WO2022012038A1 - 图像处理方法及装置、电子设备、存储介质和程序产品 - Google Patents

图像处理方法及装置、电子设备、存储介质和程序产品 Download PDF

Info

Publication number
WO2022012038A1
WO2022012038A1 PCT/CN2021/075633 CN2021075633W WO2022012038A1 WO 2022012038 A1 WO2022012038 A1 WO 2022012038A1 CN 2021075633 W CN2021075633 W CN 2021075633W WO 2022012038 A1 WO2022012038 A1 WO 2022012038A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
image
name
pixel
sub
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/CN2021/075633
Other languages
English (en)
French (fr)
Inventor
李卓威
夏清
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 JP2022505271A priority Critical patent/JP2022543547A/ja
Priority to KR1020227002923A priority patent/KR20220028011A/ko
Publication of WO2022012038A1 publication Critical patent/WO2022012038A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • 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
    • 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
    • G06T7/10—Segmentation; Edge detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/30048—Heart; Cardiac

Definitions

  • the present disclosure relates to the technical field of image processing, and in particular, to an image processing method and apparatus, an electronic device, a storage medium, and a program product.
  • Cardiovascular and cerebrovascular disease is one of the diseases with the highest mortality rate, among which coronary heart disease has the highest incidence.
  • Coronary heart disease is a series of clinical diseases caused by atherosclerosis leading to narrowing of the coronary lumen, resulting in insufficient blood supply to the myocardium.
  • the embodiments of the present disclosure provide an image processing method and apparatus, an electronic device, a storage medium, and a program product.
  • an image processing method including:
  • the segmenting the target image to obtain the name of at least one pixel in the target image, as the segmentation result includes: inputting the target image into a neural network; The output of the neural network is used to determine the name of at least one pixel in the target image as the segmentation result; wherein, the neural network is trained by a training image including the target object, and the target in the training image is trained. Objects are labeled by the name of at least one target sub-object. Therefore, the target image can be segmented through a neural network, thereby effectively improving the stability, efficiency and accuracy of segmentation, and then improving the stability, efficiency and accuracy of image processing.
  • the neural network can flexibly adjust the structure and implementation method according to the actual situation of the name of the target sub-object included in the training image, the flexibility of key point detection can be improved, and the flexibility of the image processing method can be improved.
  • the neural network is trained by using a training image including the target object, including: determining labels of at least some pixels in the training image according to labels of the target object in the training image ; train the neural network through the training images including the labels of the at least part of the pixels.
  • the neural network can be trained according to the training images containing at least part of the labels of the pixels.
  • the neural network is trained by using a training image including the target object, including: within the target object in the training image, determining a target sub-object closest to the first target pixel , wherein the first target pixel is at least one pixel other than the target object in the training image; the determined label of the target sub-object is used as the label of the first target pixel; The neural network is trained according to the training image including the label of the first target pixel and the label of the target object.
  • the training image with dense labels and complete information can be obtained based on the training image containing the target object annotation to train the neural network, so as to improve the accuracy of the trained neural network without increasing the difficulty of labeling, and then Improve the accuracy of the segmentation results and the names of the final target sub-objects. Improve the accuracy and convenience of image processing.
  • the determining the name of at least one target sub-object in the target object according to the segmentation result includes: determining the name of at least one second target pixel point according to the segmentation result, Wherein, the second target pixel is the pixel included in the target sub-object; the name of each second target pixel in the target sub-object is counted to obtain a statistical result, and the number of Up to name, as the name of the target child object.
  • determining the name of at least one second target pixel point according to the segmentation result includes: using the segmentation result corresponding to each second target pixel point as each second target pixel point The name of the target pixel; or, based on the segmentation result of at least one pixel within a preset range of each second target pixel, the name of each second target pixel is determined.
  • the segmentation results of these auxiliary function pixels can be effectively introduced to determine the second target pixel , improve the accuracy of the determined name of the second target pixel point, and then improve the accuracy of the determined target sub-object name, thereby improving the accuracy of image processing.
  • the method further includes: processing the target object according to the name of at least one of the target sub-objects to obtain a processing result.
  • the target object can be further optimized according to the name of the target sub-object effectively, even if the segmentation result obtained by segmenting the target image before is accurate The rate is low, resulting in a certain error in the determined name, which can also be corrected through processing to obtain more accurate processing results and improve the accuracy and robustness of image processing.
  • the processing of the target object according to the name of at least one of the target sub-objects includes: extracting target sub-objects with the same name in the target object; and/ Or, modify the name of at least one target sub-object in the target object according to the name of the adjacent target sub-object.
  • the target image includes: a cardiac coronary mask image, or a cardiac coronary mask image and a cardiac mask image; the target object includes a coronary centerline.
  • segmentation can be performed based on the mask image, which simplifies the input environment for the subsequent segmentation process compared to segmentation based on the original coronary artery image. , reducing the noise in the image processing process; it can also retain the information of each chamber of the heart through the heart mask image, so that the tree structure information and blood supply position information of the coronary artery can be preserved as much as possible at the same time. Difficulty of segmentation, improve the accuracy of segmentation results and finalized names.
  • an image processing apparatus including:
  • a target image acquisition module configured to acquire a target image including a target object; a segmentation module, configured to segment the target image to obtain the name of at least one pixel in the target image as a segmentation result; a naming module, configured as According to the segmentation result, the name of at least one target sub-object in the target object is determined.
  • an electronic device including:
  • processor configured to invoke the instructions stored in the memory to execute the above image processing method.
  • a computer-readable storage medium having computer program instructions stored thereon, the computer program instructions implementing the above-mentioned image processing method when executed by a processor.
  • a computer program product comprising computer readable code, when the computer readable code is executed in an electronic device, a processor in the electronic device executes the above image Approach.
  • the name of at least one pixel in the target image is obtained by acquiring a target image including the target object and segmenting the target image, and the name of at least one pixel in the target image is obtained as the segmentation result, so as to determine the name of at least one target sub-object in the target object according to the segmentation result. name.
  • FIG. 1 shows a flowchart of an image processing method according to an embodiment of the present disclosure.
  • FIG. 2 shows a schematic diagram of extracting a target object according to an embodiment of the present disclosure.
  • FIG. 3 shows a schematic diagram of an application example according to the present disclosure.
  • FIG. 4 shows a block diagram of an image processing apparatus according to an embodiment of the present disclosure.
  • FIG. 5 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • FIG. 6 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • FIG. 1 shows a flowchart of an image processing method according to an embodiment of the present disclosure.
  • the method may be applied to an image processing apparatus, and the image processing apparatus may be a terminal device, a server, or other processing devices.
  • the terminal device may be user equipment (User Equipment, UE), mobile device, user terminal, terminal, cellular phone, cordless phone, Personal Digital Assistant (PDA), handheld device, computing device, in-vehicle device, available wearable devices, etc.
  • UE user equipment
  • PDA Personal Digital Assistant
  • the image processing method may be implemented by the processor calling computer-readable instructions stored in the memory. As shown in Figure 1, the image processing method may include:
  • Step S11 acquiring a target image including the target object.
  • Step S12 segment the target image to obtain the name of at least one pixel in the target image as the segmentation result.
  • Step S13 according to the segmentation result, determine the name of at least one target sub-object in the target object.
  • the target object can be any object with naming requirements.
  • the target object may be an object that needs to be named by region or segment. Since the target object may contain multiple segments or areas that need to be named, in a possible implementation manner, the target object may be divided into multiple target sub-objects according to the naming requirements of multiple segments or multiple areas.
  • the image processing method proposed in the embodiment of the present disclosure can be used to name each segment of the coronary centerline of the heart.
  • the target object may be the centerline of the coronary artery of the heart, ie the coronary centerline, in which case the target object may contain multiple target sub-objects, and each target sub-object may be tree-like One of the branches of the coronary centerline structure or one of the branches, that is, the coronary centerline segment; in an example, the target object can also be other parts of the coronary artery that have a naming requirement, or the center of the coronary artery of the heart Part of the line, etc., can be flexibly selected according to the actual situation.
  • the methods proposed in the embodiments of the present disclosure may also be applied to naming other organs or tissues containing multiple segments or regions.
  • the target object may be Organs or tissues with multi-regional or multi-segment naming requirements.
  • the subsequent disclosed embodiments are described by taking the target object as the coronary centerline of the heart as an example.
  • the processing method can be flexibly extended according to the method proposed in the embodiments of the present disclosure, and no further examples are provided. .
  • the name of the pixel in the target image which can be the name of the target sub-object corresponding to the pixel.
  • the name of the target sub-object in the target object can also be flexibly determined according to the actual situation.
  • the name of the target sub-object may be the names of multiple coronary artery centerline segments into which the coronary artery centerline can be divided, such as the right coronary artery (RCA, Right Coronary Artery), right coronary origin posterior descending artery (R-PDA, Right-Posterior Descending Artery), left coronary origin posterior descending artery (L-PDA, Left-Posterior Descending Artery), right coronary origin posterior collateral branch (R-PLB, Right-Posterior Lateral Branch), Left-Posterior Lateral Branch (L-PLB, Left-Posterior Lateral Branch), Left Main (LM, Left Main), Mid-Distal Circular Branch (LCX, Leaky Coaxial Cable) , first and second obtuse marginal branch (OM1-2, Obtuse Marginal 1 and 2), left anterior descending branch (LAD, Left Anterior Descending), first and second diagonal branch (D1-2, Diagonal 1
  • the target image can be any image that contains the target object.
  • the target image can be an image that contains all the target objects.
  • the target image can also be a partial image.
  • the image of the target object can be selected flexibly according to the actual situation.
  • the target image may be a mask image, and the type of the mask image may be flexibly determined according to the actual situation.
  • the target object may include the coronary centerline, and accordingly, the target image may be an image including the coronary centerline or may be used to extract the coronary centerline line image. Therefore, in one possible implementation, the target image may include a cardiac coronary mask image.
  • the coronary artery mask image may be a binary image, for example, two values of 0 and 1 can be used to distinguish whether the pixels in the image belong to the coronary artery or the background irrelevant to the coronary artery.
  • the target image may include a cardiac coronary mask image and a cardiac mask image.
  • the form of the cardiac coronary mask image can be as described in the above disclosed embodiments, which will not be repeated here.
  • the heart mask image may be a mask image formed by the heart to which the coronary arteries of the heart belong, and various parts or chambers of the heart are divided by different values.
  • the target image may be a mask image formed by superimposing a coronary artery mask image and a cardiac mask image. During the superposition process, pixels in the cardiac mask image that do not belong to the coronary artery may be marked as Background values in the coronary mask image of the heart.
  • the target image is a mask image obtained by superimposing a cardiac mask image and a cardiac coronary mask image as an example for description, and other possible implementations can be expanded with reference to the subsequent disclosed embodiments. No more enumerating.
  • segmentation can be performed based on the mask image, which simplifies the input environment for the subsequent segmentation process compared to segmentation based on the original coronary artery image. , reducing the noise in the image processing process; it can also retain the information of each chamber of the heart through the heart mask image, so that the tree structure information and blood supply position information of the coronary artery can be preserved as much as possible at the same time. Difficulty of segmentation, improve the accuracy of segmentation results and finalized names.
  • step S12 may be used to segment the target image including the target object to obtain the name of at least one pixel in the target image as the segmentation result.
  • the segmentation may be to classify each pixel in the target image, so as to determine the category of each pixel in the target image; in a possible implementation, the segmentation may also be to classify each pixel in the target image.
  • Each foreground pixel in the target image (for example, the pixel belonging to the coronary artery in the coronary artery mask image) is classified, so as to determine the category of each foreground pixel in the target image, and the like.
  • the implementation manner of the segmentation can be flexibly determined according to the actual situation, and reference may be made to the following disclosed embodiments, which will not be expanded here.
  • the name of each target sub-object of the target object in the target image can also be determined according to the classification result of each pixel in the target image.
  • the name of each target sub-object in the target object may be determined according to the segmentation result; in a possible implementation manner, only the names of some target sub-objects in the target object may be determined according to requirements , the number of the determined target sub-objects and which target sub-objects to select for name determination can be flexibly selected according to the actual situation, which are not limited in the embodiments of the present disclosure. Specifically how to determine the name of the target sub-object according to the classification results of these pixel points, the implementation method can be flexibly selected according to the actual situation, please refer to the subsequent disclosed embodiments for details, and will not be expanded here.
  • a target image including the target object is acquired, and a segmentation result is obtained by segmenting the target image, so that the name of at least one target sub-object in the target object is determined according to the segmentation result.
  • the segmentation can be performed based on the mask image, thereby simplifying the input environment for segmentation and reducing noise in the image processing process.
  • the realization form of the target image can be flexibly determined according to the actual situation. Therefore, the manner of acquiring the target image including the target object can also be flexibly determined according to the actual situation.
  • the method of obtaining the target image may be: performing blood vessel segmentation on the original coronary image of the heart to obtain a coronary mask image, and then extracting the centerline of the coronary mask image to obtain the coronary artery mask image.
  • the coronary artery centerline image is used as the target object, and then the coronary artery centerline and the coronary artery mask image are superimposed to obtain a target image containing the target object.
  • the method of acquiring the target image may also be: performing blood vessel segmentation on the original coronary image of the heart to obtain a mask image of the coronary artery, and then extracting the centerline of the mask image of the coronary artery to obtain the coronary artery mask image.
  • the blood vessel centerline image is used as the target object, and the original heart image is segmented to obtain the heart mask image, and then the coronary centerline, the coronary artery mask image and the heart mask image are superimposed to obtain the target object. target image.
  • the remaining possible implementation manners of step S11 can be flexibly selected according to the actual situation of the target image, and will not be illustrated one by one.
  • the target image may be segmented through step S12 to obtain a segmentation result.
  • the name of the target sub-object to which at least some of the pixels in the target image are closest to or belong to may be determined according to the relative positions of the pixels in the target image and the target object, as the segmentation result.
  • the segmentation result of the target image may also be obtained through a neural network, in this case, step S12 may include:
  • the name of the pixel output by the neural network can be the name of each pixel in the target image, or the name of some pixels in the target image, which can be flexibly determined according to the actual situation.
  • the output of the neural network may be the name of each foreground pixel in the target image, and the name of each foreground pixel is used as the segmentation result.
  • the realization form of the training image may refer to the realization form of the above target image, which will not be repeated here.
  • the manner of labeling the target object of the training image is not limited in the embodiments of the present disclosure, and can be flexibly determined according to the function implemented by the neural network.
  • each pixel in the target object in the training image can also be marked separately, or each pixel in the training image can be marked separately.
  • multiple target sub-objects in the training image can be labeled separately.
  • the target object may be divided into multiple target sub-objects, and then some target sub-objects or the name of each target sub-object may be marked, so that the pixels covered by the target sub-object may be uniformly marked. Which labeling method is selected can be flexibly selected according to the actual situation, which is not limited in the embodiments of the present disclosure.
  • the implementation form and training method of the neural network are not limited in the embodiments of the present disclosure, which can be flexibly selected according to the actual situation, and the initial model of the neural network can be arbitrarily selected according to the actual situation.
  • -Vnet as an implementation of a neural network.
  • the loss function adopted for training can also be flexibly selected according to the actual situation.
  • the dice loss can be used as the loss function to train the neural network.
  • the target image can be segmented through a neural network, because the neural network for segmenting the target image can be trained by training images including the target object, and the training The target object in the image is annotated by the name of at least one target sub-object. Therefore, based on the neural network obtained from the above training image, the pixels belonging to the target object in the target image can be segmented, so as to obtain the segmentation result of each segmented pixel, and the segmentation result can be the corresponding pixel in the target image.
  • the possible name types can be referred to the above disclosed embodiments, which will not be repeated here.
  • the neural network is trained through the training images marked with the names of the target sub-objects, so that the trained neural network is used to realize the segmentation of the target image, and the segmentation results are obtained.
  • the neural network can be effectively used to achieve pixel-level segmentation of the target image, and the segmentation results can be obtained more conveniently, thereby effectively reducing the difficulty of image processing and improving the practicability and generalization ability of image processing.
  • the neural network is trained by the training images including the target object, which may include: according to the target in the training image The labeling of the object determines the labels of at least part of the pixel points in the training image; the neural network is trained through the training image including the labels of at least part of the pixel points.
  • the labeling form of the target object in the training image can be flexibly selected according to the actual situation.
