and other magnetic substances.
  • the density of these products is higher or lower than the surrounding tissues.
  • Contrast agents can change the relaxation rate of water protons in local tissues in the body, improve the imaging contrast and resolution of non-lesion areas and lesion areas, and provide more information for the location and diagnosis of lesion areas.
  • the content of the contrast agent in the tissue changes with the flow of blood.
  • the blood circulation in the non-lesion area is smooth, the content of the contrast agent will decrease rapidly, and the brightness of the area corresponding to the non-lesion area in the magnetic resonance image gradually increases.
  • the contrast agent content decreases slowly, and the brightness of the area corresponding to the lesion area in the MRI image changes slowly, which in turn forms a clear contrast with the brightness of the non-lesion area. Therefore, the present application can collect medical images to be processed at multiple time points on the same tissue cross-section based on this feature of the contrast agent.
  • the server may obtain one or more medical image sets to be processed.
  • the server may perform medical image segmentation on 2D images;
  • each medical image set to be processed corresponds to a cross section, and each medical image set to be processed corresponds to the medical image to be processed at the same time point to form a three-dimensional medical image.
  • the 3D image can be Medical image segmentation.
  • the embodiments of the present application mainly introduce medical image segmentation of 3D images.
  • each medical image set to be processed includes multiple medical image sets to be processed corresponding to different time points, that is, each medical image set to be processed
  • the medical images to be processed in the same medical image set to be processed are the images obtained by collecting information on the same cross-section at different time points.
  • the medical image set is the image obtained when the MRI machine scans different cross-sections of the breast during the scanning process; the multiple medical images to be processed in the same medical image set to be processed are the images obtained by the MRI machine during the scanning process.
  • the time points corresponding to the multiple medical images to be processed may be consecutive time points, and a medical image sequence can be formed through the multiple medical image sets to be processed.
  • Figure 3 shows a schematic diagram of the structure of the medical image to be processed corresponding to multiple different time points in the medical image set to be processed.
  • the cross section marked i contains n+1 medical images to be processed.
  • the medical images to be processed corresponding to the same time point in each medical image to be processed may form a three-dimensional medical image.
  • the medical images to be processed are corresponding to the same
  • the images composed of medical images to be processed at a time point are 3D MRI images. Since each medical image set to be processed contains medical images to be processed at multiple time points, the multiple medical image sets to be processed can form a 3D MRI image sequence.
  • dynamic contrast-enhanced magnetic resonance imaging can be performed on breast tumors to obtain 3D DCE-MRI images.
  • the medical image sequence formed by multiple sets of medical images to be processed is three-dimensional dynamic contrast-enhanced magnetic resonance imaging (3D DCE-MRI).
  • MRI three-dimensional dynamic contrast-enhanced magnetic resonance imaging
  • the medical image sequence may also be a two-dimensional image sequence. Since most of the medical images are 3D medical images in clinical diagnosis, the following description mainly takes a 3D DCE-MRI image sequence as an example.
  • time-dimensional processing is performed on the to-be-processed medical image set to obtain a time dynamic image.
  • the medical image set to be processed can be processed in the time dimension according to the medical images to be processed and their corresponding time points to obtain Time dynamic image. It is worth noting that before the time-dimensional processing of the medical image set to be processed, the coordinate plane and the cross-sectional coordinate axis corresponding to the cross section can be determined according to the three dimensions of the three-dimensional medical image.
  • the coordinate system corresponding to the three-dimensional medical image can be It is a three-dimensional Cartesian coordinate system.
  • the coordinate plane perpendicular to the cross-sectional coordinate axis is the coordinate plane corresponding to the cross-section.
  • 4 shows a schematic diagram of a cross-sectional distribution.
  • the z-axis can be defined as the cross-sectional coordinate axis, that is, the set of medical images to be processed is distributed along the z-axis.
  • the xy coordinate plane is the coordinate plane corresponding to the cross section, that is, any medical image to be processed is an image in the xy coordinate system.
  • FIG. 5 shows a schematic diagram of the flow of acquiring time dynamic images.
  • the medical image set to be processed is processed in the time dimension according to the medical image to be processed and its corresponding time point.
  • the process of acquiring time dynamic images specifically includes the following S510 to S540, which are described in detail as follows:
  • each cross-section corresponds to the medical image to be processed at multiple time points, and the coordinates corresponding to each cross-section are different. Therefore, firstly, according to the medical image data to be processed corresponding to the cross-section Determine the first, second, and cross-sectional coordinates, where the first, second, and cross-sectional coordinates are perpendicular to each other, and the first and second coordinates form the coordinate plane corresponding to the cross-section; then determine according to the time point Time coordinates; Finally, a four-dimensional coordinate axis is formed according to the first coordinate, the second coordinate, the cross-sectional coordinate and the time coordinate, and the four-dimensional data is determined according to the four-dimensional coordinate axis.
  • the first coordinate may specifically be the x-axis
  • the second coordinate may specifically be the y-axis
  • the cross-sectional coordinate may specifically be the z-axis.
  • the time coordinate t can be determined according to the time point corresponding to the medical image to be processed, and then a four-dimensional coordinate system (x, y, z, t) can be determined according to the first coordinate, the second coordinate, the cross-sectional coordinate and the time coordinate.
  • FIG. 6 shows a schematic flow chart of obtaining the time component corresponding to each cross section.
  • the process of obtaining the time component corresponding to each cross section includes the following S610 to S630 , The detailed description is as follows:
  • a target cross section is determined according to the cross section coordinates, and first image data corresponding to the target cross section is acquired.
  • the first image data includes the first coordinates, the second coordinates, and the The time coordinate.
  • a coordinate value i can be determined from the z-axis as the cross-sectional coordinate, and the corresponding target cross-section can be obtained according to the cross-sectional coordinate.
  • the target cross-section contains multiple data corresponding to different time points.
  • the first image data can be determined by counting the image data of the medical images to be processed.
  • the first image data includes the first coordinates, second coordinates, and time coordinates corresponding to each medical image to be processed, that is to say ,
  • a multi-dimensional analysis is performed on the first image data to obtain a time component corresponding to the target cross section.
  • the first image data in order to process the medical image to be processed in the time dimension, may be analyzed in a multi-dimensional manner to obtain the components of the first image data in the time dimension.
  • the three-dimensional Clifford algebra may be used to calculate the multi-dimensional analytical signal ⁇ (x, y, t) of the first image data f(x, y, t) to split the original signal into components of different dimensions.
  • Clifford algebra also known as geometric algebra (Geometric algebra)
  • the multi-dimensional analytical signal ⁇ (x,y,t) obtained through the calculation of the three-dimensional Clifford algebra is as follows:
  • the formula consists of 8 components, the direction of each component is determined by e 1 , e 2 , and e 3 , and these 8 components are mutually orthogonal in the Clifford algebraic space .
  • e 1 corresponds to the information in the x direction in the space of the first image data f(x, y, t)
  • e 2 corresponds to the information in the y direction in the space of the first image data f(x, y, t)
  • e 3 It corresponds to the information in the t direction in the space of the first image data f(x, y, t).
  • the information in the t direction is what we are concerned about, so the component f(x,y,t)*** ⁇ (x) can be extracted from the multidimensional signal ⁇ (x,y,t) ) ⁇ (y)e 3 /( ⁇ t) ⁇
  • S610 and S620 may be repeated to obtain the time components corresponding to each cross section.
  • a target time component corresponding to the plurality of medical image sets to be processed is determined according to the time component corresponding to each cross section.
  • the target time components may be post-processed to obtain a three-dimensional time dynamic image.
  • the target time component I'(x, y, z, t) is post-processed in other ways, which will not be repeated in this application.
  • the specific process of calculating the average value of the target time component I'(x, y, z, t) along the t axis includes: according to the target time component, respectively determining each time point at each time point The number of sub-time components corresponding to the cross-section is the same as the number of time points; then the sub-time components corresponding to each cross-section are respectively added and averaged to obtain the target average value; finally according to the target average Value build time dynamic image.
  • the specific calculation formula of time dynamic image is shown in formula (1):
  • the specific process of calculating the maximum value of the target time component I'(x, y, z, t) along the t axis includes: according to the target time component, the specific process of calculating the target time component at each time point The number of sub-time components corresponding to the cross-section is the same as the number of time points; then the maximum value of the sub-time components corresponding to each cross-section is obtained respectively; and finally the maximum value of the sub-time components is constructed Time dynamic image.
  • the specific expression of the time dynamic image is shown in formula (2):
  • the specific process of calculating the difference between the maximum value and the minimum value of the target time component I'(x, y, z, t) along the t axis includes: according to the target time component, the specific process The number of sub-time components corresponding to each cross-section at each time point is the same as the number of time points; the maximum and minimum values of the sub-time components corresponding to each cross-section are obtained, and the maximum The difference between the value and the minimum value is the target difference; finally, the time dynamic image is constructed according to the target difference.
  • the specific calculation formula of the time dynamic image is shown in formula (3):
  • I t (x,y,z) Max(I'(x,y,z,t))
  • t t 1 ,t 2 ,...,t n -Min(I'(x,y,z ,t))
  • t t 1 ,t 2 ,...,t n (3)
  • a three-dimensional time dynamic image I t (x, y, z) can be obtained according to the calculation result, specifically, the change of pixel brightness of multiple 3D DCE-MRI images at different times is obtained. Since the difference between the maximum value and the minimum value of the sub-time component reflects the difference between the maximum value and the minimum value obtained at different time points of the 3D DCE-MRI image sequence at the same spatial position, it can display this space to the maximum Therefore, in the embodiment of the present application, the target difference value may be used as the criterion, and the time dynamic image can be constructed according to the target difference value to improve the efficiency of segmentation of medical images.
  • a medical image segmentation model is used to extract target area features from the temporal dynamic image to obtain the target area.
  • the temporal dynamic image after acquiring the temporal dynamic image, can be input to the medical image segmentation model, and the target area feature is extracted from the temporal dynamic image through the medical image segmentation model to obtain the target area.
  • the time dynamic image is a three-dimensional time dynamic image.
  • the medical image segmentation model may be a trained deep learning segmentation model.
  • the medical image segmentation model may be a deep learning segmentation model specially used for processing three-dimensional images, such as The 3D Unet model, the 3D vnet model, the fully convolutional neural network model, etc., the type of the deep learning segmentation model is not specifically limited in the embodiment of this application.
  • the target area is a diseased area, such as a tumor area, a calcified area, and so on. Medical workers can regard the target area as the area of interest and perform further analysis on the area of interest in order to formulate the best treatment plan.
  • Figures 7A-7C show DCE-MRI images at a certain time point after the injection of the contrast agent in a related technology, as shown in Figures 7A-7C, and Figure 7A is a cross-sectional image of the breast.
  • Figure 7B is a longitudinal section of the upper and lower parts;
  • Figure 7B is a sagittal image of the breast, which is a longitudinal section that divides the body into left and right parts;
  • Figure 7C is a coronal image of the breast, which divides the body into front and rear Two parts of the longitudinal section, where the rectangular frame in Figures 7A-7C is the lesion area, and the elliptical frame is the non-lesion area.
  • both the lesion area and the non-lesion area are presented as highlighted pixels, and the medical image segmentation model cannot distinguish the lesion area from the non-lesion area, and thus cannot accurately segment the lesion area.
  • Figures 8A-8C show three-dimensional time dynamic images after the injection of the contrast agent in this application, as shown in Figures 8A-8C,
  • Figure 8A is a cross-sectional image of the breast
  • Figure 8B is a sagittal image of the breast
  • Figure 8C It is a coronal image of the breast, where the rectangular frame in Figures 8A-8C is the lesion area, and the elliptical frame is the non-lesion area.
  • the pixel brightness of the lesion area in Figure 8A-8C is higher, the pixel brightness of the non-lesion area is lower, and the contrast between the pixels of the lesion area and the non-lesion area is more obvious .
  • the medical image segmentation model can quickly distinguish the lesion area from the non-lesion area, and then can accurately segment the lesion area.
  • the medical image segmentation model before the temporal dynamic image is input to the medical image segmentation model, and the target region feature is extracted from the temporal dynamic image through the medical image segmentation model to obtain the target region, the medical image segmentation model can also be trained Training is performed to obtain a medical image segmentation model for subsequent image segmentation of the medical image to be processed.
  • FIG. 9 shows a schematic diagram of the training process of the medical image segmentation model to be trained.
  • the process of training the medical image segmentation model to be trained includes the following S910 to S920, which are described in detail as follows:
  • a temporal dynamic image sample and a target area annotation sample corresponding to the temporal dynamic image sample are acquired.
  • the method for acquiring the temporal dynamic image sample is the same as the method for acquiring the temporal dynamic image in the foregoing embodiment, and will not be repeated here.
  • the target area (lesion area) in the temporal dynamic image can be marked manually by means of manual labeling, so as to obtain the target area labeled sample corresponding to the temporal dynamic image sample.
  • multiple samples may be used to train the medical image segmentation model to be trained to obtain the medical image to be trained
  • the optimal parameters of the segmentation model Specifically, 3D DCE-MRI image data of 244 cases of malignant tumor patients can be selected, of which 221 cases of data are used as training data for training the medical image segmentation model to be trained; 23 cases of data are used as test data for training medical images.
  • the image segmentation model is tested to determine whether it has reached a stable state.
  • the number of training data and test data in this application includes but is not limited to the above examples, which is not specifically limited in this application.
  • the medical image segmentation model to be trained is trained according to the temporal dynamic image sample and the target region label sample to obtain the medical image segmentation model.
  • the temporal dynamic image sample can be input to the medical image segmentation model to be trained to obtain the extracted medical image segmentation model.
  • Target area then compare the extracted target area with the target area labeled samples corresponding to the input dynamic image samples to determine the segmentation accuracy of the medical image segmentation model to be trained.
