WO2021184600A1 - 一种图像分割方法及装置、设备及计算机可读存储介质 - Google Patents

一种图像分割方法及装置、设备及计算机可读存储介质 Download PDF

Info

Publication number
WO2021184600A1
WO2021184600A1 PCT/CN2020/100711 CN2020100711W WO2021184600A1 WO 2021184600 A1 WO2021184600 A1 WO 2021184600A1 CN 2020100711 W CN2020100711 W CN 2020100711W WO 2021184600 A1 WO2021184600 A1 WO 2021184600A1
Authority
WO
WIPO (PCT)
Prior art keywords
cerebral
point
image
prediction
area
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/100711
Other languages
English (en)
French (fr)
Inventor
宋涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Sensetime Intelligent Technology Co Ltd
Original Assignee
Shanghai Sensetime Intelligent Technology Co Ltd
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 Shanghai Sensetime Intelligent Technology Co Ltd filed Critical Shanghai Sensetime Intelligent Technology Co Ltd
Priority to KR1020217020478A priority Critical patent/KR20210118389A/ko
Priority to JP2021538462A priority patent/JP2022529094A/ja
Publication of WO2021184600A1 publication Critical patent/WO2021184600A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
    • A61B6/5217—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data extracting a diagnostic or physiological parameter from medical diagnostic data
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
    • A61B6/03—Computed tomography [CT]
    • A61B6/032—Transmission computed tomography [CT]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
    • A61B6/501—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of the head, e.g. neuroimaging or craniography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
    • A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
    • A61B6/507—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for determination of haemodynamic parameters, e.g. perfusion CT
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06T2207/30016—Brain
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
    • G06T2207/30104—Vascular flow; Blood flow; Perfusion