  • the method of determining at least part of the pixel point labels in step S21 can also occur accordingly.
  • the label of the target object in the training image can be directly used as the label of the corresponding pixel in the target image.
  • the target object can be divided into multiple target sub-objects, and each target sub-object can be named by naming each segment. Therefore, the labeling of the pixels covered by the named target sub-objects can be used as these target sub-objects.
  • the label of the pixel point is flexibly selected according to the actual situation.
  • the method of determining at least part of the pixel point labels in step S21 can also occur accordingly. Variety.
  • the label of the target object in the training image can be directly used as the label of the corresponding pixel in the target image.
  • the target object can be divided into multiple target sub-objects, and each target sub-object can be named by naming each segment. Therefore, the labeling of the pixels
  • the neural network can be trained according to the training images containing at least part of the labels of the pixels.
  • a neural network for segmenting named functions to improve the convenience of image processing.
  • the target object may be the coronary artery centerline
  • the coronary artery centerline is a part of the cardiac coronary artery mask
  • there are still some pixels in the coronary mask image which do not belong to the target object, namely the coronary centerline, but belong to the coronary arteries. Therefore, the names corresponding to these pixels can help determine the name of the coronary centerline.
  • each pixel in the training image may be labelled, or each pixel in the target object in the training image may be labelled. Pixels are labeled. Therefore, these pixels, which help determine the name of the coronary centerline, may not contain labels.
  • the neural network is trained by the training image including the target object, which may include: in the target object in the training image, determining the target sub-object closest to the first target pixel, wherein, The first target pixel is at least one pixel other than the target object in the training image; the label of the determined target sub-object is used as the label of the first target pixel; according to the label including the first target pixel and the target object Annotated training images to train the neural network.
  • the first target pixel point may be at least one pixel point in the training image that does not belong to the target object.
  • pixel point to select as the first target pixel point it can be flexibly determined according to the actual situation.
  • the pixels in the training image that are located outside the target object but can assist in naming the target object can be used as the first target pixel.
  • the target object as the coronary centerline as an example
  • the pixel points in the training image that are located outside the coronary centerline and belong to the coronary artery of the heart can be used as the first target pixel point; since the realization form of the training image can be the same as that of the target image, and the above disclosures are implemented.
  • the target image may be a mask image. Therefore, in one example, the pixels in the training image that do not belong to the background and do not belong to the target object may be all or part of the first target pixels.
  • the label of the first target pixel can be determined according to the label of at least some of the pixels in the training image, and the determination method of the first target pixel can be flexibly selected according to the actual situation, which is not limited to the present disclosure Examples. It can be seen from the above content that, in a possible implementation manner, in the target object of the training image, the label of the target sub-object closest to the first target pixel can be used as the label of the first target pixel. In an example, the target sub-object closest to the first target pixel may be the target sub-object to which the pixel closest to the first target pixel in the target object belongs.
  • the label of the target object in the training image can also be directly used as the label of the corresponding pixel in the target image by the method described in the above disclosed embodiments, so as to obtain dense labels training images to train the neural network.
  • the above process is described by taking the target object as the coronary centerline as an example.
  • the labels of the pixels on the coronary centerline can be used as the labels of these pixels, and then neither can The pixels that belong to the background and do not belong to the coronary center line are regarded as the first target pixel.
  • the label of the first target pixel point, and then a training image containing a dense label is obtained, and a better training result can be obtained by training the neural network through the training image containing the dense label.
  • the training image with dense labels and complete information can be obtained based on the training image containing the target object annotation to train the neural network, so as to improve the accuracy of the trained neural network without increasing the difficulty of labeling, and then Improve the accuracy of the segmentation results and the names of the final target sub-objects. Improve the accuracy and convenience of image processing.
  • step S13 After obtaining the segmentation result of the target image through any of the above disclosed embodiments, the name of at least one target sub-object in the target object may be determined based on the segmentation result in step S13.
  • the implementation manner of step S13 is not limited.
  • the segmentation result of any pixel point included in the target sub-object may be used as the name of the target sub-object.
  • step S13 may include:
  • Step S131 determine the name of at least one second target pixel, wherein the second target pixel is a pixel included in the target sub-object;
  • Step S132 count the names of the second target pixel points in the target sub-object, obtain a statistical result, and use the name with the largest number in the statistical result as the name of the target sub-object.
  • the second target pixel point may be a pixel point included in the target sub-object, that is, a pixel point covered by the target sub-object.
  • the segmentation result of the second target pixel included in the same target sub-object may be accurate, that is, the same as the corresponding target sub-object.
  • the name of the object is the same or may be inaccurate, that is, it is different from the name of the corresponding target sub-object; in this case, if the segmentation result of any second target pixel contained in the target sub-object is used as the name of the target sub-object , it is possible to get inaccurate naming results.
  • step S131 may be used to obtain the name of at least one second target pixel in the target sub-object according to the segmentation result.
  • the number of second target pixels in the target sub-object obtained in step S131 can be flexibly determined according to the actual situation.
  • each second target pixel included in the target sub-object may be obtained, and in a possible implementation, random sampling may be performed on the second target pixel included in the target sub-object to obtain How to sample and the number of samples to obtain part of the second target pixels can be flexibly determined according to the actual situation, which is not limited in the embodiments of the present disclosure.
  • the name of the obtained second target pixels needs to be determined according to the segmentation result.
  • the way of determining the name can be flexibly determined according to the actual situation.
  • the name of at least one second target pixel is determined, which may include:
  • the segmentation result of the second target pixel point may be directly used as the name of the second target pixel point.
  • the segmentation result of the pixel points in the target image may not be completely accurate.
  • the segmentation result of the second target pixel point is used as the name of the second target pixel point, there may also be Therefore, in a possible implementation manner, the name of the second target pixel point may be determined according to the segmentation result of at least one pixel point within the preset range of the second target pixel point.
  • the preset range of the second target pixel can be flexibly set according to the actual situation, such as 9 neighborhoods, 16 neighborhoods, etc. around the second target pixel, which are not limited in the embodiments of the present disclosure.
  • the pixels near the second target pixel have a high probability of belonging to the same target sub-object as the second target pixel. Therefore, through these pixels, a more accurate name of the second target pixel can be determined.
  • the segmentation result of each pixel included in the preset range of the second target pixel may be obtained, and then, from the obtained segmentation results, the segmentation result with the highest ratio or the largest number is selected, or the The segmentation result whose proportion or quantity exceeds the set threshold (for example, the proportion exceeds 50%, etc.) is used as the name of the second target pixel; in an example, the pixels included in the preset range of the second target pixel can also be Perform sampling, and then select the segmentation results with the highest proportion or the largest number from the segmentation results corresponding to the pixels obtained by these sampling, or select the segmentation results whose proportion or number exceeds the set threshold (for example, the proportion exceeds 50%, etc.) , as the name of the second target pixel.
  • the set threshold for example, the proportion exceeds 50%, etc.
  • the target image there may be some pixels in the target image. Although they do not belong to the target object, they can assist in determining the name of the target sub-object. Therefore, by using at least one pixel within a preset range based on the second target pixel The segmentation result of the point, to determine the name of the second target pixel point, can effectively introduce the segmentation results of these auxiliary function pixels to determine the name of the second target pixel point, and improve the accuracy of the determined name of the second target pixel point , and then improve the accuracy of the determined target sub-object name, thereby improving the accuracy of image processing.
  • the name of the target sub-object may be obtained based on the statistical result of the names of the second target pixel in step S132.
  • the implementation of step S132 can be flexibly determined according to the actual situation. For example, after counting the names of each second target pixel or part of the second target pixels in the target sub-object, the names with the largest number of the counted names can be counted.
  • the name is used as the name of the target sub-object, or the name with the highest proportion of the names of these statistics is used as the name of the target sub-object, or the number of the names of these statistics exceeds the preset threshold or the proportion exceeds the preset threshold.
  • the value of the preset threshold is not limited in the embodiments of the present disclosure, and can be flexibly set according to the actual situation.
  • each coronary centerline segment of the coronary artery centerline can be sequentially determined. For each coronary artery centerline segment, it can be traversed first. Each pixel it contains is the second target pixel, and the names of these second target pixels are determined. The way of determination can be directly determined according to its segmentation result, or it can be determined jointly according to the segmentation results of the surrounding pixels. For determination, reference may be made to the above disclosed embodiments, which will not be repeated here. After the names of the second target pixel points are determined, the name with the largest number or the highest proportion in the coronary artery center line segment can be counted as the name of the coronary artery center line segment.
  • the name of at least one second target pixel is determined according to the segmentation result, and the name of the determined second target pixel is counted, so as to obtain the name of the target sub-object according to the statistical result.
  • the image processing method proposed by the embodiment of the present disclosure may further include: step S14 , processing the target object according to the name of at least one target sub-object to obtain a processing result.
  • the target object in the target image can be further processed and optimized based on the determined name according to actual requirements, and the final image processing results. How to process according to the name of the target sub-object and what kind of processing result is obtained can be flexibly determined according to the actual image processing requirements. For details, please refer to the following disclosed embodiments, which will not be expanded here.
  • the target object By processing the target object according to the name of at least one target sub-object to obtain the processing result, the target object can be further optimized according to the name of the target sub-object, even if the accuracy of the segmentation result obtained by segmenting the target image before is relatively low It can also be corrected through processing to obtain more accurate processing results and improve the accuracy and robustness of image processing.
  • step S14 may include: extracting target sub-objects with the same name in the target object; and/or, according to the names of adjacent target sub-objects, extracting at least one target sub-object in the target object The name of the object is corrected.
  • the manner of extracting target sub-objects with the same name in the target object is not limited.
  • the target object is the coronary centerline
  • the coronary artery centerline may be a tree-like structure, for the coronary artery centerline segments belonging to the same name, the inaccurate segmentation may cause the The coronary centerline segment contains multiple branches.
  • the coronary artery centerline segment containing multiple branches under the same name may be extracted, and the most complete or longest branch among them may be selected as the extraction result.
  • 2 shows a schematic diagram of extracting a target object according to an embodiment of the present disclosure. As can be seen from FIG.
  • the name of each coronary artery center line segment in the coronary artery center line of the target image 21 before extraction has been determined, And the coronary center line segment of multiple branches is included under the same name; the coronary center line segment of the extracted target image 22 contains only one branch coronary center line segment under the same name, such as the coronary center line segment named R-PLB
  • R-PLB the coronary center line segment
  • some coronary artery centerline segments in the coronary artery centerline may be as follows: that is, the coronary artery centerline segment may be located in two coronary artery centerline segments with the same name between the other coronary centerline segments, and the name of the coronary artery centerline segment is different from the names of the two other coronary artery centerline segments. Due to the continuity of the coronary centerline, the name of the coronary centerline segment is likely to be inaccurate. Therefore, the name of the current coronary centerline segment can be modified based on the names of other adjacent coronary centerline segments, so that the coronary The vein centerline remains continuous.
  • the above-mentioned extraction process and correction process may be included at the same time, or only a certain process among them may be included according to the actual situation, or other correction processes may also be included, and these processes may be included in the implementation process.
  • the execution order can also be flexibly selected according to the actual situation, and how to realize it can be flexibly determined according to the actual situation, and is not limited to the above disclosed embodiments.
  • Cardiovascular and cerebrovascular disease is one of the diseases with the highest mortality rate, among which coronary heart disease has the highest incidence.
  • Coronary heart disease is a series of clinical diseases caused by atherosclerosis leading to narrowing of the coronary lumen, resulting in insufficient blood supply to the myocardium, including angina pectoris, myocardial infarction, myocardial failure, arrhythmia and sudden death. Therefore, coronary stenosis, plaque detection and other results have guiding significance for diagnosis and subsequent treatment, and coronary centerline extraction based on cardiac coronary angiography (CTA, Computed Tomography Angiography) images is a prerequisite.
  • CTA Computed Tomography Angiography
  • the naming methods in the related art are mainly based on two categories: the first category is based on knowledge and modeling, and this category of methods is mainly based on matching the target blood vessels with a general model based on statistics; the second category is learning-based classification algorithms, This class of methods is based on the extraction of artificially designed features for centerline classification and naming.
  • Another scheme uses a tree bidirectional long short term memory network (LSTM, Long Short Term) to learn tree structure information for classification and naming. Whether it is the naming of centerlines based on modeling criticism or the naming of treelines based on rules, these two related schemes have at least the problems of complex design and insufficient generalization ability.
  • LSTM long short term memory network
  • the main challenge of the vessel naming task lies in the huge differences and variability among individuals.
  • the clinical basis of vessel naming is the blood supply site of the blood vessel, so the core learning information should be the blood supply site, that is, the position of the blood vessel relative to the heart. Any peeling or loss
  • the methods based on the relative position of the heart have a certain degree of information loss, resulting in low accuracy and weak generalization ability.
  • the application example of the present disclosure proposes a segmentation-based end-to-end naming processing method.
  • Most of the related technologies are based on the workflow of segmenting blood vessels first, then extracting the centerline, and finally naming the centerline (implemented by classifying the centerline).
  • the application example of the present disclosure is to reconstruct the problem: the problem of naming the centerline is reconstructed into the instance segmentation problem of the blood vessel mask image, that is, after the centerline is extracted, the centerline is not directly named, but the blood vessel mask image (each pixel point There are certain label names) for segmentation, and vote on the segmentation results to determine the name of the centerline.
  • Pixel-based segmentation model training will provide more intensive annotation and information, and in order to preserve the blood supply position information of blood vessels, the mask images of each chamber of the heart are superimposed on the mask images of coronary vessels as input. Such a process completely preserves all information and reduces noise compared to the original image, does not require manual feature extraction, does not require rule-based design, and is easy to embed into existing workflows.
  • the application example of the present disclosure includes a training process and a prediction process, wherein the training process includes the following steps: the first step is to prepare the input: obtain the cardiac coronary mask image and the whole cardiac mask image through the segmentation model, superimpose the two mask images and process as The final input; the second step is to prepare the supervision data: first obtain the correctly named coronary centerline, and then associate each pixel label of the coronary mask image with each centerline closest to the coronary centerline; the first Three-step multi-label segmentation training using Res-Vnet type and dice loss.
  • the prediction process includes the following steps: the first step is the same as the preparation input step of the training process, that is, the mask image as the final input and the extracted coronary centerline are obtained; the second step is segmented by the trained neural network; the third step is based on The segmentation votes and names the extracted centerline segments; the fourth step extracts the target blood vessels concerned by doctors through post-processing and corrects possible errors.
  • the application example of the present disclosure constructs the blood vessel centerline naming task into a blood vessel mask image classification task, thereby reducing the learning difficulty and improving the robustness of the centerline naming.
  • the application example of the present disclosure takes the blood vessel mask image and the heart mask image together as input, which reduces the difficulty of direct learning of the original image, and retains all necessary information required for all naming, so that the learning-based method is compared with the traditional method based on The regular tree line naming algorithm has better generalization ability.
  • FIG. 3 shows a schematic diagram of an application example according to the present disclosure. As shown in FIG. 3 , an image processing method is proposed in an embodiment of the present disclosure. The image processing process can be roughly divided into four steps.
  • the coronary artery mask map 31 and the heart mask map 32 are obtained respectively, and the two mask maps are merged as a combined mask map.
  • the centerline of the coronary artery mask is extracted, and the coronary artery centerline composed of multiple coronary artery centerline segments (target sub-objects) is obtained as the target object.
  • the merged mask image is then superimposed with the target object to obtain a target image 33 including the target object (coronary center line).
  • the target image is segmented through the trained neural network 34, and a segmentation result 35 of each foreground pixel point (ie, the pixel point belonging to the coronary artery) in the target image is obtained.
  • the name 36 of each coronary artery center line segment of the coronary artery center line in the target image is determined by voting for naming.
  • the voting naming process can be as follows:
  • For each coronary artery center line segment determine the pixels contained in it as the second target pixel point, then count the names corresponding to the segmentation results of these second target pixel points, and select the name with the largest number as the coronary artery center line segment. The name.