  • the preset threshold can be set according to actual needs, for example, set to 95% and so on. After the training is completed, the trained medical image segmentation model can be tested through test data to determine whether it is widely applicable to any time dynamic image.
  • the segmentation accuracy of the lesion area obtained through the technical solution of the embodiment of the present application is compared with that of using the original DCE-MRI data to train the medical image segmentation model to segment the medical image to be processed to obtain the segmentation of the lesion area.
  • the accuracy has been greatly improved.
  • Table 1 shows the experimental results of using original DCE-MRI data and using three-dimensional time dynamic images, as shown in Table 1:
  • FIGS 10A-10C show a schematic diagram of an interface for performing tumor segmentation on background-enhanced medical images, as shown in Figures 10A-10C, and Figure 10A shows the labeling results of background-enhanced medical images; Figure 10B shows The segmentation result of the background-enhanced medical image through the three-dimensional temporal dynamic image is shown, and the segmentation accuracy of this method reaches 87%; Figure 10C shows the segmentation result of the background-enhanced medical image through the original DCE-MRI data. The segmentation accuracy of this method is 69%. It is thus explained that the technical solution of the embodiment of the present application can improve the accuracy of medical image segmentation, can segment various types of medical images, and has a wider application range.
  • the technical solutions of the above-mentioned embodiments of the present application can extract a piece of 3D data (three-dimensional temporal dynamic image) from multiple 3D DCE-MRI data, and directly use it for labeling the target area and training the medical image segmentation model, so that the medical image segmentation model Training is more convenient; it also avoids the need for doctors to choose which time point to read the 3D DCE-MRI image first when doing MRI image diagnosis, instead, they can observe the 3D image obtained by watching the technical solution of the embodiment of this application. For the lesion area, 3D DCE-MRI images at certain time points are further selected to further improve the diagnosis efficiency.
  • FIG. 11 shows a flowchart of the image segmentation method.
  • S1110 a set of images to be processed is obtained.
  • the image set includes a plurality of images to be processed corresponding to different time points; in S1120, according to the time points corresponding to the image to be processed and the image to be processed, the image set to be processed is processed in the time dimension to obtain the time Dynamic image;
  • the target area feature is extracted from the temporal dynamic image through the image segmentation model to obtain the target area.
  • the image segmentation method is similar to the medical image segmentation method in the above embodiment, but this method can not only segment medical images, but also can be used to segment any other types of images, such as segmentation of sample images in biological experiments. , Segmentation of the image in the metal processing process, segmentation of the damage location in the pipeline, etc., as long as the characteristics of some areas in the image are different from those of other areas over time, this can be used.
  • the image segmentation method in the application embodiment is used for segmentation. Further, the image segmentation method can be implemented in the specific implementation manner of the medical image segmentation method in the embodiment of the present application, and therefore, the details are not described herein again.
  • the images to be processed corresponding to the same time point in each of the image sets to be processed can form a three-dimensional image, while multiple images to be processed corresponding to different time points can be A three-dimensional image sequence is formed, and the three-dimensional image sequence can be identified and segmented by the image segmentation method in the embodiment of the present application, and the target region therein can be obtained.
  • Fig. 12 schematically shows a block diagram of a medical image segmentation device according to an embodiment of the present application.
  • the medical image segmentation device 1200 includes: an acquisition module 1201, a processing module 1202, and a segmentation module 1203.
  • the acquiring module 1201 is used to acquire a set of medical images to be processed, and the set of medical images to be processed includes a plurality of medical images to be processed corresponding to different time points;
  • the time point corresponding to the medical image to be processed is processed in the time dimension on the medical image set to be processed to obtain a time dynamic image;
  • the segmentation module 1203 is used to extract a target area from the time dynamic image through a medical image segmentation model Features to obtain the target area.
  • each of the to-be-processed medical image sets corresponds to a cross section
  • each of the to-be-processed medical image sets corresponds to all the sets of images at the same time point.
  • the medical image to be processed constitutes a three-dimensional medical image; based on the foregoing solution, the medical image segmentation device 1200 includes: a cross-section determination module 1204 for determining the coordinates corresponding to the cross-section according to the three dimensions of the three-dimensional medical image The plane and the cross-sectional coordinate axis, the cross-sectional coordinate axis is perpendicular to the coordinate plane.
  • the processing module 1202 includes: a four-dimensional data acquisition unit configured to determine, according to the medical image data to be processed corresponding to the cross section and the time point, the data corresponding to the cross section Four-dimensional data; a first time component acquisition unit for analyzing the four-dimensional data corresponding to the cross section to obtain the time component corresponding to each of the cross sections; a second time component acquisition unit for analyzing each of the cross sections The time component corresponding to the cross section determines the target time component corresponding to the plurality of medical image sets to be processed; the post-processing unit is configured to perform post-processing on the target time component to obtain the time dynamic image.
  • the four-dimensional data acquisition unit includes: a spatial coordinate determination unit configured to determine the first coordinate, the second coordinate, and the cross-sectional coordinate according to the medical image data to be processed corresponding to the cross-section;
  • the first coordinate, the second coordinate, and the cross-sectional coordinate are perpendicular to each other;
  • a time coordinate determining unit is configured to determine a time coordinate according to the time point;
  • a four-dimensional data determining unit is configured to determine a time coordinate according to the first coordinate, The second coordinates, the cross-sectional coordinates, and the time coordinates construct a four-dimensional coordinate axis, and the four-dimensional data is determined according to the four-dimensional coordinate axis.
  • the first time component obtaining unit includes: a first image data obtaining unit, configured to determine a target cross section according to the cross section coordinates, and obtain the first time component corresponding to the target cross section An image data, the first image data includes the first coordinate, the second coordinate, and the time coordinate; a multi-dimensional analysis unit for multi-dimensional analysis of the first image data to obtain the target cross section The time component corresponding to the plane; repeat the above steps until the time component corresponding to each cross-section is obtained.
  • the second time component acquiring unit is configured to: according to the time component corresponding to each cross section, the first coordinate, the second coordinate and the cross section corresponding to each cross section The surface coordinates determine the target time component.
  • the multi-dimensional analysis unit is configured to perform multi-dimensional analysis of the first image data through three-dimensional Clifford algebra to obtain a time component corresponding to the target cross section.
  • the post-processing unit is configured to: according to the target time component, respectively determine the sub-time components corresponding to each of the cross sections at each time point, and the sub-time components
  • the number of time components is the same as the number of time points; the sub-time components corresponding to each cross section are respectively added and averaged to obtain the target average value; the time is constructed according to the target average value Dynamic images.
  • the post-processing unit is configured to: according to the target time component, respectively determine the sub-time components corresponding to each of the cross sections at each time point, and the sub-time components
  • the number of time components is the same as the number of time points; the maximum value of the sub-time components corresponding to each cross section is obtained respectively; the time dynamic image is constructed according to the maximum value of the sub-time components .
  • the post-processing unit is configured to: according to the target time component, respectively determine the sub-time components corresponding to each of the cross sections at each time point, and the sub-time components
  • the number of time components is the same as the number of time points; obtaining the maximum value and the minimum value of the sub-time components corresponding to each of the cross sections, and making the difference between the maximum value and the minimum value, To obtain a target difference; construct the time dynamic image according to the target difference.
  • the medical image segmentation device 1200 further includes: a sample acquisition module 1205 and a training module 1206.
  • the sample acquisition module 1205 is used to obtain time dynamic image samples and target area annotation samples corresponding to the time dynamic samples;
  • the training module 1206 is used to obtain training based on the time dynamic image samples and the target area annotation samples
  • the medical image segmentation model is trained to obtain the medical image segmentation model.
  • the three-dimensional medical image is a three-dimensional dynamic contrast enhanced magnetic resonance imaging image.
  • FIG. 13 shows a medical image segmentation system.
  • the medical image segmentation system 1300 includes a detection device 1301 and an electronic device 1302.
  • the detection device 1301 is used to scan and detect the detection object to obtain a set of medical images to be processed, the set of medical images to be processed includes a plurality of medical images to be processed corresponding to different time points; the electronic device 1302, the electronic The device is connected to the detection device, and the electronic device includes a storage device and a processor, wherein the storage device is used to store one or more programs, and when the one or more programs are executed by the processor, The processor is enabled to implement the above-mentioned medical image segmentation method.
  • the detection device 1301 may be a scanning device used to obtain scanned images in a CT device, and the scanning device includes a radiation source, a detector, and a scanning frame; it may be used in a nuclear magnetic resonance imaging device to obtain Scanning device for scanning images.
  • the scanning device includes a magnet part, a magnetic resonance spectrometer part and a scanning bed; it can also be a scanning device for acquiring scanned images in a fluoroscopy device.
  • the scanning device includes a radiation source and a detector. It may also be other detection equipment, as long as it can be used to scan the detection object to obtain a scanned image, which is not specifically limited in this application.
  • the medical image sets to be processed can be sent to the storage device 1302a and/or the processor 1302b in the electronic device 1302, and the storage device 1302a also stores one or more programs.
  • the processor 1302b can execute one or more programs stored in the storage device 1302a on the medical image set to be processed, that is, the processor 1302b can perform image segmentation on the medical image set to be processed according to the technical solution in the embodiment of the present application to
  • the target area is acquired.
  • the processor 1302b may also send an image containing the target area to a display device (not shown) connected to the electronic device 1302 for display, so that the doctor can observe and determine the lesion and formulate a treatment plan.
  • FIG. 14 shows a schematic structural diagram of a computer system suitable for implementing an electronic device 1302 of an embodiment of the present application.
  • the computer system 1400 includes a central processing unit (Central Processing Unit, CPU) 1401, which can be loaded into a random system according to a program stored in a read-only memory (Read-Only Memory, ROM) 1402 or from the storage part 1408. Access to the program in the memory (Random Access Memory, RAM) 1403 to execute various appropriate actions and processing. In RAM 1403, various programs and data required for system operation are also stored.
  • the CPU 1401, the ROM 1402, and the RAM 1403 are connected to each other through a bus 1404.
  • An Input/Output (I/O) interface 1405 is also connected to the bus 1404.
  • the following components are connected to the I/O interface 1405: input part 1406 including keyboard, mouse, etc.; output part 1407 including cathode ray tube (Cathode Ray Tube, CRT), liquid crystal display (LCD), etc., and speakers, etc. ; A storage part 1408 including a hard disk, etc.; and a communication part 1409 including a network interface card such as a LAN (Local Area Network) card and a modem.
  • the communication section 1409 performs communication processing via a network such as the Internet.
  • the driver 1410 is also connected to the I/O interface 1405 as needed.
  • a removable medium 1411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1410 as needed, so that the computer program read from it is installed into the storage portion 1408 as needed.
  • the process described below with reference to the flowchart can be implemented as a computer software program.
  • the embodiments of the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart.
  • the computer program may be downloaded and installed from the network through the communication part 1409, and/or installed from the removable medium 1411.
  • CPU central processing unit
  • the computer-readable medium shown in the embodiments of the present application may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above.
  • Computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable Erasable Programmable Read Only Memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical storage device, magnetic storage device, or any suitable of the above The combination.
  • the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.
  • a computer-readable signal medium may include a data signal propagated in a baseband or as a part of a carrier wave, and a computer-readable program code is carried therein.
  • This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.
  • the computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium.
  • the computer-readable medium may send, propagate, or transmit the program for use by or in combination with the instruction execution system, apparatus, or device .
  • the program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.
  • each block in the flowchart or block diagram may represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more for realizing the specified logic function Executable instructions.
  • the functions marked in the block may also occur in a different order from the order marked in the drawings. For example, two blocks shown in succession can actually be executed substantially in parallel, or they can sometimes be executed in the reverse order, depending on the functions involved.
  • each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be It is realized by a combination of dedicated hardware and computer instructions.
  • the units involved in the embodiments described in the present application can be implemented in software or hardware, and the described units can also be provided in a processor. Among them, the names of these units do not constitute a limitation on the unit itself under certain circumstances.
  • this application also provides a computer-readable medium.
  • the computer-readable medium may be included in the electronic device described in the above-mentioned embodiments; or it may exist alone without being assembled into the electronic device. in.
  • the foregoing computer-readable medium carries one or more programs, and when the foregoing one or more programs are executed by an electronic device, the electronic device realizes the method described in the foregoing embodiment.
  • modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory.
  • the features and functions of two or more modules or units described above may be embodied in one module or unit.
  • the features and functions of a module or unit described above can be further divided into multiple modules or units to be embodied.
  • the exemplary embodiments described herein can be implemented by software, or can be implemented by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (can be a CD-ROM, U disk, mobile hard disk, etc.) or on the network , Including several instructions to make a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) execute the method according to the embodiment of the present application.
  • a computing device which can be a personal computer, a server, a touch terminal, or a network device, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Quality & Reliability (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

