WO2022062590A1 - 图像识别方法及装置、设备、存储介质和程序 - Google Patents

图像识别方法及装置、设备、存储介质和程序 Download PDF

Info

Publication number
WO2022062590A1
WO2022062590A1 PCT/CN2021/106479 CN2021106479W WO2022062590A1 WO 2022062590 A1 WO2022062590 A1 WO 2022062590A1 CN 2021106479 W CN2021106479 W CN 2021106479W WO 2022062590 A1 WO2022062590 A1 WO 2022062590A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
network
medical images
sample
recognized
Prior art date
Application number
PCT/CN2021/106479
Other languages
English (en)
French (fr)
Inventor
陈翼男
Original Assignee
上海商汤智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海商汤智能科技有限公司 filed Critical 上海商汤智能科技有限公司
Publication of WO2022062590A1 publication Critical patent/WO2022062590A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour

Definitions

  • the present disclosure relates to the technical field of artificial intelligence, and relates to, but is not limited to, an image recognition method and apparatus, electronic device, computer storage medium and computer program.
  • CT Computed Tomography
  • MRI Magnetic Resonance Imaging
  • the scan image categories often include time series-related pre-contrast scan, early arterial phase, late arterial phase, portal venous phase, delayed phase, etc.
  • the scan image category can also include scan parameters related to the scan image. T1-weighted inverse imaging, T1-weighted in-phase imaging, T2-weighted imaging, diffusion-weighted imaging, surface diffusion coefficient imaging, and more.
  • a radiologist is usually required to identify the scanned image category of the scanned medical image to ensure that the required medical image is obtained; or, during inpatient or outpatient diagnosis and treatment, a doctor is usually required to identify the scanned medical image. , judge the scanned image category of each medical image, and then read the image.
  • Embodiments of the present application provide an image recognition method and apparatus, an electronic device, a computer storage medium, and a computer program.
  • the embodiment of the present application provides an image recognition method, including: acquiring a plurality of medical images to be recognized; extracting a style feature representation of each medical image to be recognized; classifying the style feature representations of the plurality of medical images to be recognized, Obtain the scanned image category of each medical image to be recognized.
  • the style feature representation of the plurality of medical images to be recognized can be classified and processed.
  • the differences in the respective style features of the medical images to be recognized can further improve the accuracy of the recognized scanned image categories, and because the style feature representations of multiple medical images to be recognized can be classified and processed, and each medical image to be recognized can be obtained. Therefore, the scanning image categories of multiple medical images to be recognized can be obtained at one time, thereby improving the efficiency of image recognition. Therefore, the above solution can improve the efficiency and accuracy of image recognition.
  • classifying the style feature representations of a plurality of medical images to be recognized, and obtaining the scanned image category of each medical image to be recognized includes: performing a first step on the style feature representations of the plurality of medical images to be recognized. Fusion processing is performed to obtain the final style feature representation; the final style feature representation is classified to obtain the scanned image category of each medical image to be recognized.
  • the first fusion processing is performed on the style feature representations of the multiple unrecognized medical images to obtain the final style feature representation, so the final style feature representation can represent each
  • the difference between the style feature representation of a medical image to be recognized and the style feature representation of other to-be-recognized medical images, so using the final style feature representation for classification processing can improve the accuracy of the recognized scanned image category.
  • the image recognition method further includes at least one of the following: The recognized medical images are sorted according to their scanned image categories; at least one to-be-recognized medical image sorted according to the scanned image categories is displayed on the same screen; if the scanned image categories of the to-be-recognized medical images are duplicated, the first warning information is output, to prompt the scanning personnel; if there is no preset scanning image category in the scanned image categories of the medical images to be identified, output second warning information to prompt the scanning personnel; if the classification confidence of the scanned image categories of the medical images to be identified If it is less than the preset reliability threshold, the third warning information is output to remind the scanning personnel.
  • the at least one to-be-identified medical image is sorted according to its scanned image category, which can improve the convenience of doctor reading;
  • the last at least one medical image to be recognized is displayed on the same screen, which can avoid back-and-forth comparison when the doctor reads the medical image to be recognized, thereby improving the efficiency of the doctor's reading.
  • an early warning message to prompt the scanner, when there is no preset scan image category in the scanned image category of at least one medical image to be recognized, output second early warning information to prompt the scanner, in the scanned image category of the medical image to be recognized
  • the third warning information is output to remind the scanning personnel, and the image quality control can be realized during the scanning process, so that when it is inconsistent with the actual situation, the error can be corrected in time and the second registration of the patient can be avoided.
  • the above method before extracting the style feature representation of each to-be-recognized medical image, the above method further includes: preprocessing each to-be-recognized medical image, wherein the preprocessing includes at least one of the following: The image size of the recognized medical image is adjusted to a preset size, and the image intensity of the medical image to be recognized is normalized to a preset range.
  • the image data of each target area is preprocessed, and the preprocessing includes at least one of the following: adjusting the image size of the target area to a preset size, and normalizing the image intensity of the target area It can help to improve the accuracy of subsequent image recognition.
  • the image recognition method further includes: extracting the content feature representation of each to-be-recognized medical image respectively; performing lesion identification on the content-feature representations of a plurality of to-be-recognized medical images, and obtaining the content feature representation of each to-be-recognized medical image. the lesion area.
  • the lesion area in each to-be-recognized medical image can be obtained, and each to-be-recognized medical image can be obtained after obtaining the lesion area.
  • the lesion area is determined while scanning the image category of the medical image, which can help improve the overall reading efficiency, and can help eliminate the interference caused by the lesion to the category recognition of the scanned image, thereby improving the accuracy of image recognition.
  • performing lesion identification on the content feature representations of multiple medical images to be identified, and obtaining a lesion area in each to-be-recognized medical image includes: performing a second step on the content feature representations of the multiple to-be-recognized medical images.
  • the fusion processing is performed to obtain the final content feature representation; the lesion identification is performed on the final content feature representation to obtain the lesion area in each medical image to be identified.
  • the second fusion processing is performed on the content feature representations of multiple medical images to be recognized to obtain the final content feature representation, which can help to make the final content feature representation compensate for inconspicuous lesions or motion interference that may exist in a single medical image to be recognized. Therefore, when using the final content feature representation for lesion identification, the accuracy of lesion identification can be improved.
  • the image recognition method further includes: prompting the currently displayed lesion area of the medical image to be recognized.
  • performing a second fusion process on the content feature representations of the plurality of medical images to be recognized includes any one of the following: performing a splicing process on the content feature representations of the plurality of medical images to be recognized; The content feature representations of the medical images to be recognized are added, wherein the final content feature representation has the same dimension as the content feature representations of the multiple medical images to be recognized.
  • a final content feature representation is obtained, and the final content feature representation is
  • the representation has the same dimension as the content feature representation of multiple medical images to be recognized, and the final content feature representation can be obtained in various ways, thereby improving the robustness of image recognition.
  • extracting the style feature representation of each to-be-recognized medical image separately includes: using the style coding sub-network of the recognition network to separately extract the style-feature representation of each to-be-recognized medical image;
  • the style feature representation of the image is classified and processed to obtain the scanned image category of each to-be-recognized medical image, including: using the classification processing sub-network of the recognition network to classify and process the style feature representations of a plurality of to-be-recognized medical images to obtain each to-be-recognized medical image.
  • the style coding sub-network of the recognition network to extract the style feature representation of each medical image to be recognized, and use the classification processing sub-network of the recognition network to classify and process the style feature representations of multiple medical images to be recognized, and obtain each to-be-recognized medical image. Identify the scanned image categories of medical images, use the content coding sub-network of the recognition network to extract the content feature representation of each medical image to be recognized, and use the region segmentation sub-network of the recognition network to perform focus on the content feature representations of multiple medical images to be recognized.
  • Recognition can obtain the lesion area in each medical image to be recognized, and the recognition network can be used to perform tasks such as extraction of style feature representation, classification processing, extraction of content feature representation, and lesion identification, so it can help improve the efficiency of image recognition.
  • the image recognition method before extracting the style feature representation of each to-be-recognized medical image, the image recognition method further includes: acquiring a plurality of sample medical images, wherein the plurality of sample medical images are marked with their real scanned image categories and The real lesion area; use the style coding sub-network to extract the sample style feature representation of each sample medical image, and use the content coding sub-network to separately extract the sample content feature representation of each sample medical image; use the classification processing sub-network to classify multiple samples.
  • the sample style feature representation of medical images is classified and processed to obtain the predicted scan image category of each sample medical image, and the region segmentation sub-network is used to identify the sample content feature representation of multiple sample medical images to obtain each sample medical image.