  • post-processing can be performed to retain the most complete or longest branch of the coronary artery centerline segment containing multiple branches with the same name as the final result. Correct the discontinuous coronary center line segment. For example, the parent node (parent line segment) and child node (child line segment) of a center line segment have the same name, but the coronary artery center line segment is inconsistent with their names. The coronary centerline segment is modified to the same name as its parent and child nodes.
  • the training process of the neural network used for segmenting the target image in the second step may be as follows: first, prepare a training image, and the training image may be in the same form as the target image, that is, the mask image of the coronary artery of the heart. , The image obtained by superimposing the heart mask image and the coronary centerline is used as the training image.
  • the generation method of the supervision data can be as follows: firstly, the coronary artery centerline can be labeled by segment naming, so that the label of the pixel point in each coronary artery centerline segment is the corresponding coronary artery centerline segment the label name.
  • the pixels that belong to the foreground and do not belong to the coronary center line segment are respectively used as the first target pixel points, and then the label (also the label) of the nearest pixel point to the first target pixel point on the coronary center line can be used. as the label of the first target pixel.
  • a training image containing labels can be obtained, and a trained neural network can be obtained by inputting the training images containing labels into a neural network for training.
  • the neural network can use the Res-Vnet neural network as the base model, and use the dice loss as the loss function to perform multi-label segmentation training to obtain the final training result.
  • the coronary artery centerline segmentation naming task can be constructed as an instance segmentation task of the blood vessel mask image; at the same time, the image processing method proposed in the application example of the present disclosure can use the cardiac coronary mask image and the cardiac mask image as the Input, compared with the original image of the coronary artery as input, simplifies the input environment, reduces noise, and retains all tree structure information and blood supply position information; in addition, in the naming process, the segmentation results are voted.
  • the naming of the coronary centerline segment depends on the segmentation labels that appear the most in the segment of pixels.
  • the naming results are further post-processed. Since the segmentation results depend on the training data, post-processing can maintain the robustness of the final processing results even when the training data is small.
  • the image processing method of the application example of the present disclosure proposes a new problem architecture, that is, the centerline segmentation naming task is constructed as an instance segmentation task of a blood vessel mask image; at the same time, a new input idea is provided.
  • this technology combines two A kind of mask image is used as input, which simplifies the input environment, reduces noise and preserves all tree structure information and blood supply position information; from the result of segmentation back to the centerline naming, the voting mechanism is adopted, and the naming of the blood vessel segment depends on the segment Click on the category with the most segmentation labels, so that even if the segmentation results are not good, as long as the segmentation results are generally correct, the centerline naming will be correct; post-processing based on prior knowledge, the learning algorithm depends on the training data, and the post-processing is based on the training data. It can also maintain the robustness of the final result when it is small.
  • the innovative points of the application examples of the present disclosure include at least: a centerline naming strategy based on coronary multi-label segmentation; a voting strategy for returning from mask image segmentation to centerline naming; and a post-processing method based on segmentation results and prior knowledge .
  • the solution provided by the application example of the present disclosure is based on an end-to-end design, and has the advantages of clear process, higher accuracy, and strong robustness. At the same time, the solution provided by the application example of the present disclosure maintains a high speed, and each patient needs about 3 to 5 seconds.
  • the application examples of the present disclosure can at least be applied to the processing of cardiac coronary images in aspects such as assisted diagnosis of cardiovascular and cerebrovascular diseases, telemedicine diagnosis, cloud platform assisted intelligent diagnosis, and medical labeling platform.
  • the cardiologist obtains the patient's cardiac coronary angiography data to extract the centerline, the centerline is segmented and named for the detection and positioning of vascular plaque and stenosis, thereby generating a structured report.
  • the image processing method in the embodiment of the present disclosure is not limited to be applied to the above-mentioned processing of coronary images of the heart, and may be applied to any image processing, which is not limited in the embodiment of the present disclosure.
  • a product for defining a centerline named as a coronary mask image segmentation task, or including a whole heart mask image and a cardiac coronary mask image as product input, etc. fall within the scope of protection of the embodiments of the present disclosure.
  • the embodiments of the present disclosure also provide image processing apparatuses, electronic devices, computer-readable storage media, and program products, all of which can be used to implement any image processing method provided by the embodiments of the present disclosure.
  • image processing apparatuses electronic devices, computer-readable storage media, and program products, all of which can be used to implement any image processing method provided by the embodiments of the present disclosure.
  • program products all of which can be used to implement any image processing method provided by the embodiments of the present disclosure.
  • FIG. 4 shows a block diagram of an image processing apparatus according to an embodiment of the present disclosure.
  • the image processing apparatus may be a terminal device, a server, or other processing devices.
  • the terminal device may be UE, mobile device, user terminal, terminal, cellular phone, cordless phone, PDA, handheld device, computing device, vehicle-mounted device, wearable device, etc.
  • the image processing apparatus may be implemented by a processor invoking computer-readable instructions stored in a memory.
  • the image processing apparatus 40 may include:
  • the target image acquisition module 41 is configured to acquire a target image including the target object.
  • the segmentation module 42 is configured to segment the target image to obtain the name of at least one pixel in the target image as a segmentation result.
  • the naming module 43 is configured to determine the name of at least one target sub-object in the target object according to the segmentation result.
  • the segmentation module is configured to: input the target image into the neural network; determine the name of at least one pixel in the target image according to the output of the neural network as the segmentation result; wherein, the neural network includes the target image
  • the training image of the object is used for training, and the target object in the training image is annotated by the name of at least one target sub-object.
  • the neural network is trained on the training image including the target object, including: determining the label of at least part of the pixel points in the training image according to the label of the target object in the training image; Label the training images to train the neural network.
  • the neural network is trained on the training image including the target object, including: in the target object in the training image, determining the target sub-object closest to the first target pixel, wherein the first target The pixel point is at least one pixel point other than the target object in the training image; the label of the determined target sub-object is used as the label of the first target pixel point; according to the training including the label of the first target pixel point and the label of the target object images to train the neural network.
  • the naming module is configured to: determine the name of at least one second target pixel point according to the segmentation result, where the second target pixel point is a pixel point included in the target sub-object; The name of the second target pixel is obtained from the statistical result, and the name with the largest number in the statistical result is taken as the name of the target sub-object.
  • the naming module is further configured to: use the segmentation result corresponding to the second target pixel as the name of the second target pixel; or, based on at least one pixel within a preset range of the second target pixel The segmentation result of the point determines the name of the second target pixel point.
  • the image processing apparatus 40 further includes a processing module, and the processing module is configured to: process the target object according to the name of at least one target sub-object to obtain a processing result.
  • the processing module is further configured to: extract target sub-objects with the same name in the target object; and/or, according to the names of adjacent target sub-objects, extract at least one target object in the target object The names of the sub-objects are corrected.
  • the target image includes: a cardiac coronary mask image, or a cardiac coronary mask image and a cardiac mask image; the target object includes a coronary centerline.
  • Embodiments of the present disclosure further provide a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium.
  • An embodiment of the present disclosure further provides an electronic device, comprising: a processor; a memory configured to store instructions executable by the processor; wherein the processor is configured to invoke the instructions stored in the memory to execute the above method.
  • the electronic device may be provided as a terminal, server or other form of device.
  • Embodiments of the present disclosure also provide a computer program product, including computer-readable codes.
  • a processor in the device executes a method configured to implement the image processing method provided in any of the above embodiments. instruction.
  • Embodiments of the present disclosure further provide another computer program product for storing computer-readable instructions, which, when executed, cause the computer to perform the operations of the image processing method provided by any of the foregoing embodiments.
  • FIG. 5 shows a block diagram of an electronic device 800 according to an embodiment of the present disclosure.
  • electronic device 800 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, fitness device, and personal digital assistant, among other terminals.
  • an electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812 , sensor component 814 , and communication component 816 .
  • the processing component 802 generally controls the overall operations of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 802 may include one or more processors 820 to execute instructions to perform all or some of the steps of the methods described above. Additionally, processing component 802 may include one or more modules that facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.
  • Memory 804 is configured to store various types of data to support operation at electronic device 800 . Examples of such data include instructions for any application or method configured to operate on electronic device 800, contact data, phonebook data, messages, pictures, videos, and the like.
  • Memory 804 may be implemented by any type of volatile or non-volatile storage device or combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM) , Electrically Erasable Programmable Read-Only Memory), Erasable Programmable Read-Only Memory (EPROM, Erasable Programmable Read-Only Memory), Programmable Read-Only Memory (PROM, Programmable Read-Only Memory), Read-Only Memory (ROM, Read Only Memory), magnetic memory, flash memory, magnetic disk or optical disk.
  • SRAM Static Random-Access Memory
  • EEPROM Electrically Erasable Programmable Read-Only Memory
  • EPROM Erasable Programmable Read-Only Memory
  • PROM Programmable Read-Only Memory
  • Power supply assembly 806 provides power to various components of electronic device 800 .
  • Power supply components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800 .
  • Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD, Liquid Crystal Display) and a touch panel (TP, Touch Panel).
  • the screen may be implemented as a touch screen to receive input signals from a user.
  • the touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe action, but also detect the duration and pressure associated with the touch or swipe action.
  • the multimedia component 808 includes a front-facing camera and/or a rear-facing camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data.
  • Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.
  • Audio component 810 is configured to output and/or input audio signals.
  • the audio component 810 includes a microphone (MIC, Microphone) that is configured to receive external audio signals when the electronic device 800 is in operating modes, such as calling mode, recording mode, and voice recognition mode.
  • the received audio signal may be further stored in memory 804 or transmitted via communication component 816 .
  • audio component 810 also includes a speaker configured to output audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.
  • Sensor assembly 814 includes one or more sensors configured to provide status assessment of various aspects of electronic device 800 .
  • the sensor assembly 814 can detect the open/closed state of the electronic device 800 and the relative positioning of the components, such as the display and the keypad of the electronic device 800, and the sensor assembly 814 can also detect the electronic device 800 or one of the electronic device 800. Changes in the position of components, presence or absence of user contact with the electronic device 800 , orientation or acceleration/deceleration of the electronic device 800 and changes in the temperature of the electronic device 800 .
  • Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact.
  • Sensor assembly 814 may also include a light sensor, such as a Complementary Metal-Oxide-Semiconductor (CMOS) or Charge Coupled Device (CCD) image sensor, configured for use in imaging applications.
  • CMOS Complementary Metal-Oxide-Semiconductor
  • CCD Charge Coupled Device
  • the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices.
  • the electronic device 800 can access a wireless network based on a communication standard, such as Wireless Fidelity (Wi-Fi, Wireless Fidelity), the second generation mobile communication technology (2G, The 2nd Generation,) or the third generation mobile communication technology (3G, The 3nd Generation,) or a combination thereof.
  • the communication component 816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 816 further includes a Near Field Communication (NFC, Near Field Communication) module to facilitate short-range communication.
  • NFC Near Field Communication
  • the NFC module can be based on Radio Frequency Identification (RFID, Radio Frequency Identification) technology, Infrared Data Association (IrDA, Infrared Data Association) technology, Ultra Wide Band (UWB, Ultra Wide Band) technology, Bluetooth (BT, Blue Tooth) technology and other technologies to achieve.
  • RFID Radio Frequency Identification
  • IrDA Infrared Data Association
  • UWB Ultra Wide Band
  • Bluetooth Bluetooth
  • the electronic device 800 may be implemented by one or more of an Application Specific Integrated Circuit (ASIC, Application Specific Integrated Circuit), a Digital Signal Processor (DSP, Digital Signal Processor), a Digital Signal Processing Device (DSPD, Digital Signal Processing Device), Programmable Logic Device (PLD, Programmable Logic Device), Field Programmable Gate Array (FPGA, Field Programmable Gate Array), controller, microcontroller, microprocessor or other electronic component implementation, configured to perform the above method.
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • DSPD Digital Signal Processing Device
  • PLD Programmable Logic Device
  • FPGA Field Programmable Gate Array
  • controller microcontroller, microprocessor or other electronic component implementation, configured to perform the above method.
  • a non-volatile computer-readable storage medium such as a memory 804 comprising computer program instructions executable by the processor 820 of the electronic device 800 to perform the above method is also provided.
  • FIG. 6 shows a block diagram of an electronic device 1900 according to an embodiment of the present disclosure.
  • the electronic device 1900 may be provided as a server.
  • electronic device 1900 includes a processing component 1922, which may include one or more processors, and a memory resource represented by memory 1932 configured to store instructions executable by processing component 1922, such as an application program.
  • An application program stored in memory 1932 may include one or more modules, each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above-described methods.
  • the electronic device 1900 may also include a power supply assembly 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an I/O interface 1958.
  • Electronic device 1900 may operate based on an operating system stored in memory 1932, such as Windows Server TM , Mac OS X TM , Unix TM , LinuxTM, FreeBSD TM, or similar systems.
  • a non-volatile computer-readable storage medium such as memory 1932 comprising computer program instructions executable by processing component 1922 of electronic device 1900 to perform the above-described method.
  • the present disclosure may be a system, method and/or computer program product.
  • the computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the present disclosure.
  • a computer-readable storage medium may be a tangible device that can hold and store instructions for use by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Computer-readable storage media may include: portable computer disks, hard disks, random access memory (RAM, Random Access Memory), read-only memory, erasable programmable read-only memory (EPROM or flash memory), static random access memory, Portable Compact Disc Read-Only Memory (CD-ROM, Compact Disc Read-Only Memory), Digital Versatile Disc (DVD, Digital Video Disc), memory stick, floppy disk, mechanical coding device, such as a punch card on which instructions are stored Or the protruding structure in the groove, and any suitable combination of the above.
  • RAM Random Access Memory
  • EPROM or flash memory erasable programmable read-only memory
  • static random access memory Portable Compact Disc Read-Only Memory
  • CD-ROM Compact Disc Read-Only Memory
  • DVD Digital Versatile Disc
  • memory stick floppy disk
  • mechanical coding device such as a punch card on which instructions are stored Or the protruding structure in the groove, and any suitable combination of the above.
  • Computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, light pulses through fiber optic cables), or through electrical wires transmitted electrical signals.
  • the computer readable program instructions described herein may be downloaded to various computing/processing devices from a computer readable storage medium, or to an external computer or external storage device over a network such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • a network adapter card or network interface in each computing/processing device receives computer-readable program instructions from a network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
  • Computer program instructions for performing operations of embodiments of the present disclosure may be assembly instructions, Industry Standard Architecture (ISA) instructions, machine instructions, machine-related instructions, pseudocode, firmware instructions, state setting data, or in a form of Source or object code written in any combination of programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., as well as conventional procedural programming languages such as C or similar programming languages.
  • the computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement.
  • the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or Wide Area Network (WAN), or may be connected to an external computer (eg, using the Internet service provider to connect via the Internet).
  • LAN Local Area Network
  • WAN Wide Area Network
  • electronic circuits such as programmable logic circuits, field programmable gate arrays, or programmable logic arrays, that can execute computer readable program instructions are personalized by utilizing state information of computer readable program instructions , thereby implementing various aspects of the present disclosure.
  • These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus to produce a machine that causes the instructions when executed by the processor of the computer or other programmable data processing apparatus , resulting in means for implementing the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • These computer readable program instructions can also be stored in a computer readable storage medium, these instructions cause a computer, programmable data processing apparatus and/or other equipment to operate in a specific manner, so that the computer readable medium on which the instructions are stored includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • Computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , thereby causing instructions executing on a computer, other programmable data processing apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams.
  • each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more functions for implementing the specified logical function(s) executable instructions.
  • the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions.
  • the computer program product can be implemented in hardware, software or a combination thereof.
  • the computer program product may be embodied as a computer storage medium, and in other embodiments, the computer program product may be embodied as a software product, such as a software development kit (SDK, Software Development Kit) and the like.
  • SDK software development kit
  • a target image including a target object is acquired; the target image is segmented to obtain the name of at least one pixel in the target image as a segmentation result; according to the segmentation result, at least one of the target objects is determined The name of a target subobject.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Medical Informatics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