本申请提供了一种医学图像分割方法、图像分割方法及相关装置。该医学图像分割方法包括：获取待处理医学图像集，待处理医学图像集包括多个对应不同时间点的待处理医学图像；根据待处理医学图像和待处理医学图像对应的时间点，对待处理医学图像集进行时间维度的处理，得到时间动态图像；通过医学图像分割模型从时间动态图像中提取目标区域特征，以获取目标区域。本申请的技术方案在保证医学图像分割精度的情况下，降低了标注的工作量，使得医学图像分割模型训练更加简便，提高了诊断效率。

Description

医学图像分割方法、图像分割方法及相关装置、系统
本申请要求于2019年2月15日提交中国专利局、申请号201910116353.0、申请名称为“医学图像分割方法、装置、系统及图像分割方法”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机技术领域，具体而言，涉及图像分割技术。
背景技术
图像分割是把图像分成若干个特定的、具有独特性质的区域并提取出感兴趣目标的技术和过程。随着计算机技术和医疗分析技术的发展，医学图像分割成为医疗分析技术的重中之重，医学图像分割是决定医学图像在临床诊疗中能否提供可靠依据的关键。医学图像分割技术的发展不仅影响到医学图像处理中其他相关技术的发展，如可视化、三维重建等，而且在生物医学图像的分析中也占有极其重要的地位。
由于病变区域和非病变区域在医疗试剂的作用下随着时间变化会出现不同的变化，相关技术中通过采用全部时间点的医学图像训练深度学习网络模型，然后再通过深度学习网络模型对医学图像上的病变区域进行分割。
但是通过深度学习网络模型分割病变区域虽然准确率高，但是在训练模型是需要对每个时间点的医学图像进行标注，标注的工作量较大，提高了训练模型的复杂度。
发明内容
本申请的实施例提供了一种医学图像分割方法、图像分割方法及相关装置，进而至少在一定程度上降低标注的工作量，使得医学图像分割模型训练更加简便，提高医生的诊断效率。
本申请的其他特性和优点将通过下面的详细描述变得显然，或部分地通过本申请的实践而习得。
根据本申请实施例的一个方面，提供了一种医学图像分割方法，包括：获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像；通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
根据本申请实施例的一个方面，提供了一种医学图像分割装置，包括：获取模块，用于获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；处理模块，用于根据所述待处理医学图像和所述待处理医学图像对应的时间点，对 所述待处理医学图像集进行时间维度的处理，得到时间动态图像；分割模块，用于通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
在本申请的一些实施例中，若获取到多个所述待处理图像集，各所述待处理医学图像集分别对应一横断面，并且各所述待处理医学图像集中对应同一时间点的所述待处理医学图像组成三维医学图像；基于前述方案，所述医学图像分割装置包括：横断面确定模块，用于根据所述三维医学图像的三个维度确定所述横断面对应的坐标面和横断面坐标轴，所述横断面坐标轴与所述坐标面垂直。
在本申请的一些实施例中，基于前述方案，所述处理模块包括：四维数据获取单元，用于根据所述横断面对应的待处理医学图像数据和所述时间点确定与所述横断面对应的四维数据；第一时间分量获取单元，用于对所述横断面对应的四维数据进行解析，获得各所述横断面对应的时间分量；第二时间分量获取单元，用于根据各所述横断面对应的时间分量确定与多个所述待处理医学图像集对应的目标时间分量；后处理单元，用于对所述目标时间分量进行后处理，得到所述时间动态图像。
在本申请的一些实施例中，基于前述方案，所述四维数据获取单元包括：空间坐标确定单元，用于根据所述横断面对应的三维医学图像数据确定第一坐标、第二坐标、横断面坐标，所述第一坐标、所述第二坐标和所述横断面坐标相互垂直；时间坐标确定单元，用于根据所述时间点确定时间坐标；四维数据确定单元，用于根据所述第一坐标、所述第二坐标、所述横断面坐标和所述时间坐标构建四维坐标轴，并根据所述四维坐标轴确定所述四维数据。
在本申请的一些实施例中，基于前述方案，所述第一时间分量获取单元包括：第一图像数据获取单元，用于根据所述横断面坐标确定目标横断面，获取与所述目标横断面对应的第一图像数据，所述第一图像数据包括所述第一坐标、所述第二坐标和所述时间坐标；多维解析单元，用于对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量；重复上述步骤，直至获取与各所述横断面对应的时间分量。
在本申请的一些实施例中，基于前述方案，所述第二时间分量获取单元配置为：根据各所述横断面对应的时间分量、与各所述横断面对应的第一坐标、第二坐标和横断面坐标确定所述目标时间分量。
在本申请的一些实施例中，基于前述方案，所述多维解析单元配置为：通过三维克利福德代数对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量。
在本申请的一些实施例中，基于前述方案，所述后处理单元配置为：根据所述目标时间分量，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；将与各所述横断面对应的所述子时间分量分别进行加和平均，以获取目标平均值；根据所述目标平均值构建所述时间动态图像。
在本申请的一些实施例中，基于前述方案，所述后处理单元配置为：根据所述目标时间分量，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；分别获取与各所述横断面对应的所述子时间分量中的最大值；根据所述子时间分量中的最大值构建所述时间动态图像。
在本申请的一些实施例中，基于前述方案，所述后处理单元配置为：根据所述目标时间分量，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；获取与各横断面对应的所述子时间分量中的最大值和最小值，并将所述最大值和所述最小值作差得到目标差值；根据所述目标差值构建所述时间动态图像。
在本申请的一些实施例中，基于前述方案，所述的医学图像分割装置还包括：样本获取模块，用于获取时间动态图像样本和与所述时间动态样本对应的目标区域标注样本；训练模块，用于根据所述时间动态图像样本和所述目标区域标注样本对待训练医学图像分割模型进行训练，得到所述医学图像分割模型。
在本申请的一些实施例中，基于前述方案，所述三维医学图像为三维动态对比度增强磁共振成像图像。
根据本申请实施例的一个方面，提供了一种医学图像分割系统，包括：检测设备，用于对检测对象进行扫描检测，以获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；电子设备，所述电子设备与所述检测设备连接，并且所述电子设备包括存储装置和处理器，其中所述存储装置用于存储一个或多个程序，当所述一个或多个程序被所述处理器执行时，使得所述处理器实现上述的医学图像分割方法。
根据本申请实施例的一个方面，提供了一种图像分割方法，包括：获取待处理图像集，所述待处理图像集包括多个对应不同时间点的待处理图像；根据所述待处理图像和所述待处理图像对应的时间点，对所述待处理图像集进行时间维度的处理，得到时间动态图像；通过图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
在本申请的一些实施例中，基于前述方案，若获取到多个所述待处理图像集，各所述待处理图像集中对应相同时间点的待处理图像形成一三维图像，并且所述多个对应不同时间点的待处理图像形成三维图像序列。
在本申请的一些实施例所提供的技术方案中，通过获取待处理医学图像集，待处理医学图像集包括多个对应不同时间点的待处理医学图像；然后根据待处理医学图像和对应的时间点对待处理医学图像集进行时间维度的处理，得到时间动态图像。时间动态图像是从待处理图像集中提取得到的，可以体现不同时间点的图像数据变化情况，故可以通过医学图像分割模型对时间动态图像进行处理以从所述时间动态图像中提取目标区域。可见，该方法从待处理图像集中提取一个时间动态图像，通过时间动态图像代替多个对应不同时间点的待处理医学图像，减少了图像的数量在保证医学图像分割精度，有效帮助医生制定最 准确的治疗方案的前提下，降低了标注的工作量，使得医学图像分割模型训练更加简便，提高了诊断效率。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本申请。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本申请的实施例，并与说明书一起用于解释本申请的原理。显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。在附图中：
图1示出了可以应用本申请实施例的技术方案的示例性系统架构的示意图；
图2示意性示出了根据本申请的一个实施例的医学图像分割方法的流程图；
图3示意性示出了根据本申请的一个实施例的待处理医学图像集中对应多个不同时间点的待处理医学图像的结构示意图；
图4示意性示出了根据本申请的一个实施例的横断面的分布示意图；
图5示意性示出了根据本申请的一个实施例的获取时间动态图像的流程图；
图6示意性示出了根据本申请的一个实施例的挖掘目标文本数据中包含的主题词和观点词，得到置信度高于预定值的目标主题词和目标观点词的流程图；
图7A-7C示意性示出了相关技术中注射造影剂后的某个时间点的DCE-MRI图像；
图8A-8C示意性示出了根据本申请的一个实施例的注射造影剂后的三维的时间动态图像；
图9示意性示出了根据本申请的一个实施例的对待训练医学图像分割模型进行训练的流程图；
图10A-10C示意性示出了据本申请的一个实施例的对背景强化类型的医学图像进行肿瘤分割的界面示意图；
图11示意性示出了根据本申请的一个实施例的图像分割方法的流程图；
图12示意性示出了根据本申请的一个实施例的医学图像分割装置的框图；
图13示意性示出了根据本申请的一个实施例的医学图像分割系统的结构示意图；
图14示出了适于用来实现本申请实施例的电子设备的计算机系统的结构示意图。
具体实施方式
现在将参考附图更全面地描述示例实施方式。然而，示例实施方式能够以多种形式实施，且不应被理解为限于在此阐述的范例；相反，提供这些实施方式使得本申请将更加全面和完整，并将示例实施方式的构思全面地传达给本领域的技术人员。
此外，所描述的特征、结构或特性可以以任何合适的方式结合在一个或更多实施例中。在下面的描述中，提供许多具体细节从而给出对本申请的实施例的充分理解。然而，本领域技术人员将意识到，可以实践本申请的技术方案而没有特定细节中的一个或更多，或者可以采用其它的方法、组元、装置、步骤等。在其它情况下，不详细示出或描述公知方法、装置、实现或者操作以避免模糊本申请的各方面。
附图中所示的方框图仅仅是功能实体，不一定必须与物理上独立的实体相对应。即，可以采用软件形式来实现这些功能实体，或在一个或多个硬件模块或集成电路中实现这些功能实体，或在不同网络和/或处理器装置和/或微控制器装置中实现这些功能实体。
附图中所示的流程图仅是示例性说明，不是必须包括所有的内容和操作/步骤，也不是必须按所描述的顺序执行。例如，有的操作/步骤还可以分解，而有的操作/步骤可以合并或部分合并，因此实际执行的顺序有可能根据实际情况改变。
图1示出了可以应用本申请实施例的技术方案的示例性系统架构的示意图。
如图1所示，系统架构100可以包括终端设备(如图1中所示台式计算机101、平板电脑102和便携式计算机103中的一种或多种，当然也可以是其它的具有显示屏幕的终端设备等等)、网络104和服务器105。网络104用以在终端设备和服务器105之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线通信链路、无线通信链路等等。