  • the predicted lesion area in the model using the difference between the real scan image category and the predicted scan image category, adjust the network parameters of the style coding sub-network and the classification processing sub-network, and use the difference between the real and predicted lesion areas to adjust the content coding sub-network. and network parameters of the region segmentation sub-network.
  • the training of the content coding sub-network and the region segmentation sub-network can be added, so that the lesion identification ability of the region segmentation sub-network can be improved, and the content coding can be improved at the same time.
  • the degree of acquisition of the content features related to the lesions by the sub-network can help make the style coding sub-network not respond to the features related to the lesions, so that the subsequent classification will not be affected by the features related to the lesions, so it can improve the robustness of image recognition. .
  • the image recognition method further includes: acquiring the sample data distribution represented by the sample style feature of each sample medical image; and adjusting the network parameters of the style coding sub-network by using the difference between the sample data distributions.
  • the distribution of sample data is obtained at the same time, and the difference between the distribution of sample data is used to adjust the network parameters of the style coding sub-network, which can help to make the subsequent extraction of style feature representation independent of each other. Therefore, the accuracy of the recognized scanned image category can be improved.
  • the image recognition method further includes: using a sample style feature representation and a content feature representation, constructing a reconstructed image corresponding to the sample style feature representation; using the reconstructed image and the corresponding sample style feature representation belong to Differences between sample medical images, adjusting the network parameters of the style-coding sub-network and the content-coding sub-network.
  • a sample style feature representation and a content feature representation are used at the same time to construct a reconstructed image corresponding to the sample style feature representation, and the reconstructed image and the corresponding sample style feature represent the relationship between the sample medical images to which they belong.
  • Adjust the network parameters of the style coding sub-network and the content coding sub-network so that the style coding sub-network can extract as complete and accurate style features as possible, while the content coding sub-network can extract as complete and accurate style features as possible. , which can help to improve the classification of subsequent scanned images and the accuracy of lesion identification.
  • the style encoding sub-network includes: a sequentially connected downsampling layer and a global pooling layer; and/or, the content encoding sub-network includes any one of the following: a sequentially connected downsampling layer and a residual block , sequentially connected convolutional and pooling layers.
  • the style encoding sub-network to include sequentially connected downsampling layers and global pooling layers, it is possible to facilitate network training while simplifying the network structure; by setting the content encoding sub-network to include any of the following: Sequentially connected downsampling layers and residual blocks, and sequentially connected convolutional layers and pooling layers can help simplify network training while simplifying the network structure.
  • the embodiment of the present application also provides an image recognition device, including an image acquisition module, a style extraction module, and a classification processing module.
  • the image acquisition module is configured to acquire a plurality of medical images to be recognized;
  • the style extraction module is configured to separately extract each to-be-recognized medical image.
  • the classification processing module is configured to classify and process the style feature representation of a plurality of to-be-recognized medical images to obtain the scanned image category of each to-be-recognized medical image.
  • An embodiment of the present application further provides an electronic device, including a memory and a processor coupled to each other, and the processor is configured to execute program instructions stored in the memory, so as to implement any one of the above image recognition methods.
  • the embodiments of the present application also provide a computer-readable storage medium, which stores program instructions, and when the program instructions are executed by a processor, any one of the above image recognition methods is implemented.
  • Embodiments of the present application further provide a computer program, including computer-readable codes.
  • a processor in the electronic device executes any one of the image recognition described above. method.
  • the style feature representation of the plurality of medical images to be recognized is classified and processed. , considering the differences in the respective style features of multiple medical images to be identified, thereby improving the accuracy of the recognized scanned image categories, and because the style feature representations of multiple medical images to be identified can be classified and processed, and each One scanned image category of the medical image to be recognized, so multiple scanned image categories of the medical image to be recognized can be obtained at one time, thereby improving the efficiency of image recognition. Therefore, the embodiment of the present application can improve the efficiency and accuracy of image recognition.
  • FIG. 1 is a schematic flowchart of an embodiment of an image recognition method of the present application
  • FIG. 2 is a schematic flowchart of an embodiment of a training recognition network
  • FIG. 3 is a schematic state diagram of an embodiment of a training recognition network
  • FIG. 4 is a schematic diagram of a framework of an embodiment of an image recognition apparatus of the present application.
  • FIG. 5 is a schematic diagram of a framework of an embodiment of an electronic device of the present application.
  • FIG. 6 is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application.
  • system and “network” are often used interchangeably herein.
  • the term “and/or” in this article is only an association relationship to describe the associated objects, indicating that there can be three kinds of relationships, for example, A and/or B, which can mean: A alone exists, A and B exist simultaneously, and A and B exist independently B these three cases.
  • the character "/” in this document generally indicates that the related objects are an “or” relationship.
  • “multiple” herein means two or more than two.
  • FIG. 1 is a schematic flowchart of an embodiment of an image recognition method of the present application. Specifically, the following steps can be included:
  • Step S11 Acquire a plurality of medical images to be recognized.
  • the medical images to be identified may include CT images and MR images, which are not limited herein.
  • the medical image to be recognized may be obtained by scanning regions such as the abdomen, chest, etc., which may be specifically set according to the actual application, which is not limited herein. For example, when the liver, spleen, and kidney are the organs that need diagnosis and treatment, the abdomen can be scanned to obtain the medical image to be identified; Recognition of medical images, other situations can be deduced by analogy, and no examples will be given here.
  • the scanning mode may be a flat scanning, an enhanced scanning, or the like, which is not limited herein.
  • the medical image to be recognized may be a three-dimensional image, which is not limited herein.
  • multiple medical images to be recognized may be obtained by scanning the same object.
  • Step S12 Extract the style feature representation of each medical image to be recognized, respectively.
  • the style feature represents the style used to describe the medical image to be recognized, eg, the degree of enhancement of a contrast agent injected into a blood vessel represented in the medical image to be recognized.
  • the degree of enhancement of angiographic contrast agents such as veins and portal veins of the liver in the medical images to be identified in different scanning image categories is different.
  • the portal vein has been enhanced in the to-be-identified medical image whose image category is the late arterial stage, and the portal vein has been fully enhanced in the unidentified medical image whose scan image category is the portal venous phase, the liver blood vessels have been enhanced by forward blood flow, and the liver soft cells are in the The peak value has been reached under the marker, and the scanning image category is the delayed phase in the medical image to be identified, in which the portal vein and artery are in an enhanced state and are weaker than the portal phase, and the liver soft cell tissue is in an enhanced state and weaker than the portal phase.
  • Other scans Image categories are not listed one by one here.
  • the style feature representation may be represented by a vector, and the size of the vector may be set according to the actual situation.
  • the size of the vector may be set to 8 bits, which is not limited herein.
  • a recognition network in order to improve the convenience of extracting style feature representations, can be pre-trained, and the recognition network includes a style coding sub-network, and the style coding sub-network of the recognition network is used to extract each medical item to be recognized separately.
  • the style encoding sub-network may include a sequential connection of a downsampling layer and a global pooling layer, so that after the medical image to be recognized is subjected to downsampling processing, the global pooling layer is used for pooling processing , to get the style feature representation.
  • each medical image to be recognized may also be preprocessed.
  • the image size of the medical image to be recognized is adjusted to a preset size (eg, 32*256*256).
  • the preprocessing may further include normalizing the image intensity of the medical image to be recognized to a preset range (for example, a range of 0 to 1).
  • a preset ratio For example, the gray value corresponding to 99.9% is used as the normalized clamping value, so that the contrast of the medical image to be recognized can be enhanced, and the accuracy of subsequent image recognition can be improved.
  • Step S13 Perform classification processing on the style feature representations of a plurality of medical images to be recognized, to obtain a scanned image category of each medical image to be recognized.
  • the scanned image category can be set according to the actual situation.
  • the scan image category may include time-series-related pre-contrast scan, early arterial phase, late arterial phase, portal phase, and delayed phase; or, the scan image category may also include It can include T1-weighted inverse imaging, T1-weighted in-phase imaging, T2-weighted imaging, diffusion-weighted imaging, and surface diffusion coefficient imaging related to scanning parameters.
  • the early arterial phase may indicate that the portal vein has not been enhanced
  • the late arterial phase may indicate that the portal vein has been enhanced
  • the portal venous phase may indicate that the portal vein is sufficiently enhanced and the liver vessels have been enhanced by forward blood flow, and liver parenchyma tissue is under the marker.
  • the peak value has been reached, and the delay period can indicate that the portal vein and arteries are in an enhanced state and weaker than the portal venous phase, and the liver parenchyma tissue is in an enhanced state and weaker than the portal venous phase.
  • Other scan image categories are not listed here.
  • the medical image to be identified is a medical image obtained by scanning other organs, it can be deduced by analogy, and no examples will be given here.