本公开实施例涉及一种图像处理方法及装置、电子设备、存储介质和程序产品。所述图像处理方法包括：获取包括目标对象的目标图像；对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果；根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称。

Description

图像处理方法及装置、电子设备、存储介质和程序产品
相关申请的交叉引用
本公开基于申请号为202010674675.X、申请日为2020年07月14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以全文引入的方式引入本公开。
技术领域
本公开涉及图像处理技术领域，尤其涉及一种图像处理方法及装置、电子设备、存储介质和程序产品。
背景技术
心脑血管疾病是当前致死率最高的疾病之一，其中冠心病发病率最高。冠心病是由于动脉粥样硬化导致冠状动脉管腔狭窄，导致心肌供血不足而引起一系列临床病症。
在对冠心病的分析过程中，病变的定位以及医疗报告中的说明分析往往需要依赖于冠脉中心线的命名。
发明内容
本公开实施例提出了一种图像处理方法及装置、电子设备、存储介质和程序产品。
根据本公开实施例的一方面，提供了一种图像处理方法，包括：
获取包括目标对象的目标图像；对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果；根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称。
在一种可能的实现方式中，所述对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果，包括：将所述目标图像输入至神经网络；根据所述神经网络的输出，确定所述目标图像中至少一个像素点的名称，作为所述分割结果；其中，所述神经网络通过包括所述目标对象的训练图像进行训练，所述训练图像中的目标对象通过至少一个目标子对象的名称进行标注。因此，可以通过神经网络实现对目标图像进行分割，从而有效提升分割的稳定性、效率和精度，继而提升图像处理的稳定性、效率和精度。同时由于神经网络可以根据训练图像中包括的目标子对象的名称的实际情况灵活调整结构和实现方式，因此，可以提升关键点检测的灵活性，继而提升图像处理方法实现的灵活性。
在一种可能的实现方式中，所述神经网络通过包括所述目标对象的训练图像进行训练，包括：根据所述训练图像中目标对象的标注，确定所述训练图像中至少部分像素点的标签；通过包括所述至少部分像素点的标签的训练图像，对所述神经网络进行训练。这样，在确定了训练图像中至少部分像素点的标签后，可以根据包含至少部分像素点标签的训练图像对神经网络训练，通过上述过程，可以基于标注的训练图像有效地对神经网络进行训练，得到具有分割命名功能的神经网络，提升图像处理的便捷性。
在一种可能的实现方式中，所述神经网络通过包括所述目标对象的训练图像进行训练，包括：在所述训练图像中的目标对象内，确定距离第一目标像素点最近的目标子对象，其中，所述第一目标像素点为所述训练图像中除所述目标对象以外的至少一个像素点；将确定的所述目标子对象的标注，作为所述第一目标像素点的标签；根据包括所述第一目标像素点的标签和所述目标对象的标注的训练图像，对所述神经网络进行训练。通过上述过程，可以基于包含目标对象标注的训练图像，得到具有密集标签和完整信息的训练图像对神经网络进行训练，从而在不增加标注难度的基础上，提升训练得到的神经网络的精度，继而提升分割结果以及最终得到的目标子对象的名称的准确度。提升图像处理的准确性和便捷性。
在一种可能的实现方式中，所述根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称，包括：根据所述分割结果，确定至少一个第二目标像素点的名称，其中，所述第二目标像素点为所述目标子对象包含的像素点；统计所述目标子对象中每一所述第二目标像素点的名称，得到统计结果，将所述统计结果中数量最多的名称，作为所述目标子对象的名称。通过上述过程，即使目标图像中存在部分像素点的分割结果不准确，也可以基于比例得到较为准确的目标子对象的名称，从而可以提升最终确定的目标子对象的名称的准确性，继而提升图像处理的准确度和精度。
在一种可能的实现方式中，根据所述分割结果，确定至少一个第二目标像素点的名称，包括：将每一所述第二目标像素点对应的分割结果，作为每一所述第二目标像素点的名称；或者，基于每一所述第二目标像素点预设范围内至少一个像素点的分割结果，确定每一所述第二目标像素点的名 称。因此，通过基于第二目标像素点预设范围内至少一个像素点的分割结果，确定第二目标像素点的名称，可以有效地引入这些具有辅助功能像素点的分割结果来确定第二目标像素点的名称，提升确定的第二目标像素点的名称的准确性，继而提升确定的目标子对象名称的准确性，从而提升图像处理准确度。
在一种可能的实现方式中，所述方法还包括：根据至少一个所述目标子对象的名称，对所述目标对象进行处理，得到处理结果。这样，通过根据至少一个目标子对象的名称对目标对象进行处理，得到处理结果，可以有效地根据目标子对象的名称对目标对象进行进一步地优化，即使之前对目标图像进行分割得到的分割结果准确率较低，从而导致确定的名称存在一定的误差，也可以通过处理进行修正，得到更为准确的处理结果，提升图像处理的精度和鲁棒性。
在一种可能的实现方式中，所述根据至少一个所述目标子对象的名称，对所述目标对象进行处理，包括：对所述目标对象中具有相同名称的目标子对象进行提取；和/或，根据相邻的目标子对象的名称，对所述目标对象中至少一个目标子对象的名称进行修正。这样，可以进一步减小由于分割不准确导致的命名结果不准确的情况的发生，提升最终得到的处理结果的准确性，提升图像处理的精度和鲁棒性。
在一种可能的实现方式中，所述目标图像包括：心脏冠脉掩模图像，或者，心脏冠脉掩模图像以及心脏掩模图像；所述目标对象包括冠脉中心线。在目标图像同时包含心脏冠脉掩模图像以及心脏掩模图像的情况下，可以基于掩模图像进行分割，相比直接基于原始的心脏冠脉图像进行分割来说，简化后续分割过程的输入环境，降低了图像处理过程中的噪音；还可以通过心脏掩模图像，保留心脏各个腔室的信息，从而便于同时可以尽可能地保留冠脉的树状结构信息以及供血位置信息，减小了后续分割的难度，提升分割结果以及最终确定的名称的准确程度。
根据本公开实施例的一方面，提供了一种图像处理装置，包括：
目标图像获取模块，配置为获取包括目标对象的目标图像；分割模块，配置为对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果；命名模块，配置为根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称。
根据本公开实施例的一方面，提供了一种电子设备，包括：
处理器；配置为存储处理器可执行指令的存储器；其中，所述处理器被配置为调用所述存储器存储的指令，以执行上述图像处理方法。
根据本公开实施例的一方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述图像处理方法。
根据本公开实施例的一方面，提供了一种计算机程序产品，包括计算机可读代码，在所述计算机可读代码在电子设备中运行的情况下，所述电子设备中的处理器执行上述图像处理方法。
在本公开实施例中，通过获取包括目标对象的目标图像，并对目标图像进行分割得到目标图像中至少一个像素点的名称作为分割结果，从而根据分割结果确定目标对象中至少一个目标子对象的名称。通过上述过程，可以将对目标图像中包含多个目标子对象的目标对象的命名过程，转化为像素级别的分割过程，有效地减小命名过程实现的难度以及提升命名的准确度，从而提升图像处理过程的鲁棒性。
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，而非限制本公开实施例。
根据下面参考附图对示例性实施例的详细说明，本公开实施例的其它特征及方面将变得清楚。
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，这些附图示出了符合本公开实施例的实施例，并与说明书一起用于说明本公开实施例的技术方案。
图1示出根据本公开一实施例的图像处理方法的流程图。
图2示出根据本公开一实施例的对目标对象进行提取的示意图。
图3示出根据本公开一应用示例的示意图。
图4示出根据本公开一实施例的图像处理装置的框图。
图5示出根据本公开实施例的一种电子设备的框图。
图6示出根据本公开实施例的一种电子设备的框图。
实施方式
以下将参考附图详细说明本公开实施例的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不 必按比例绘制附图。
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
另外，为了更好地说明本公开实施例，在下文的实施方式中给出了众多的细节。本领域技术人员应当理解，没有某些细节，本公开实施例同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开实施例的主旨。
图1示出根据本公开一实施例的图像处理方法的流程图，该方法可以应用于图像处理装置，图像处理装置可以为终端设备、服务器或者其他处理设备等。其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字处理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。
在一些可能的实现方式中，该图像处理方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。如图1所示，所述图像处理方法可以包括：
步骤S11，获取包括目标对象的目标图像。
步骤S12，对目标图像进行分割，得到目标图像中至少一个像素点的名称，作为分割结果。
步骤S13，根据分割结果，确定目标对象中至少一个目标子对象的名称。
其中，目标对象可以是任何具有命名需求的对象。在一种可能的实现方式中，目标对象可以是需要进行分区域或分段命名的对象。由于目标对象可能包含有多个需要命名的段或区域，因此，在一种可能的实现方式中，目标对象可以根据多段或多区域的命名需求，被划分为多个目标子对象。
在一种可能的实现方式中，本公开实施例中提出的图像处理方法，可以用于对心脏冠脉中心线的各段进行命名。因此，在一个示例中，目标对象可以是心脏冠脉的中心线，即冠脉中心线，在这种情况下，目标对象可以包含有多个目标子对象，每个目标子对象可以是树状冠脉中心线结构的其中一个分支或是分支中的其中一段，即冠脉中心线段；在一个示例中，目标对象也可以是心脏冠脉中其他具有命名需求的部分，或是心脏冠脉中心线的部分区域等，根据实际情况灵活选择即可。在一种可能的实现方式中，本公开实施例中提出的方法也可以应用于对其他包含多段或多区域的器官或组织进行命名，相应的，在一些可能的实现方式中，目标对象可以是具有多区域或多段命名需求的器官或组织。后续各公开实施例均以目标对象为心脏冠脉中心线为例进行说明，在目标对象为其他形式的情况下，其处理方法可以根据本公开实施例提出的方法进行灵活扩展，不再举例说明。
目标图像中像素点的名称，可以是与该像素点对应的目标子对象的名称。随着目标对象实现方式的不同，目标对象中目标子对象的名称，其实现方式也可以根据实际情况灵活决定。在一种可能的实现方式中，在目标对象为冠脉中心线的情况下，目标子对象的名称可以为冠脉中心线可以被划分为的多个冠脉中心线段的名称，如右冠状动脉(RCA，Right Coronary Artery)、右冠起源后降支(R-PDA，Right-Posterior Descending Artery)、左冠起源后降支(L-PDA，Left-Posterior Descending Artery)、右冠起源后侧支(R-PLB，Right-Posterior Lateral Branch)、左冠起源后侧支(L-PLB，Left-Posterior Lateral Branch)、左主干(LM，Left Main)、回旋支中远段(LCX，Leaky Coaxial Cable)、第一及第二钝缘支(OM1-2，Obtuse Marginal 1 and 2)、左前降支(LAD，Left Anterior Descending)、第一及第二对角支(D1-2，Diagonal 1 and 2)或是其他(Others)等。
目标图像可以是包含有目标对象的任意图像，在一种可能的实现方式中，目标图像可以是包含有全部目标对象的图像，在一种可能的实现方式中，目标图像也可以是包含有部分目标对象的图像，根据实际情况灵活选择即可。从上述公开实施例可以看出，在一种可能的实现方式中，目标图像可以是掩模图像(mask)，至于是何种掩模图像，可以根据实际情况灵活决定。如上述各公开实施例所述，在一种可能的情况下，目标对象可以包括冠脉中心线，相应地，目标图像可以是包含了冠脉中心线的图像或是可以用于提取冠脉中心线的图像。因此，在一种可能的实现方式中，目标图像可以包括心脏冠脉掩模图像。在一个示例中，心脏冠脉掩模图像可以是一个二值图像，比如可以通过0和1两个值区分图中的像素点是属于心脏冠脉还是与心脏冠脉无关的背景。