应该理解，图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。比如服务器105可以是多个服务器组成的服务器集群等。
在本申请的一个实施例中，用户可以利用台式计算机101(也可以是平板电脑102或便携式计算机103)向服务器105上传待处理医学图像集，该待处理医学图像集中包括的待处理医学图像可以是任意检验检测图像，例如可以是电子计算机断层扫描(Computed Tomography，CT)，即图像、可以是磁共振成像(Magnetic Resonance Imaging，MRI)图像，也可以是其它的图像信息随时间变化而改变的检验检测图像。服务器105在获取到待处理医学图像集之后，根据其中的待处理医学图像和对应的时间点对待处理医学图像集进行时间维度的处理，以将原始的待处理医学图像集转换为时间动态图像，该时间动态图像能够有效地反映病变区域与非病变区域的区别；接着将时间动态图像输入至训练好的医学图像分割模型中，通过医学图像分割模型提取时间动态图像中目标区域特征，进而获取目标区域。
该实施例的技术方案通过将多个待处理医学图像集进行时间维度的处理，获取时间动态图像，使得时间动态图像能够清晰地呈现病灶区域，避免了医生需要对多幅不同时间点 获取的医学图像进行分析确定病灶区域，进一步提升了医生的诊断效率，并且通过医学图像分割模型对时间动态图像中的目标区域进行分割，能够提高病变区域的分割精度，为医生的临床诊断提供支持。由于时间动态图像代替了多幅不同时间点获取的医学图像，减少了标注的工作量，使得医学图像分割模型的训练更加简便。
需要说明的是，本申请实施例所提供的医学图像分割方法一般由服务器105执行，相应地，医学图像分割装置一般设置于服务器105中。但是，在本申请的其它实施例中，终端设备也可以与服务器具有相似的功能，从而执行本申请实施例所提供的医学图像分割方案。
在本领域的相关技术中，为了提高诊断的准确度和效率，通常需要采集多个时间点的医学图像，如进行乳腺磁共振检查时，需要获取多个时间点的动态对比度增强(Dynamic contrast enhanced，DCE)-MRI图像，然后用每个时间点的DCE-MRI图像减去注射造影剂之前的图像，获得剪影图像；最后通过医生临床检查，观察多个医学图像以确定组织上的病变区域，或者通过训练好的深度学习网络模型对医学图像进行病变区域的提取。但是相关技术存在相应的缺陷，临床诊断方面：医生一般需要观察3至5个时间点的医学图像，有时甚至需要观察几十个时间点的医学图像，图像信息的数量巨大，例如5个时间点的三维(3Dimensions，3D)数据有5×70帧＝350个2D图像，对于20个时间点的数据库，一次检查有20×128帧＝2560个2D图像，医生从这么多图像中获取信息的难度非常大，并且诊断效率很低；计算机深度学习分割算法方面：如果用全部时间点的医学图像训练深度学习网络模型，为了避免在不同时间段病人有位移，对一个时间点的医学图像进行标注，很难精确对应其它时间点的病变区域，因此需要针对每个时间点的医学图像进行标注，或者对全部时间点的医学图像进行3D配准，使得深度学习网络模型的训练难度较高。
鉴于相关技术中存在的问题，本申请实施例首先提出了一种医学图像分割方法，以下对本申请实施例的技术方案的实现细节进行详细阐述：
图2示意性示出了根据本申请的一个实施例的医学图像分割方法的流程图，该医学图像分割方法可以由服务器来执行，该服务器可以是图1中所示的服务器。参照图2所示，该医学图像分割方法至少包括S210至S230，详细介绍如下：
在S210中，服务器获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像。
在本申请的一个实施例中，待处理医学图像集可以是医院各科室对从患者身体中取得的样本进行样本检测或对患者进行身体检查获取的图像集，比如该待处理医学图像集中的待处理医学图像可以是CT扫描图像、可以是MRI图像、可以是X光图像，也可以是其它的能够检测到病变区域性状随时间变化而变化的图像，本申请对待处理医学图像的种类不做具体限定。
为了便于理解本申请，下文中将以MRI图像作为待处理医学图像，并具体以对乳腺肿瘤的MRI图像进行分割为例行说明，当然该MRI图像还可以是对肺部肿瘤、胃部肿瘤、肝部肿瘤等病变区域进行磁共振成像获得的核磁共振成像图像。
在本申请的一个实施例中，对乳腺肿瘤进行磁共振成像时，首先需要给被检查人注射造影剂；然后让被检查人俯卧在检查床上，将乳房放置于特制的线圈中；最后随着检查床的移动，核磁共振仪对乳房进行从脚部至头部的层层扫描。造影剂是为增强影像观察效果而注入(或服用)到人体组织或器官的化学制品，如铁、锰等带有磁性的物质，这些制品的密度高于或低于周围组织，当这些制品接近共振中的氢原子时，能有效地改变质子所处的磁场，造成T1(纵向驰豫)和T2(横向驰豫)驰豫时间明显缩短。造影剂能改变体内局部组织中水质子的驰豫速率，提高非病变区域和病变区域的成像对比度和分辨率，为病变区域的定位和诊断提供更多的信息。具体地，造影剂在组织中的含量随血液的流动而发生变化，例如非病灶区域的血液流通顺畅，造影剂含量会迅速减少，磁共振图像中与非病灶区域对应的区域的亮度逐渐变高，而病灶区域的血液流通不畅，造影剂含量减少速度较慢，磁共振图像中与病灶区域对应的区域的亮度变化缓慢，进而与非病灶区域的亮度形成明显对比。因此本申请根据造影剂的该特征可以对同一组织横断面采集多个时间点的待处理医学图像。
需要说明的是，在本申请实施例中，服务器可以获取一个或多个待处理医学图像集，当服务器获取一个待处理医学图像集时，可以是对2D图像进行医学图像分割；当服务器获取多个待处理医学图像集时，各待处理医学图像集分别对应一横断面，并且各待处理医学图像集中对应同一时间点的待处理医学图像组成三维医学图像，此时，可以是对3D图像进行医学图像分割。本申请实施例主要以对3D图像进行医学图像分割进行介绍。
在本申请的一个实施例中，若服务器获取多个待处理医学图像集，并且各待处理医学图像集包括多个对应不同时间点的待处理医学图像，也就是说，各个待处理医学图像集对应不同的横断面，同一个待处理医学图像集中的待处理医学图像是对同一横断面在不同时间点进行信息采集所获得的图像，例如，当采集乳腺肿瘤的MRI图像时，多个待处理医学图像集是核磁共振成像仪在扫描过程中，扫描到乳腺不同横断面时获得的图像；同一个待处理医学图像集中的多个待处理医学图像则是核磁共振成像仪在扫描过程中，于多个时间点扫描乳腺的某一横断面时所产生的图像。进一步的，多个待处理医学图像对应的时间点可以是连续的时间点，进而通过多个待处理医学图像集可以形成医学图像序列。
图3示出了待处理医学图像集中对应多个不同时间点的待处理医学图像的结构示意图，如图3所示，标记为i的横断面中包含n+1个待处理医学图像，该n+1个待处理医学图像对应不同的时间点t＝0、1……n。
在本申请的一个实施例中，各个待处理医学图像集中对应同一时间点的待处理医学图像可以组成三维医学图像，例如在对乳腺肿瘤进行磁共振成像时，由各个待处理医学图像集中对应同一时间点的待处理医学图像组成的图像即为3D MRI图像，由于各待处理医学 图像集中包含多个时间点的待处理医学图像，因此该多个待处理医学图像集可以形成3D MRI图像序列。进一步的，可以对乳腺肿瘤进行动态对比增强磁共振成像，即可获取3D DCE-MRI图像，通过多个待处理医学图像集形成的医学图像序列即为三维动态对比度增强磁共振成像(3D DCE-MRI)图像序列。当然，医学图像序列还可以是二维图像序列，由于临床诊断中，大部分医学图像为3D医学图像，因此下文主要以3D DCE-MRI图像序列为例进行说明。
在S220中，根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像。
在本申请的一个实施例中，获取到多个对应不同时间点的待处理医学图像后，可以根据待处理医学图像及其对应的时间点，对待处理医学图像集进行时间维度的处理，以获取时间动态图像。值得说明的是，可以在对待处理医学图像集进行时间维度的处理之前，根据三维医学图像的三个维度确定横断面所对应的坐标面和横断面坐标轴，该三维医学图像对应的坐标系可以为三维笛卡尔坐标系，在确定坐标面和横断面坐标轴时，可以选择任意的坐标轴作为横断面坐标轴，与该横断面坐标轴垂直的坐标面作为横断面对应的坐标面，图4示出了一种横断面的分布示意图，在三维笛卡尔坐标系(x,y,z)中，可以将z轴定义为横断面坐标轴，即待处理医学图像集沿着z轴分布，则x-y坐标面即为横断面对应的坐标面，即任一待处理医学图像都是x-y坐标系的图像。
在本申请的一个实施例中，图5示出了获取时间动态图像的流程示意图，如图5所示，根据待处理医学图像和其对应的时间点对待处理医学图像集进行时间维度的处理，获取时间动态图像的过程，具体包括如下S510至S540，详细介绍如下：
在S510中，根据所述横断面对应的待处理医学图像数据和所述时间点确定与所述横断面对应的四维数据。
在本申请的一个实施例中，每一横断面对应多个时间点的待处理医学图像，并且每个横断面对应的坐标不同，因此首先可以根据横断面对应的待处理医学图像数据确定第一坐标、第二坐标和横断面坐标，其中第一坐标、第二坐标和横断面坐标相互垂直，并且第一坐标和第二坐标组成横断面对应的坐标面；然后根据时间点确定时间坐标；最后根据第一坐标、第二坐标、横断面坐标和时间坐标形成四维坐标轴，并根据四维坐标轴确定四维数据。继续参照图4示出的三维笛卡尔坐标系，第一坐标具体可以是x轴，第二坐标具体可以是y轴，横断面坐标具体可以是z轴。根据待处理医学图像对应的时间点可以确定时间坐标t，进而可以根据第一坐标、第二坐标、横断面坐标和时间坐标确定一四维坐标系(x,y,z,t)，进一步的可以根据四维坐标系确定与多个待处理医学图像集对应的四维数据I(x,y,z,t)，那么横断面坐标为z＝i处的横断面对应的四维数据即为I(x,y,i,t)。
在S520中，对所述横断面对应的四维数据进行解析，获得各所述横断面对应的时间分量。
在本申请的一个实施例中，图6示出了获取各横断面对应的时间分量的流程示意图，如图6所示，获取各横断面对应的时间分量的过程，包括如下S610至S630，详细说明如下：
在S610中，根据所述横断面坐标确定目标横断面，获取与所述目标横断面对应的第一图像数据，所述第一图像数据包括所述第一坐标、所述第二坐标和所述时间坐标。
在本申请的一个实施例中，可以从z轴中确定一个坐标值i作为横断面坐标，根据该横断面坐标获取与其对应的目标横断面，该目标横断面中包含多个对应不同时间点的待处理医学图像，通过统计该些待处理医学图像的图像数据可以确定第一图像数据，该第一图像数据包括各个待处理医学图像对应的第一坐标、第二坐标和时间坐标，也就是说，横断面坐标z＝i处的目标横断面对应的第一图像数据为f(x,y,t)。
在S620中，对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量。
在本申请的一个实施例中，为了对待处理医学图像进行时间维度的处理，可以将第一图像数据进行多维解析，获取第一图像数据在时间维度上的分量。
具体地，可以采用三维克利福德代数计算第一图像数据f(x,y,t)的多维解析信号ψ(x,y,t)，以将原始信号拆分成不同维度的分量。克利福德代数(Clifford algebra)，又称几何代数(Geometric algebra)，综合了内积和外积两种运算，是复数代数、四元数代数和外代数的推广，在几何和物理中有应用广泛。通过三维克利福德代数计算获得的多维解析信号ψ(x,y,t)如下所示：
Figure PCTCN2020074712-appb-000001
其中，“***”代表三维卷积计算，e 1、e 2、e 3是三维克利福德代数虚数单位的三个生成元，σ(·)是狄拉克函数。