  • the respective style feature representations of the respective style feature representations are classified to obtain the scanned image category of each medical image to be recognized.
  • the above-mentioned recognition network further includes a classification processing sub-network, so that the classification processing sub-network can be used to classify and process the style feature representations of multiple medical images to be recognized, and obtain each The scanned image category of the medical image to be recognized is obtained, and the scanning image category to which the medical image to be recognized belongs can be obtained by classifying the style feature representation by the classification processing sub-network, so the convenience of the classification processing can be improved.
  • the classification processing sub-network may include sequentially connected fully connected layers and softmax layers, which are not limited herein.
  • the style feature representations of multiple medical images to be recognized can also be subjected to a first fusion process to obtain a final style feature representation, and the final style feature representation is classified to obtain each A scanned image class of the medical image to be identified.
  • the operation of the first fusion process may be to splicing the style feature representations of multiple medical images to be recognized, so as to obtain the final style feature representation; or, the operation of the first fusion process may also be to combine the multiple The style feature representations of the medical images are stacked to obtain the final style feature representation, which is not limited here.
  • the final style feature representation obtained by performing the first fusion process on the style feature representations of multiple medical images to be recognized can represent the relationship between the style feature representation of each to-be-recognized medical image and the style feature representations of other to-be-recognized medical images Therefore, using the final style feature representation for classification processing can improve the accuracy of the recognized scanned image categories.
  • the content feature representation of each to-be-recognized medical image can also be extracted, and the content feature representation of multiple to-be-recognized medical images can be used for lesion identification.
  • the lesion area can be determined while obtaining the scanned image category of each medical image to be identified, so it can help to improve the overall reading performance, and at the same time, it can be beneficial to Eliminate the interference caused by the lesions to the classification recognition of scanned images, thereby improving the accuracy of image recognition.
  • the content feature represents the content in the medical image to be recognized, for example, the anatomical features of an organ in the medical image to be recognized.
  • the content feature representation can describe the physiological positional relationship between the liver and its adjacent organs (such as spleen, kidney), the shape characteristics of the liver, the texture (such as soft, hard, etc.) characteristics, the composition (such as spleen, kidney, etc.) , water-containing, fat-containing, etc.) characteristics, etc., are not limited here.
  • the lesions may include tumors, thrombus, nodules, etc., which may be set according to actual conditions, and are not limited herein.
  • the recognition network may further include a content coding sub-network, so that the content feature representation of each medical image to be recognized can be extracted separately by using the content coding sub-network of the recognition network.
  • the content coding sub-network can use sequential connected downsampling layers and residual blocks (resblocks), the number of residual blocks can be set according to the actual situation, and the network depth can be improved by setting the residual blocks.
  • the content coding sub-network can also use sequentially connected convolutional layers and pooling layers, and the number of groups of convolutional layers and pooling layers can be set according to the actual situation. For example, a set of sequentially connected convolutional layers and pooling layers, two sets of sequentially connected convolutional layers and pooling layers, three sets of sequentially connected convolutional layers and pooling layers, etc., can be used, which are not limited here. .
  • the identification network may further include a region segmentation sub-network, so as to use the region segmentation sub-network of the identification network to perform lesion identification on the content feature representations of multiple medical images to be identified , to obtain the lesion area in each medical image to be identified.
  • the area segmentation sub-network may adopt Unet, Vnet, etc., which is not limited herein.
  • the lesion area may include an area surrounding the lesion, for example, the outline of the lesion, etc., which is not limited herein.
  • the content feature representation may be represented by a tensor, for example, the content feature representation may be represented by a low-resolution tensor, which is not limited herein.
  • a second fusion process may be performed on the content feature representations of a plurality of medical images to be identified to obtain a final content feature representation, and the final content feature representation is subjected to lesion identification, The lesion area in each medical image to be recognized is obtained, so that the final content feature can be used to compensate for problems such as inconspicuous lesions or artifacts caused by motion interference that may exist in a single medical image to be recognized. Indicates that the accuracy of lesion identification can be improved when performing lesion identification.
  • the second fusion process may specifically be performing a concatenation process on the content feature representations of a plurality of medical images to be recognized to implement a fusion process on the content feature representations.
  • the final content feature representation can be obtained through several simple convolutional layers without pooling; here, splicing the content features can be regarded as a concatenation operation of tensors; or, it can be the content of multiple medical images to be recognized.
  • the feature representation is added (add) to realize the fusion processing of the content feature representation, which can be regarded as a summation operation of tensors;
  • a convolution kernel such as 1*1 performs a convolution operation on the stacked content feature representations to achieve fusion processing of the content feature representations;
  • the content feature representations of multiple medical images to be recognized can also be weighted, here Not limited.
  • the final content feature representation and the content feature representation of the plurality of medical images to be recognized have the same dimension.
  • the currently displayed lesion area of the medical image to be recognized may be prompted.
  • preset lines eg, bold lines, dot-dash lines, double solid lines, etc.
  • preset colors eg, yellow, red, green, etc.
  • Symbols for example, arrows pointing to the lesion area, etc. represent the lesion area, which can be set according to the actual situation, which is not limited here.
  • the above-mentioned trained identification network can be set in image post-processing workstations, imaging workstations, computer-aided reading systems, telemedicine diagnosis scenarios, cloud platform-assisted intelligent diagnosis scenarios, etc. Automatic recognition of medical images to improve recognition efficiency.
  • At least one medical image to be recognized is obtained by scanning the same object, so in order to facilitate the doctor to read the image, after obtaining the scanned image category to which each medical image to be recognized belongs, the at least one medical image to be recognized can also be scanned. Images are sorted by their scan image category, for example, T1-weighted inverse imaging, T1-weighted in-phase imaging, pre-contrast, early arterial, late arterial, portal venous phase, delayed phase, T2-weighted imaging, diffusion-weighted imaging, The preset order of the surface diffusion coefficient imaging is sorted. In addition, the preset order can also be set according to the doctor's habits, which is not limited here, so as to improve the convenience of the doctor's reading.
  • At least one medical image to be recognized sorted according to the scanned image category can also be displayed on the same screen.
  • the number of medical images to be recognized is 5.
  • the medical images to be recognized can be displayed in the five display windows respectively. Therefore, it is possible to reduce the time for doctors to review multiple medical images to be recognized and compare them back and forth, and improve the efficiency of reading images.
  • At least one medical image to be recognized is obtained by scanning the same object. Therefore, in order to perform quality control during the scanning process, after obtaining the scanned image category to which each medical image to be recognized belongs, it is also possible to determine the medical image to be recognized. Identify whether the scanned image category of the medical image is duplicated, and when duplicated, output first warning information to prompt the scanning personnel. For example, if there are two medical images to be identified with both scanned image categories as "delay period", it can be considered that the scanning quality is not compliant during the scanning process. Therefore, in order to remind the scanning personnel, the first warning information can be output. For example Optionally, the cause of the warning can be output (for example, there are medical images to be identified with duplicate types of scanned images, etc.).
  • the preset scanned image category is "portal phase". If there is no image with the scanned image category of "portal phase" in at least one of the medical images to be identified, it can be considered that the scanning quality is not compliant during the scanning process. , so in order to prompt the scanning personnel, the second warning information can be output, for example, the warning reason (eg, there is no portal phase image in the medical image to be identified, etc.) can be output.
  • the classification confidence may be predicted by the classification processing sub-network when the classification processing is performed. For example, when the classification processing sub-network performs classification processing, it predicts the scanned image category and the corresponding classification confidence of each medical image to be recognized.
  • the third warning information may be output, for example, the cause of the warning may be output (for example, the medical image to be recognized may have a problem of poor scanning quality). Therefore, the image quality control can be realized during the scanning process, so that when it is contrary to the actual situation, the error can be corrected in time and the second registration of the patient can be avoided.
  • the style feature representation of the plurality of medical images to be recognized is classified and processed, so it is possible to consider many
  • the differences in the respective style features of the medical images to be recognized can further improve the accuracy of the recognized scanned image categories, and because the style feature representations of a plurality of medical images to be recognized can be classified and processed, and each of the medical images to be recognized can be obtained. Scanning image categories of medical images, so multiple scanned image categories of medical images to be recognized can be obtained at one time, thereby improving the efficiency of image recognition. Therefore, the above solution can improve the efficiency and accuracy of image recognition.
  • the identification network includes the content coding sub-network, the style coding sub-network, the classification processing sub-network and the region segmentation sub-network in the foregoing embodiments, and the specific process is as follows:
  • Step S21 acquiring multiple sample medical images, wherein the multiple sample medical images are marked with their real scanned image categories and real lesion areas.
  • the multiple sample medical images used in a training process can be obtained by scanning the same object.