Definitions

  • the embodiments of the present application relate to the field of medical imaging technology, and in particular, to an image segmentation method and device, equipment, and computer-readable storage medium.
  • Computed Tomography (CT) perfusion imaging of the brain is an imaging technique used to analyze intracranial hemodynamics and is widely used to diagnose ischemic stroke.
  • CT perfusion imaging is to observe the dynamic changes of brain tissue density during intravenous bolus injection of iodine contrast agent, and calculate the local cerebral blood volume, local cerebral blood flow, average transit time and peak time according to different mathematical models. Generally, the peak is reached.
  • Time greater than 6 seconds (s, second) is the hypoperfusion area
  • the local cerebral blood flow on the lesion side is less than 30% than the local cerebral blood flow on the upper normal side as the cerebral infarct core area
  • the rest of the infarct core area is excluded from the hypoperfusion area
  • the segmentation of the infarct core area and the ischemic penumbra area is very important for the formulation of the treatment plan.
  • the embodiments of the present application provide an image segmentation method and device, equipment, and computer-readable storage medium.
  • An embodiment of the present application provides an image segmentation method, the method includes: acquiring a first brain image to be segmented; positioning the brain image through a deep learning positioning network to obtain prediction points of the anterior cerebral artery and prediction of cerebral veins Point; using the anterior cerebral artery prediction point and the cerebral vein prediction point to determine the perfusion parameter map corresponding to the brain image; using the perfusion parameter map to determine the cerebral perfusion infarct core area.
  • the method before locating the brain image through the deep learning positioning network to obtain the anterior cerebral artery prediction point and the cerebral vein prediction point, the method further includes: obtaining a brain sample image, the brain sample image being Preliminarily annotated images of anterior cerebral artery points and cerebral venous points; using brain sample images to train the initial deep learning positioning network to obtain the deep learning positioning network.
  • the positioning the brain image through the deep learning positioning network to obtain the anterior cerebral artery prediction point and the cerebral vein prediction point includes: positioning the brain image through the deep learning positioning network to obtain the anterior cerebral artery Candidate points and cerebral vein candidate points; using a local search algorithm to determine the anterior cerebral artery prediction point corresponding to the anterior cerebral artery candidate point and the cerebral vein prediction point corresponding to the cerebral vein candidate point.
  • using a local search algorithm to determine the anterior cerebral artery prediction point corresponding to the anterior cerebral artery candidate point and the cerebral venous prediction point corresponding to the cerebral venous candidate point includes: taking the anterior cerebral artery candidate point as the center of the circle , The area within the preset first radius is determined as the anterior cerebral artery point candidate area, the area within the preset second radius is determined as the cerebral venous point candidate area with the cerebral venous candidate point as the center of the circle; The local search algorithm respectively determines the anterior cerebral artery prediction point from the anterior cerebral artery point candidate area, and determines the cerebral venous prediction point from the cerebral venous point selection area.
  • said using said local search algorithm to respectively determine said anterior cerebral artery prediction point from said anterior cerebral artery point candidate area and to determine said cerebral vein prediction point from said cerebral venous point selection area respectively includes: Determine multiple first tissue density values corresponding to the maximum time value from the candidate area of the anterior cerebral artery point, and determine multiple second tissue density values corresponding to the maximum time value from the selected area of the cerebral vein; The maximum value of the tissue density value among the plurality of first tissue density values is determined as the prediction point of the anterior cerebral artery, and the maximum value of the tissue density value among the plurality of second tissue density values is determined as the cerebral vein prediction point.
  • the perfusion parameter map includes: a regional cerebral blood volume map, a regional cerebral blood flow map, an average transit time map, and a peak time map.
  • using the anterior cerebral artery prediction point and the cerebral venous prediction point to determine the perfusion parameter map corresponding to the brain image includes: obtaining a function curve corresponding to the anterior cerebral artery prediction point and the cerebral venous prediction point Corresponding function curve; determining the function curve corresponding to the anterior cerebral artery prediction point as the first arterial input function, and determining the function curve corresponding to the cerebral vein prediction point as the venous output function; using the venous output function pair
  • the first arterial input function is modified to obtain a second arterial input function; the second arterial input function is used to determine the regional cerebral blood volume map, the regional cerebral blood flow map, and the regional cerebral blood flow map through a deconvolution algorithm.
  • the average transit time graph and the peak time graph is used to determine the regional cerebral blood volume map, the regional cerebral blood flow map, and the regional cerebral blood flow map through a deconvolution algorithm.
  • using the perfusion parameter map to calculate the cerebral perfusion infarct core region includes: acquiring a second brain image to be segmented; determining an area greater than a preset time in the peak time map as low Irrigation area; using the regional cerebral blood volume image and the regional cerebral blood flow image using a convolutional neural network algorithm to obtain the infarct core corresponding to the second brain image to be segmented from the low-irrigation area area.
  • the calculation of the core area of cerebral perfusion infarction by using the perfusion parameter map includes: acquiring a second brain image to be segmented; Region; using the regional cerebral blood volume image, the regional cerebral blood flow image and the average transit time map to obtain the second brain image to be segmented from the low-irrigation region using a convolutional neural network algorithm Corresponding to the core area of the infarct.
  • the method further includes: determining an ischemic penumbra area based on the hypoperfusion area and the infarct core area; wherein the infarct core area plus the ischemic penumbra area is equal to the hypoperfusion area and the ischemic penumbra area area.
  • An embodiment of the application provides an image segmentation device, which includes: an acquisition module configured to acquire a first brain image to be segmented; a positioning module configured to locate the brain image through a deep learning positioning network to obtain the brain image Arterial prediction points and cerebral venous prediction points; Perfusion parameter map acquisition module, configured to use the anterior cerebral artery prediction points and cerebral venous prediction points to determine the perfusion parameter map corresponding to the brain image; Infarct core region segmentation module, configured to use the perfusion parameter map Determine the core area of the cerebral perfusion infarct.
  • An embodiment of the present application provides an image segmentation device, including a processor and a memory that are coupled to each other, wherein the memory is configured to store program instructions for implementing the image segmentation method described in any one of the above; the processor is configured to Execute the program instructions stored in the memory.
  • An embodiment of the present application provides a computer-readable storage medium that stores a program file, and the program file can be executed to implement the image segmentation method described in any one of the above.
  • the image segmentation method, device, device, and computer-readable storage medium provided by the embodiments of the present application, when the first brain image to be segmented is acquired, the brain image is located through a deep learning positioning network to obtain The anterior cerebral artery prediction point and the cerebral venous prediction point, thus improving the positioning accuracy of the anterior cerebral artery prediction point and the cerebral venous prediction point, and then using the anterior cerebral artery prediction point and the cerebral venous prediction point to determine the corresponding brain image Perfusion parameter map, the cerebral perfusion infarct core area is calculated by using the perfusion parameter map.
  • the positioning accuracy of the anterior cerebral artery prediction point and the cerebral venous prediction point is improved, it is determined by the anterior cerebral artery prediction point and the cerebral venous prediction point
  • the perfusion parameter map corresponding to the brain image is more accurate, thereby improving the segmentation accuracy and robustness of the infarct core region.
  • Figure 1a is a schematic diagram of a network architecture according to an embodiment of the application.
  • FIG. 1b is a schematic diagram of another network architecture according to an embodiment of the application.
  • FIG. 2 is a schematic flowchart of an image segmentation method provided by an embodiment of this application.
  • step S22 is a schematic diagram of the implementation flow of step S22 in an image segmentation method provided by an embodiment of the application;
  • step S32 is a schematic diagram of the implementation flow of step S32 in an image segmentation method provided by an embodiment of the application;
  • FIG. 5 is a schematic diagram of an implementation flow of step S42 in an image segmentation method provided by an embodiment of the application.
  • FIG. 6 is a schematic diagram of the implementation flow of step S23 in an image segmentation method provided by an embodiment of the application.
  • FIG. 7 is a schematic diagram of an implementation flow of step S24 in an image segmentation method provided by an embodiment of the application.
  • FIG. 8 is a schematic diagram of another implementation flow of step S24 in an image segmentation method provided by an embodiment of the application.
  • FIG. 9 is a schematic diagram of another implementation process of an image segmentation method provided by an embodiment of the application.
  • FIG. 10 is a schematic flowchart of still another image segmentation method provided by an embodiment of this application.
  • FIG. 11 is a schematic structural diagram of an image segmentation device provided by an embodiment of this application.
  • FIG. 12 is a schematic structural diagram of an image segmentation device provided by an embodiment of this application.
  • FIG. 13 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of this application.
  • the image segmentation method provided by the embodiments of the application uses deep learning algorithms to locate the anterior cerebral artery point and the cerebral venous point, which can effectively solve the shortcomings of automatic or manual positioning of the positions of the cerebral artery point and venous point. Calculations provide protection.
  • the image segmentation method provided by the embodiments of the present application does not need to rely on the contrast between the normal side and the lesion side, and can prevent the problem that abnormalities cannot be detected by the ratio when the normal side and the lesion side are both abnormal, and further improve the segmentation of the infarct core area. The accuracy and robustness.
  • the embodiments of the present application will be described in detail below with reference to the drawings and embodiments.
  • Figure 1a is a schematic diagram of the network architecture of an embodiment of the application.
  • the network architecture includes a CT machine 11 and a computer device 12, where the CT machine 11 is used to collect brain images to be segmented.
  • the CT machine 11 establishes a communication connection with the computer equipment 12, and the CT machine 11 can send the obtained brain image to be segmented to the computer equipment 12.
  • the computer device 12 stores brain sample images.
  • the computer device 12 can train a deep learning positioning network based on the brain sample images.
  • the computer device 12 inputs the brain image to be segmented into the deep learning positioning network to obtain the prediction points of the anterior cerebral artery And cerebral vein prediction points; using the anterior cerebral artery prediction points and the cerebral vein prediction points to determine the perfusion parameter map corresponding to the brain image; using the perfusion parameter map to determine the cerebral perfusion infarct core area.
  • Fig. 1b is a schematic diagram of another network architecture according to an embodiment of the application.
  • the network architecture includes a CT machine 11, a computer device 12, and a server 13, where the CT machine 11 is used to collect the brain to be divided image.
  • the CT machine 11 establishes a communication connection with the computer equipment 12, and the CT machine 11 can send the obtained brain image to be segmented to the computer equipment 12.
  • the server 13 stores brain sample images, and the server 13 can train a deep learning positioning network based on the brain sample images.
  • the computer device 12 and the server 13 have also established a communication connection.
  • the computer device 12 can obtain the deep learning positioning network from the server 13.
  • the computer device 12 inputs the brain image to be segmented into the deep learning positioning network to obtain the prediction points of the anterior cerebral artery and Cerebral vein prediction point; using the anterior cerebral artery prediction point and the cerebral vein prediction point to determine the perfusion parameter map corresponding to the brain image; using the perfusion parameter map to determine the cerebral perfusion infarct core area.
  • FIGS. 1a and 1b With reference to the schematic diagrams of the application scenarios shown in FIGS. 1a and 1b, the following describes various embodiments of an image segmentation method and device, equipment, and computer-readable storage medium. It should be noted that, in the embodiment of the present application, the CT machine and the computer device 12 may be integrated.
  • FIG. 2 is a schematic flowchart of an image segmentation method provided by an embodiment of the application, and the method includes:
  • Step S21 Obtain the first brain image to be segmented.
  • the brain image may be a brain CT image, which is obtained by CT imaging technology.
  • X X-ray
  • the detector receives the X-rays that pass through this layer and converts it into visible light. It is converted into an electrical signal by a photoelectric converter, and then converted into a digital signal by an analog/digital converter, which is input to a computer for processing to obtain a CT image.
  • the processing of CT image formation includes: dividing the selected layer into a number of rectangular parallelepipeds with the same volume, which are called voxels; the information obtained by scanning is calculated to obtain the X-ray attenuation coefficient or absorption coefficient of each voxel,
  • the X-ray attenuation coefficients or absorption coefficients are arranged in a matrix, that is, a digital matrix.
  • the digital matrix can be stored in a magnetic disk or an optical disk.
  • the digital/analog converter converts each number in the digital matrix into small squares with varying gray levels from black to white, that is, pixels, and arrange them in a matrix to form a CT image.