在一种可能的实现方式中，目标图像可以包括心脏冠脉掩模图像以及心脏掩模图像。心脏冠脉掩模图像的形式可以如上述公开实施例所述，在此不再赘述。心脏掩模图像可以是心脏冠脉所属的心脏所构成的一个掩模图像，通过不同的值划分出心脏的各个部位或腔室。在一个示例中，目标图 像可以是心脏冠脉掩模图像和心脏掩模图像叠加而成的掩模图像，在叠加的过程中，心脏掩模图像中不属于心脏冠脉的像素点可以标记成心脏冠脉掩模图像中的背景值。后续各公开实施例中，均以目标图像为心脏掩模图像和心脏冠脉掩模图像叠加而成的掩模图像为例进行说明，其余可能的实现方式可以参考后续各公开实施例进行扩展，不再一一列举。
在目标图像同时包含心脏冠脉掩模图像以及心脏掩模图像的情况下，可以基于掩模图像进行分割，相比直接基于原始的心脏冠脉图像进行分割来说，简化后续分割过程的输入环境，降低了图像处理过程中的噪音；还可以通过心脏掩模图像，保留心脏各个腔室的信息，从而便于同时可以尽可能地保留冠脉的树状结构信息以及供血位置信息，减小了后续分割的难度，提升分割结果以及最终确定的名称的准确程度。
在得到了包括目标对象的目标图像后，可以通过步骤S12，对包括目标对象的目标图像进行分割，来得到目标图像中至少一个像素点的名称，作为分割结果。在一种可能的实现方式中，分割可以是对目标图像中的每个像素点进行分类，从而确定目标图像中每个像素点的类别；在一种可能的实现方式中，分割也可以是对目标图像中的每个前景像素点(比如心脏冠脉掩模图像中属于心脏冠脉的像素点)进行分类，从而确定目标图像中每个前景像素点的类别等。在本公开实施例中，分割的实现方式可以根据实际情况灵活决定，可以参见下述各公开实施例，在此先不做展开。
由于分割可以对目标图像中的像素点进行分类，因此，还可以根据目标图像中每个像素点的分类结果，确定目标图像中目标对象的各个目标子对象的名称。在一种可能的实现方式中，可以根据分割结果确定目标对象中每个目标子对象的名称；在一种可能的实现方式中，也可以根据需求，仅确定目标对象中部分目标子对象的名称，被确定的目标子对象的数量以及选择哪些目标子对象进行名称确定均可以根据实际情况灵活选择，在本公开实施例中不做限制。具体如何根据这些像素点的分类结果来确定目标子对象的名称，其实现方式可以根据实际情况灵活选择，详见后续各公开实施例，在此先不做展开。
在本公开实施例中，通过获取包括目标对象的目标图像，并对目标图像进行分割得到分割结果，从而根据分割结果确定目标对象中至少一个目标子对象的名称。通过上述过程，可以将对目标图像中包含多个目标子对象的目标对象的命名过程，转化为掩模分割过程，有效地减小命名过程实现的难度以及提升命名的准确度，从而提升图像处理过程的鲁棒性。
在一种可能的实现方式中，由于目标图像可以包括掩模图像，可以使得分割基于掩模图像进行，从而简化分割的输入环境，降低图像处理过程的噪音。
如上述各公开实施例所述，目标图像的实现形式可以根据实际情况灵活决定。因此获取包括目标对象的目标图像的方式也可以根据实际情况灵活决定。在一种可能的实现方式中，获取目标图像的方式可以为，对心脏冠脉原始图像进行血管分割，得到心脏冠脉掩模图像，然后对心脏冠脉掩模图像进行中心线提取，得到冠脉中心线图像作为目标对象，然后将冠脉中心线与心脏冠脉掩模图像进行叠加，从而得到包含目标对象的目标图像。在一种可能的实现方式中，获取目标图像的方式也可以为：对心脏冠脉原始图像进行血管分割得到心脏冠脉掩模图像，然后对心脏冠脉掩模图像进行中心线提取，得到冠脉中心线图像作为目标对象，同时对心脏原始图像进行血管分割，得到心脏掩模图像，然后将冠脉中心线、心脏冠脉掩模图像和心脏掩模图像进行叠加，从而得到包含目标对象的目标图像。步骤S11其余的可能实现方式可以根据目标图像的实际情况灵活选择，不再一一举例说明。
在得到了包括目标对象的目标图像后，可以通过步骤S12对目标图像进行分割，得到分割结果。在一种可能的实现方式中，可以根据目标图像中像素点与目标对象的相对位置，确定目标图像中至少部分像素点最接近或属于的目标子对象的名称，作为分割结果。在一种可能的实现方式中，也可以通过神经网络来得到目标图像的分割结果，在这种情况下，步骤S12可以包括：
将目标图像输入至神经网络；根据神经网络的输出，确定目标图像中至少一个像素点的名称，作为分割结果；其中，神经网络通过包括目标对象的训练图像进行训练，训练图像中的目标对象通过至少一个目标子对象的名称进行标注。
其中，神经网络输出的像素点的名称，可以是目标图像中的每个像素点的名称，也可以是目标图像中的部分像素点的名称，根据实际情况灵活确定即可。在一种可能的实现方式中，神经网络输出的可以是目标图像中每个前景像素点的名称，将每个前景像素点的名称作为分割结果。
训练图像的实现形式可以参考上述目标图像的实现形式，在此不再赘述。对训练图像的目标对象进行标注的方式在本公开实施例中不做限制，可以根据神经网络实现的功能灵活决定。在一种可能的实现方式中，也可以对训练图像中目标对象中的每个像素点分别进行标注，或是对训练图像中 的每个像素点分别进行标注，在一种可能的实现方式中，可以分别对训练图像中的多个目标子对象进行标注。在一个示例中，可以将目标对象划分为多个目标子对象，然后标注其中部分目标子对象或每个目标子对象的名称，从而可以对目标子对象覆盖的像素点统一进行标注。选择哪种标注方式，可以根据实际情况灵活选择，在本公开实施例中不做限制。
神经网络的实现形式以及训练方式在本公开实施例中均不作限制，可以根据实际情况灵活选择，神经网络的初始模型可以根据实际情况任意选择，在一个示例中，可以将残差虚拟网络(Res-Vnet)作为神经网络的实现形式。训练采取的损失函数也可以根据实际情况灵活选择，在一个示例中，可以将骰子损失(dice loss)作为损失函数，对神经网络进行训练。
通过上述公开实施例可以看出，在一种可能的实现方式中，可以通过神经网络对目标图像进行分割，由于对目标图像进行分割的神经网络可以通过包括目标对象的训练图像进行训练，且训练图像中的目标对象通过至少一个目标子对象的名称进行标注。因此，基于上述训练图像得到的神经网络，可以对目标图像中属于目标对象的像素点进行分割，从而得到每个被分割的像素点的分割结果，且分割结果可以是目标图像包含的像素点对应的目标子对象的名称，可能的名称种类可以参考上述各公开实施例，在此不再赘述。
通过以目标子对象的名称进行标注的训练图像，对神经网络训练，从而利用训练好的神经网络实现对目标图像的分割，得到分割结果。通过上述过程，可以有效地利用神经网络对目标图像实现像素级的分割，较为便捷地得到分割结果，从而有效地降低了图像处理的难度，提升了图像处理的实用性和泛化能力。
神经网络通过包括目标对象的训练图像进行训练的方式，可以根据实际情况灵活决定，在一种可能的实现方式中，神经网络通过包括目标对象的训练图像进行训练，可以包括：根据训练图像中目标对象的标注，确定训练图像中至少部分像素点的标签；通过包括至少部分像素点的标签的训练图像，对神经网络进行训练。
如上述各公开实施例所述，训练图像中目标对象的标注形式可以根据实际情况灵活选择，相应地，随着标注的形式不同，步骤S21中确定至少部分像素点标签的方式也可以随之发生变化。在一种可能的实现方式中，可以将训练图像中目标对象的标注，直接作为目标图像中对应像素点的标签。如上所述，在一个示例中，可以通过将目标对象划分为多段目标子对象，对每段目标子对象进行命名的方式实现标注，因此被命名的目标子对象覆盖的像素点的标注可以作为这些像素点的标签。
在确定了训练图像中至少部分像素点的标签后，可以根据包含至少部分像素点标签的训练图像对神经网络训练，通过上述过程，可以基于标注的训练图像有效地对神经网络进行训练，得到具有分割命名功能的神经网络，提升图像处理的便捷性。
如上述各公开实施例所述，在一种可能的实现方式中，目标对象可以是冠脉中心线，而对于心脏冠脉掩模图像来说，冠脉中心线为心脏冠脉掩模的一部分，心脏冠脉掩模图像中还存在一部分像素点，其不属于目标对象即冠脉中心线，但是属于冠脉，因此这些像素点对应的名称，可以有助于确定冠脉中心线的名称。如上述各公开实施例所述，在对训练图像中的目标对象进行标注的过程中，可以是对训练图像中的每个像素点进行标注，也可以是对训练图像的目标对象中的每个像素点进行标注。因此，这些有助于确定冠脉中心线名称的像素点，可能不包含标注。在一种可能的实现方式中，为了提高训练得到的神经网络的精度，可以考虑根据标注确定这些像素点的标签。
因此，在一种可能的实现方式中，神经网络通过包括目标对象的训练图像进行训练，可以包括：在训练图像中的目标对象内，确定距离第一目标像素点最近的目标子对象，其中，第一目标像素点为训练图像中除目标对象以外的至少一个像素点；将确定的目标子对象的标注，作为第一目标像素点的标签；根据包括第一目标像素点的标签和目标对象的标注的训练图像，对神经网络进行训练。
其中，第一目标像素点可以是训练图像中不属于目标对象的至少一个像素点，至于选择哪些像素点作为第一目标像素点，可以根据实际情况灵活决定。在一种可能的实现方式中，可以将训练图像中位于目标对象以外，但是可以辅助对目标对象进行命名的像素点作为第一目标像素点，以目标对象为冠脉中心线为例，在一个示例中，可以将训练图像中位于冠脉中心线以外且属于心脏冠脉的像素点，作为第一目标像素点；由于训练图像的实现形式可以与目标图像的实现形式相同，而上述各公开实施例已经提到，目标图像可以是掩模图像，因此，在一个示例中，可以将训练图像中不属于背景且不属于目标对象的像素点，均作为或是部分作为第一目标像素点。
在一种可能的实现方式中，可以根据训练图像中至少部分像素点的标签，确定第一目标像素点的标签，第一目标像素点的确定方式可以根据实际情况灵活选择，不局限于本公开各实施例。通过上述内容可以看出，在一种可能的实现方式中，可以将训练图像的目标对象中，距离第一目标像素 点最近的目标子对象的标注，作为第一目标像素点的标签。在一个示例中，距离第一目标像素点最近的目标子对象，可以是目标对象中距离第一目标像素点最近的像素点所属的目标子对象。
在得到了第一目标像素点的标签后，还可以通过如上述公开实施例所述的方式，将训练图像中目标对象的标注，直接作为目标图像中对应像素点的标签，从而得到具有密集标签的训练图像，对神经网络进行训练。
以目标对象为冠脉中心线为例对上述过程进行说明，在一个示例中，可以在训练图像中，首先将冠脉中心线上的像素点的标注作为这些像素点的标签，然后将既不属于背景也不属于冠脉中心线的像素点，作为第一目标像素点，对每个第一目标像素点，将冠脉中心线上距离该第一目标像素点最近的像素点的标注，作为该第一目标像素点的标签，继而得到一个包含密集标签的训练图像，通过此包含密集标签的训练图像对神经网络进行训练，可以得到较好的训练结果。
通过上述过程，可以基于包含目标对象标注的训练图像，得到具有密集标签和完整信息的训练图像对神经网络进行训练，从而在不增加标注难度的基础上，提升训练得到的神经网络的精度，继而提升分割结果以及最终得到的目标子对象的名称的准确度。提升图像处理的准确性和便捷性。
通过上述任意公开实施例得到目标图像的分割结果后，可以通过步骤S13，基于分割结果来确定目标对象中至少一个目标子对象的名称。步骤S13的实现方式不受限定，在一种可能的实现方式中，可以将目标子对象包含的任意像素点的分割结果，作为目标子对象的名称。在一种可能的实现方式中，步骤S13可以包括：
步骤S131，根据分割结果，确定至少一个第二目标像素点的名称，其中，第二目标像素点为目标子对象包含的像素点；
步骤S132，统计目标子对象中第二目标像素点的名称，得到统计结果，将统计结果中数量最多的名称，作为目标子对象的名称。
其中，第二目标像素点可以是目标子对象包含的像素点，即目标子对象所覆盖的像素点。在一种可能的实现方式中，由于得到的分割结果可能不是完全准确，因此，对于同一个目标子对象来说，其包含的第二目标像素点的分割结果可能准确，即与对应的目标子对象的名称相同，也可能不准确，即与对应的目标子对象的名称不同；在这种情况下，如果将目标子对象包含的任意一个第二目标像素点的分割结果作为目标子对象的名称，很有可能得到不准确的命名结果。
基于上述原因，在一种可能的实现方式中，对于任意需要确定名称的目标子对象，可以通过步骤S131，来根据分割结果，获得该目标子对象中至少一个第二目标像素点的名称。
步骤S131中获取目标子对象中第二目标像素点的数量，可以根据实际情况灵活决定。在一种可能的实现方式中，可以获取目标子对象包含的每个第二目标像素点，在一种可能的实现方式中，可以对目标子对象包含的第二目标像素点进行随机采样，来获得部分第二目标像素点，如何采样以及采样的数量可以根据实际情况灵活决定，在本公开实施例中不做限制。
无论获得的是部分第二目标像素点还是全部第二目标像素点，均需要根据分割结果来确定获得的第二目标像素点的名称。名称确定的方式可以根据实际情况灵活决定，在一种可能的实现方式中，步骤S131中根据分割结果，确定至少一个第二目标像素点的名称，可以包括：
将每一第二目标像素点对应的分割结果，作为每一第二目标像素点的名称；或者，基于每一第二目标像素点预设范围内至少一个像素点的分割结果，确定每一第二目标像素点的名称。
通过上述公开实施例可以看出，在一种可能的实现方式中，可以直接将第二目标像素点的分割结果作为第二目标像素点的名称。
上述公开实施例已经提出，目标图像中像素点的分割结果可能不是完全准确的，在这种情况下，如果将第二目标像素点的分割结果作为第二目标像素点的名称，也可能会出现差错，因此，在一种可能的实现方式中，可以根据第二目标像素点预设范围内至少一个像素点的分割结果，来确定第二目标像素点的名称。