从上述公式中可以看出，该公式由8个分量组成，每个分量的方向由e 1、e 2、e 3来确定，并且这8个分量在克利福德代数空间中是相互正交的。其中，e 1对应第一图像数据f(x,y,t)空间中x方向上的信息，e 2对应第一图像数据f(x,y,t)空间中y方向上的信息，e 3对应第一图像数据f(x,y,t)空间中t方向上的信息。对于第一图像数据，其在t方向上的信 息是我们所关注的，因此可以从多维信号ψ(x,y,t)中抽取分量f(x,y,t)***{σ(x)σ(y)e 3/(πt)}作为目标横断面对应的时间分量，可以记为时间分量f’(x,y,t)＝f(x,y,t)***{σ(x)σ(y)e 3/(πt)}。
在S630中，重复上述步骤，直至获取与各所述横断面对应的时间分量。
在本申请的一个实施例中，可以重复S610和S620，以获得各个横断面对应的时间分量。具体地，可以根据S610中的方法获取横断面坐标轴上所有的横断面坐标z＝i及各横断面坐标i所对应的横断面的第一图像数据fi(x,y,t)，并根据S620中的方法对各横断面的第一图像数据fi(x,y,t)进行解析，以获取与各横断面对应的时间分量fi’(x,y,t)。
在S530中，根据各所述横断面对应的时间分量确定与多个所述待处理医学图像集对应的目标时间分量。
在本申请的一个实施例中，获得各个横断面对应的时间分量后，可以根据横断面坐标z＝i和相应的时间分量fi’(x,y,t)确定与多个横断面(待处理医学图像集)对应的四维数据I(x,y,z,t)所对应的目标时间分量I’(x,y,z,t)，其中，当z＝i时，I’(x,y,z＝i,t)＝fi’(x,y,t)。
在S540中，对所述目标时间分量进行后处理，得到所述时间动态图像。
在本申请的一个实施例中，在获取了与多个待处理医学图像集对应的目标时间分量后，可以对目标时间分量进行后处理，以获取三维的时间动态图像。在本申请的实施例中，后处理的方法有多种，例如可以计算目标时间分量I’(x,y,z,t)沿着t轴的平均值，可以计算目标时间分量I’(x,y,z,t)沿着t轴的最大值，也可以计算目标时间分量I’(x,y,z,t)沿着t轴的最大值与最小值的差值，当然也可以通过其它方式对目标时间分量I’(x,y,z,t)进行后处理，本申请在此不再赘述。
在本申请的一个实施例中，计算目标时间分量I’(x,y,z,t)沿着t轴的平均值的具体流程包括：根据目标时间分量，分别确定在每个时间点时各横断面对应的子时间分量，该子时间分量的数量与时间点的数量相同；然后将与各横断面对应的子时间分量分别进行加和平均，以获取目标平均值；最后根据目标平均值构建时间动态图像。时间动态图像的具体计算公式如公式(1)所示：
I t(x,y,z)＝[I′(x,y,z,t＝t 1)+I′(x,y,z,t＝t 2)+…+I′(x,y,z,t＝t n)]/n    (1)
在本申请的一个实施例中，计算目标时间分量I’(x,y,z,t)沿着t轴的最大值的具体流程包括：根据目标时间分量，分别确定在每个时间点时各横断面对应的子时间分量，该子时间分量的数量与时间点的数量相同；然后分别获取与各横断面对应的子时间分量中的最大值；最后根据子时间分量中的最大值构建时间动态图像。时间动态图像的具体表达式如式(2)所示：
I t(x,y,z)＝Max(I'(x,y,z,t))|t＝t 1,t 2,...,t n      (2)
在本申请的一个实施例中，计算目标时间分量I’(x,y,z,t)沿着t轴的最大值与最小值的差值的具体流程包括：根据目标时间分量，分别确定在每个时间点时各横断面对应的子时间分量，该子时间分量的数量与时间点的数量相同；获取与各横断面对应的子时间分量中的最大值和最小值，并将最大值和最小值作差得到目标差值；最后根据目标差值构建时间动态图像。时间动态图像的具体计算公式如公式(3)所示：
I t(x,y,z)＝Max(I'(x,y,z,t))|t＝t 1,t 2,...,t n-Min(I'(x,y,z,t))|t＝t 1,t 2,...,t n      (3)
根据上述方法，可以根据计算结果获得三维的时间动态图像I t(x,y,z)，具体而言，就是获取了多个3D DCE-MRI图像在不同时间内像素亮度的变化情况。由于子时间分量的最大值和最小值的差值反映了3D DCE-MRI图像序列在同一个空间位置上不同时间点下获得的最大值和最小值的差异，其能够最大限度的显示这个空间上的点的亮度变化，因此在本申请的实施例中可以以目标差值为准，根据目标差值构建时间动态图像，以提高对医学图像的分割效率。
继续参照图2所示，在S230中通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
在本申请的一个实施例中，在获取时间动态图像后，可以将时间动态图像输入至医学图像分割模型，通过医学图像分割模型从时间动态图像中提取目标区域特征，以获取目标区域。其中，若S210中获取到多个待处理医学图像集，即对3D医学图像进行分割，则时间动态图像为三维时间动态图像。该医学图像分割模型可以是一个经过训练的深度学习分割模型，对于本申请中的三维时间动态图像而言，该医学图像分割模型可以是专门用于对三维图像进行处理的深度学习分割模型，如3D Unet模型、3D vnet模型、全卷积神经网络模型等等，对于深度学习分割模型的种类，本申请实施例不做具体限定。其中，该目标区域即为病变区域，如肿瘤区域、钙化区域等等。医疗工作者可以将目标区域作为感兴趣区域，并对感兴趣区域进行进一步的分析，以便制定最优的治疗方案。
图7A-7C示出了一种相关技术中注射造影剂后的某个时间点的DCE-MRI图像，如图7A-7C所示，图7A为乳腺的横截面图像，横截面是将身体分为上下两部分的纵切面；图7B为乳腺的矢状面图像，矢状面是将身体分为左右两部分的纵切面；图7C为乳腺的冠状面图像，冠状面是将身体分为前后两部分的纵切面，其中图7A-7C中的矩形框为病灶区域，椭圆框为非病灶区域。从图7A-7C可以发现，病灶区域和非病灶区域都呈现为高亮像素，医学图像分割模型无法将病灶区域和非病灶区域区别开，进而无法精确地分割出病灶区域。
图8A-8C示出了本申请中注射造影剂后的三维的时间动态图像，如图8A-8C所示，图8A为乳腺的横截面图像；图8B为乳腺的矢状面图像；图8C为乳腺的冠状面图像，其中图8A-8C中的矩形框为病灶区域，椭圆框为非病灶区域。与图7A-7C所示的DCE-MRI图像相比，图8A-8C中的病灶区域的像素亮度更高，非病灶区域的像素亮度更低，病灶区域 和非病灶区域的像素的对比度更明显，医学图像分割模型能够迅速将病灶区域和非病灶区域区别开，进而能够精确地分割出病灶区域。
在本申请的一个实施例中，在将时间动态图像输入至医学图像分割模型，通过医学图像分割模型从时间动态图像中提取目标区域特征，以获取目标区域之前，还可以对待训练医学图像分割模型进行训练，以获取后续对待处理医学图像进行图像分割的医学图像分割模型。
图9示出了对待训练医学图像分割模型进行训练的流程示意图，如图9所示，训练待训练医学图像分割模型的过程，包括如下S910至S920，详细说明如下：
在S910中，获取时间动态图像样本和与所述时间动态图像样本对应的目标区域标注样本。
在本申请的一个实施例中，获取时间动态图像样本的方法与上述实施例中获取时间动态图像的方法相同，在此不再赘述。在获得时间动态图像样本后，可以通过人工标注的方式对时间动态图像中的目标区域(病灶区域)进行标注，以获得与时间动态图像样本对应的目标区域标注样本。
在本申请的一个实施例中，为了提高医学图像分割模型的稳定性，使医学图像分割模型的损失函数收敛，可以采用多个样本对待训练医学图像分割模型进行训练，以获取该待训练医学图像分割模型的最优参数。具体地，可以选取244例恶性肿瘤病人的3D DCE-MRI图像数据，其中221例数据作为训练数据，用于训练待训练医学图像分割模型；23例数据作为测试数据，用于对训练后的医学图像分割模型进行测试，判断其是否达到稳定状态。当然，本申请中训练数据和测试数据的数量包括但不限于上述的举例，本申请对此不做具体限定。
在S920中，根据所述时间动态图像样本和所述目标区域标注样本对待训练医学图像分割模型进行训练，得到所述医学图像分割模型。
在本申请的一个实施例中，确定好时间动态图像样本和对应的目标区域标注样本后，可以将时间动态图像样本输入至待训练医学图像分割模型，以获取该待训练医学图像分割模型提取的目标区域；然后将提取的目标区域和与输入的动态图像样本对应的目标区域标注样本进行对比，判断待训练医学图像分割模型的分割准确度，若分割准确度度大于或等于预设阈值，则说明该待训练医学图像分割模型达到了稳定状态，可以作为后续进行医学图像分割的医学图像分割模型；若分割准确度度未达到预设阈值，则继续调整该待训练医学图像分割模型的参数，以使输出的分割图像得到分割准确度达到或大于预设阈值，该预设阈值可以根据实际需要进行设定，如设置为95％等等。在完成训练后，可以通过测试数据对训练得到的医学图像分割模型进行测试，判断其是否广泛适用于任意的时间动态图像。
在本申请的一个实施例中，通过本申请实施例的技术方案分割得到病灶区域的分割精度相较于使用原始的DCE-MRI数据训练医学图像分割模型对待处理医学图像进行分割得 到病灶区域的分割精度有了大幅度的提升，表1示出了使用原始DCE-MRI数据和使用三维的时间动态图像的实验结果，如表1所示：
表1
Figure PCTCN2020074712-appb-000002
通过对表1进行分析可知，使用本申请实施例中的三维的时间动态图像的分割精度较使用原始DCE-MRI数据的分割精度平均高10％，也就是说，本申请实施例的技术方案能够有效提升医生的诊断效率和医学图像的分割精度。
在本申请的一个实施例中，使用本申请实施例的技术方案中的三维时间动态图像还能够提高对较复杂的背景强化类型的医学图像的分割结果。图10A-10C示出了一种对背景强化类型的医学图像进行肿瘤分割的界面示意图，如图10A-10C所示，图10A示出了对背景强化类型的医学图像的标注结果；图10B示出了通过三维时间动态图像对背景强化类型的医学图像的分割结果，该方法的分割精度达到87％；图10C示出了通过原始的DCE-MRI数据对背景强化类型的医学图像的分割结果，该方法的分割精度为69％。由此说明，本申请实施例的技术方案能够提高医学图像分割的精度，并且能够对各种类型的医学图像进行分割，适用范围更广。
本申请上述实施例的技术方案能够从多个3D DCE-MRI数据中提取一个3D数据(三维时间动态图像)，直接用于目标区域的标注和医学图像分割模型的训练，使得医学图像分割模型的训练更简便；也避免了医生在做MRI影像诊断时，需要选择先阅读哪一个时间点的3D DCE-MRI图像，而是可以先通过观看本申请实施例的技术方案所获取的3D图像来观察病灶区域，再进一步选择某些时间点的3D DCE-MRI图像，进一步提高了诊断效率。
在本申请的一个实施例中，还提供了一种图像分割方法，图11示出了图像分割方法的流程图，如图11所示，在S1110中，获取待处理图像集，所述待处理图像集包括多个对应不同时间点的待处理图像；在S1120中，根据所述待处理图像和所述待处理图像对应的时间点，对所述待处理图像集进行时间维度的处理，得到时间动态图像；在S1130中，通过所述图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
该图像分割方法与上述实施例中的医学图像分割方法类似，但是该方法不仅可以对医学图像进行分割，还可以用于对其他任意类型的图像进行分割，例如对生物实验中的样本图像进行分割、对金属加工过程中的图像进行分割、对管道中损伤位置的分割，等等，只要图像中部分区域的特征随着时间推移，其变化趋势不同于其它区域特征的变化趋势，都可以采用本申请实施例中的图像分割方法进行分割。进一步地，该图像分割方法可以以本申请实施例中医学图像分割方法的具体实施方式实施，因此本申请在此不再赘述。
在本申请的一个实施例中，若获取到多个待处理图像集，各待处理图像集中对应相同时间点的待处理图像能够形成一三维图像，同时多个对应不同时间点的待处理图像可以形成三维图像序列，通过本申请实施例中的图像分割方法可以对三维图像序列进行识别分割，获取其中的目标区域。