  • Step S22 Position the brain image through the deep learning positioning network to obtain the prediction points of the anterior cerebral artery and the prediction points of the cerebral vein.
  • the brain image is located through a deep learning positioning network, and training is required to obtain the deep learning positioning network before obtaining the prediction points of the anterior cerebral artery and the prediction points of the cerebral veins.
  • the process of training the deep learning positioning network includes: acquiring brain sample images, initially marking the anterior cerebral artery points and cerebral venous points in the brain sample images; using the initially marked brain sample images to locate the initial deep learning network Perform training to get a deep learning positioning network.
  • manual marking may be used.
  • the labeling may also be performed in other ways, such as machine recognition labeling.
  • the deep learning positioning network trained in the above manner performs positioning in the brain image to obtain the prediction points of the anterior cerebral artery and the prediction points of the cerebral vein respectively.
  • calculations may be performed through convergence, deconvolution, and the like.
  • FIG. 3 is a schematic diagram of the implementation process of step S22 in an image segmentation method provided by an embodiment of the application.
  • step S22 positions the brain image through a deep learning positioning network to obtain a prediction of the anterior cerebral artery Points and brain vein prediction points" can be achieved through the following steps:
  • Step S31 Position the brain image through a deep learning positioning network to obtain candidate points of the anterior cerebral artery and candidate points of the cerebral vein.
  • Step S32 Use the local search algorithm to determine the anterior cerebral artery prediction point corresponding to the anterior cerebral artery candidate point and the cerebral venous prediction point corresponding to the cerebral vein candidate point.
  • the local search algorithm is a simple greedy search algorithm. The algorithm selects an optimal solution from the adjacent solution space of the current solution as the current solution each time until a local optimal solution is reached.
  • a local search algorithm is used to obtain the corresponding anterior cerebral artery prediction points from the anterior cerebral artery candidate points, and the corresponding cerebral venous prediction points are obtained from the cerebral venous candidate points, so that more accurate anterior cerebral artery prediction points can be obtained.
  • brain vein prediction points As well as brain vein prediction points.
  • the deep learning positioning network is used to improve the accuracy and robustness of determining the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein.
  • a local search algorithm is also used to determine the anterior cerebral artery prediction point and the cerebral vein prediction point according to the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein.
  • FIG. 4 is a schematic diagram of the implementation flow of step S32 in an image segmentation method provided by an embodiment of the application.
  • step S32 "uses a local search algorithm to calculate the prediction points of the anterior cerebral artery corresponding to the candidate points of the anterior cerebral artery and The brain vein prediction point corresponding to the brain vein candidate point" can be achieved through the following steps:
  • Step S41 The candidate point of the anterior cerebral artery is determined as the center of the circle, the area within the preset first radius is determined as the candidate area of the anterior cerebral artery point, the candidate cerebral vein is determined as the center of the circle, and the area within the preset second radius is determined as the brain Vein point candidate area.
  • the candidate point of the anterior cerebral artery obtained by the deep learning algorithm is the center of the circle, and the area within the first radius is preset as the candidate area of the anterior cerebral artery point.
  • the brain vein candidate point obtained by the deep learning algorithm is the center of the circle, and the area within the second radius is preset as the brain vein point candidate area.
  • the preset first radius and the preset second radius may be 1mm, 2mm, 2.5mm, 10mm, etc., and the preset first radius and the preset second radius may have the same or different values, which are not limited here. .
  • Step S42 Use the local search algorithm to determine the prediction points of the anterior cerebral artery from the candidate regions of the anterior cerebral artery points, and determine the prediction points of the cerebral vein from the selected regions of the cerebral veins.
  • the local search algorithm is used to search the obtained anterior cerebral artery point candidate area to obtain the anterior cerebral artery prediction point, and the cerebral vein prediction point is searched from the cerebral venous point selection area.
  • the local search algorithm is a simple greedy search algorithm.
  • the algorithm selects an optimal solution from the adjacent solution space of the current solution as the current solution each time until a local optimal solution is reached.
  • the method described in the embodiment of the present application uses the local search algorithm combined with the deep learning algorithm to obtain the anterior cerebral artery prediction point and the cerebral venous prediction point, which improves the accuracy and robustness of the determined anterior cerebral artery prediction point and the cerebral venous prediction point sex.
  • FIG. 5 is a schematic diagram of the implementation process of step S42 in an image segmentation method provided by an embodiment of the application.
  • step S42 "Using a local search algorithm to respectively determine the anterior cerebral artery prediction point from the anterior cerebral artery point candidate area , And determine the cerebral vein prediction point from the selected area of the cerebral vein” can be achieved through the following steps:
  • Step S51 Determine multiple first tissue density values corresponding to the maximum time from the candidate area of the anterior cerebral artery point, and determine multiple second tissue density values corresponding to the maximum time from the selected area of the cerebral vein.
  • the abscissa of the several curves is the time coordinate
  • the ordinate is the tissue density coordinate.
  • the first corresponding to the maximum value of the time coordinate is obtained from the several curves.
  • a value of tissue density How many curves there are, that is, how many first tissue density values are there. For example, if there are 30 curves, the first tissue density value obtained at this time is 30.
  • the abscissa of the several curves is the time coordinate and the ordinate is the tissue density coordinate.
  • the second corresponding to the maximum value of the time coordinate is obtained from the several curves.
  • Tissue density value how many curves there are, that is, how many second tissue density values there are. For example, if there are 30 curves, the second tissue density value obtained at this time is 30.
  • Step S52 Determine the maximum value of the tissue density value among the plurality of first tissue density values as the anterior cerebral artery prediction point, and determine the maximum value of the tissue density value among the plurality of second tissue density values as the cerebral vein prediction point.
  • the obtained first tissue density values are 30, compare the obtained 30 first tissue density values, and use the coordinates corresponding to the maximum value of the first tissue density values as the prediction point of the anterior cerebral artery.
  • the maximum time value is recorded as Tm1
  • the maximum value of the first tissue density value is recorded as Hm1.
  • the predicted point of the anterior cerebral artery corresponds to the position (Tm1, Hm1).
  • the obtained second tissue density values are 30, compare the sizes of the obtained 30 second tissue density values, and use the coordinates corresponding to the maximum value of the second tissue density values as the cerebral vein prediction points.
  • the maximum time value is recorded as Tm2, and the maximum value of the second tissue density value is recorded as Hm2.
  • Tm2 the maximum time value
  • Hm2 the maximum value of the second tissue density value
  • the method shown in the embodiment of the application first uses the deep learning network model to detect the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein, and then uses the local search algorithm to determine the predicted point of the anterior cerebral artery and the predicted point of the cerebral vein, so that the determined anterior cerebral artery
  • the prediction points and the cerebral venous prediction points are the maximum values of the corresponding multiple tissue density values, thereby improving the accuracy and robustness of the obtained anterior cerebral artery prediction points and cerebral venous prediction points.
  • Step S23 Use the prediction points of the anterior cerebral artery and the prediction points of the cerebral vein to determine the perfusion parameter map corresponding to the brain image.
  • the perfusion parameter map corresponding to the brain image is further calculated.
  • Fig. 6 is a schematic diagram of the implementation process of step S23 in an image segmentation method provided by an embodiment of the application.
  • step S23 "uses the prediction points of the anterior cerebral artery and the prediction points of the cerebral veins to calculate the perfusion corresponding to the brain image Parameter map" can be achieved through the following steps:
  • Step S61 Obtain the function curve corresponding to the prediction point of the anterior cerebral artery and the function curve corresponding to the prediction point of the cerebral vein.
  • the curves of the cerebral venous prediction points and the anterior cerebral artery prediction points are respectively used as the function curve corresponding to the cerebral venous prediction point and the function curve corresponding to the anterior cerebral artery prediction point.
  • Step S62 Use the function curve corresponding to the prediction point of the anterior cerebral artery as the first artery input function, and use the function curve corresponding to the prediction point of the cerebral vein as the venous output function.
  • the function curve corresponding to the prediction point of the anterior cerebral artery is used as the first artery input function
  • the function curve corresponding to the prediction point of the cerebral vein is used as the venous output function.
  • the venous output function is used as the accumulation of the first arterial input function, so the value of the venous output is greater than the value of the arterial input.
  • Step S63 Use the venous output function to correct the first arterial input function to obtain the second arterial input function.
  • the venous output function is used to perform convergence correction on the first arterial input function to obtain the second arterial input function.
  • Step S64 Use the second arterial input function to determine the regional cerebral blood volume map, the regional cerebral blood flow map, the average transit time map, and the peak time map through the deconvolution algorithm.
  • the corrected second arterial input function obtains the perfusion parameter map of the brain image through the deconvolution calculation method.
  • the perfusion parameter map includes a regional cerebral blood volume map (rCBV, Regional Cerebral Blood Volume), a regional cerebral blood flow map (rCBF, renal cortical blood flow), a mean transit time map (MTT), and a peak time map. (Tmax).
  • the method provided by the embodiment of the application first uses deep learning to detect the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein, and then uses the predicted points of the anterior cerebral artery and the predicted points of the cerebral vein obtained by the local search algorithm to obtain the predicted points of the anterior cerebral artery And the accuracy and robustness of brain vein prediction points, which in turn makes the calculated regional cerebral blood volume map, regional cerebral blood flow map, average transit time map, and peak time map more accurate.
  • Step S24 Use the perfusion parameter map to determine the core area of cerebral perfusion infarction.
  • the local cerebral blood flow on the side of the lesion is smaller than the preset threshold value as the regional cerebral blood flow on the upper normal side as the core area of cerebral infarction.
  • the blood flow is less than 30% as the core area of cerebral infarction.
  • FIG. 7 is a schematic diagram of an implementation flow of step S24 in an image segmentation method provided by an embodiment of the application. As shown in FIG. 7, step S24 can be implemented through the following steps:
  • Step S71 Obtain a second brain image to be segmented.
  • the second image to be segmented is a Diffusion-Weighted Imaging (DWI) image
  • the second brain image to be segmented in the embodiment of the present application is an image with an infarct core region marked.
  • the second brain image to be segmented and the first brain image to be segmented are brain images of the same patient.
  • the second brain image to be segmented and the first brain image to be segmented in the embodiment of the present application may also be brain images of different patients.
  • Step S72 Determine an area in the peak time map that is greater than the preset time as a low-irrigation area.
  • the preset time is 6 seconds, and an area with a time greater than 6s in the peak time map (Tmax) is determined as a low-irrigation area.
  • Step S73 Use the regional cerebral blood volume image and the regional cerebral blood flow image to obtain the infarct core region corresponding to the second brain image to be segmented from the low-perfusion region using a convolutional neural network algorithm.
  • the infarct core area is obtained from the regional cerebral blood volume image and the regional cerebral blood flow image according to the second brain image to be segmented with the infarct core area marked.
  • the core area of the infarct can be obtained by using the convolutional neural network algorithm by overlapping, de-distortion and other methods.
  • the perfusion parameter images used when calculating the infarct core region are the regional cerebral blood volume image and the regional cerebral blood flow image, which further improves the accuracy of segmentation of the infarct core region.
  • FIG. 8 is a schematic diagram of another implementation process of step S24 in an image segmentation method provided by an embodiment of the application. As shown in FIG. 8, step S24 "calculate the core area of cerebral perfusion infarction using the perfusion parameter map" can go through the following steps accomplish:
  • Step S81 Obtain a second brain image to be segmented.
  • the second image to be segmented is a DWI image
  • the second brain image to be segmented in the embodiment of the present application is an image with an infarct core region marked.
  • the second brain image to be segmented and the first brain image to be segmented are brain images of the same patient.
  • the second brain image to be segmented and the first brain image to be segmented in the embodiment of the present application may also be brain images of different patients.
  • Step S82 Determine an area in the peak time map that is greater than the preset time as a low-irrigation area.
  • the preset time is 6s
  • the area in the peak-to-peak time map with the time greater than 6s is regarded as the low-irrigation area.
  • Step S83 Use the regional cerebral blood volume image, the regional cerebral blood flow image, and the average transit time map to obtain the second brain image to be segmented corresponding to the second brain image to be segmented from the low-irrigation region using a convolutional neural network algorithm The core area of the infarct.
  • the convolutional neural network algorithm is used to obtain the infarct core area from the regional cerebral blood volume image, the regional cerebral blood flow image and the average transit time map according to the second to-be segmented brain image marked with the infarct core area.