其中，第二目标像素点的预设范围，可以根据实际情况灵活设定，如第二目标像素点周围的9邻域、16邻域等等，在本公开实施例中不做限制。一般来说，第二目标像素点附近的像素点，与第二目标像素点属于同一目标子对象的概率较大，因此通过这些像素点，可以确定第二目标像素点较为准确的名称。在一个示例中，可以获取第二目标像素点预设范围内包含的每个像素点的分割结果，然后从这些获取到的分割结果中，选择比例最高或是数量最多的分割结果，或是选择比例或是数量超过设定阈值(如比例超过50％等)的分割结果，作为第二目标像素点的名称；在一个示例中，也可以对第二目标像素点预设范围内包含的像素点进行采样，然后从这些采样得到的像素点对应的分割结果中，选择比例最高或是数量最多的分割结果，或是选择比例或是数量超过设定阈值(如比例 超过50％等)的分割结果，作为第二目标像素点的名称。
如上述各公开实施例所述，目标图像中可能存在一些像素点，虽然不属于目标对象，但是可以辅助确定目标子对象的名称，因此，通过基于第二目标像素点预设范围内至少一个像素点的分割结果，确定第二目标像素点的名称，可以有效地引入这些具有辅助功能像素点的分割结果来确定第二目标像素点的名称，提升确定的第二目标像素点的名称的准确性，继而提升确定的目标子对象名称的准确性，从而提升图像处理准确度。
在通过上述任意方式确定至少一个第二目标像素点的名称后，可以通过步骤S132，基于第二目标像素点名称的统计结果，来得到目标子对象的名称。其中，步骤S132的实现方式可以根据实际情况灵活决定，比如在统计目标子对象中的每个第二目标像素点或部分第二目标像素点的名称后，可以将这些统计的名称中数量最多的名称作为目标子对象的名称，或是将这些统计的名称中出现比例最高的名称作为目标子对象的名称，或是将这些统计的名称中数量超过预设阈值或是比例超过预设阈值的名称，作为目标子对象的名称，预设阈值的数值在本公开实施例中不做限定，根据实际情况灵活设定即可。
以目标对象为冠脉中心线为例对上述过程进行说明，在一个示例中，可以依次确定冠脉中心线每个冠脉中心线段的名称，对于每个冠脉中心线段来说，可以先遍历其包含的每个像素点即第二目标像素点，确定这些第二目标像素点的名称，确定的方式可以是直接根据其分割结果进行确定，也可以是根据其周围的像素点的分割结果共同确定，参考上述各公开实施例即可，在此不再赘述。在确定了这些第二目标像素点的名称后，可以统计该冠脉中心线段中数量最多或是比例最高的名称，作为该冠脉中心线段的名称。
通过根据分割结果确定至少一个第二目标像素点的名称，并统计确定的第二目标像素点的名称，从而根据统计结果得到目标子对象的名称，通过上述过程，即使目标图像中存在部分像素点的分割结果不准确，也可以基于比例得到较为准确的目标子对象的名称，从而可以提升最终确定的目标子对象的名称的准确性，继而提升图像处理的准确度和精度。
在一种可能的实现方式中，本公开实施例提出的图像处理方法还可以包括：步骤S14，根据至少一个目标子对象的名称，对目标对象进行处理，得到处理结果。
通过上述公开实施例可以看出，在得到了目标图像中至少一个目标子对象的名称后，还可以基于确定的名称，根据实际需求对目标图像中的目标对象进行进一步地处理和优化，得到最终的图像处理结果。如何根据目标子对象的名称进行处理，得到何种处理结果，可以根据实际的图像处理需求灵活决定，详见下述各公开实施例，在此先不做展开。
通过根据至少一个目标子对象的名称对目标对象进行处理，得到处理结果，可以有效地根据目标子对象的名称对目标对象进行进一步地优化，即使之前对目标图像进行分割得到的分割结果准确率较低，从而导致确定的名称存在一定的误差，也可以通过处理进行修正，得到更为准确的处理结果，提升图像处理的精度和鲁棒性。
在一种可能的实现方式中，步骤S14可以包括：对目标对象中具有相同名称的目标子对象进行提取；和/或，根据相邻的目标子对象的名称，对目标对象中至少一个目标子对象的名称进行修正。
其中，对目标对象中具有相同名称的目标子对象进行提取的方式不受限定。在一种可能的实现方式中，在目标对象为冠脉中心线的情况下，由于冠脉中心线可以为树状结构，对于属于同一名称的冠脉中心线段，由于分割的不准确可能导致该冠脉中心线段包含有多个分支。为了得到较为准确的冠脉中心线段，在一个示例中，可以对同一名称下包含多个分支的冠脉中心线段进行提取，选定其中最完整或是最长的一支作为提取结果。图2示出根据本公开一实施例的对目标对象进行提取的示意图，从图2中可以看出，提取前的目标图像21的冠脉中心线中每个冠脉中心线段的名称已确定，且同一名称下包含多个分支的冠脉中心线段；提取后的目标图像22的冠脉中心线中同一名称下仅包含一个分支的冠脉中心线段，例如名称为R-PLB的冠脉中心线段在提取前存在两个分支，经过提取，较短的分支被删除，而仅保留了较长的分支作为提取结果。
根据目标子对象的名称对目标对象进行修正的方式在本公开实施例中也不做限定。在一种可能的实现方式中，在目标对象为冠脉中心线的情况下，冠脉中心线中的部分冠脉中心线段可能存在如下情况：即该冠脉中心线段可能位于两个具有同样名称的其他冠脉中心线段之间，且该冠脉中心线段的名称与这两个其他冠脉中心线段的名称不同。由于冠脉中心线的连续性，该冠脉中心线段的名称很有可能是不准确的，因此，可以基于相邻的其他冠脉中心线段的名称，修改当前冠脉中心线段的名称，使冠脉中心线保持连续性。
步骤S14在实现的过程中，可以同时包含有上述的提取过程和修正过程，也可以根据实际情况 仅包含其中的某个过程，或者也可以包含有其他的修正过程，这些过程在实现过程中的执行顺序也可以根据实际情况灵活选择，如何实现可以根据实际情况灵活决定，不局限于上述各公开实施例。
通过对目标对象中具有相同名称的目标子对象进行提取，和/或，根据相邻的目标子对象的名称对目标对象中至少一个目标子对象的名称进行修正，可以进一步减小由于分割不准确导致的命名结果不准确的情况的发生，提升最终得到的处理结果的准确性，提升图像处理的精度和鲁棒性。
心脑血管疾病是当前致死率最高的疾病之一，其中冠心病发病率最高。冠心病是由于动脉粥样硬化导致冠状动脉管腔狭窄，导致心肌供血不足而引起一系列临床病症，包括心绞痛、心肌梗死、心肌衰竭、心律失常和猝死。因此冠脉狭窄，斑块检测等结果对诊断以及后续治疗有着指导性意义，而基于心脏冠状动脉造影(CTA，Computed Tomography Angiography)图像的冠脉中心线提取是前提。在对冠心病的分析过程中，病变的定位以及医疗报告中的说明分析往往需要依赖于冠脉中心线的命名。
相关技术中的命名方法主要是基于分为两类：第一类基于知识以及建模，该类方法主要基于将目标血管匹配基于统计的通用模型给出命名；第二类基于学习的分类算法，该类方法根据提取人为设计的特征进行中心线分类命名。还有一种方案使用树状双向长短期记忆网络(LSTM，Long Short Term)去学习树状结构信息进行分类命名。无论是基于建模的批评的中心线命名，还是基于规则的树状线命名，这两种相关方案都至少存在设计复杂、泛化能力不足的问题。
血管命名任务主要挑战在于个体之间的巨大差异以及巨大变异度，然而血管命名的临床依据是血管的供血部位，所以核心学习信息应该在于供血部位也就是相对于心脏的血管位置，任何剥离或者损失了心脏相对位置的方法都有一定程度的信息损失从而导致准确率不高、泛化能力弱。
本公开应用示例提出一种基于分割的端到端的命名处理方法。相关技术中大多基于先进行血管分割，再提取中心线，最后进行中心线命名(通过对中心线分类实现)的工作流程。本公开应用示例重新构建问题：将中心线命名的问题重构成血管掩膜图像的实例分割问题，即提取中心线后不直接对中心线进行命名，而是对血管掩膜图像(每一个像素点有确定的标签命名)进行分割，对于分割结果进行投票确定中心线的名称。基于像素点的分割模型训练会提供更为密集的标注及信息，而且为了保存血管的供血位置信息，将心脏各个腔室的掩膜图像叠加在冠脉血管的掩膜图像上面一起作为输入。这样的流程完全保留了所有信息又相比较原图减少了噪音，不需要手动提取特征，不需要基于规则的设计且易于嵌入现有工作流当中。
本公开应用示例包括训练过程和预测过程，其中训练过程包括以下步骤：第一步准备输入：通过分割模型获取心脏冠脉掩膜图像以及全心脏掩膜图像，叠加两个掩膜图像并处理作为最终输入；第二步准备监督数据：首先获取已经正确命名的冠脉中心线，然后将冠脉掩膜图像的每一个像素点标签与距离该冠脉中心线最近的每一段中心线关联；第三步使用Res-Vnet型以及骰子损失进行多标签分割训练。预测过程包括以下步骤：第一步同训练过程的准备输入步骤，即获取作为最终输入的掩膜图像以及提取的冠脉中心线；第二步通过训练好的神经网络进行分割；第三步基于分割对已经提取的中心线段进行投票命名；第四步通过后处理提取出医生关注的目标血管并修正可能的错误。
本公开应用示例将血管中心线命名任务构建成血管掩膜图像分类任务，从而降低了学习难度，提升了中心线命名的鲁棒性。本公开应用示例将血管掩膜图像以及心脏掩膜图像一起作为输入，降低了原图直接学习的难度，并且保留了全部命名所需的全部必要信息，使得基于学习的方法相比较于传统的基于规则的树状线命名算法具有更优的泛化能力。
图3示出根据本公开一应用示例的示意图，如图3所示，本公开实施例提出了一种图像处理方法，这一处理方法可以通过图像处理实现对冠脉中心线的分段命名，该图像处理的过程可以大致分为四个步骤。
第一步，分别获取心脏冠脉掩模图31和心脏掩模图32，并将两个掩模图进行合并，作为合并掩模图。同时并对心脏冠脉掩模图进行中心线提取，得到由多个冠脉中心线段(目标子对象)构成的冠脉中心线作为目标对象。然后将合并掩模图与目标对象进行叠加，得到包括目标对象(冠脉中心线)的目标图像33。
第二步，通过训练好的神经网络34，对目标图像进行分割，得到目标图像中每个前景像素点(即属于心脏冠脉的像素点)的分割结果35。
第三步，根据第二步得到的分割结果，通过投票命名的方式，确定目标图像中冠脉中心线每个冠脉中心线段的名称36，投票命名过程可以为：
对每个冠脉中心线段，确定其包含的像素点作为第二目标像素点，然后统计这些第二目标像素点的分割结果所对应的名称，选择其中数量最多的名称，作为该冠脉中心线段的名称。
第四步，在对每个冠脉中心线段完成命名后，可以通过后处理，对其中包含多个具有同名分支的冠脉中心线段，保留其最完整或长度最长的一支作为最终结果，并对不连续的冠脉中心线段进行修正，比如某中心线段的父节点(父线段)和子节点(子线段)具有相同的名称，但是该冠脉中心线段与他们的名称不一致，则可以将该冠脉中心线段修改为与其父节点和子节点一致的名称。通过上述后处理，得到最终的处理结果37。
在一些可能的实施例中，第二步中用于对目标图像分割的神经网络，其训练过程可以为：首先准备训练图像，训练图像可以与目标图像的形式相同，即将心脏冠脉掩模图像、心脏掩模图像与冠脉中心线进行叠加得到的图像，作为训练图像。
由于训练图像用于对神经网络进行训练，其还需要有监督数据(ground truth)，即训练图像中的像素点还需要准备标签。在本公开应用示例中，监督数据的生成方式可以为：首先可以对冠脉中心线通过分段命名进行标注，这样，每个冠脉中心线段中像素点的标签即为对应的冠脉中心线段的标注名称。
然后将训练图像中，属于前景且不属于冠脉中心线段的像素点分别作为第一目标像素点，则可以将冠脉中心线上距离第一目标像素点最近像素点的标签(也是标注)，作为该第一目标像素点的标签。
通过上述过程，可以得到包含有标签的训练图像，将这一包含有标签的训练图像输入到神经网络中进行训练，则可以得到训练好的神经网络。在一个示例中，神经网络可以将Res-Vnet神经网络作为基础模型，通过骰子损失作为损失函数进行多标签分割训练，从而得到最终的训练结果。
通过上述过程，可以将冠脉中心线分段命名任务构建成为血管掩膜图像的实例分割任务；同时本公开应用示例中提出的图像处理方法可以将心脏冠脉掩模图像和心脏掩模图像作为输入，比起将心脏冠脉的原图作为输入，简化了输入环境，降低了噪音并保留了所有树状结构信息以及供血位置信息；另外，在命名过程中，对分割结果采用投票机制，使得冠脉中心线段的命名依赖于该段像素点中出现最多的分割标签，即使部分像素点的分割结果不准确，只要正确的分割结果占多数比例，即可得到正确的冠脉中心线命名结果；最后，在基于分割结果进行命名后还进一步对命名结果进行后处理，由于分割结果依赖于训练数据，通过后处理可以在训练数据较少的时候也能保持最终处理结果的鲁棒性。
本公开应用示例的图像处理方法提出新的问题架构，即中心线分段命名任务构建成为血管掩膜图像的实例分割任务；同时提供新的输入思路，比起原图作为输入，该技术将两种掩膜图像作为输入，简化了输入环境，降低了噪音并保留了所有树状结构信息以及供血位置信息；从分割后的结果回到中心线命名，采用投票机制，血管段命名依赖于该段点所述分割标签最多的一类，这样即使分割结果不好只要大体是正确的，中心线命名就会正确；基于先验知识的后处理，学习算法依赖于训练数据，后处理在训练数据较少的时候也能保持最终结果的鲁棒性。
因此，本公开应用示例的创新点至少包括：基于冠脉多标签分割的中心线命名策略；从掩膜图像分割回归到中心线命名的投票策略；以及基于分割结果以及先验知识的后处理方法。本公开应用示例提供的方案为基于端到端的设计，具有流程清晰、准确度更高、鲁棒性强等优点。同时本公开应用示例提供的方案保持了较高的速度，每个患者需要时间约3到5秒。
本公开应用示例至少可以应用于心脑血管疾病辅助诊断、远程医疗诊断、云平台辅助智能诊断、医疗标注平台等方面对心脏冠脉图像的处理。心内科医生在得到患者的心脏冠状动脉造影数据提取中心线后，进行中心线分段命名以用于血管斑块、狭窄检测定位，从而生成结构化报告。
需要说明的是，本公开实施例的图像处理方法不限于应用在上述心脏冠脉图像的处理中，可以应用于任意的图像处理，本公开实施例对此不作限定。例如使用于定义中心线命名为冠脉掩膜图像分割任务的产品，或者包含全心脏掩膜图像以及心脏冠脉掩膜图像作为产品输入等情况属于本公开实施例所保护的范围。