  • the sample medical image used in a certain training may be obtained by scanning object A
  • the sample medical image used in another training may be obtained by scanning object B.
  • the sample medical images may also include CT images and MR images, which are not limited herein.
  • CT images and MR images which are not limited herein.
  • the real scanned image category and real lesion area marked by the sample medical image can be marked by clinicians, radiologists and other personnel with medical imaging knowledge.
  • the scan image category can be set according to the actual situation. For example, if the sample medical image is obtained by scanning the liver, the scan image category can specifically include the timing-related pre-contrast scan, early arterial, late arterial, portal venous phase, Delay period; alternatively, the scan image category may also include T1-weighted inverse imaging, T1-weighted in-phase imaging, T2-weighted imaging, diffusion-weighted imaging, and surface diffusion coefficient imaging related to scanning parameters. For details, please refer to the relevant steps in the foregoing embodiments. , and will not be repeated here.
  • the real lesion area may be marked with polygons, for example, the contour of the lesion may be marked with polygons, etc., which is not limited herein.
  • sample medical image 1, sample medical image 2, sample medical image 3, ..., sample medical image n can be obtained, where the value of n can be set according to the actual situation, for example , when the scan image category includes time-related pre-contrast scan, early arterial, late arterial, portal venous phase, and delayed phase, the value of n can be set to an integer less than or equal to 5, for example, 5, 4, 3, etc.
  • the value of n can be set to a value less than or equal to 5 Integer, for example, 5, 4, 3, etc.; or, when the scan image category includes both T1-weighted inverse imaging, T1-weighted in-phase imaging, T2-weighted imaging, diffusion-weighted imaging, and surface diffusion coefficient imaging related to scanning parameters, also When including the time-series-related pre-contrast scan, early arterial, late arterial, portal venous phase, and delayed phase, the value of n can be set to an integer less than or equal to 10, for example, 10, 9, 8, etc. Set according to the actual situation, which is not limited here.
  • Step S22 Using the style coding sub-network to extract the sample style feature representation of each sample medical image respectively, and using the content coding sub-network to separately extract the sample content feature representation of each sample medical image.
  • Fig. 3 in conjunction with the sample medical image 1, the sample medical image 2, the sample medical image 3, ... and the sample medical image n after the style feature extraction by the style coding sub-network, the sample style feature representation 1, sample Style feature representation 2, sample style feature representation 3, ..., sample style feature representation n; after content feature extraction by the content coding sub-network, sample content feature representation 1, sample content feature representation 2, sample content feature representation can be obtained respectively 3, ..., the sample content feature representation n.
  • sample style feature representation and the sample content feature representation reference may be made to the style feature representation and the content feature representation in the foregoing embodiments, which will not be repeated here.
  • Step S23 Use the classification processing sub-network to classify and process the sample style feature representations of the multiple sample medical images to obtain the predicted scanning image category of each sample medical image, and use the region segmentation sub-network to classify the sample content of the multiple sample medical images.
  • the feature representation is used for lesion identification, and the predicted lesion area in each sample medical image is obtained.
  • the sample style feature representations of multiple sample medical images can be spliced to obtain the final sample style feature representation, and the final sample style feature representation can be classified by using the classification processing sub-network to obtain each sample medical image.
  • the predicted scan image category of the image so that the final sample style feature representation can represent the difference between the sample style feature representation of each sample medical image and the sample style feature representation of other sample medical images, so the classification processing sub-network is used to classify the final sample style.
  • the feature representation is used for classification processing, the accuracy of the classification processing can be improved.
  • sample medical image 1 the sample medical image 2, the sample medical image 3, .
  • the sample style features of the image are subjected to splicing processing (or stacking processing and other processing methods, for details, please refer to the aforementioned disclosed embodiments, which will not be repeated here) to obtain the final sample style feature representation, so that the classification processing sub-network is used to classify the final sample style features.
  • splicing processing or stacking processing and other processing methods, for details, please refer to the aforementioned disclosed embodiments, which will not be repeated here
  • classification processing sub-network is used to classify the final sample style features.
  • Indicates that classification processing is performed to obtain the predicted scanning image categories of sample medical image 1, sample medical image 2, sample medical image 3, ..., and sample medical image n.
  • the sample content feature representations of multiple sample medical images can be fused to obtain the final sample content feature representation, and the region segmentation sub-network can be used to identify lesions on the final sample content feature representation to obtain each sample medical image.
  • the predicted lesion area in the image can help to make the final sample content feature representation to compensate for problems such as inconspicuous lesions or artifacts caused by motion interference that may exist in a single sample medical image.
  • Feature representation can improve the accuracy of lesion identification when performing lesion identification.
  • the operation of the above fusion processing may include any one of the following: performing a splicing process on the sample content feature representations of multiple sample medical images, and using at least one convolution layer to characterize the sample content feature representations of the multiple sample medical images. extraction, and the final sample content feature representation has the same dimension as the content feature representation of multiple sample medical images. For details, reference may be made to the relevant steps in the foregoing embodiments, which will not be repeated here.
  • sample medical image 1 when using the region segmentation sub-network to perform lesion identification on the sample content feature representation of sample medical image 1, sample medical image 2, sample medical image 3, . . . , sample medical image n, the above sample medical image can be
  • the sample content feature representation of the image is subjected to fusion processing (eg, splicing processing, addition processing, etc., for details, please refer to the aforementioned disclosed embodiments, which will not be repeated here) to obtain the final sample content feature representation.
  • the feature of the sample content indicates that the lesion is identified, and the predicted lesion area in the sample medical image 1, the sample medical image 2, the sample medical image 3, . . . , and the sample medical image n is obtained.
  • Step S24 Adjust the network parameters of the style coding sub-network and the classification processing sub-network by using the difference between the real scanned image category and the predicted scanned image category, and use the difference between the real lesion area and the predicted lesion area to adjust the content coding sub-network and area segmentation Network parameters for the subnet.
  • the real scanned image category and the predicted scanned image category can be used to calculate the first loss value of the style encoding sub-network and the classification processing sub-network, and use the first loss value to adjust the style encoding sub-network and the classification processing sub-network.
  • Network parameters of the network In an implementation scenario, a cross entropy loss (cross entropy loss) or a softmax logistic loss (logistic softmax loss) or the like may be used to calculate the first loss value, which is not limited herein.
  • the actual lesion area and the predicted lesion area may be used to obtain the second loss value of the content coding sub-network and the region segmentation sub-network, and the second loss value may be used to adjust the content coding sub-network and the region segmentation sub-network.
  • Network parameters may be used to calculate a binary cross-entropy loss (binary cross-entropy loss) or a dice coefficient loss may be used to calculate the second loss value, which is not limited herein.
  • the dice coefficient loss is a set similarity measure function, which is usually used to calculate the similarity between two samples (the range is 0 to 1), and can be calculated by formula (1):
  • loss dice represents the second loss value calculated by the loss of dice coefficient
  • X represents the real lesion area
  • Y represents the predicted lesion area
  • X ⁇ Y represents the intersection of the real lesion area and the predicted lesion area.
  • the sample data distribution represented by the sample style feature of each sample medical image can also be obtained, and the difference between the sample data distributions can be used. , and adjust the network parameters of the style coding sub-network, which can help to make the subsequent extracted style feature representations independent of each other, so it can help to improve the accuracy of the identified scanned image categories.
  • the KL divergence can be used to measure the difference between the distributions of the sample data, and use it as the third loss value, so as to constrain the distribution of the style feature representation.
  • KL divergence Kullback–Leibler divergence
  • relative entropy relative entropy
  • ⁇ (X) and ⁇ (X) respectively represent the distribution of sample data represented by two sample style features
  • E ⁇ represents the probability distribution expectation represented by one of the sample style features
  • ⁇ (x) and ⁇ (x) respectively represent the distribution probability of the element x in the two sample style feature representations in the respective sample data distributions.
  • a Gaussian distribution function can be used to obtain the sample data distribution represented by the sample style features, so that through the above training, the style features obtained by the style coding sub-network subsequently extracted can be expressed as a Gaussian distribution with the same center and anisotropy , and thus can help to improve the accuracy of the recognized scanned image category.
  • a sample style feature in order to enable the style encoding sub-network to extract as complete and accurate style features as possible, and the content encoding sub-network to extract as complete and accurate style features as possible, a sample style feature can also be used to represent the same This content feature representation, constructs a reconstructed image corresponding to the sample style feature representation, and uses the difference between the reconstructed image and the corresponding sample style feature representation to the sample medical image to which it belongs, and adjusts the style coding sub-network and the content coding sub-network network parameters, so that the reconstructed image and the corresponding sample medical image can be as identical as possible during the training process, so that the style coding sub-network can extract as complete and accurate style features as possible, and the content coding sub-network can extract as much as possible.