  • the core area of the infarct can be obtained by using the convolutional neural network algorithm by overlapping, de-distortion and other methods.
  • the perfusion parameter images used when calculating the infarct core area are the regional cerebral blood volume image, the regional cerebral blood flow image, and the average transit time image, so that the infarct core area obtained in the embodiment of the application is Higher accuracy and higher robustness.
  • the image segmentation method described in the embodiment of the application uses deep learning to detect the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein, and uses the local search algorithm to obtain the predicted points of the anterior cerebral artery and the predicted points of the cerebral vein, which improves the obtained anterior cerebral artery.
  • the accuracy and robustness of the prediction points and the prediction points of the cerebral veins which in turn makes the calculated regional cerebral blood volume map, regional cerebral blood flow image, average transit time map, and peak time map more accurate.
  • the method described in the embodiments of the present application does not need to use the ratio of the lesion side to the normal side as a reference when segmenting the infarct core area, which excludes the inaccurate segmentation accuracy when both the lesion side and the normal side are abnormal.
  • the segmentation accuracy and robustness of the infarct core area are improved.
  • FIG. 9 is a schematic diagram of another implementation process of an image segmentation method provided by an embodiment of this application. As shown in FIG. 9, in the method shown in the embodiment of this application, steps S91 to S94 are the same as steps S21 to S24. Compared with the image segmentation method shown in FIG. 2, the embodiment of the present application further includes:
  • Step S95 Determine the ischemic penumbra area according to the hypoperfusion area and the infarct core area.
  • the infarct core area plus the ischemic penumbra area is equal to the hypoperfusion area.
  • the remaining area in the hypoperfusion area except the infarct core area is the ischemic penumbra area.
  • the method described in the embodiments of the present application uses deep learning to detect the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein, and uses the local search algorithm to obtain the predicted points of the anterior cerebral artery and the predicted points of the cerebral vein, thereby improving the obtained predicted points of the anterior cerebral artery And the accuracy and robustness of the predicted points of the cerebral veins, which in turn makes the calculated regional cerebral blood volume map, regional cerebral blood flow image, average transit time map, and peak time map more accurate.
  • the method described in the embodiments of the present application does not need to use the ratio of the lesion side to the normal side as a reference when segmenting the infarct core area, which excludes the inaccurate segmentation accuracy when both the lesion side and the normal side are abnormal.
  • the segmentation accuracy and robustness of the infarct core area are improved. Since the infarct core area plus the ischemic penumbra area is equal to the hypoperfusion area, the method of the embodiment of the present application also improves the segmentation of the ischemic penumbra area under the premise that the segmentation accuracy and robustness of the infarct core area are higher. Accuracy and robustness.
  • FIG. 10 is a schematic flowchart of another image segmentation method provided by an embodiment of the application. As shown in FIG. 10, the method includes:
  • step S101 the anterior cerebral artery point and the cerebral venous point are located.
  • the deep learning positioning network is used to locate the two candidate points of the anterior cerebral artery point and the cerebral venous point, and then the accurate anterior cerebral artery point and the cerebral venous point are determined by the local search algorithm, and then the arterial input function AIF and the venous output function VOF( Volume Fluent).
  • Step S102 perfusion calculation.
  • step S103 the ischemic penumbra and the infarct core are segmented.
  • Tmax is greater than 6s as the hypoperfusion area
  • four parameter maps are used as input to segment the infarct core, and the two mismatches are used as the segmentation result of the ischemic penumbra.
  • the method provided in the embodiments of the present application automatically detects the anterior cerebral artery point and the cerebral venous point through the deep learning positioning network and the local search algorithm, which improves the algorithm accuracy and robustness of the acquired anterior cerebral artery point and cerebral venous point, and utilizes depth Learning to segment the infarct core improves its accuracy and robustness, and can solve the special case of infarct core regions on both sides of the brain.
  • FIG. 11 is a schematic structural diagram of an image segmentation device provided by an embodiment of the application.
  • the image segmentation device includes: an acquisition module 111, a positioning module 112, an attention parameter map acquisition module 113, and infarct core region segmentation Module 114.
  • the acquiring module 111 is configured to acquire the first brain image to be segmented
  • the positioning module 112 is configured to locate the brain image through a deep learning positioning network to obtain the anterior cerebral artery prediction point and the cerebral vein prediction point.
  • the perfusion parameter map acquisition module 113 is configured to calculate the perfusion parameter map corresponding to the brain image by using the prediction points of the anterior cerebral artery and the prediction points of the cerebral vein.
  • the infarct core region segmentation module 114 is configured to use the perfusion parameter map to calculate the cerebral perfusion infarct core region.
  • the obtaining module 111 is further configured to:
  • the brain sample image is an image preliminarily marking the anterior cerebral artery point and the cerebral venous point;
  • the initial deep learning positioning network is trained by using the brain sample image to obtain the deep learning positioning network.
  • the positioning module 112 is further configured to:
  • a local search algorithm is used to determine the anterior cerebral artery predicted point corresponding to the anterior cerebral artery candidate point and the cerebral venous predicted point corresponding to the cerebral venous candidate point.
  • the positioning module 112 is further configured to:
  • the determining the anterior cerebral artery predicted point corresponding to the anterior cerebral artery candidate point and the cerebral venous predicted point corresponding to the cerebral venous candidate point by using a local search algorithm includes:
  • the candidate point of the anterior cerebral artery is determined as the center of the circle, the area within the preset first radius is determined as the candidate area of the anterior cerebral artery point, the candidate point of the cerebral vein is determined as the center of the circle, and the area within the preset second radius is determined as Candidate areas of cerebral vein points;
  • the local search algorithm is used to determine the anterior cerebral artery prediction point from the anterior cerebral artery point candidate area, and the cerebral venous prediction point is determined from the cerebral venous point selection area.
  • the positioning module 112 is further configured to:
  • the perfusion parameter map includes: a regional cerebral blood volume map, a regional cerebral blood flow map, an average transit time map, and a peak time map.
  • the attention parameter map obtaining module 113 is further configured to:
  • the second input artery function is used to determine the regional cerebral blood volume map, the regional cerebral blood flow map, the average transit time map, and the peak time map through a deconvolution algorithm.
  • the infarct core region segmentation module 114 is further configured to:
  • the infarct core region corresponding to the second brain image to be segmented is acquired from the low-perfusion region using the regional cerebral blood volume image and the regional cerebral blood flow image using a convolutional neural network algorithm.
  • the infarct core region segmentation module 114 is further configured to:
  • the convolutional neural network algorithm is used to obtain the second brain image to be segmented from the low-irrigation region.
  • the core area of the infarct is used to obtain the second brain image to be segmented from the low-irrigation region.
  • the infarct core region segmentation module 114 is further configured to:
  • the infarct core area plus the ischemic penumbra area is equal to the hypoperfusion area.
  • the image segmentation device provided by the embodiments of the application can realize the location and acquisition of the anterior cerebral artery candidate points and the cerebral venous candidate points through the deep learning algorithm, and then use the local search algorithm to determine the anterior cerebral artery according to the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein. Prediction points and brain vein prediction points. The accuracy and robustness of using deep learning to detect the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein, and using the local search algorithm to obtain the prediction points of the anterior cerebral artery and the cerebral vein is more prominent.
  • the image segmentation device does not need to use the ratio of the lesion side to the normal side as a reference when segmenting the infarct core area, which excludes the inaccurate segmentation accuracy when the lesion side and the normal side are both abnormal. Circumstances, the segmentation accuracy and robustness of the infarct core region are improved.
  • FIG. 12 is a schematic structural diagram of an image segmentation device provided by an embodiment of the application. As shown in FIG. 12, the image segmentation device includes a memory 121, a processor 122, and a communication bus 123 connected to each other.
  • the memory 121 is configured to store program instructions for implementing any one of the above-mentioned image segmentation methods.
  • the processor 122 is configured to execute program instructions stored in the memory 121.
  • the communication bus 123 is configured to implement connection and communication between the memory 121 and the processor 122.
  • the processor 122 may also be referred to as a central processing unit (CPU, Central Processing Unit).
  • the processor 122 may be an integrated circuit chip with signal processing capabilities.
  • the processor 122 may also be a general-purpose processor, a digital signal processor (DSP, Digital Signal Processor), an application specific integrated circuit (ASIC, Application Specific Integrated Circuit), a field programmable gate array (FPGA, Field Programmable Gate Array), or other Programming logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • DSP digital signal processor
  • ASIC Application Specific Integrated Circuit
  • FPGA Field Programmable Gate Array
  • Programming logic devices discrete gates or transistor logic devices, discrete hardware components.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the memory 121 can be a memory stick, a flash memory card (TF card, Micro SD Card), etc., which can store all the information in the image segmentation device, including the input original data, computer programs, intermediate running results, and final running results are all stored in the memory . It stores and retrieves information according to the location specified by the controller. With the memory, the image segmentation device has the memory function to ensure normal operation.
  • the memory in the image segmentation device can be divided into main memory (memory) and auxiliary memory (external memory) according to purpose, and there are also classification methods for external memory and internal memory. External storage is usually magnetic media or optical discs, etc., which can store information for a long time.
  • Memory refers to the storage components on the motherboard, used to store the currently executing data and programs, but only used to temporarily store the programs and data, the data will be lost if the power is turned off or power off.
  • the disclosed method and device can be implemented in other ways.
  • the device implementation described above is only illustrative, for example, the division of modules or units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components can be combined or It can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the objectives of the solution of this embodiment.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.
  • the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • a computer device which may be a personal computer, a system server, or a network device, etc.
  • a processor to execute all or part of the steps of the methods in the various embodiments of the present application.
  • FIG. 13 is a schematic structural diagram of a computer-readable storage medium provided by an embodiment of this application.
  • the computer-readable storage medium of this application stores a program file 131 that can implement all the above-mentioned image segmentation methods, where:
  • the program file 131 may be stored in the above-mentioned computer-readable storage medium in the form of a software product, and includes a number of instructions to make a computer device (may be a personal computer, a server, or a network device, etc.) or a processor to execute All or part of the steps of each implementation method of this application.
  • the aforementioned storage devices include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks and other media that can store program codes.
  • terminal devices such as computers, servers, mobile phones, and tablets.
  • the image segmentation device provided in the embodiment of the present application can firstly locate and acquire the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein through the deep learning algorithm. Then the local search algorithm determines the prediction points of the anterior cerebral artery and the prediction points of the cerebral vein according to the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein. The accuracy and robustness of using deep learning to detect the candidate points of the anterior cerebral artery and the candidate points of the cerebral vein, and using the local search algorithm to obtain the prediction points of the anterior cerebral artery and the cerebral vein is more prominent.
  • the image segmentation device provided in the embodiments of the present application does not need to use the ratio of the lesion side to the normal side as a reference when segmenting the infarct core area, which excludes the inaccurate segmentation accuracy when the lesion side and the normal side are both abnormal. Circumstances, the segmentation accuracy and robustness of the infarct core region are further improved.
  • the embodiments of the present application provide an image segmentation method and device, equipment, and computer-readable storage medium.
  • the method includes: acquiring a first brain image to be segmented; and performing a deep learning positioning network on the brain image Perform positioning to obtain the anterior cerebral artery prediction point and the cerebral vein prediction point; use the anterior cerebral artery prediction point and the cerebral vein prediction point to calculate the perfusion parameter map corresponding to the brain image; use the perfusion parameter map to calculate Obtain the cerebral perfusion infarct core region, thereby improving the segmentation accuracy and robustness of the infarct core region.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Optics & Photonics (AREA)
  • Pathology (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Dentistry (AREA)
  • Quality & Reliability (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Neurosurgery (AREA)
  • Neurology (AREA)
  • Physiology (AREA)
  • Pulmonology (AREA)
  • Image Analysis (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