可以理解，本公开实施例提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开实施例不再赘述。本领域技术人员可以理解，在实施方式的上述方法中，各步骤的执行顺序应当以其功能和可能的内在逻辑确定。
此外，本公开实施例还提供了图像处理装置、电子设备、计算机可读存储介质、程序产品，上述均可用来实现本公开实施例提供的任一种图像处理方法，相应技术方案和描述和参见方法部分的相应记载，不再赘述。
图4示出根据本公开实施例的图像处理装置的框图。该图像处理装置可以为终端设备、服务器或者其他处理设备等。其中，终端设备可以为UE、移动设备、用户终端、终端、蜂窝电话、无绳电 话、PDA、手持设备、计算设备、车载设备、可穿戴设备等。在一些可能的实现方式中，该图像处理装置可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。如图4所示，所述图像处理装置40可以包括：
目标图像获取模块41，配置为获取包括目标对象的目标图像。
分割模块42，配置为对目标图像进行分割，得到目标图像中至少一个像素点的名称，作为分割结果。
命名模块43，配置为根据分割结果，确定目标对象中至少一个目标子对象的名称。
在一种可能的实现方式中，分割模块配置为：将目标图像输入至神经网络；根据神经网络的输出，确定目标图像中至少一个像素点的名称，作为分割结果；其中，神经网络通过包括目标对象的训练图像进行训练，训练图像中的目标对象通过至少一个目标子对象的名称进行标注。
在一种可能的实现方式中，神经网络通过包括目标对象的训练图像进行训练，包括：根据训练图像中目标对象的标注，确定训练图像中至少部分像素点的标签；通过包括至少部分像素点的标签的训练图像，对神经网络进行训练。
在一种可能的实现方式中，神经网络通过包括目标对象的训练图像进行训练，包括：在训练图像中的目标对象内，确定距离第一目标像素点最近的目标子对象，其中，第一目标像素点为训练图像中除目标对象以外的至少一个像素点；将确定的目标子对象的标注，作为第一目标像素点的标签；根据包括第一目标像素点的标签和目标对象的标注的训练图像，对神经网络进行训练。
在一种可能的实现方式中，命名模块配置为：根据分割结果，确定至少一个第二目标像素点的名称，其中，第二目标像素点为目标子对象包含的像素点；统计目标子对象中第二目标像素点的名称，得到统计结果，将统计结果中数量最多的名称，作为目标子对象的名称。
在一种可能的实现方式中，命名模块还配置为：将第二目标像素点对应的分割结果，作为第二目标像素点的名称；或者，基于第二目标像素点预设范围内至少一个像素点的分割结果，确定第二目标像素点的名称。
在一种可能的实现方式中，图像处理装置40还包括处理模块，处理模块配置为：根据至少一个目标子对象的名称，对目标对象进行处理，得到处理结果。
在一种可能的实现方式中，处理模块还配置为：对目标对象中具有相同名称的目标子对象进行提取；和/或，根据相邻的目标子对象的名称，对目标对象中至少一个目标子对象的名称进行修正。
在一种可能的实现方式中，目标图像包括：心脏冠脉掩模图像，或者，心脏冠脉掩模图像以及心脏掩模图像；目标对象包括冠脉中心线。
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是非易失性计算机可读存储介质。
本公开实施例还提出一种电子设备，包括：处理器；配置为存储处理器可执行指令的存储器；其中，所述处理器被配置为调用所述存储器存储的指令，以执行上述方法。电子设备可以被提供为终端、服务器或其它形态的设备。
本公开实施例还提供了一种计算机程序产品，包括计算机可读代码，当计算机可读代码在设备上运行时，设备中的处理器执行配置为实现如上任一实施例提供的图像处理方法的指令。
本公开实施例还提供了另一种计算机程序产品，用于存储计算机可读指令，指令被执行时使得计算机执行上述任一实施例提供的图像处理方法的操作。
图5示出根据本公开实施例的一种电子设备800的框图。例如，电子设备800可以是移动电话、计算机、数字广播终端、消息收发设备、游戏控制台、平板设备、医疗设备、健身设备和个人数字助理等终端。
参照图5，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(Input/Output，I/O)的接口812，传感器组件814，以及通信组件816。
处理组件802通常控制电子设备800的整体操作，诸如与显示、电话呼叫、数据通信、相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括配置为在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图 片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM，Static Random-Access Memory)，电可擦除可编程只读存储器(EEPROM，Electrically Erasable Programmable Read-Only Memory)，可擦除可编程只读存储器(EPROM，Erasable Programmable Read-Only Memory)，可编程只读存储器(PROM，Programmable Read-Only Memory)，只读存储器(ROM，Read Only Memory)，磁存储器，快闪存储器，磁盘或光盘。
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD，Liquid Crystal Display)和触摸面板(TP，Touch Panel)。在屏幕包括触摸面板的情况下，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。在电子设备800处于操作模式，如拍摄模式或视频模式的情况下，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC，Microphone)，在电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式的情况下，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，配置为输出音频信号。
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘、点击轮、按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件814包括一个或多个传感器，配置为为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态和组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如互补金属氧化物半导体(CMOS，Complementary Metal-Oxide-Semiconductor)或电荷耦合器件(CCD，Charge Coupled Device，)图像传感器，配置为在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器、陀螺仪传感器、磁传感器、压力传感器或温度传感器。
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如无线保真(Wi-Fi，Wireless Fidelity)、第二代移动通信技术(2G，The 2nd Generation，)或第三代移动通信技术(3G，The 3nd Generation，)或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC，Near Field Communication)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID，Radio Frequency Identification)技术，红外数据协会(IrDA，Infrared Data Association)技术，超宽带(UWB，Ultra Wide Band)技术，蓝牙(BT，Blue Tooth)技术和其他技术来实现。
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC，Application Specific Integrated Circuit)、数字信号处理器(DSP，Digital Signal Processor)、数字信号处理设备(DSPD，Digital Signal Processing Device)、可编程逻辑器件(PLD，Programmable Logic Device)、现场可编程门阵列(FPGA，Field Programmable Gate Array)、控制器、微控制器、微处理器或其他电子元件实现，配置为执行上述方法。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。
图6示出根据本公开实施例的一种电子设备1900的框图。例如，电子设备1900可以被提供为一服务器。参照图6，电子设备1900包括处理组件1922，可以包括一个或多个处理器，以及由存储器1932所代表的存储器资源，配置为存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有 线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个I/O接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows Server TM、Mac OS X TM、UnixTM、Linux TM、FreeBSD TM或类似系统。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是但不限于电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质可以包括：便携式计算机盘、硬盘、随机存取存储器(RAM，Random Access Memory)、只读存储器、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器、便携式压缩盘只读存储器(CD-ROM，Compact Disc Read-Only Memory)、数字多功能盘(DVD，Digital Video Disc)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本公开实施例操作的计算机程序指令可以是汇编指令、指令集架构(ISA，Industry Standard Architecture)指令、机器指令、机器相关指令、伪代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言诸如Smalltalk、C++等，以及常规的过程式编程语言例如C语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络包括局域网(LAN，Local Area Network)或广域网(WAN，Wide Area Network)连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列或可编程逻辑阵列，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实施例的系统、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、程序段或指令的一部分，所述模块、程序段或指令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。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，两个连续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，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框 的组合，可以用执行规定的功能或动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。
该计算机程序产品可以通过硬件、软件或其结合的方式实现。在一些实施例中，所述计算机程序产品可以体现为计算机存储介质，在另一些实施例中，计算机程序产品体现为软件产品，例如软件开发包(SDK，Software Development Kit)等等。
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。
工业实用性
本公开实施例获取包括目标对象的目标图像；对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果；根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称。通过上述过程，可以有效地减小命名过程实现的难度以及提升命名的准确度，从而提升图像处理过程的鲁棒性。