以下介绍本申请的装置实施例，可以用于执行本申请上述实施例中的医学图像分割方法。对于本申请装置实施例中未披露的细节，请参照本申请上述的医学图像分割方法的实施例。
图12示意性示出了根据本申请的一个实施例的医学图像分割装置的框图。
参照图12所示，根据本申请的一个实施例的医学图像分割装置1200，包括：获取模块1201、处理模块1202和分割模块1203。
其中，获取模块1201，用于获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；处理模块1202，用于根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像；分割模块1203，用于通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
在本申请的一个实施例中，若获取到多个所述待处理图像集，各所述待处理医学图像集分别对应一横断面，并且各所述待处理医学图像集中对应同一时间点的所述待处理医学图像组成三维医学图像；基于前述方案，所述医学图像分割装置1200包括：横断面确定模块1204，用于根据所述三维医学图像的三个维度确定所述横断面对应的坐标面和横断面坐标轴，所述横断面坐标轴与所述坐标面垂直。
在本申请的一个实施例中，所述处理模块1202包括：四维数据获取单元，用于根据所述横断面对应的待处理医学图像数据和所述时间点确定与所述横断面对应的四维数据；第一时间分量获取单元，用于对所述横断面对应的四维数据进行解析，获得各所述横断面对应的时间分量；第二时间分量获取单元，用于根据各所述横断面对应的时间分量确定与多个所述待处理医学图像集对应的目标时间分量；后处理单元，用于对所述目标时间分量进行后处理，得到所述时间动态图像。
在本申请的一个实施例中，所述四维数据获取单元包括：空间坐标确定单元，用于根据所述横断面对应的待处理医学图像数据确定第一坐标、第二坐标、横断面坐标，所述第一坐标、所述第二坐标和所述横断面坐标相互垂直；时间坐标确定单元，用于根据所述时间点确定时间坐标；四维数据确定单元，用于根据所述第一坐标、所述第二坐标、所述横断面坐标和所述时间坐标构建四维坐标轴，并根据所述四维坐标轴确定所述四维数据。
在本申请的一个实施例中，所述第一时间分量获取单元包括：第一图像数据获取单元，用于根据所述横断面坐标确定目标横断面，获取与所述目标横断面对应的第一图像数据，所述第一图像数据包括所述第一坐标、所述第二坐标和所述时间坐标；多维解析单元，用 于对所述第一图像数据进行多维解析，获取所述目标横断面对应的时间分量；重复上述步骤，直至获取与各所述横断面对应的时间分量。
在本申请的一个实施例中，所述第二时间分量获取单元配置为：根据各所述横断面对应的时间分量、与各所述横断面对应的第一坐标、第二坐标和横断面坐标确定所述目标时间分量。
在本申请的一个实施例中，所述多维解析单元配置为：通过三维克利福德代数对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量。
在本申请的一个实施例中，所述后处理单元配置为：根据所述目标时间分量，，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；将与各所述横断面对应的所述子时间分量分别进行加和平均，以获取目标平均值；根据所述目标平均值构建所述时间动态图像。
在本申请的一个实施例中，所述后处理单元配置为：根据所述目标时间分量，，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；分别获取与各所述横断面对应的所述子时间分量中的最大值；根据所述子时间分量中的最大值构建所述时间动态图像。
在本申请的一个实施例中，所述后处理单元配置为：根据所述目标时间分量，，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；获取与各所述横断面对应的所述子时间分量中的最大值和最小值，并将所述最大值和所述最小值作差，以获取目标差值；根据所述目标差值构建所述时间动态图像。
在本申请的一个实施例中，所述的医学图像分割装置1200还包括：样本获取模块1205和训练模块1206。
其中，样本获取模块1205，用于获取时间动态图像样本和与所述时间动态样本对应的目标区域标注样本；训练模块1206，用于根据所述时间动态图像样本和所述目标区域标注样本对待训练医学图像分割模型进行训练，得到所述医学图像分割模型。
在本申请的一些实施例中，所述三维医学图像为三维动态对比度增强磁共振成像图像。
图13示出了一种医学图像分割系统，如图13所示，医学图像分割系统1300包括检测设备1301和电子设备1302。
其中，检测设备1301，用于对检测对象进行扫描检测，以获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；电子设备1302，所述电子设备与所述检测设备连接，并且所述电子设备包括存储装置和处理器，其中所述存储装置用于存储一个或多个程序，当所述一个或多个程序被所述处理器执行时，使得所述处理器实现上述的医学图像分割方法。
在本申请的一个实施例中，检测设备1301可以是CT设备中用于获取扫描图像的扫描装置，该扫描装置包含射线发射源、探测器和扫描架；可以是核磁共振成像设备中用于获取扫描图像的扫描装置，该扫描装置包含磁体部分、磁共振波谱仪部分和扫描床；还可以是荧光透视设备中用于获取扫描图像的扫描装置，该扫描装置包含射线发射源和探测器，当然还可以是其它的检测设备，只要可以用于对检测对象进行扫描获取扫描图像即可，本申请对此不做具体限定。检测设备1301扫描得到多个待处理医学图像集后，可以将待处理医学图像集发送至电子设备1302中的存储装置1302a和/或处理器1302b，存储装置1302a还存储有一个或多个程序，供处理器1302b执行。处理器1302b可以对待处理医学图像集执行存储在存储装置1302a中的一个或多个程序，也就是说，处理器1302b能够根据本申请实施例中的技术方案对待处理医学图像集进行图像分割，以获取目标区域，进一步地，处理器1302b还可以将包含该目标区域的图像发送至与电子设备1302连接的显示设备(未示出)进行显示，以供医生观察确定病灶并制定治疗方案。
图14示出了适于用来实现本申请实施例的电子设备1302的计算机系统的结构示意图。
需要说明的是，图14示出的电子设备的计算机系统1400仅是一个示例，不应对本申请实施例的功能和使用范围带来任何限制。
如图14所示，计算机系统1400包括中央处理单元(Central Processing Unit，CPU)1401，其可以根据存储在只读存储器(Read-Only Memory，ROM)1402中的程序或者从存储部分1408加载到随机访问存储器(Random Access Memory，RAM)1403中的程序而执行各种适当的动作和处理。在RAM 1403中，还存储有系统操作所需的各种程序和数据。CPU 1401、ROM 1402以及RAM 1403通过总线1404彼此相连。输入/输出(Input/Output，I/O)接口1405也连接至总线1404。
以下部件连接至I/O接口1405：包括键盘、鼠标等的输入部分1406；包括诸如阴极射线管(Cathode Ray Tube，CRT)、液晶显示器(Liquid Crystal Display，LCD)等以及扬声器等的输出部分1407；包括硬盘等的存储部分1408；以及包括诸如LAN(Local Area Network，局域网)卡、调制解调器等的网络接口卡的通信部分1409。通信部分1409经由诸如因特网的网络执行通信处理。驱动器1410也根据需要连接至I/O接口1405。可拆卸介质1411，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器1410上，以便于从其上读出的计算机程序根据需要被安装入存储部分1408。
特别地，根据本申请的实施例，下文参考流程图描述的过程可以被实现为计算机软件程序。例如，本申请的实施例包括一种计算机程序产品，其包括承载在计算机可读介质上的计算机程序，该计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分1409从网络上被下载和安装，和/或从可拆卸介质1411被安装。在该计算机程序被中央处理单元(CPU)1401执行时，执行本申请的系统中限定的各种功能。
需要说明的是，本申请实施例所示的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是——但不限于——电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(Erasable Programmable Read Only Memory，EPROM)、闪存、光纤、便携式紧凑磁盘只读存储器(Compact Disc Read-Only Memory，CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本申请中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。而在本申请中，计算机可读的信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读的信号介质还可以是计算机可读存储介质以外的任何计算机可读介质，该计算机可读介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于：无线、有线等等，或者上述的任意合适的组合。
附图中的流程图和框图，图示了按照本申请各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，上述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图或流程图中的每个方框、以及框图或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本申请实施例中所涉及到的单元可以通过软件的方式实现，也可以通过硬件的方式来实现，所描述的单元也可以设置在处理器中。其中，这些单元的名称在某种情况下并不构成对该单元本身的限定。
作为另一方面，本申请还提供了一种计算机可读介质，该计算机可读介质可以是上述实施例中描述的电子设备中所包含的；也可以是单独存在，而未装配入该电子设备中。上述计算机可读介质承载有一个或者多个程序，当上述一个或者多个程序被一个该电子设备执行时，使得该电子设备实现上述实施例中所述的方法。
应当注意，尽管在上文详细描述中提及了用于动作执行的设备的若干模块或者单元，但是这种划分并非强制性的。实际上，根据本申请的实施方式，上文描述的两个或更多模块或者单元的特征和功能可以在一个模块或者单元中具体化。反之，上文描述的一个模块或者单元的特征和功能可以进一步划分为由多个模块或者单元来具体化。
通过以上的实施方式的描述，本领域的技术人员易于理解，这里描述的示例实施方式可以通过软件实现，也可以通过软件结合必要的硬件的方式来实现。因此，根据本申请实施方式的技术方案可以以软件产品的形式体现出来，该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM，U盘，移动硬盘等)中或网络上，包括若干指令以使得一台计算设备(可以是个人计算机、服务器、触控终端、或者网络设备等)执行根据本申请实施方式的方法。
本领域技术人员在考虑说明书及实践这里公开的发明后，将容易想到本申请的其它实施方案。本申请旨在涵盖本申请的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本申请的一般性原理并包括本申请未公开的本技术领域中的公知常识或惯用技术手段。
应当理解的是，本申请并不局限于上面已经描述并在附图中示出的精确结构，并且可以在不脱离其范围进行各种修改和改变。本申请的范围仅由所附的权利要求来限制。