本申请实施例提供一种图像分割方法及装置、设备及计算机可读存储介质，其中，所述方法包括：获取到第一待分割的脑部图像；通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点；利用所述大脑前动脉预测点和所述大脑静脉预测点计算得到所述脑部图像对应的灌注参数图；利用所述灌注参数图计算得到所述脑灌注梗死核心区域。

Description

一种图像分割方法及装置、设备及计算机可读存储介质
相关申请的交叉引用
本申请基于申请号为202010192934.5、申请日为2020年03月18日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请实施例涉及医学影像技术领域，尤其涉及一种图像分割方法及装置、设备及计算机可读存储介质。
背景技术
大脑电子计算机断层扫描(CT，Computed Tomography)灌注成像是一种用于分析颅内的血流动力学的成像技术，广泛用于诊断缺血性脑卒中。CT灌注成像是通过观察静脉内快速团注碘对比剂时脑组织密度动态变化，根据不同的数学模型计算得到局部脑血容积、局部脑血流量、平均通过时间和达峰时间，一般通过达峰时间大于6秒(s，second)为低灌注区域，病灶侧的局部脑血流量比上正常侧的局部脑血流量小于30％作为脑梗死核心区域，低灌注区域中除去梗死核心区域的其余区域为缺血半暗带，梗死核心区域和缺血半暗带区域的分割对于治疗方案的制定至关重要。
传统基于数学模型的梗死核心区域和缺血半暗带区域的分割方法存在定位不准、检测结果不准确等缺点。
发明内容
本申请实施例提供一种图像分割方法及装置、设备及计算机可读存储介质。
本申请实施例提供一种图像分割方法，所述方法包括：获取第一待分割的脑部图像；通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点；利用所述大脑前动脉预测点和所述大脑静脉预测点确定所述脑部图像对应的灌注参数图；利用所述灌注参数图确定所述脑灌注梗死核心区域。
其中，所述通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点之前，所述方法还包括：获取脑部样本图像，所述脑部样本图像为对大脑前动脉点和大脑静脉点进行初步标注的图像；利用脑部样本图像对初始深度学习定位网络进行训练，以得到所述深度学习定位网络。
其中，所述通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点，包括：通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉候选点和大脑静脉候选点；利用局部搜索算法确定所述大脑前动脉候选点对应的大脑前动脉预测点以及所述大脑静脉候选点对应的大脑静脉预测点。
其中，所述利用局部搜索算法确定所述大脑前动脉候选点对应的大脑前动脉预测点以及所述大脑静脉候选点对应的大脑静脉预测点，包括：将以所述大脑前动脉候选点为圆心，预设第一半径范围内的区域确定为大脑前动脉点候选区域，将以所 述大脑静脉候选点为圆心，预设第二半径范围内的区域确定为大脑静脉点候选区域；利用所述局部搜索算法分别从所述大脑前动脉点候选区域中确定所述大脑前动脉预测点，及从所述大脑静脉点选区域中确定所述大脑静脉预测点。
其中，所述利用所述局部搜索算法分别从所述大脑前动脉点候选区域中确定所述大脑前动脉预测点，及从所述大脑静脉点选区域中确定所述大脑静脉预测点，包括：从所述大脑前动脉点候选区域中确定时间最大值对应的多个第一组织密度值，及从所述大脑静脉点选区域中确定时间最大值对应的多个第二组织密度值；将所述多个第一组织密度值中组织密度值的最大值确定为所述大脑前动脉预测点，及将多个所述第二组织密度值中组织密度值的最大值确定为所述大脑静脉预测点。
其中，所述灌注参数图包括：局部脑血容积图、局部脑血流量图、平均通过时间图和达峰时间图。
其中，利用所述大脑前动脉预测点和所述大脑静脉预测点确定所述脑部图像对应的灌注参数图，包括：获取所述大脑前动脉预测点对应的函数曲线和所述大脑静脉预测点对应的函数曲线；将所述大脑前动脉预测点对应的函数曲线确定为第一动脉输入函数，及将所述大脑静脉预测点对应的函数曲线确定为静脉输出函数；利用所述静脉输出函数对所述第一动脉输入函数进行修正，以得到第二动脉输入函数；利用所述第二动脉输入函数通过去卷积算法确定所述局部脑血容积图、所述局部脑血流量图、所述平均通过时间图和所述达峰时间图。
其中，所述利用所述灌注参数图计算得到所述脑灌注梗死核心区域，包括：获取第二待分割的脑部图像；将所述达峰时间图中大于预设时间的区域确定为为低灌区域；利用所述局部脑血容积图像、所述局部脑血流量图像采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
其中，所述利用所述灌注参数图计算得到所述脑灌注梗死核心区域，包括：获取第二待分割的脑部图像；将所述达峰时间图中大于预设时间的区域确定为低灌区域；利用所述局部脑血容积图像、所述局部脑血流量图像及所述平均通过时间图采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
其中，所述方法还包括：根据所述低灌区域及所述梗死核心区域确定缺血半暗带区域；其中，所述梗死核心区域加上所述缺血半暗带区域等于所述低灌区域。
本申请实施例提供一种图像分割装置，包括：获取模块，配置为获取第一待分割的脑部图像；定位模块，配置为通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点；灌注参数图获取模块，配置为利用大脑前动脉预测点和大脑静脉预测点确定脑部图像对应的灌注参数图；梗死核心区域分割模块，配置为利用灌注参数图确定所述脑灌注梗死核心区域。
本申请实施例提供一种图像分割设备，包括相互藕接的处理器、存储器，其中，所述存储器配置为存储实现上述任意一项所述的图像分割方法的程序指令；所述处 理器配置为执行所述存储器存储的所述程序指令。
本申请实施例提供一种计算机可读存储介质，存储有程序文件，所述程序文件能够被执行以实现上述任意一项所述的图像分割方法。
本申请实施例提供的一种图像分割方法及装置、设备及计算机可读存储介质，在获取到第一待分割的脑部图像时，通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点，如此提升了大脑前动脉预测点和大脑静脉预测点的定位精度，再利用所述大脑前动脉预测点和所述大脑静脉预测点确定脑部图像对应的灌注参数图，利用灌注参数图计算得到所述脑灌注梗死核心区域，由于提高了大脑前动脉预测点和大脑静脉预测点的定位精度，使得通过大脑前动脉预测点和所述大脑静脉预测点确定脑部图像对应的灌注参数图更准确，从而提高了梗死核心区域的分割精度及鲁棒性。
附图说明
图1a为本申请实施例网络架构的示意图；
图1b为本申请实施例另一网络架构的示意图；
图2为本申请实施例提供的一种图像分割方法的流程示意图；
图3为本申请实施例提供的一种图像分割方法中步骤S22的实现流程示意图；
图4为本申请实施例提供的一种图像分割方法中步骤S32的实现流程示意图；
图5为本申请实施例提供的一种图像分割方法中步骤S42的实现流程示意图；
图6为本申请实施例提供的一种图像分割方法中步骤S23的实现流程示意图；
图7为本申请实施例提供的一种图像分割方法中步骤S24的一种实现流程示意图；
图8为本申请实施例提供的一种图像分割方法中步骤S24的另一种实现流程示意图；
图9为本申请实施例提供的一种图像分割方法的另一种实现流程示意图；
图10为本申请实施例提供的再一种图像分割方法的流程示意图；
图11为本申请实施例提供的一种图像分割装置的结构示意图；
图12为本申请实施例提供的一种图像分割设备的结构示意图；
图13为本申请实施例提供的一种计算机可读存储介质的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本申请的一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请实施例提供的图像分割方法，通过深度学习算法对大脑前动脉点及大脑静脉点进行定位，可以有效地解决大脑动脉点和静脉点的位置自动定位或者手动定 位不准的缺点，为后续计算提供保障。且本申请实施例提供的图像分割方法不需要依赖于正常侧及病灶侧的对比，可以防止在正常侧及病灶侧均是异常侧时，比值检测不出异常的问题，进一步提高分割梗死核心区域的精度及鲁棒性。下面结合附图和实施例对本申请实施例进行详细的说明。
图1a为本申请实施例网络架构的示意图，如图1a所示，在该网络架构中包括CT机11和计算机设备12，其中，CT机11用于采集待分割的脑部图像。CT机11与计算机设备12建立有通信连接，CT机11可以将得到的待分割的脑部图像发送给计算机设备12。计算机设备12中存储有脑部样本图像，计算机设备12可以根据脑部样本图像训练出深度学习定位网络，计算机设备12将待分割的脑部图像输入至深度学习定位网络，得到大脑前动脉预测点和大脑静脉预测点；利用所述大脑前动脉预测点和所述大脑静脉预测点确定所述脑部图像对应的灌注参数图；利用所述灌注参数图确定所述脑灌注梗死核心区域。
图1b为本申请实施例另一网络架构的示意图，如图1b所示，在该网络架构中包括CT机11、计算机设备12和服务器13，其中，CT机11用于采集待分割的脑部图像。CT机11与计算机设备12建立有通信连接，CT机11可以将得到的待分割的脑部图像发送给计算机设备12。服务器13中存储有脑部样本图像，服务器13可以根据脑部样本图像训练出深度学习定位网络。计算机设备12与服务器13同样建立有通信连接，计算机设备12可以从服务器13处获取深度学习定位网络，计算机设备12将待分割的脑部图像输入至深度学习定位网络，得到大脑前动脉预测点和大脑静脉预测点；利用所述大脑前动脉预测点和所述大脑静脉预测点确定所述脑部图像对应的灌注参数图；利用所述灌注参数图确定所述脑灌注梗死核心区域。
结合图1a和图1b所示的应用场景示意图，以下对图像分割方法及装置、设备及计算机可读存储介质的各实施例进行说明。需要说明的是，在本申请实施例中，CT机和计算机设备12可以是集成在一起的。
本申请实施例提供的一种图像分割方法，所述方法应用于图像分割设备，所述图像分割设备可以是计算机设备，本申请实施例提供的方法可以通过计算机程序来实现，该计算机程序在执行的时候，完成本申请实施例提供的图像分割方法中各个步骤。在一些实施例中，该计算机程序可以被图像分割设备的处理器执行。请参见图2，图2为本申请实施例提供的一种图像分割方法的流程示意图，所述方法包括：
步骤S21：获取到第一待分割的脑部图像。
在一实施例中，脑部图像可以为脑部CT图像，脑部CT图像由CT成像技术获得，示例性地，用X(X-ray)线束对人体检查部位一定厚度的层面进行扫描，由探测器接收透过该层面的X线，转变为可见光后，由光电转换器转变为电信号，再经模拟/数字转换器转为数字信号，输入计算机处理以得到CT图像。本申请实施例中，CT图像形成的处理包括：将选定层面分成若干个体积相同的长方体，称之为体素；扫描所得的信息经计算获得每个体素的X线衰减系数或吸收系数，将X线衰减 系数或吸收系数排列成矩阵，即数字矩阵，本申请实施例中，数字矩阵可存储于磁盘或光盘中。经数字/模拟转换器把数字矩阵中的每个数字转为由黑到白不等灰度的小方块，即像素，并按矩阵排列，即构成CT图像。以下各个实施例中图像分割方法均以脑部CT图像为例进行说明。
步骤S22：通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点。
在一些实施例中，通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点之前需要通过训练以得到深度学习定位网络。训练得到深度学习定位网络的过程包括：获取脑部样本图像，在脑部样本图像中对大脑前动脉点及大脑静脉点进行初步标注；利用经过初步标注的脑部样本图像对初始深度学习定位网络进行训练，以得到深度学习定位网络。本申请实施例中，在脑部样本图像中对大脑前动脉点及大脑静脉点进行初步标注时，可通过人工标注。在其他实施方式中，也可以通过其他方式进行标注，例如机器识别标注等。
通过上述方式训练而得的深度学习定位网络在脑部图像中进行定位，以分别获取到大脑前动脉预测点及大脑静脉预测点。本申请实施例中，通过深度学习定位网络在脑部图像中进行定位时，可通过收敛、去卷积等进行计算。
图3为本申请实施例提供的一种图像分割方法中步骤S22的实现流程示意图，如图3所示，步骤S22“通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点”可以通过以下步骤实现：
步骤S31：通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉候选点和大脑静脉候选点。
在通过深度学习定位网络得到大脑前动脉预测点和大脑静脉预测点前，需要先进行定位以获取大脑前动脉候选点和大脑静脉候选点。
步骤S32：利用局部搜索算法确定大脑前动脉候选点对应的大脑前动脉预测点以及大脑静脉候选点对应的大脑静脉预测点。
局部搜索算法是一种简单的贪心搜索算法，该算法每次从当前解的临近解空间中选择一个最优解作为当前解，直到达到一个局部最优解。本申请实施例中使用局部搜索算法从大脑前动脉候选点中获取对应的大脑前动脉预测点，以及从大脑静脉候选点中获取对应的大脑静脉预测点，能够得到更加精准的大脑前动脉预测点以及大脑静脉预测点。
本申请实施例中，利用深度学习定位网络提升了确定大脑前动脉候选点和大脑静脉候选点的精度及鲁棒性。本申请实施例的中，还通过局部搜索算法根据大脑前动脉候选点和大脑静脉候选点确定出大脑前动脉预测点和大脑静脉预测点。
图4为本申请实施例提供的一种图像分割方法中步骤S32的实现流程示意图，如图4所示，步骤S32“利用局部搜索算法计算得到大脑前动脉候选点对应的大脑前动脉预测点以及大脑静脉候选点对应的大脑静脉预测点”可以通过以下步骤实现：
步骤S41：将大脑前动脉候选点为圆心，预设第一半径范围内的区域确定为大脑前动脉点候选区域，将大脑静脉候选点为圆心，预设第二半径范围内的区域确定为大脑静脉点候选区域。
通过深度学习算法获得到的大脑前动脉候选点为圆心，预设第一半径范围内的区域作为大脑前动脉点候选区域。通过深度学习算法获得到的大脑静脉候选点为圆心，预设第二半径范围内的区域作为大脑静脉点候选区域。
其中，预设第一半径及预设第二半径可以为1mm、2mm、2.5mm、10mm等，预设第一半径与预设第二半径的值可以相同，也可以不同，在此不做限定。
步骤S42：利用局部搜索算法分别从大脑前动脉点候选区域中确定大脑前动脉预测点，及从大脑静脉点选区域中确定大脑静脉预测点。
采用局部搜索算法从获取到的大脑前动脉点候选区域中搜索得到大脑前动脉预测点，从大脑静脉点选区域中搜索得到大脑静脉预测点。
局部搜索算法是一种简单的贪心搜索算法，该算法每次从当前解的临近解空间中选择一个最优解作为当前解，直到达到一个局部最优解。本申请实施例所述的方法，通过局部搜索算法结合深度学习算法获取到的大脑前动脉预测点及大脑静脉预测点，提升了确定的大脑前动脉预测点和大脑静脉预测点的精度及鲁棒性。
图5为本申请实施例提供的一种图像分割方法中步骤S42的实现流程示意图，如图5所示，步骤S42“利用局部搜索算法分别从大脑前动脉点候选区域中确定大脑前动脉预测点，及从大脑静脉点选区域中确定大脑静脉预测点”可以通过以下步骤实现：
步骤S51：从大脑前动脉点候选区域中确定时间最大值对应的多个第一组织密度值，及从大脑静脉点选区域中确定时间最大值对应的多个第二组织密度值。
大脑前动脉点候选区域中具有若干条曲线，在一些实施例中，假设若干条曲线的横坐标为时间坐标，纵坐标为组织密度坐标，分别从若干条曲线中获取时间坐标最大值对应的第一组织密度值。有多少条曲线，即有多少个第一组织密度值。例如，若曲线为30条，此时获取得到的第一组织密度值为30个。
大脑静脉点选区域中具有若干条曲线，在一些实施例中，假设若干条曲线的横坐标为时间坐标，纵坐标为组织密度坐标，分别从若干条曲线中获取时间坐标最大值对应的第二组织密度值，有多少条曲线，即有多少个第二组织密度值。例如，若曲线为30条，此时获取得到的第二组织密度值为30个。
步骤S52：将多个所述第一组织密度值中组织密度值的最大值确定为大脑前动脉预测点，将多个所述第二组织密度值中组织密度值的最大值确定为大脑静脉预测点。
若获取得到的第一组织密度值为30个，比较所得到的30个第一组织密度值的大小，将第一组织密度值的最大值所对应的坐标作为大脑前动脉预测点。将时间最大值记作Tm1，第一组织密度值最大值记作Hm1，此时大脑前动脉预测点对应位 (Tm1，Hm1)。
若获取得到的第二组织密度值为30个，比较所得到的30个第二组织密度值的大小，将第二组织密度值的最大值所对应的坐标作为大脑静脉预测点。将时间最大值记作Tm2，第二组织密度值最大值记作Hm2，此时大脑前动脉预测点对应位(Tm2，Hm2)。