Claims (21)

  1. 一种图像处理方法，包括：
    获取包括目标对象的目标图像；
    对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果；
    根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称。
  2. 根据权利要求1所述的方法，其中，所述对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果，包括：
    将所述目标图像输入至神经网络；
    根据所述神经网络的输出，确定所述目标图像中至少一个像素点的名称，作为所述分割结果；
    其中，所述神经网络通过包括所述目标对象的训练图像进行训练，所述训练图像中的目标对象通过至少一个目标子对象的名称进行标注。
  3. 根据权利要求2所述的方法，其中，所述神经网络通过包括所述目标对象的训练图像进行训练，包括：
    根据所述训练图像中目标对象的标注，确定所述训练图像中至少部分像素点的标签；
    通过包括所述至少部分像素点的标签的训练图像，对所述神经网络进行训练。
  4. 根据权利要求2或3所述的方法，其中，所述神经网络通过包括所述目标对象的训练图像进行训练，包括：
    在所述训练图像中的目标对象内，确定距离第一目标像素点最近的目标子对象，其中，所述第一目标像素点为所述训练图像中除所述目标对象以外的至少一个像素点；
    将确定的所述目标子对象的标注，作为所述第一目标像素点的标签；
    根据包括所述第一目标像素点的标签和所述目标对象的标注的训练图像，对所述神经网络进行训练。
  5. 根据权利要求1至4中任意一项所述的方法，其中，所述根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称，包括：
    根据所述分割结果，确定至少一个第二目标像素点的名称，其中，所述第二目标像素点为所述目标子对象包含的像素点；
    统计所述目标子对象中每一所述第二目标像素点的名称，得到统计结果，将所述统计结果中数量最多的名称，作为所述目标子对象的名称。
  6. 根据权利要求5所述的方法，其中，所述根据所述分割结果，确定至少一个第二目标像素点的名称，包括：
    将每一所述第二目标像素点对应的分割结果，作为每一所述第二目标像素点的名称；或者，
    基于每一所述第二目标像素点预设范围内至少一个像素点的分割结果，确定每一所述第二目标像素点的名称。
  7. 根据权利要求1至6中任意一项所述的方法，其中，所述方法还包括：
    根据至少一个所述目标子对象的名称，对所述目标对象进行处理，得到处理结果。
  8. 根据权利要求7所述的方法，其中，所述根据至少一个所述目标子对象的名称，对所述目标对象进行处理，包括：
    对所述目标对象中具有相同名称的目标子对象进行提取；和/或，
    根据相邻的所述目标子对象的名称，对所述目标对象中至少一个所述目标子对象的名称进行修正。
  9. 根据权利要求1至8中任意一项所述的方法，其中，所述目标图像包括：心脏冠脉掩模图像，或者，心脏冠脉掩模图像以及心脏掩模图像；
    所述目标对象包括冠脉中心线。
  10. 一种图像处理装置，包括：
    目标图像获取模块，配置为获取包括目标对象的目标图像；
    分割模块，配置为对所述目标图像进行分割，得到所述目标图像中至少一个像素点的名称，作为分割结果；
    命名模块，配置为根据所述分割结果，确定所述目标对象中至少一个目标子对象的名称。
  11. 根据权利要求10所述的装置，其中，所述分割模块还配置为
    将所述目标图像输入至神经网络；
    根据所述神经网络的输出，确定所述目标图像中至少一个像素点的名称，作为所述分割结果；
    其中，所述神经网络通过包括所述目标对象的训练图像进行训练，所述训练图像中的目标对象通过至少一个目标子对象的名称进行标注。
  12. 根据权利要求11所述的装置，其中，所述神经网络通过包括所述目标对象的训练图像进行训练，包括：
    根据所述训练图像中目标对象的标注，确定所述训练图像中至少部分像素点的标签；
    通过包括所述至少部分像素点的标签的训练图像，对所述神经网络进行训练。
  13. 根据权利要求10或11所述的装置，其中，所述神经网络通过包括所述目标对象的训练图像进行训练，包括：
    在所述训练图像中的目标对象内，确定距离第一目标像素点最近的目标子对象，其中，所述第一目标像素点为所述训练图像中除所述目标对象以外的至少一个像素点；
    将确定的所述目标子对象的标注，作为所述第一目标像素点的标签；
    根据包括所述第一目标像素点的标签和所述目标对象的标注的训练图像，对所述神经网络进行训练。
  14. 根据权利要求10至13任意一项所述的装置，其中，所述命名模块还配置为：
    根据所述分割结果，确定至少一个第二目标像素点的名称，其中，所述第二目标像素点为所述目标子对象包含的像素点；
    统计所述目标子对象中每一所述第二目标像素点的名称，得到统计结果，将所述统计结果中数量最多的名称，作为所述目标子对象的名称。
  15. 根据权利要求14所述的方法，其中，所述命名模块还配置为：
    将每一所述第二目标像素点对应的分割结果，作为每一所述第二目标像素点的名称；或者，
    基于每一所述第二目标像素点预设范围内至少一个像素点的分割结果，确定每一所述第二目标像素点的名称。
  16. 根据权利要求10至15中任意一项所述的装置，其中，所述装置还包括：
    处理模块，配置为根据至少一个所述目标子对象的名称，对所述目标对象进行处理，得到处理结果。
  17. 根据权利要求16所述的装置，其中，所述处理模块，还配置为
    对所述目标对象中具有相同名称的目标子对象进行提取；和/或，
    根据相邻的所述目标子对象的名称，对所述目标对象中至少一个所述目标子对象的名称进行修正。
  18. 根据权利要求10至17中任意一项所述的装置，其中，所述目标图像包括：心脏冠脉掩模图像，或者，心脏冠脉掩模图像以及心脏掩模图像；所述目标对象包括冠脉中心线。
  19. 一种电子设备，包括：
    处理器；
    配置为存储处理器可执行指令的存储器；
    其中，所述处理器被配置为调用所述存储器存储的指令，以执行权利要求1至9中任意一项所述的方法。
  20. 一种计算机可读存储介质，其上存储有计算机程序指令，其中，所述计算机程序指令被处理器执行时实现权利要求1至9中任意一项所述的方法。
  21. 一种计算机程序产品，包括计算机可读代码，在所述计算机可读代码在电子设备中运行的情况下，所述电子设备中的处理器执行如权利要求1至9中任意一项所述的方法。
PCT/CN2021/075633 2020-07-14 2021-02-05 图像处理方法及装置、电子设备、存储介质和程序产品 Ceased WO2022012038A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2022505271A JP2022543547A (ja) 2020-07-14 2021-02-05 画像処理方法及び装置、電子機器、記憶媒体並びにプログラム製品
KR1020227002923A KR20220028011A (ko) 2020-07-14 2021-02-05 이미지 처리 방법 및 장치, 전자 기기, 저장 매체 및 프로그램 제품

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010674675.X 2020-07-14
CN202010674675.XA CN111882527B (zh) 2020-07-14 2020-07-14 图像处理方法及装置、电子设备和存储介质

Publications (1)

Publication Number Publication Date
WO2022012038A1 true WO2022012038A1 (zh) 2022-01-20

Family

ID=73151298

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/075633 Ceased WO2022012038A1 (zh) 2020-07-14 2021-02-05 图像处理方法及装置、电子设备、存储介质和程序产品

Country Status (5)

Country Link
JP (1) JP2022543547A (zh)
KR (1) KR20220028011A (zh)
CN (2) CN114494221B (zh)
TW (1) TW202203151A (zh)
WO (1) WO2022012038A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119889670A (zh) * 2025-03-31 2025-04-25 东莞市东南部中心医院(东莞市东南部中医医疗服务中心、广东医科大学附属东莞第一医院) 基于深度学习的急性脑卒中早期智能筛查预警方法及系统

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114494221B (zh) * 2020-07-14 2025-01-10 上海商汤善萃医疗科技有限公司 图像处理方法及装置、电子设备和存储介质
CN114299289B (zh) * 2021-12-24 2026-01-13 上海商汤善萃医疗科技有限公司 图像处理方法及装置、电子设备和存储介质
CN115171174B (zh) * 2022-07-05 2026-03-03 推想医疗科技股份有限公司 训练样本的生成方法、装置、设备和存储介质
CN118172362B (zh) * 2024-05-14 2024-08-16 柏意慧心(杭州)网络科技有限公司 心脏冠脉血管分段命名方法、装置、设备和存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100076296A1 (en) * 2008-09-22 2010-03-25 Siemens Corporate Research, Inc. Method and System for Automatic Detection of Coronary Stenosis in Cardiac Computed Tomography Data
CN106204587A (zh) * 2016-05-27 2016-12-07 孔德兴 基于深度卷积神经网络和区域竞争模型的多器官分割方法
CN108717695A (zh) * 2018-04-25 2018-10-30 数坤（北京）网络科技有限公司 心脏冠脉血管自动分段命名方法
CN109522898A (zh) * 2018-09-18 2019-03-26 平安科技（深圳）有限公司 手写样本图片标注方法、装置、计算机设备及存储介质
CN111161301A (zh) * 2019-12-31 2020-05-15 上海商汤智能科技有限公司 图像分割方法及装置、电子设备和存储介质
CN111882527A (zh) * 2020-07-14 2020-11-03 上海商汤智能科技有限公司 图像处理方法及装置、电子设备和存储介质

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3636332B2 (ja) * 1993-12-24 2005-04-06 セイコーエプソン株式会社 画像処理方法および画像処理装置
JPH0991032A (ja) * 1995-09-21 1997-04-04 Fuji Electric Co Ltd 制御中の工程名称を表示するプログラムコントローラ
US8107726B2 (en) * 2008-06-18 2012-01-31 Samsung Electronics Co., Ltd. System and method for class-specific object segmentation of image data
JP5846368B2 (ja) * 2011-11-16 2016-01-20 学校法人東京理科大学 医療画像処理装置、方法、及びプログラム
AU2016236323A1 (en) * 2015-03-20 2017-08-10 Ventana Medical Systems, Inc. System and method for image segmentation
US9965863B2 (en) * 2016-08-26 2018-05-08 Elekta, Inc. System and methods for image segmentation using convolutional neural network
CN106651886A (zh) * 2017-01-03 2017-05-10 北京工业大学 一种基于超像素聚类优化cnn的云图分割方法
US10170600B2 (en) * 2017-01-12 2019-01-01 Semiconductor Energy Laboratory Co., Ltd. Method for manufacturing semiconductor device
US11126894B2 (en) * 2017-06-05 2021-09-21 Siemens Aktiengesellschaft Method and apparatus for analysing an image
CN108230343B (zh) * 2018-01-05 2020-06-05 厦门华联电子股份有限公司 一种图像处理方法及装置
JP7153264B2 (ja) * 2018-08-02 2022-10-14 三菱重工業株式会社 画像解析システム、画像解析方法及び画像解析プログラム
CN110223279B (zh) * 2019-05-31 2021-10-08 上海商汤智能科技有限公司 一种图像处理方法和装置、电子设备
CN111079797B (zh) * 2019-11-25 2022-02-25 贝壳技术有限公司 一种图像分类的方法、装置和存储介质
CN111161240B (zh) * 2019-12-27 2024-03-05 上海联影智能医疗科技有限公司 血管分类方法、装置、计算机设备和可读存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100076296A1 (en) * 2008-09-22 2010-03-25 Siemens Corporate Research, Inc. Method and System for Automatic Detection of Coronary Stenosis in Cardiac Computed Tomography Data
CN106204587A (zh) * 2016-05-27 2016-12-07 孔德兴 基于深度卷积神经网络和区域竞争模型的多器官分割方法
CN108717695A (zh) * 2018-04-25 2018-10-30 数坤（北京）网络科技有限公司 心脏冠脉血管自动分段命名方法
CN109522898A (zh) * 2018-09-18 2019-03-26 平安科技（深圳）有限公司 手写样本图片标注方法、装置、计算机设备及存储介质
CN111161301A (zh) * 2019-12-31 2020-05-15 上海商汤智能科技有限公司 图像分割方法及装置、电子设备和存储介质
CN111882527A (zh) * 2020-07-14 2020-11-03 上海商汤智能科技有限公司 图像处理方法及装置、电子设备和存储介质

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119889670A (zh) * 2025-03-31 2025-04-25 东莞市东南部中心医院(东莞市东南部中医医疗服务中心、广东医科大学附属东莞第一医院) 基于深度学习的急性脑卒中早期智能筛查预警方法及系统

Also Published As

Publication number Publication date
CN114494221B (zh) 2025-01-10
CN111882527B (zh) 2021-12-21
TW202203151A (zh) 2022-01-16
CN111882527A (zh) 2020-11-03
KR20220028011A (ko) 2022-03-08
CN114494221A (zh) 2022-05-13
JP2022543547A (ja) 2022-10-13

Similar Documents

Publication Publication Date Title
CN110047078B (zh) 图像处理方法及装置、电子设备和存储介质
CN111882527B (zh) 图像处理方法及装置、电子设备和存储介质
CN112967291B (zh) 图像处理方法及装置、电子设备和存储介质
CN114820584B (zh) 肺部病灶定位装置
CN113222038B (zh) 基于核磁图像的乳腺病灶分类和定位方法及装置
TWI755175B (zh) 圖像分割方法、電子設備和儲存介質
WO2022127071A1 (zh) 网络训练方法、图像分割方法、装置、设备、介质及产品
CN111310764A (zh) 网络训练、图像处理方法及装置、电子设备和存储介质
TWI767614B (zh) 圖像處理方法及電子設備、電腦可讀儲存介質
CN113034491B (zh) 一种冠脉钙化斑块检测方法及装置
TW202209254A (zh) 圖像分割方法、電子設備和電腦可讀儲存介質
CN113012816B (zh) 脑分区风险预测方法及装置、电子设备和存储介质
CN113012166A (zh) 颅内动脉瘤分割方法及装置、电子设备和存储介质
CN115423819B (zh) 肺血管的分割方法及装置、电子设备和存储介质
CN113012146A (zh) 血管信息获取方法及装置、电子设备和存储介质
CN111369512A (zh) 图像处理方法及装置、电子设备和存储介质
TW202219831A (zh) 圖像處理方法、電子設備和電腦可讀儲存介質
CN118542689A (zh) 心膈角定位方法及装置、程序产品、医疗设备和应用
CN114418931A (zh) 术后剩余肺叶的提取方法及装置、电子设备和存储介质
WO2025086476A1 (zh) 神经影像的病灶识别方法、装置、电子设备和存储介质
US20250139771A1 (en) Method and device for nidus recognition in neuroimages, electronic apparatus, and storage medium
WO2023050690A1 (zh) 图像处理方法、装置、电子设备、存储介质和程序
WO2021259394A2 (zh) 一种图像处理方法及装置、电子设备和存储介质
CN112686867A (zh) 医学图像的识别方法及装置、电子设备和存储介质
CN118542688A (zh) 心胸比的确定方法及装置、程序产品及医疗设备

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022505271

Country of ref document: JP

Kind code of ref document: A

ENP Entry into the national phase

Ref document number: 20227002923

Country of ref document: KR

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: 21842109

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: 21842109

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 1205A DATED 11.07.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21842109

Country of ref document: EP

Kind code of ref document: A1