  • the difference between the reconstructed image and the sample medical image to which the corresponding sample style feature representation belongs can be used as the fourth loss. value.
  • the sample style feature representation and the sample content feature representation of each sample medical image can be used to obtain an intra-domain reconstructed image of the corresponding sample medical image, and the difference between each sample medical image and its intra-domain reconstructed image can be used to adjust The network parameters of the style encoding sub-network and the content encoding sub-network, so as to ensure that the decomposed sample content feature representation and sample style feature representation can stably reconstruct themselves and prevent midway variation.
  • sample style feature representation of each sample medical image and the sample content feature representation (or final sample content feature representation) of any other sample medical image can also be used to obtain the cross-domain reconstruction of the corresponding sample medical image image, and use the differences of each sample medical image and its cross-domain reconstructed images to adjust the network parameters of the style coding sub-network and the content coding sub-network, so as to ensure that the extracted content feature representation is the basis for the true coincidence between medical images. feature set.
  • the generator can be used to achieve reconstruction, and the discriminator can be used to identify whether the reconstructed image is a real sample medical image or a reconstructed image, so a generative adversarial network loss (GAN loss) can be used to The loss value of the above-mentioned cross-domain reconstruction is measured, and the L1 norm loss is used to measure the loss value of the above-mentioned intra-domain reconstruction, and details are not repeated here.
  • GAN loss generative adversarial network loss
  • the sample style feature representation 1 corresponding to n, the sample style feature representation 2, the sample style feature representation 3, ..., the sample style feature representation n and the final sample content feature representation are reconstructed to obtain reconstructed image 1, reconstructed image 2, and reconstructed image 3 , ..., reconstruct the image n, so as to realize the intra-domain reconstruction, and use the L1 norm loss to measure the loss value of the above-mentioned intra-domain reconstruction.
  • sample style feature representation 1 corresponding to the sample medical image 1 and the sample content feature representations corresponding to other sample medical images can also be used for cross-domain reconstruction, and the other sample medical images can be deduced by analogy.
  • Generative adversarial loss GAN loss is used to measure the loss value of the above cross-domain reconstruction.
  • the above-mentioned first loss value, second loss value, third loss value and fourth loss value can also be calculated at the same time, and the network parameters of the identification network are adjusted according to these loss values, so as to improve the content coding
  • the acquisition degree of the sub-network for the content features related to the lesions makes the style coding sub-network not respond to the features related to the lesions, improves the robustness of image recognition, and makes the subsequent extracted style feature representations independent of each other, and makes the style
  • the coding sub-network can extract the complete and accurate style feature representation, and the content encoding sub-network can extract the complete and accurate content feature representation, thereby improving the accuracy of image recognition.
  • the training of the content coding sub-network and the region segmentation sub-network is added, so that the lesion identification ability of the region segmentation sub-network can be improved at the same time.
  • improve the acquisition of content features related to lesions by the content coding sub-network which can help make the style coding sub-network not respond to features related to lesions, so that subsequent classification is not affected by features related to lesions, so it can improve image recognition. robustness.
  • FIG. 4 is a schematic frame diagram of an embodiment of an image recognition apparatus 40 of the present application.
  • the image recognition device 40 includes an image acquisition module 41, a style extraction module 42 and a classification processing module 43.
  • the image acquisition module 41 is configured to acquire a plurality of medical images to be recognized;
  • the style extraction module 42 is configured to extract the style of each medical image to be recognized respectively.
  • the classification processing module 43 is configured to perform classification processing on the style feature representations of a plurality of medical images to be recognized, so as to obtain the scanned image category of each medical image to be recognized.
  • the style feature representation of the plurality of medical images to be recognized is classified and processed, so it is possible to consider many
  • the differences in the respective style features of the medical images to be recognized can further improve the accuracy of the recognized scanned image categories, and because the style feature representations of a plurality of medical images to be recognized can be classified and processed, and each of the medical images to be recognized can be obtained. Scanning image categories of medical images, so multiple scanned image categories of medical images to be recognized can be obtained at one time, thereby improving the efficiency of image recognition. Therefore, the above solution can improve the efficiency and accuracy of image recognition.
  • the classification processing module 43 includes a first fusion processing sub-module, which is configured to perform a first fusion processing on the style feature representations of a plurality of medical images to be recognized to obtain a final style feature representation; the classification processing module 43 includes a classification processing The sub-module is configured to perform classification processing on the final style feature representation to obtain the scanned image category of each medical image to be recognized.
  • the style feature representations of multiple medical images to be recognized are classified and processed, the style feature representations of the multiple unidentified medical images are subjected to a first fusion process to obtain the final style feature representation. Therefore, the final style feature representation is The representation can represent the difference between the style feature representation of each to-be-recognized medical image and the style feature representations of other to-be-recognized medical images, so using the final style feature representation for classification processing can improve the accuracy of the recognized scanned image category.
  • the image recognition device 40 further includes at least one of an image exclusion module, an image display module, a first early warning module, a second early warning module and a third early warning module;
  • the image exclusion module is configured to sort a plurality of medical images to be identified according to their scanned image categories; the image display module is configured to display on the same screen at least one medical image to be identified sorted according to the scanned image categories
  • the first warning module is configured to output first warning information to prompt the scanning personnel when there are repetitions of the scanned image categories of the medical images to be identified;
  • the second warning module is configured to display multiple medical images to be identified When the preset scanned image category does not exist in the scanned image category of the medical image to be identified, output the second early warning information to prompt the scanning personnel;
  • the third early warning module is configured to be configured when the classification confidence of the scanned image category of the medical image to be recognized is less than When the reliability threshold is preset, the third warning information is output to remind the scanning personnel.
  • At least one medical image to be recognized is sorted according to its scanning image category, which can improve the convenience of doctor reading; After the image categories are sorted, at least one to-be-recognized medical image is displayed on the same screen, which can avoid back-and-forth comparison of the medical image to be recognized by the doctor, thereby improving the efficiency of the doctor's image reading; there are duplications in the scanned image categories of the to-be-recognized medical image.
  • output the second early warning information to remind the scanning personnel that it can be scanned during the scanning process. Realize image quality control, so that when it is contrary to reality, it can correct errors in time and avoid the second registration of patients.
  • the image recognition apparatus 40 further includes a preprocessing module configured to perform preprocessing on each medical image to be recognized, wherein the preprocessing includes at least one of the following: adjusting the image size of the medical image to be recognized to a predetermined size. Set the size and normalize the image intensity of the medical image to be recognized to a preset range.
  • the image data of each target area is preprocessed, and the preprocessing includes at least one of the following: adjusting the image size of the target area to a preset size, and adjusting the image size of the target area to a preset size.
  • the image intensity is normalized to a preset range, which can help to improve the accuracy of subsequent image recognition.
  • the image recognition device 40 further includes a content extraction module, configured to extract the content feature representation of each medical image to be recognized, respectively; the image recognition device 40 further includes a lesion recognition module, configured to extract a plurality of medical images to be recognized.
  • the content feature of indicates that the lesion is identified, and the lesion area in each medical image to be identified is obtained.
  • the lesion area in each to-be-recognized medical image can be obtained.
  • the lesion area in it is determined, so it can help improve the overall reading performance, and at the same time, it can help to eliminate the interference caused by the lesion to the scanning image category recognition, thereby improving image recognition. accuracy.
  • the lesion identification module includes a second fusion processing sub-module configured to perform a second fusion process on the content feature representations of multiple medical images to be identified to obtain a final content feature representation; the lesion identification module includes a lesion identification submodule , which is configured to perform lesion identification on the final content feature representation to obtain the lesion area in each medical image to be identified.
  • the second fusion processing is performed on the content feature representations of multiple medical images to be recognized to obtain a final content feature representation, which can help to make the final content feature representation compensate for the inconspicuous lesions that may exist in a single to-be-recognized medical image. or artifacts caused by motion interference, so that when using the final content feature representation for lesion identification, the accuracy of lesion identification can be improved.
  • the lesion identification module further includes a lesion prompting sub-module configured to prompt the lesion area of the currently displayed medical image to be identified.
  • the doctor's reading experience can be improved.
  • the second fusion processing sub-module is specifically configured to perform any one of the following: perform splicing processing on the content feature representations of a plurality of medical images to be recognized; and add the content feature representations of a plurality of medical images to be recognized processing; wherein, the final content feature representation has the same dimension as the content feature representation of multiple medical images to be identified.
  • the final content feature representation is obtained by performing a splicing process on the content feature representations of a plurality of medical images to be recognized, or performing an addition process on the content feature representations of a plurality of medical images to be recognized.
  • the final content feature representation has the same dimension as the content feature representation of multiple medical images to be recognized, and the final content feature representation can be obtained in various ways, thereby improving the robustness of image recognition.