Claims (17)

  1. 一种医学图像分割方法，所述方法应用于电子设备，包括：
    获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；
    根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像；
    通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
  2. 根据权利要求1所述的医学图像分割方法，若获取到多个所述待处理图像集，各所述待处理医学图像集分别对应一横断面，并且各所述待处理医学图像集中对应同一时间点的所述待处理医学图像组成三维医学图像；
    所述根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像之前，所述方法还包括：
    根据所述三维医学图像的三个维度确定所述横断面对应的坐标面和横断面坐标轴，所述横断面坐标轴与所述坐标面垂直。
  3. 根据权利要求2所述的医学图像分割方法，所述根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像，包括：
    根据所述横断面对应的待处理医学图像数据和所述时间点确定与所述横断面对应的四维数据；
    对所述横断面对应的四维数据进行解析，获得各所述横断面对应的时间分量；
    根据各所述横断面对应的时间分量确定与多个所述待处理医学图像集对应的目标时间分量；
    对所述目标时间分量进行后处理，得到所述时间动态图像。
  4. 根据权利要求3所述的医学图像分割方法，所述根据所述横断面对应的待处理医学图像数据和所述时间点确定与所述横断面对应的四维数据，包括：
    根据所述横断面对应的待处理医学图像数据确定第一坐标、第二坐标、横断面坐标，所述第一坐标、所述第二坐标和所述横断面坐标相互垂直；
    根据所述时间点确定时间坐标；
    根据所述第一坐标、所述第二坐标、所述横断面坐标和所述时间坐标构建四维坐标轴，并根据所述四维坐标轴确定所述四维数据。
  5. 根据权利要求4所述的医学图像分割方法，所述对所述横断面对应的四维数据进行解析，获得各所述横断面对应的时间分量，包括：
    根据所述横断面坐标确定目标横断面，获取与所述目标横断面对应的第一图像数据，所述第一图像数据包括所述第一坐标、所述第二坐标和所述时间坐标；
    对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量；
    重复上述步骤，直至获取与各所述横断面对应的时间分量。
  6. 根据权利要求4所述的医学图像分割方法，所述根据各所述横断面对应的时间分量确定与多个所述待处理医学图像集对应的目标时间分量，包括：
    根据各所述横断面对应的时间分量、与各所述横断面对应的第一坐标、第二坐标和横断面坐标确定所述目标时间分量。
  7. 根据权利要求5所述的医学图像分割方法，所述对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量，包括：
    通过三维克利福德代数对所述第一图像数据进行多维解析，获取与所述目标横断面对应的时间分量。
  8. 根据权利要求3-7任一项所述的医学图像分割方法，所述对所述目标时间分量进行后处理，得到所述时间动态图像，包括：
    根据所述目标时间分量，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；
    将与各所述横断面对应的所述子时间分量分别进行加和平均，以获取目标平均值；
    根据所述目标平均值构建所述时间动态图像。
  9. 根据权利要求3-7任一项所述的医学图像分割方法，所述对所述目标时间分量进行后处理，得到所述时间动态图像，包括：
    根据所述目标时间分量，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；
    分别获取与各所述横断面对应的所述子时间分量中的最大值；
    根据所述子时间分量中的最大值构建所述时间动态图像。
  10. 根据权利要求3-7任一项所述的医学图像分割方法，所述对所述目标时间分量进行后处理，得到所述时间动态图像，包括：
    根据所述目标时间分量，分别确定在每个所述时间点时各所述横断面对应的子时间分量，所述子时间分量的数量与所述时间点的数量相同；
    获取与各所述横断面对应的所述子时间分量中的最大值和最小值，并将所述最大值和所述最小值作差得到目标差值；
    根据所述目标差值构建所述时间动态图像。
  11. 根据权利要求1所述的医学图像分割方法，所述通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域之前，所述方法还包括：
    获取时间动态图像样本和与所述时间动态图像样本对应的目标区域标注样本；
    根据所述时间动态图像样本和所述目标区域标注样本对待训练医学图像分割模型进行训练，得到所述医学图像分割模型。
  12. 根据权利要求2所述的医学图像分割方法，所述三维医学图像为三维动态对比度增强磁共振成像图像。
  13. 一种医学图像分割装置，包括：
    获取模块，用于获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；
    处理模块，用于根据所述待处理医学图像和所述待处理医学图像对应的时间点，对所述待处理医学图像集进行时间维度的处理，得到时间动态图像；
    分割模块，用于通过医学图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
  14. 一种医学图像分割系统，包括：
    检测设备，用于对检测对象进行扫描检测，以获取待处理医学图像集，所述待处理医学图像集包括多个对应不同时间点的待处理医学图像；
    电子设备，所述电子设备与所述检测设备连接，并且所述电子设备包括存储装置和处理器，其中所述存储装置用于存储一个或多个程序，当所述一个或多个程序被所述处理器执行时，使得所述处理器实现如权利要求1至12中任一项所述的医学图像分割方法。
  15. 一种图像分割方法，包括：
    获取待处理图像集，所述待处理图像集包括多个对应不同时间点的待处理图像；
    根据所述待处理图像和所述待处理图像对应的时间点，对所述待处理图像集进行时间维度的处理，得到时间动态图像；
    通过图像分割模型从所述时间动态图像中提取目标区域特征，以获取目标区域。
  16. 根据权利要求15所述的图像分割方法，若获取到多个所述待处理图像集，各所述待处理图像集中对应相同时间点的待处理图像形成一三维图像，并且所述多个对应不同时间点的待处理图像形成三维图像序列。
  17. 一种计算机可读存储介质，其上存储有计算机可读指令，当所述计算机可读指令被计算机的处理器执行时，使计算机执行权利要求1-12或15-16中的任一项所述的方法。
PCT/CN2020/074712 2019-02-15 2020-02-11 医学图像分割方法、图像分割方法及相关装置、系统 WO2020164468A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP20755319.9A EP3926537A4 (en) 2019-02-15 2020-02-11 METHOD OF SEGMENTING A MEDICAL IMAGE, METHOD OF SEGMENTING AN IMAGE, RELATED DEVICE AND SYSTEM
US17/239,532 US11954864B2 (en) 2019-02-15 2021-04-23 Medical image segmentation method, image segmentation method, and related apparatus and system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910116353.0A CN109872312B (zh) 2019-02-15 2019-02-15 医学图像分割方法、装置、系统及图像分割方法
CN201910116353.0 2019-02-15