本申请实施例所示的方法，先利用深度学习网络模型检测大脑前动脉候选点及大脑静脉候选点，再利用局部搜索算法确定大脑前动脉预测点和大脑静脉预测点，使得确定的大脑前动脉预测点和大脑静脉预测点为对应的多个组织密度值的最大值，进而提升了获取的大脑前动脉预测点和大脑静脉预测点的精度及鲁棒性。
步骤S23：利用大脑前动脉预测点和大脑静脉预测点确定脑部图像对应的灌注参数图。
在得到大脑静脉预测点及大脑前动脉预测点后进一步计算获取脑部图像对应的灌注参数图。
图6为本申请实施例提供的一种图像分割方法中步骤S23的实现流程示意图，如图6所示，步骤S23“利用大脑前动脉预测点和大脑静脉预测点计算得到脑部图像对应的灌注参数图”可以通过以下步骤实现：
步骤S61：获取大脑前动脉预测点对应的函数曲线和大脑静脉预测点对应的函数曲线。
在得到大脑静脉预测点及大脑前动脉预测点后，将大脑静脉预测点及大脑前动脉预测点所在的曲线分别作为大脑静脉预测点对应的函数曲线及大脑前动脉预测点对应的函数曲线。
步骤S62：将大脑前动脉预测点对应的函数曲线作为第一动脉输入函数，及将大脑静脉预测点对应的函数曲线作为静脉输出函数。
将大脑前动脉预测点对应的函数曲线作为第一动脉输入函数，将大脑静脉预测点对应的函数曲线作为静脉输出函数。在一些实施例中，静脉输出函数作为第一动脉输入函数的积累，因此静脉输出的值大于动脉输入的值。
步骤S63：利用静脉输出函数对第一动脉输入函数进行修正，以得到第二动脉输入函数。
为增加鲁棒性，在获取到第一动脉输入函数后，利用静脉输出函数对第一动脉输入函数进行收敛修正，以得到第二动脉输入函数。
步骤S64：利用第二动脉输入函数通过去卷积算法确定局部脑血容积图、局部脑血流量图、平均通过时间图和达峰时间图。
修正后得到的第二动脉输入函数通过去卷积的计算方法得到脑部图像的灌注参数图。在一实施例中，灌注参数图包括局部脑血容积图(rCBV，Regional Cerebral Blood Volume)、局部脑血流量图(rCBF，renal cortical blood flow)、平均通过时间图(MTT)和达峰时间图(Tmax)。
本申请实施例提供的方法，首先利用深度学习检测大脑前动脉候选点及大脑静脉候选点，再利用局部搜索算法获取的大脑前动脉预测点及大脑静脉预测点，使得获取的大脑前动脉预测点及大脑静脉预测点精度及鲁棒性，进而使得计算所得的局部脑血容积图、局部脑血流量图、平均通过时间图和达峰时间图的精度更高。
步骤S24：利用灌注参数图确定脑灌注梗死核心区域。
本申请实施例中，病灶侧的局部脑血流量比上正常侧的局部脑血流量小于预设阈值作为脑梗死核心区域，示例性地，病灶侧的局部脑血流量比上正常侧的局部脑血流量小于30％作为脑梗死核心区域。
图7为本申请实施例提供的一种图像分割方法中步骤S24的一种实现流程示意图，如图7所示，步骤S24可以通过以下步骤实现：
步骤S71：获取第二待分割的脑部图像。
在一些实施例中，第二待分割图像为磁共振(DWI，Diffusion-Weighted Imaging)图像，本申请实施例中的第二待分割的脑部图像为标注了梗死核心区域的图像。本申请实施例中的第二待分割的脑部图像与第一待分割的脑部图像为同一患者的脑部图像。在其他实施例中，本申请实施例中的第二待分割的脑部图像与第一待分割的脑部图像还可以为不同患者的脑部图像。
步骤S72：将达峰时间图中大于预设时间的区域确定为低灌区域。
示例性地，预设时间为6秒，将达峰时间图(Tmax)中时间大于6s的区域确定为低灌区域。
步骤S73：利用局部脑血容积图像、局部脑血流量图像采用卷积神经网络算法从低灌区域中获取到所述第二待分割的脑部图像对应的梗死核心区域。
通过卷积神经网络算法根据标注了梗死核心区域的第二待分割脑部图像从局部脑血容积图像、局部脑血流量图像中得到梗死核心区域。可采用重叠、去畸变等方式利用卷积神经网络算法得到梗死核心区域。
本申请实施例提供的方法，在计算得到梗死核心区域时采用的灌注参数图像为局部脑血容积图像、局部脑血流量图像，进一步提高了对梗死核心区域的分割的准确性。
图8为本申请实施例提供的一种图像分割方法中步骤S24的另一种实现流程示意图，如图8所示，步骤S24“利用灌注参数图计算得到脑灌注梗死核心区域”可以通过以下步骤实现：
步骤S81：获取第二待分割的脑部图像。
在一些实施例中，第二待分割图像为DWI图像，本申请实施例中的第二待分割的脑部图像为标注了梗死核心区域的图像。本申请实施例中的第二待分割的脑部图像与第一待分割的脑部图像为同一患者的脑部图像。在其他实施例中，本申请实施例中的第二待分割的脑部图像与第一待分割的脑部图像还可以为不同患者的脑部图像。
步骤S82：将达峰时间图中大于预设时间的区域确定为低灌区域。
示例性地，预设时间为6s，将达峰时间图中时间大于6s的区域作为低灌区域。
步骤S83：利用所述局部脑血容积图像、局部脑血流量图像及所述平均通过时间图采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
通过卷积神经网络算法根据标注了梗死核心区域的第二待分割脑部图像从局部脑血容积图像、局部脑血流量图像及平均通过时间图中得到梗死核心区域。可采用重叠、去畸变等方式利用卷积神经网络算法得到梗死核心区域。
本申请实施例所述的方法，在计算得到梗死核心区域时采用的灌注参数图像为局部脑血容积图像、局部脑血流量图像及平均通过时间图像，使得本申请实施例的得到的梗死核心区域精度更高和鲁棒性更高。
本申请实施例所述的图像分割方法，利用深度学习检测大脑前动脉候选点及大脑静脉候选点，及利用局部搜索算法获取大脑前动脉预测点及大脑静脉预测点，提升了获取的大脑前动脉预测点及大脑静脉预测点的精度及鲁棒性，进而使得计算所得的局部脑血容积图、局部脑血流量图像、平均通过时间图和达峰时间图的精度更高。本申请实施例所述的方法在分割梗死核心区域时，不需要以病灶侧与正常侧的比值作参考，排除了在病灶侧与正常侧均为异常侧时，分割精度不准的情况，进一步提高了梗死核心区域的分割精度及鲁棒性。
图9为本申请实施例提供的一种图像分割方法的另一种实现流程示意图，如图9所示，本申请实施例所示的方法，步骤S91至步骤S94与步骤S21至步骤S24相同。本申请实施例与图2所示的图像分割方法相比，所述方法还包括：
步骤S95：根据低灌区域及梗死核心区域确定缺血半暗带区域。
梗死核心区域加上缺血半暗带区域等于低灌区域。经由上述实施例得到梗死核心区域后，低灌区域中除梗死核心区域外的其余区域为缺血半暗带区域。
本申请实施例所述的方法，利用深度学习检测大脑前动脉候选点及大脑静脉候选点，及利用局部搜索算法获取大脑前动脉预测点及大脑静脉预测点，提升了获取的大脑前动脉预测点及大脑静脉预测点的精度及鲁棒性，进而使得计算所得的局部脑血容积图、局部脑血流量图像、平均通过时间图和达峰时间图的精度更高。
本申请实施例所述的方法在分割梗死核心区域时，不需要以病灶侧与正常侧的比值作参考，排除了在病灶侧与正常侧均为异常侧时，分割精度不准的情况，进一步提高了梗死核心区域的分割精度及鲁棒性。由于梗死核心区域加缺血半暗带区域等于低灌区域，在梗死核心区域的分割精度及鲁棒性更高的前提下，本申请实施例的方法还提高了缺血半暗带区域的分割精度及鲁棒性。
基于前述的各个实施例，本申请实施例提供再一种图像分割方法，图10为本申请实施例提供的再一种图像分割方法的流程示意图，如图10所示，所述方法包括：
步骤S101，大脑前动脉点和大脑静脉点定位。
首先利用深度学习定位网络定位出来大脑前动脉点和大脑静脉点两个候选点，再通过局部搜索算法确定精准的大脑前动脉点和大脑静脉点，进而确定动脉输入函数AIF和静脉输出函数VOF(Volume Fluent)。
步骤S102，灌注计算。
利用AIF通过去卷积的SVD算法计算出四个灌注参数图rCBV、rCBF、MTT和Tmax。
步骤S103，缺血半暗带和梗死核心分割。
最后通过Tmax大于6s作为低灌区域，利用四个参数图作为输入分割梗死核心，两者错配(mismatch)作为缺血半暗带的分割结果。
本申请实施例提供的方法，通过深度学习定位网络和局部搜索算法自动检测大脑前动脉点和大脑静脉点，提升了获取的大脑前动脉点和大脑静脉点的算法精度和鲁棒性，利用深度学习分割梗死核心，提升了其精度和鲁棒性，且可以解决大脑两侧都有梗死核心区域的特殊情况。
图11为本申请实施例提供的一种图像分割装置的结构示意图，如图11所示，所述图像分割装置包括：获取模块111、定位模块112、关注参数图获取模块113及梗死核心区域分割模块114。
其中，获取模块111配置为获取到第一待分割的脑部图像，定位模块112配置为通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点。灌注参数图获取模块113配置为利用所述大脑前动脉预测点和所述大脑静脉预测点计算得到所述脑部图像对应的灌注参数图。梗死核心区域分割模块114配置为利用灌注参数图计算得到脑灌注梗死核心区域。
在一些实施例中，获取模块111还配置为：
获取脑部样本图像，所述脑部样本图像为对大脑前动脉点和大脑静脉点进行初步标注的图像；
利用所述脑部样本图像对初始深度学习定位网络进行训练，得到所述深度学习定位网络。
在一些实施例中，定位模块112还配置为：
通过所述深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉候选点和大脑静脉候选点；
利用局部搜索算法确定所述大脑前动脉候选点对应的大脑前动脉预测点以及所述大脑静脉候选点对应的大脑静脉预测点。
在一些实施例中，定位模块112还配置为：
所述利用局部搜索算法确定所述大脑前动脉候选点对应的大脑前动脉预测点以及所述大脑静脉候选点对应的大脑静脉预测点，包括：
将所述大脑前动脉候选点为圆心，预设第一半径范围内的区域确定为大脑前动脉点候选区域，将所述大脑静脉候选点为圆心，预设第二半径范围内的区域确定为大脑静脉点候选区域；
利用所述局部搜索算法分别从所述大脑前动脉点候选区域中确定所述大脑前动脉预测点，及从所述大脑静脉点选区域中确定所述大脑静脉预测点。
在一些实施例中，定位模块112还配置为：
从所述大脑前动脉点候选区域中确定时间最大值对应的多个第一组织密度值，及从所述大脑静脉点选区域中确定时间最大值对应的多个第二组织密度值；
将所述多个第一组织密度值中组织密度值的最大值确定为所述大脑前动脉预测点，将所述第二组织密度值中组织密度值的最大值确定为所述大脑静脉预测点。
在一些实施例中，所述灌注参数图包括：局部脑血容积图、局部脑血流量图、平均通过时间图和达峰时间图。
在一些实施例中，关注参数图获取模块113还配置为：
获取所述大脑前动脉预测点对应的函数曲线和所述大脑静脉预测点对应的函数曲线；
将所述大脑前动脉预测点对应的函数曲线作为第一动脉输入函数，及将所述大脑静脉预测点对应的函数曲线作为静脉输出函数；
利用所述静脉输出函数对所述第一动脉输入函数进行修正，以得到第二动脉输入函数；
利用所述第二输入动脉函数通过去卷积算法确定所述局部脑血容积图、所述局部脑血流量图、所述平均通过时间图和所述达峰时间图。
在一些实施例中，梗死核心区域分割模块114还配置为：
获取第二待分割的脑部图像；
将所述达峰时间图中大于预设时间的区域确定为低灌区域；
利用所述局部脑血容积图像、所述局部脑血流量图像采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
在一些实施例中，梗死核心区域分割模块114还配置为：
获取第二待分割的脑部图像；
将所述达峰时间图中大于预设时间的区域确定为低灌区域；
利用所述局部脑血容积图像、所述局部脑血流量图像及所述平均通过时间图采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
在一些实施例中，梗死核心区域分割模块114还配置为：
根据所述低灌区域及所述梗死核心区域确定缺血半暗带区域；
其中，所述梗死核心区域加上所述缺血半暗带区域等于所述低灌区域。
本申请实施例提供的图像分割装置能够实现通过深度学习算法定位获取到大 脑前动脉候选点及大脑静脉候选点，再通过局部搜索算法根据大脑前动脉候选点及大脑静脉候选点确定出大脑前动脉预测点及大脑静脉预测点。利用深度学习检测大脑前动脉候选点及大脑静脉候选点，及利用局部搜索算法获取大脑前动脉预测点及大脑静脉预测点的精度及鲁棒性更加突出。
另外，本申请实施例提供的图像分割装置在分割梗死核心区域时，不需要以病灶侧与正常侧的比值作参考，排除了在病灶侧与正常侧均为异常侧时，分割精度不准的情况，提高了梗死核心区域的分割精度及鲁棒性。
图12为本申请实施例提供的一种图像分割设备的结构示意图，如图12所示，图像分割设备包括相互连接的存储器121、处理器122和通信总线123。
存储器121配置为存储实现上述任意一项的图像分割方法的程序指令。
处理器122配置为执行存储器121存储的程序指令。
通信总线123配置为实现存储器121和处理器122之间的连接通信。其中，处理器122还可以称为中央处理单元(CPU，Central Processing Unit)。处理器122可能是一种集成电路芯片，具有信号的处理能力。处理器122还可以是通用处理器、数字信号处理器(DSP，Digital Signal Processor)、专用集成电路(ASIC，Application Specific Integrated Circuit)、现场可编程门阵列(FPGA，Field Programmable Gate Array)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
存储器121可以为内存条、闪存储器卡(TF卡，Micro SD Card)等，可以存储图像分割设备中全部信息，包括输入的原始数据、计算机程序、中间运行结果和最终运行结果都保存在存储器中。它根据控制器指定的位置存入和取出信息。有了存储器，图像分割设备才有记忆功能，才能保证正常工作。图像分割设备中的存储器按用途存储器可分为主存储器(内存)和辅助存储器(外存),也有分为外部存储器和内部存储器的分类方法。外存通常是磁性介质或光盘等，能长期保存信息。内存指主板上的存储部件，用来存放当前正在执行的数据和程序，但仅用于暂时存放程序和数据，关闭电源或断电，数据会丢失。
在本申请所提供的几个实施例中，应该理解到，所揭露的方法和装置，可以通过其它的方式实现。例如，以上所描述的装置实施方式仅仅是示意性的，例如，模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施 方式方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个计算机可读存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，系统服务器，或者网络设备等)或处理器(processor)执行本申请各个实施方式方法的全部或部分步骤。
图13为本申请实施例提供的一种计算机可读存储介质的结构示意图，如图13所示，本申请的计算机可读存储介质存储有能够实现上述所有图像分割方法的程序文件131，其中，该程序文件131可以以软件产品的形式存储在上述计算机可读存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本申请各个实施方式方法的全部或部分步骤。而前述的存储装置包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质，或者是计算机、服务器、手机、平板等终端设备。
本申请实施例提供的图像分割设备能够实现先通过深度学习算法定位获取到大脑前动脉候选点及大脑静脉候选点。再通过局部搜索算法根据大脑前动脉候选点及大脑静脉候选点确定出大脑前动脉预测点及大脑静脉预测点。利用深度学习检测大脑前动脉候选点及大脑静脉候选点，及利用局部搜索算法获取大脑前动脉预测点及大脑静脉预测点的精度及鲁棒性更加突出。
另外，本申请实施例提供的图像分割设备在分割梗死核心区域时，不需要以病灶侧与正常侧的比值作参考，排除了在病灶侧与正常侧均为异常侧时，分割精度不准的情况，进一步提高了梗死核心区域的分割精度及鲁棒性。
以上仅为本申请实施例的实施方式，并非因此限制本申请实施例的专利范围，凡是利用本申请实施例说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请实施例的专利保护范围内。
工业实用性
本申请实施例提供一种图像分割方法及装置、设备及计算机可读存储介质，其中，所述方法包括：获取到第一待分割的脑部图像；通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点；利用所述大脑前动脉预测点和所述大脑静脉预测点计算得到所述脑部图像对应的灌注参数图；利用所述灌注参数图计算得到所述脑灌注梗死核心区域，以此提高梗死核心区域的分割精度及鲁棒性。