  • the style extraction module 42 is specifically configured to use the style coding sub-network of the recognition network to extract the style feature representation of each medical image to be recognized;
  • the classification processing module 43 is specifically configured to use the classification processing sub-network of the recognition network to The style feature representation of a plurality of medical images to be recognized is classified and processed to obtain the scanned image category of each medical image to be recognized;
  • the content extraction module is specifically configured to use the content coding sub-network of the recognition network to extract the content of each medical image to be recognized.
  • the lesion identification module is specifically configured to use the region segmentation sub-network of the identification network to perform lesion identification on the content feature representation of a plurality of medical images to be identified, and obtain the lesion area in each medical image to be identified.
  • the style feature representation of each medical image to be recognized is extracted by the style coding sub-network of the recognition network, and the style feature representation of a plurality of medical images to be recognized is classified and processed by the classification processing sub-network of the recognition network, Obtain the scanned image category of each medical image to be recognized, use the content coding sub-network of the recognition network to extract the content feature representation of each medical image to be recognized, and use the regional segmentation sub-network of the recognition network to identify the content of multiple medical images to be recognized.
  • the feature representation is used to identify lesions, and to obtain the lesion area in each medical image to be recognized
  • the recognition network can be used to perform tasks such as extraction of style feature representation, classification processing, extraction of content feature representation, and lesion recognition, so it can help improve image recognition. s efficiency.
  • the image recognition apparatus 40 further includes a sample acquisition module configured to acquire a plurality of sample medical images, wherein the plurality of sample medical images are marked with their real scanned image categories and real lesion areas; the image recognition apparatus 40 further includes The feature extraction module is configured to use the style coding sub-network to extract the sample style feature representation of each sample medical image respectively, and use the content coding sub-network to separately extract the sample content feature representation of each sample medical image; the image recognition device 40 also includes a recognition The processing module is configured to use the classification processing sub-network to classify and process the sample style feature representation of the multiple sample medical images, obtain the predicted scanning image category of each sample medical image, and use the region segmentation sub-network to classify the multiple sample medical images.
  • the feature of the sample content indicates that the lesion is identified, and the predicted lesion area in each sample medical image is obtained; the image recognition device 40 also includes a first adjustment module configured to utilize the difference between the real scan category and the predicted scan category to adjust the style coding sub-network and The network parameters of the classification processing sub-network, and the network parameters of the content coding sub-network and the region segmentation sub-network are adjusted by using the difference between the real lesion area and the predicted lesion area.
  • the training of the content coding sub-network and the region segmentation sub-network is added, so that the lesion identification ability of the region segmentation sub-network can be improved at the same time.
  • improve the acquisition of content features related to lesions by the content coding sub-network which can help make the style coding sub-network not respond to features related to lesions, so that subsequent classification is not affected by features related to lesions, so it can improve image recognition. robustness.
  • the image recognition apparatus 40 further includes a distribution acquisition module configured to acquire the distribution of sample data represented by the sample style features of each sample medical image; the image recognition apparatus 40 further includes a second adjustment module configured to utilize the samples The difference between the data distributions, adjust the network parameters of the style encoding sub-network.
  • the distribution of sample data is obtained at the same time, and the difference between the distribution of sample data is used to adjust the network parameters of the style coding sub-network, so it can be beneficial to the subsequent extraction of style feature representation. They are independent of each other, which can help to improve the accuracy of the recognized scanned image categories.
  • the image recognition apparatus 40 further includes an image reconstruction module configured to use a sample style feature representation and a content feature representation to construct a reconstructed image corresponding to the sample style feature representation; the image recognition apparatus 40 further includes a third The adjustment module is configured to adjust the network parameters of the style coding sub-network and the content coding sub-network by using the difference between the reconstructed image and the corresponding sample medical image to which the corresponding sample style feature representation belongs.
  • a sample style feature representation and a content feature representation are used to construct a reconstructed image corresponding to the sample style feature representation, and the reconstructed image and the corresponding sample style feature are used to represent the sample to which they belong.
  • the differences between medical images adjust the network parameters of the style coding sub-network and the content coding sub-network, so that the style coding sub-network can extract as complete and accurate style features as possible, and the content coding sub-network can extract the complete and accurate style features as much as possible.
  • Accurate style features can help to improve the classification of subsequent scanned images and the accuracy of lesion identification.
  • the style encoding sub-network includes: sequentially connected downsampling layers and a global pooling layer; and/or, the content encoding sub-network includes any of the following: sequentially connected downsampling layers and residual blocks, sequential Connected convolutional and pooling layers.
  • the style encoding sub-network by setting the style encoding sub-network to include a sequentially connected downsampling layer and a global pooling layer, it can facilitate network training while simplifying the network structure; by setting the content encoding sub-network to include the following: Either: sequentially connected downsampling layers and residual blocks, sequentially connected convolutional layers and pooling layers, can facilitate network training while simplifying the network structure.
  • FIG. 5 is a schematic diagram of a framework of an embodiment of an electronic device 50 of the present application.
  • the electronic device 50 includes a memory 51 and a processor 52 coupled to each other, and the processor 52 is configured to execute program instructions stored in the memory 51 to implement the steps of any of the image recognition method embodiments described above.
  • the electronic device 50 may include, but is not limited to, a microcomputer and a server.
  • the electronic device 50 may also include mobile devices such as a notebook computer and a tablet computer, which are not limited herein.
  • the processor 52 is configured to control itself and the memory 51 to implement the steps of any of the image recognition method embodiments described above.
  • the processor 52 may also be referred to as a central processing unit (Central Processing Unit, CPU).
  • the processor 52 may be an integrated circuit chip with signal processing capability.
  • the processor 52 may also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the processor 52 may be jointly implemented by an integrated circuit chip.
  • the above solution can improve the efficiency and accuracy of image recognition.
  • FIG. 6 is a schematic diagram of a framework of an embodiment of a computer-readable storage medium 60 of the present application.
  • the computer-readable storage medium 60 stores program instructions 601 that can be executed by the processor, and the program instructions 601 are used to implement any of the above-mentioned image recognition methods.
  • the above solution can improve the efficiency and accuracy of image recognition.
  • an embodiment of the present application further provides a computer program, including computer-readable codes, when the computer-readable codes are executed in an electronic device, the processor in the electronic device executes any one of the above-mentioned codes. an image recognition method.
  • the disclosed method and apparatus may be implemented in other manners.
  • the device implementations described above are only illustrative.
  • the division of modules or units is only a logical function division. In actual implementation, there may be other divisions.
  • units or components may be combined or integrated. to another system, or some features can be ignored, or not implemented.
  • the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, which may be in electrical, mechanical or other forms.
  • Units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed over network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this implementation manner.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.
  • the integrated unit if implemented as a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium.
  • the technical solutions of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, and the computer software products are stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the methods of the various embodiments of the present application.
  • the aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .
  • the embodiments of the present application disclose an image recognition method and device, electronic equipment, computer storage medium and computer program.
  • the image recognition method includes: acquiring a plurality of medical images to be recognized; separately extracting style features of each of the medical images to be recognized Representation; classifying and processing the style feature representations of the plurality of medical images to be recognized, to obtain a scanned image category of each of the medical images to be recognized.