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/239,532 Continuation US11954864B2 (en) 2019-02-15 2021-04-23 Medical image segmentation method, image segmentation method, and related apparatus and system

Publications (1)

Publication Number Publication Date
WO2020164468A1 true WO2020164468A1 (zh) 2020-08-20

Family

ID=66918681

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/074712 WO2020164468A1 (zh) 2019-02-15 2020-02-11 医学图像分割方法、图像分割方法及相关装置、系统

Country Status (5)

Country Link
US (1) US11954864B2 (zh)
EP (1) EP3926537A4 (zh)
CN (2) CN109872312B (zh)
TW (1) TWI750583B (zh)
WO (1) WO2020164468A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114332132A (zh) * 2021-12-31 2022-04-12 联影智能医疗科技(成都)有限公司 图像分割方法、装置和计算机设备
CN116543001A (zh) * 2023-05-26 2023-08-04 广州工程技术职业学院 彩色图像边缘检测方法及装置、设备、存储介质

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109872312B (zh) * 2019-02-15 2022-12-20 腾讯科技（深圳）有限公司 医学图像分割方法、装置、系统及图像分割方法
CN110445954B (zh) * 2019-07-26 2022-04-26 腾讯医疗健康(深圳)有限公司 图像采集方法、装置及电子设备
CN111429474B (zh) * 2020-02-27 2023-04-07 西北大学 基于混合卷积的乳腺dce-mri图像病灶分割模型建立及分割方法
WO2022011617A1 (zh) 2020-07-15 2022-01-20 北京肿瘤医院(北京大学肿瘤医院) 一种利用光学体表运动信号合成实时图像的方法及系统
CN113160253B (zh) * 2020-12-29 2024-01-30 南通大学 基于稀疏标记的三维医学图像分割方法及存储介质
US11580646B2 (en) * 2021-03-26 2023-02-14 Nanjing University Of Posts And Telecommunications Medical image segmentation method based on U-Net
CN113838020A (zh) * 2021-09-17 2021-12-24 上海仰和华健人工智能科技有限公司 一种基于钼靶影像的病变区域量化方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101334895A (zh) * 2008-08-07 2008-12-31 清华大学 一种针对动态增强乳腺磁共振影像序列的影像分割方法
CN106600621A (zh) * 2016-12-08 2017-04-26 温州医科大学 基于婴幼儿脑瘤多模态mri图的时空协同分割方法
CN107563378A (zh) * 2017-07-31 2018-01-09 上海联影医疗科技有限公司 体数据中提取感兴趣区域的方法及其系统
CN109872312A (zh) * 2019-02-15 2019-06-11 腾讯科技（深圳）有限公司 医学图像分割方法、装置、系统及图像分割方法

Family Cites Families (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6961454B2 (en) * 2001-10-04 2005-11-01 Siemens Corporation Research, Inc. System and method for segmenting the left ventricle in a cardiac MR image
WO2007035687A2 (en) * 2005-09-16 2007-03-29 The Ohio State University Method and apparatus for detecting interventricular dyssynchrony
EP1780651A1 (en) * 2005-10-25 2007-05-02 Bracco Imaging, S.P.A. Method and system for automatic processing and evaluation of images, particularly diagnostic images
JP5399225B2 (ja) * 2008-12-15 2014-01-29 富士フイルム株式会社 画像処理装置および方法並びにプログラム
EP2336979B1 (de) * 2009-11-05 2014-03-12 TomTec Imaging Systems GmbH Verfahren und Vorrichtung zur Segmentierung von medizinischen Bilddaten
US20110200227A1 (en) * 2010-02-17 2011-08-18 Siemens Medical Solutions Usa, Inc. Analysis of data from multiple time-points
US9529508B2 (en) * 2010-10-19 2016-12-27 Koninklijke Philips N.V. Medical image system
CN103236058B (zh) * 2013-04-25 2016-04-13 内蒙古科技大学 获取四维心脏图像感兴趣体积的方法
CN104143035B (zh) * 2013-05-10 2016-01-20 上海联影医疗科技有限公司 一种分割乳腺病灶的方法
CN103426169B (zh) * 2013-07-26 2016-12-28 西安华海盈泰医疗信息技术有限公司 一种医学图像的分割方法
US9652871B2 (en) * 2015-01-28 2017-05-16 Impac Medical Systems, Inc. Three dimensional localization of a moving target for adaptive radiation therapy
CN104809723B (zh) * 2015-04-13 2018-01-19 北京工业大学 基于超体素和图割算法的三维肝脏ct图像自动分割方法
WO2017039663A1 (en) * 2015-09-03 2017-03-09 Siemens Healthcare Gmbh Multi-view, multi-source registration of moving anatomies and devices
KR101718868B1 (ko) * 2015-09-21 2017-03-22 한국과학기술연구원 자동 의료영상 분할에 의한 3차원 악안면 모델 형성 방법, 이를 수행하는 자동 영상 분할과 모델 형성 서버 및 이를 저장하는 기록매체
CN109069859B (zh) * 2016-02-02 2021-04-27 医科达有限公司 放射治疗系统和确定解剖区域的精确运动的成像方法
CN106056610A (zh) * 2016-06-02 2016-10-26 南方医科大学 基于图割的肺4d‑ct多相位肿瘤联合分割方法
CN106228601B (zh) * 2016-07-21 2019-08-06 山东大学 基于小波变换的多尺度锥束ct图像快速三维重建方法
US11039757B2 (en) * 2016-11-22 2021-06-22 Cedars-Sinai Medical Center Method and system for cardiac motion corrected MR exam using deformable registration
TWI756365B (zh) * 2017-02-15 2022-03-01 美商脫其泰有限責任公司 圖像分析系統及相關方法
CN108509830B (zh) * 2017-02-28 2020-12-01 华为技术有限公司 一种视频数据处理方法及设备
EP3410393A1 (en) * 2017-06-01 2018-12-05 Siemens Healthcare GmbH Comparing medical images
TW201903708A (zh) 2017-06-06 2019-01-16 國立陽明大學 數位減影血管攝影圖像的分析方法與系統
US10219768B2 (en) * 2017-06-08 2019-03-05 Emass Llc Method for standardizing target lesion selection and tracking on medical images
CN107808377B (zh) * 2017-10-31 2019-02-12 北京青燕祥云科技有限公司 一种肺叶中病灶的定位装置
EP3714467A4 (en) * 2017-11-22 2021-09-15 Arterys Inc. CONTENT-BASED IMAGE RECOVERY FOR LESION ANALYSIS
CN108038848B (zh) * 2017-12-07 2020-08-11 上海交通大学 基于医学影像序列斑块稳定性指标的快速计算方法及系统
CN108109170B (zh) * 2017-12-18 2022-11-08 上海联影医疗科技股份有限公司 医学图像扫描方法及医学影像设备
CN108537803B (zh) * 2018-03-30 2019-08-23 北京灵医灵科技有限公司 一种ct图像交互分割方法及装置
EP3791316A1 (en) * 2018-06-13 2021-03-17 Siemens Healthcare GmbH Localization and classification of abnormalities in medical images
CN109242863B (zh) * 2018-09-14 2021-10-26 北京市商汤科技开发有限公司 一种缺血性脑卒中图像区域分割方法及装置
CN109215764B (zh) * 2018-09-21 2021-05-04 苏州瑞派宁科技有限公司 一种医学图像四维可视化的方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101334895A (zh) * 2008-08-07 2008-12-31 清华大学 一种针对动态增强乳腺磁共振影像序列的影像分割方法
CN106600621A (zh) * 2016-12-08 2017-04-26 温州医科大学 基于婴幼儿脑瘤多模态mri图的时空协同分割方法
CN107563378A (zh) * 2017-07-31 2018-01-09 上海联影医疗科技有限公司 体数据中提取感兴趣区域的方法及其系统
CN109872312A (zh) * 2019-02-15 2019-06-11 腾讯科技（深圳）有限公司 医学图像分割方法、装置、系统及图像分割方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3926537A4

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114332132A (zh) * 2021-12-31 2022-04-12 联影智能医疗科技(成都)有限公司 图像分割方法、装置和计算机设备
CN116543001A (zh) * 2023-05-26 2023-08-04 广州工程技术职业学院 彩色图像边缘检测方法及装置、设备、存储介质
CN116543001B (zh) * 2023-05-26 2024-01-12 广州工程技术职业学院 彩色图像边缘检测方法及装置、设备、存储介质

Also Published As

Publication number Publication date
EP3926537A4 (en) 2022-04-06
CN109872312B (zh) 2022-12-20
CN109872312A (zh) 2019-06-11
US20210264613A1 (en) 2021-08-26
TW202032577A (zh) 2020-09-01
EP3926537A1 (en) 2021-12-22
CN110490851B (zh) 2021-05-11
CN110490851A (zh) 2019-11-22
US11954864B2 (en) 2024-04-09
TWI750583B (zh) 2021-12-21

Similar Documents

Publication Publication Date Title
WO2020164468A1 (zh) 医学图像分割方法、图像分割方法及相关装置、系统
US11937962B2 (en) Systems and methods for automated and interactive analysis of bone scan images for detection of metastases
US20190099147A1 (en) Systems and methods for determining hepatic function from liver scans
US9858665B2 (en) Medical imaging device rendering predictive prostate cancer visualizations using quantitative multiparametric MRI models
CN107886508B (zh) 差分减影方法和医学图像处理方法及系统
Ikhsan et al. An analysis of x-ray image enhancement methods for vertebral bone segmentation
US8655040B2 (en) Integrated image registration and motion estimation for medical imaging applications
US10460508B2 (en) Visualization with anatomical intelligence
Zhao et al. Automated 3D fetal brain segmentation using an optimized deep learning approach
Heydarheydari et al. Auto-segmentation of head and neck tumors in positron emission tomography images using non-local means and morphological frameworks
CN111340825A (zh) 一种纵膈淋巴结分割模型的生成方法及系统
US20140225926A1 (en) Method and system for generating an image as a combination of two existing images
TW202033159A (zh) 圖像處理方法、裝置及系統、電子設備及電腦可讀儲存媒體
US11495346B2 (en) External device-enabled imaging support
JP2022179433A (ja) 画像処理装置及び画像処理方法
CN111035403A (zh) 一种扫描时机确定方法、装置、设备及存储介质
US20230326580A1 (en) Information processing apparatus, information processing method, and information processing program
US20220076796A1 (en) Medical document creation apparatus, method and program, learning device, method and program, and trained model
US20230260141A1 (en) Deep learning for registering anatomical to functional images
WO2022070528A1 (ja) 医用画像処理装置、方法およびプログラム
CN117455931A (zh) 基于深度学习mr体部脂肪组织的分割和定量测量方法
CN117457142A (zh) 用于报告生成的医学影像处理系统及方法
Whelan et al. Computer assisted diagnosis (CAD) for the rapid automated measurement of body fat tissue from whole body MRI
Kizuka et al. Development of the MI-Viewer KIT for medical image viewer
Rangaiah et al. Vbir-Based Assessment of Radiographic-Divergence Agent Attention in Prostate Melanoma Patients

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20755319

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2020755319

Country of ref document: EP

Effective date: 20210915