Claims (14)

  1. 一种图像分割方法，所述方法包括：
    获取第一待分割的脑部图像；
    通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点；
    利用所述大脑前动脉预测点和所述大脑静脉预测点确定所述脑部图像对应的灌注参数图；
    利用所述灌注参数图确定所述脑灌注梗死核心区域。
  2. 根据权利要求1所述的图像分割方法，所述通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点之前，所述方法还包括：
    获取脑部样本图像，所述脑部样本图像为对大脑前动脉点和大脑静脉点进行初步标注的图像；
    利用所述脑部样本图像对初始深度学习定位网络进行训练，得到所述深度学习定位网络。
  3. 根据权利要求2所述的图像分割方法，所述通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点，包括：
    通过所述深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉候选点和大脑静脉候选点；
    利用局部搜索算法确定所述大脑前动脉候选点对应的大脑前动脉预测点以及所述大脑静脉候选点对应的大脑静脉预测点。
  4. 根据权利要求3所述的图像分割方法，所述利用局部搜索算法确定所述大脑前动脉候选点对应的大脑前动脉预测点以及所述大脑静脉候选点对应的大脑静脉预测点，包括：
    将以所述大脑前动脉候选点为圆心，预设第一半径范围内的区域确定为大脑前动脉点候选区域，将以所述大脑静脉候选点为圆心，预设第二半径范围内的区域确定为大脑静脉点候选区域；
    利用所述局部搜索算法分别从所述大脑前动脉点候选区域中确定所述大脑前动脉预测点，及从所述大脑静脉点选区域中确定所述大脑静脉预测点。
  5. 根据权利要求4所述的图像分割方法，所述利用所述局部搜索算法分别从所述大脑前动脉点候选区域中确定所述大脑前动脉预测点，及从所述大脑静脉点选区域中确定所述大脑静脉预测点，包括：
    从所述大脑前动脉点候选区域中确定时间最大值对应的多个第一组织密度值，及从所述大脑静脉点选区域中确定时间最大值对应的多个第二组织密度值；
    将所述多个第一组织密度值中组织密度值的最大值确定为所述大脑前动脉预测点，将所述第二组织密度值中组织密度值的最大值确定为所述大脑静脉预测点。
  6. 根据权利要求1所述的图像分割方法，所述灌注参数图包括：局部脑血容积图、局部脑血流量图、平均通过时间图和达峰时间图。
  7. 根据权利要求6所述的图像分割方法，所述利用所述大脑前动脉预测点和所述大脑静脉预测点确定所述脑部图像对应的灌注参数图，包括：
    获取所述大脑前动脉预测点对应的函数曲线和所述大脑静脉预测点对应的函数曲线；
    将所述大脑前动脉预测点对应的函数曲线确定为第一动脉输入函数，及将所述大脑静脉预测点对应的函数曲线确定为静脉输出函数；
    利用所述静脉输出函数对所述第一动脉输入函数进行修正，以得到第二动脉输入函数；
    利用所述第二输入动脉函数通过去卷积算法确定所述局部脑血容积图、所述局部脑血流量图、所述平均通过时间图和所述达峰时间图。
  8. 根据权利要求7所述的图像分割方法，所述利用所述灌注参数图计算得到所述脑灌注梗死核心区域，包括：
    获取第二待分割的脑部图像；
    将所述达峰时间图中大于预设时间的区域确定为低灌区域；
    利用所述局部脑血容积图像、所述局部脑血流量图像采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
  9. 根据权利要求7所述的图像分割方法，所述利用所述灌注参数图确定所述脑灌注梗死核心区域，包括：
    获取第二待分割的脑部图像；
    将所述达峰时间图中大于预设时间的区域确定为低灌区域；
    利用所述局部脑血容积图像、所述局部脑血流量图像及所述平均通过时间图采用卷积神经网络算法从所述低灌区域中获取到所述第二待分割的脑部图像对应的所述梗死核心区域。
  10. 根据权利要求8或9所述的图像分割方法，所述方法还包括：
    根据所述低灌区域及所述梗死核心区域确定缺血半暗带区域；
    其中，所述梗死核心区域加上所述缺血半暗带区域等于所述低灌区域。
  11. 一种图像分割装置，包括：
    获取模块，配置为获取第一待分割的脑部图像；
    定位模块，配置为通过深度学习定位网络对所述脑部图像进行定位，得到大脑前动脉预测点和大脑静脉预测点；
    灌注参数图获取模块，配置为利用所述大脑前动脉预测点和所述大脑静脉预测点计算得到所述脑部图像对应的灌注参数图；
    梗死核心区域分割模块，配置为利用所述灌注参数图计算得到所述脑灌注梗死核心区域。
  12. 一种图像分割设备，所述设备包括：相互藕接的处理器、存储器，其中，
    所述存储器配置为存储实现如权利要求1至10任意一项所述的图像分割方法的程序指令；
    所述处理器配置为执行所述存储器存储的所述程序指令。
  13. 一种计算机可读存储介质，存储有程序文件，所述程序文件能够被执行以实现如权利要求1至10任意一项所述的图像分割方法。
  14. 一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现权利要求1至权利要求10任意一项所述的方法。
PCT/CN2020/100711 2020-03-18 2020-07-07 一种图像分割方法及装置、设备及计算机可读存储介质 Ceased WO2021184600A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
KR1020217020478A KR20210118389A (ko) 2020-03-18 2020-07-07 이미지 분할 방법 및 장치, 기기 및 컴퓨터 판독 가능한 저장 매체
JP2021538462A JP2022529094A (ja) 2020-03-18 2020-07-07 画像セグメンテーション方法及び装置、機器並びにコンピュータ可読記憶媒体