  • the above solution can improve the efficiency and accuracy of image recognition.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Image Analysis (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本申请实施例公开了一种图像识别方法及装置、电子设备、计算机存储介质和计算机程序，图像识别方法包括：获取多个待识别医学图像；分别提取每一所述待识别医学图像的风格特征表示；对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别。

Description

图像识别方法及装置、设备、存储介质和程序
相关申请的交叉引用
本申请基于申请号为202011018760.7、申请日为2020年9月24日、名称为“图像识别方法及相关装置、设备”的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本公开涉及人工智能技术领域，涉及但不限于一种图像识别方法及装置、电子设备、计算机存储介质和计算机程序。
背景技术
计算机断层扫描(Computed Tomography，CT)和核磁共振扫描(Magnetic Resonance Imaging，MRI)等医学图像在临床具有重要意义。为了使医学图像应用于临床，一般需要扫描得到至少一种扫描图像类别的医学图像。以与肝脏相关的临床为例，扫描图像类别往往包括与时序有关的造影前平扫、动脉早期、动脉晚期、门脉期、延迟期等等，此外，扫描图像类别还可以包含与扫描参数有关的T1加权反相成像、T1加权同相成像、T2加权成像、扩散加权成像、表面扩散系数成像等等。
在扫描过程中，通常需要放射科医师鉴别扫描得到的医学图像的扫描图像类别，以确保获取所需要的医学图像；或者，在住院或门诊诊疗时，通常需要医生对扫描得到的医学图像进行识别，判断每一医学图像的扫描图像类别，再进行阅片。
发明内容
本申请实施例提供一种图像识别方法及装置、电子设备、计算机存储介质和计算机程序。
本申请实施例提供了一种图像识别方法，包括：获取多个待识别医学图像；分别提取每一待识别医学图像的风格特征表示；对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
因此，通过获取多个待识别医学图像，并提取每一待识别医学图像的风格特征表示，从而对多个待识别医学图像的风格特征表示进行分类处理，故能够在分类处理时，考虑多个待识别医学图像在各自风格特征上的差异，进而能够提高识别得到的扫描图像类别的准确性，且由于能够对多个待识别医学图像的风格特征表示进行分类处理，并得到每一待识别医学图像的扫描图像类别，故能够一次得到多个待识别医学图像的扫描图像类别，从而能够提高图像识别的效率，故此，上述方案能够提高图像识别的效率和准确性。
本申请的一些实施例中，对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别包括：将多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示；对最终风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
因此，在对多个待识别医学图像的风格特征表示进行分类处理时，将多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示，故最终风格特征表示能够表示每一待识别医学图像的风格特征表示与其他待识别医学图像的风格特征表示之间的差异，故利用最终风格特征表示进行分类处理能够提高识别得到的扫描图像类别的准确性。
本申请的一些实施例中，对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别之后，图像识别方法还包括以下至少一者：将多 个待识别医学图像按照其扫描图像类别进行排序；将按照扫描图像类别进行排序后的至少一个待识别医学图像进行同屏显示；若待识别医学图像的扫描图像类别存在重复，则输出第一预警信息，以提示扫描人员；若多个待识别医学图像的扫描图像类别中不存在预设扫描图像类别，则输出第二预警信息，以提示扫描人员；若待识别医学图像的扫描图像类别的分类置信度小于预设置信度阈值，则输出第三预警信息，以提示扫描人员。
因此，在确定得到每一待识别医学图像所属的扫描图像类别之后，执行将至少一个待识别医学图像按照其扫描图像类别进行排序，能够提高医生阅片的便捷性；将按照扫描图像类别进行排序后的至少一个待识别医学图像进行同屏显示，能够免去医生翻阅待识别医学图像时来回对照，从而能够提高医生阅片的效率；在待识别医学图像的扫描图像类别存在重复时，输出第一预警信息，以提示扫描人员，在至少一个待识别医学图像的扫描图像类别中不存在预设扫描图像类别时，输出第二预警信息，以提示扫描人员，在待识别医学图像的扫描图像类别的分类置信度小于预设置信度阈值时，输出第三预警信息，以提示扫描人员，能够在扫描过程中实现图像质控，以在与实际相悖时，能够及时纠错，避免病人二次挂号。
本申请的一些实施例中，分别提取每一待识别医学图像的风格特征表示之前，上述方法还包括：对每一待识别医学图像进行预处理，其中，预处理包括以下至少一种：将待识别医学图像的图像尺寸调整至预设尺寸，将待识别医学图像的图像强度归一化至预设范围。
因此，在提取风格特征表示之前，对每一目标区域的图像数据进行预处理，且预处理包括以下至少一种：将目标区域的图像尺寸调整至预设尺寸，将目标区域的图像强度归一化至预设范围，故能够有利于提高后续图像识别的准确性。
本申请的一些实施例中，图像识别方法还包括：分别提取每一待识别医学图像的内容特征表示；对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
因此，通过提取每一待识别医学图像的内容特征表示，并对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，能够在得到每一待识别医学图像的扫描图像类别的同时，确定其中的病灶区域，故能够有利于提高整体阅片效能，同时能够有利于消除病灶对扫描图像类别识别带来的干扰，从而能够提高图像识别的准确性。
本申请的一些实施例中，对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域包括：将多个待识别医学图像的内容特征表示进行第二融合处理，得到最终内容特征表示；对最终内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
因此，将多个待识别医学图像的内容特征表示进行第二融合处理，得到最终内容特征表示，能够有利于使最终内容特征表示补偿单一待识别医学图像中可能存在的病灶不明显或运动干扰产生的伪影等问题，从而在利用最终内容特征表示进行病灶识别时，能够提高病灶识别的准确性。
本申请的一些实施例中，图像识别方法还包括：提示当前显示的待识别医学图像的病灶区域。
因此，通过提示当前显示的待识别医学图像的病灶区域，能够提升医生阅片体验。
本申请的一些实施例中，将所述多个待识别医学图像的内容特征表示进行第二融合处理，包括以下任一者：将多个待识别医学图像的内容特征表示进行拼接处理；将多个待识别医学图像的内容特征表示进行相加处理；其中，最终内容特征表示和多个待识别医学图像的内容特征表示的维度相同。
因此，通过将多个待识别医学图像的内容特征表示进行拼接处理，或者将多个待识 别医学图像的内容特征表示进行相加处理中的任一者，得到最终内容特征表示，且最终内容特征表示和多个待识别医学图像的内容特征表示的维度相同，能够通过多种方式得到最终内容特征表示，从而能够提高图像识别的鲁棒性。
本申请的一些实施例中，分别提取每一待识别医学图像的风格特征表示，包括：利用识别网络的风格编码子网络分别提取每一待识别医学图像的风格特征表示；对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别，包括：利用识别网络的分类处理子网络对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别；分别提取每一待识别医学图像的内容特征表示，包括：利用识别网络的内容编码子网络分别提取每一待识别医学图像的内容特征表示；对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，包括：利用识别网络的区域分割子网络对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
因此，利用识别网络的风格编码子网络分别提取每一待识别医学图像的风格特征表示，利用识别网络的分类处理子网络对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别，利用识别网络的内容编码子网络分别提取每一待识别医学图像的内容特征表示，利用识别网络的区域分割子网络对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，能够利用识别网络执行风格特征表示的提取、分类处理、内容特征表示的提取以及病灶识别等任务，故能够有利于提高图像识别的效率。
本申请的一些实施例中，分别提取每一待识别医学图像的风格特征表示之前，图像识别方法还包括：获取多个样本医学图像，其中，多个样本医学图像标注有其真实扫描图像类别和真实病灶区域；利用风格编码子网络分别提取每一样本医学图像的样本风格特征表示，并利用内容编码子网络分别提取每一样本医学图像的样本内容特征表示；利用分类处理子网络对多个样本医学图像的样本风格特征表示进行分类处理，得到每一样本医学图像的预测扫描图像类别，并利用区域分割子网络对多个样本医学图像的样本内容特征表示进行病灶识别，得到每一样本医学图像中的预测病灶区域；利用真实扫描图像类别和预测扫描图像类别的差异，调整风格编码子网络和分类处理子网络的网络参数，以及利用真实病灶区域和预测病灶区域的差异，调整内容编码子网络和区域分割子网络的网络参数。
因此，在对风格编码子网络和分类处理子网络进行训练的同时，加入对内容编码子网络和区域分割子网络的训练，从而能够在提高区域分割子网络的病灶识别能力的同时，提高内容编码子网络对于病灶相关的内容特征的获取度，进而能够有利于使得风格编码子网络不响应与病灶相关的特征，使后续分类时不受病灶相关特征的影响，故能够提高图像识别的鲁棒性。