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010192934.5A CN111489360A (zh) 2020-03-18 2020-03-18 一种图像分割方法及相关设备
CN202010192934.5 2020-03-18

Publications (1)

Publication Number Publication Date
WO2021184600A1 true WO2021184600A1 (zh) 2021-09-23

Family

ID=71798168

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/100711 Ceased WO2021184600A1 (zh) 2020-03-18 2020-07-07 一种图像分割方法及装置、设备及计算机可读存储介质

Country Status (5)

Country Link
JP (1) JP2022529094A (zh)
KR (1) KR20210118389A (zh)
CN (1) CN111489360A (zh)
TW (1) TWI767492B (zh)
WO (1) WO2021184600A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114170440A (zh) * 2021-12-31 2022-03-11 上海联影智能医疗科技有限公司 图像特征点的确定方法、装置、计算机设备和存储介质
CN114648514A (zh) * 2022-03-30 2022-06-21 中国人民解放军总医院第二医学中心 一种脑动脉定位提取方法、装置、电子设备及存储介质
CN115731583A (zh) * 2022-11-17 2023-03-03 珠海格力电器股份有限公司 基于指静脉的手指识别方法和装置、电子设备和存储介质
CN115830014A (zh) * 2023-02-03 2023-03-21 脑玺(苏州)智能科技有限公司 缺血半暗带的误标记识别方法、装置、设备及存储介质
CN119559199A (zh) * 2025-01-26 2025-03-04 北京邮电大学 垂直枕束亚支分割方法及装置
CN120580243A (zh) * 2025-04-25 2025-09-02 强联智创（北京）科技有限公司 用于对脑灌注图像序列分割模型进行训练的方法及产品

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112419340B (zh) * 2020-12-09 2024-06-28 东软医疗系统股份有限公司 脑脊液分割模型的生成方法、应用方法及装置
CN113034438B (zh) * 2021-03-01 2021-11-23 北京安德医智科技有限公司 一种ct灌注参数预测方法及装置、电子设备和存储介质
CN115908443B (zh) * 2021-08-16 2026-03-24 武汉中科医疗科技工业技术研究院有限公司 用于灌注成像的动脉输入函数的确定方法和装置
CN113827263B (zh) * 2021-11-08 2025-02-07 上海联影智能医疗科技有限公司 灌注影像处理方法、系统、电子设备及存储介质
CN114121236A (zh) * 2021-11-22 2022-03-01 上海商汤智能科技有限公司 图像检测方法及相关装置、电子设备和存储介质
CN114266789B (zh) * 2021-12-28 2023-05-23 沈阳东软智能医疗科技研究院有限公司 基于脑部影像的梗死区域分割方法、装置、设备及介质
CN115546159B (zh) * 2022-10-13 2023-07-25 数坤(上海)医疗科技有限公司 图像处理方法、装置、设备和存储介质
CN115546201B (zh) * 2022-11-09 2025-11-11 首都医科大学附属北京天坛医院 动静脉点确定方法、装置、电子装置和存储介质
CN116798612A (zh) * 2023-08-23 2023-09-22 北京超数时代科技有限公司 一种用于神经类疾病的数字诊疗系统
CN117975012B (zh) * 2024-02-07 2024-08-09 浙江大学 人体脉络丛分割方法、系统、存储介质及设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103839249A (zh) * 2012-11-23 2014-06-04 上海联影医疗科技有限公司 Ct肝灌注的图像后处理方法和ct肝灌注方法
CN104287764A (zh) * 2014-09-11 2015-01-21 沈阳东软医疗系统有限公司 一种ct灌注成像方法和设备
CN105997128A (zh) * 2016-08-03 2016-10-12 上海联影医疗科技有限公司 利用灌注成像识别病灶的方法及系统
CN109035255A (zh) * 2018-06-27 2018-12-18 东南大学 一种基于卷积神经网络的ct图像中带夹层主动脉分割方法
US20190015061A1 (en) * 2015-12-21 2019-01-17 The Regents Of The University Of California Perfusion digital subtraction angiography
CN110287956A (zh) * 2019-06-13 2019-09-27 北京理工大学 血管中心线自动匹配方法及装置

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5243865B2 (ja) * 2008-07-07 2013-07-24 浜松ホトニクス株式会社 脳疾患診断システム
RU2541126C2 (ru) * 2009-03-26 2015-02-10 Конинклейке Филипс Электроникс Н.В. Визуализация перфузии
CN102028494B (zh) * 2011-01-24 2012-05-09 海纳医信（北京）软件科技有限责任公司 脑灌注图像序列处理方法及系统
CN103186895A (zh) * 2013-04-15 2013-07-03 武汉大学 一种基于神经网络模型的ct灌注医疗图像智能融合方法
US10037603B2 (en) * 2015-05-04 2018-07-31 Siemens Healthcare Gmbh Method and system for whole body bone removal and vascular visualization in medical image data
CN105809670B (zh) * 2016-02-29 2019-07-19 上海联影医疗科技有限公司 灌注分析方法
CN109952598B (zh) * 2016-09-18 2023-08-04 耶达研究及发展有限公司 基于荧光照明来生成3d图像的系统和方法
WO2018119366A1 (en) * 2016-12-23 2018-06-28 Heartflow, Inc. Machine learning of anatomical model parameters
CN107451455A (zh) * 2017-07-29 2017-12-08 广东欧珀移动通信有限公司 解锁控制方法及相关产品
WO2019061202A1 (en) * 2017-09-28 2019-04-04 Shenzhen United Imaging Healthcare Co., Ltd. SYSTEM AND METHOD FOR PROCESSING COLON IMAGE DATA
CN108053446A (zh) * 2017-12-11 2018-05-18 北京奇虎科技有限公司 基于点云的定位方法、装置及电子设备
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
CN109726753A (zh) * 2018-12-25 2019-05-07 脑玺(上海)智能科技有限公司 基于时间信号曲线的灌注动态影像的分割方法及系统
CN110223285B (zh) * 2019-06-13 2021-02-02 中南大学湘雅二医院 基于动物迁移和神经网络模型的影像结果预测方法及系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103839249A (zh) * 2012-11-23 2014-06-04 上海联影医疗科技有限公司 Ct肝灌注的图像后处理方法和ct肝灌注方法
CN104287764A (zh) * 2014-09-11 2015-01-21 沈阳东软医疗系统有限公司 一种ct灌注成像方法和设备
US20190015061A1 (en) * 2015-12-21 2019-01-17 The Regents Of The University Of California Perfusion digital subtraction angiography
CN105997128A (zh) * 2016-08-03 2016-10-12 上海联影医疗科技有限公司 利用灌注成像识别病灶的方法及系统
CN109035255A (zh) * 2018-06-27 2018-12-18 东南大学 一种基于卷积神经网络的ct图像中带夹层主动脉分割方法
CN110287956A (zh) * 2019-06-13 2019-09-27 北京理工大学 血管中心线自动匹配方法及装置

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114170440A (zh) * 2021-12-31 2022-03-11 上海联影智能医疗科技有限公司 图像特征点的确定方法、装置、计算机设备和存储介质
WO2023125828A1 (en) * 2021-12-31 2023-07-06 Shanghai United Imaging Intelligence Co., Ltd. Systems and methods for determining feature points
CN114648514A (zh) * 2022-03-30 2022-06-21 中国人民解放军总医院第二医学中心 一种脑动脉定位提取方法、装置、电子设备及存储介质
CN115731583A (zh) * 2022-11-17 2023-03-03 珠海格力电器股份有限公司 基于指静脉的手指识别方法和装置、电子设备和存储介质
CN115830014A (zh) * 2023-02-03 2023-03-21 脑玺(苏州)智能科技有限公司 缺血半暗带的误标记识别方法、装置、设备及存储介质
CN119559199A (zh) * 2025-01-26 2025-03-04 北京邮电大学 垂直枕束亚支分割方法及装置
CN120580243A (zh) * 2025-04-25 2025-09-02 强联智创（北京）科技有限公司 用于对脑灌注图像序列分割模型进行训练的方法及产品

Also Published As

Publication number Publication date
CN111489360A (zh) 2020-08-04
KR20210118389A (ko) 2021-09-30
TW202137132A (zh) 2021-10-01
JP2022529094A (ja) 2022-06-17
TWI767492B (zh) 2022-06-11

Similar Documents

Publication Publication Date Title
TWI767492B (zh) 一種圖像分割方法、設備及電腦可讀儲存介質
US10076299B2 (en) Systems and methods for determining hepatic function from liver scans
CN109934812B (zh) 图像处理方法、装置、服务器及存储介质
CN109410221B (zh) 脑灌注图像分割方法、装置、服务器和存储介质
US20150141818A1 (en) Vascular imaging method and device
CN104881568A (zh) 一种基于云计算的肿瘤治疗疗效早期评价的系统和方法
US12608814B2 (en) Systems and methods for brain identifier localization
CN115330775B (zh) 一种脑卒中ct和mri影像征象定量评估方法及系统
JP2021084036A (ja) 磁気共鳴イメージで正中矢状平面を決定するための方法、イメージ処理デバイス及び格納媒体
CN112085730A (zh) 感兴趣区域成分分析方法、装置、电子设备和介质
WO2021078040A1 (zh) 一种病灶的定位方法及装置
KR20240096862A (ko) 전안부 octa용 아이 트래킹 방법, 장치, 설비 및 저장 매체
CN119295300A (zh) 一种基于CycleGan的多模态医学影像弹性配准系统
CN119991664A (zh) 颅内动脉狭窄功能学评估方法、装置、设备和存储介质
CN113850794A (zh) 一种图像处理方法及装置
CN111626974B (zh) 冠状动脉造影图像序列的质量评分方法和装置
Zheng et al. Precise segmentation of the left atrium in C-arm CT volumes with applications to atrial fibrillation ablation
CN114266759B (zh) 图像分析方法、装置及计算机设备
US12387358B2 (en) Apparatus and method for configuring scan parameters
CN116491968A (zh) 基于单时相cta的急性缺血性卒中侧支循环评估方法
CN114764807B (zh) 冠脉分割方法、装置、计算机设备和存储介质
CN116563270A (zh) 一种基于ct影像随机区域的肺气肿量化分析方法
CN116071250A (zh) 一种脑动脉狭窄闭塞的asl图像处理系统、设备及终端
CN116030071A (zh) 血管关键点位置确定方法、装置、计算机设备和存储介质
HK40025723A (zh) 一种图像分割方法及相关设备

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2021538462

Country of ref document: JP

Kind code of ref document: A

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

Ref document number: 20925293

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20925293

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 28.03.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20925293

Country of ref document: EP

Kind code of ref document: A1