本申请的一些实施例中，图像识别方法还包括：获取每一样本医学图像的样本风格特征表示的样本数据分布情况；利用样本数据分布情况之间的差异，调整风格编码子网络的网络参数。
因此，在训练过程中，同时获取样本数据分布情况，并利用样本数据分布情况之间的差异，调整风格编码子网络的网络参数，故能够有利于使后续提取到风格特征表示之间相互独立，从而能够有利于提高识别到的扫描图像类别的准确性。
本申请的一些实施例中，图像识别方法还包括：利用一样本风格特征表示和一内容特征表示，构建得到与样本风格特征表示对应的重建图像；利用重建图像与对应的样本风格特征表示所属的样本医学图像之间的差异，调整风格编码子网络和内容编码子网络的网络参数。
因此，在训练过程中，同时利用一样本风格特征表示和一内容特征表示，构建得到 与样本风格特征表示对应的重建图像，并利用重建图像与对应的样本风格特征表示所属的样本医学图像之间的差异，调整风格编码子网络和内容编码子网络的网络参数，从而能够使风格编码子网络尽可能地提取到完整准确的风格特征，而内容编码子网络尽可能地提取到完整准确的风格特征，进而能够有利于提高后续扫描图像类别以及病灶识别的准确性。
本申请的一些实施例中，风格编码子网络包括：顺序连接的下采样层和全局池化层；和/或，内容编码子网络包括以下任一者：顺序连接的下采样层和残差块、顺序连接的卷积层和池化层。
因此，通过将风格编码子网络设置为包括顺序连接的下采样层和全局池化层，能够有利于在简化网络结构的同时便于网络训练；通过将内容编码子网络设置为包括以下任一者：顺序连接的下采样层和残差块、顺序连接的卷积层和池化层，能够有利于在简化网络结构的同时便于网络训练。
本申请实施例还提供了一种图像识别装置，包括图像获取模块、风格提取模块和分类处理模块，图像获取模块配置为获取多个待识别医学图像；风格提取模块配置为分别提取每一待识别医学图像的风格特征表示；分类处理模块配置为对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
本申请实施例还提供了一种电子设备，包括相互耦接的存储器和处理器，处理器配置为执行存储器中存储的程序指令，以实现上述任意一种图像识别方法。
本申请实施例还提供了一种计算机可读存储介质，其上存储有程序指令，程序指令被处理器执行时实现上述任意一种图像识别方法。
本申请实施例还提供了一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现上述任意一种图像识别方法。
本申请实施例中，通过获取多个待识别医学图像，并提取每一待识别医学图像的风格特征表示，从而对多个待识别医学图像的风格特征表示进行分类处理，故能够在分类处理时，考虑多个待识别医学图像在各自风格特征上的差异，进而能够提高识别得到的扫描图像类别的准确性，且由于能够对多个待识别医学图像的风格特征表示进行分类处理，并得到每一待识别医学图像的扫描图像类别，故能够一次得到多个待识别医学图像的扫描图像类别，从而能够提高图像识别的效率，故此，本申请实施例能够提高图像识别的效率和准确性。
附图说明
图1是本申请图像识别方法一实施例的流程示意图；
图2是训练识别网络一实施例的流程示意图；
图3是训练识别网络一实施例的状态示意图；
图4是本申请图像识别装置一实施例的框架示意图；
图5是本申请电子设备一实施例的框架示意图；
图6是本申请计算机可读存储介质一实施例的框架示意图。
具体实施方式
下面结合说明书附图，对本申请实施例的方案进行详细说明。
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、接口、技术之类的具体细节，以便透彻理解本申请。
本文中术语“系统”和“网络”在本文中常被可互换使用。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，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。此外，本文中的“多”表示两个或者多于两个。
请参阅图1，图1是本申请图像识别方法一实施例的流程示意图。具体而言，可以包括如下步骤：
步骤S11：获取多个待识别医学图像。
待识别医学图像可以包括CT图像、MR图像，在此不做限定。在一个实施场景中，待识别医学图像可以是对腹部、胸部等区域进行扫描得到的，具体可以根据实际应用情况而设置，在此不做限定。例如，当肝脏、脾脏、肾脏为需要诊疗的脏器时，可以对腹部进行扫描，得到待识别医学图像；或者，当心脏、肺为需要诊疗的脏器时，可以对胸部进行扫描，得到待识别医学图像，其他情况可以以此类推，在此不再一一举例。在另一个实施场景中，扫描方式可以是平扫、增强扫描等方式，在此不做限定。在又一个实施场景中，待识别医学图像可以是三维图像，在此不做限定。在又一实施场景中，多个待识别医学图像可以对同一对象扫描得到的。
步骤S12：分别提取每一待识别医学图像的风格特征表示。
风格特征表示用于描述待识别医学图像的风格，例如，待识别医学图像中所表示的注射进血管中的造影剂的强化程度。以肝脏为例，不同扫描图像类别的待识别医学图像中肝脏的静脉、门脉等血管造影剂强化程度各不相同，其中，扫描图像类别为动脉早期的待识别医学图像中门静脉尚未增强，扫描图像类别为动脉晚期的待识别医学图像中门静脉已被增强，扫描图像类别为门脉期的待识别医学图像中门静脉已充分增强且肝脏血管已被前向性血流增强、肝脏软细胞组织在标记物下已达到峰值，扫描图像类别为延迟期的待识别医学图像中门脉和动脉处于增强状态并弱于门脉期、且肝脏软细胞组织处于增强状态并弱于门脉期，其他扫描图像类别在此不再一一举例。
在一个实施场景中，风格特征表示可以由一矢量表示，矢量的大小可以根据实际情况进行设置，例如，可以将矢量的大小设置为8比特，在此不做限定。
在一个实施场景中，为了提高风格特征表示提取的便利性，可以预先训练一识别网络，且识别网络中包含一风格编码子网络，并利用识别网络的风格编码子网络分别提取每一待识别医学图像的风格特征表示。在一个实施场景中，为了简化网络结构，风格编码子网络可以包括顺序连接的下采样层和全局池化层，从而待识别医学图像经过下采样处理之后，再利用全局池化层进行池化处理，得到风格特征表示。
在一个实施场景中，为了提高后续图像识别的准确性，在对待识别医学图像的风格特征表示进行提取之前，还可以对每一待识别医学图像进行预处理，示例性地，预处理可以包括将待识别医学图像的图像尺寸调整至预设尺寸(例如，32*256*256)。或者，预处理还可以包括将待识别医学图像的图像强度归一化至预设范围(例如，0至1的范围)，在一个实施场景中，可以采用灰度累积分布函数下预设比例(例如，99.9％)对应的灰度值作为归一化的钳位值，从而能够加强待识别医学图像的对比度，有利于提升后续图像识别的准确性。
步骤S13：对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
扫描图像类别具体可以根据实际情况进行设置。例如，仍以对肝脏进行扫描得到的待识别医学图像为例，扫描图像类别可以包括与时序有关的造影前平扫、动脉早期、动脉晚期、门脉期、延迟期；或者，扫描图像类别还可以包括与描参数有关的T1加权反相成像、T1加权同相成像、T2加权成像、扩散加权成像、表面扩散系数成像。示例性地，动脉早期可以表示门静脉尚未增强，动脉晚期可以表示门静脉已被增强，门脉期可以表示门静脉已充分增强且肝脏血管已被前向性血流增强、肝脏软细胞组织在标记物下已达到峰值，延迟期可以表示门脉和动脉处于增强状态并弱于门脉期、且肝脏软细胞组织处 于增强状态并弱于门脉期，其他扫描图像类别在此不再一一举例。当待识别医学图像为对其他脏器扫描得到的医学图像时，可以以此类推，在此不再一一举例。
由于属于不同扫描图像类别的待识别医学图像各自的风格特征表示存在差异，故对其各自的风格特征表示进行分类处理，能够得到每一待识别医学图像的扫描图像类别。在一个实施场景中，为了提高分类处理的便利性，上述识别网络中还包括分类处理子网络，从而可以利用分类处理子网络对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别，进而只需分类处理子网络对风格特征表示进行分类处理即可得到待识别医学图像分别所属的扫描图像类别，故此能够提高分类处理的便利性。在一个实施场景中，为了简化网络结构，分类处理子网络可以包括顺序连接的全连接层和softmax层，在此不做限定。
在一个实施场景中，为了提高图像识别准确性，还可以将多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示，并对最终风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。示例性地，第一融合处理的操作可以是将多个待识别医学图像的风格特征表示进行首尾拼接，从而得到最终风格特征表示；或者，第一融合处理的操作还可以是将多个待识别医学图像的风格特征表示进行堆叠处理，从而得到最终风格特征表示，在此不做限定。通过将多个待识别医学图像的风格特征表示进行第一融合处理而得到的最终风格特征表示，能够表示每一待识别医学图像的风格特征表示与其他待识别医学图像的风格特征表示之间的差异，故利用最终风格特征表示进行分类处理能够提高识别得到的扫描图像类别的准确性。
在一个实施场景中，为了提高阅片效能，以及扫描图像类别识别的准确性，还可以提取每一待识别医学图像的内容特征表示，并对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，故能够在得到每一待识别医学图像的扫描图像类别的同时，确定其中的病灶区域，故能够有利于提高整体阅片效能，同时能够有利于消除病灶对扫描图像类别识别带来的干扰，从而能够提高图像识别的准确性。内容特征表示待识别医学图像中的内容，例如，待识别医学图像中器脏的解剖特征。仍以肝脏为例，内容特征表示可以描述肝脏及其毗邻器脏(如，脾脏、肾脏)之间的生理位置关系、肝脏的外形特征、质地(如，软、硬等)特征、成分(如，含水、含脂等)特征等等，在此不做限定。此外，病灶可以包括肿瘤、血栓、结节等，具体可以根据实际情况进行设置，在此不做限定。
在一个实施场景中，为了提高提取内容特征表示的便利性，识别网络中还可以包括一内容编码子网络，从而可以利用识别网络的内容编码子网络分别提取每一待识别医学图像的内容特征表示。示例性地，为了简化网络结构，内容编码子网络可以采用顺序连接的下采样层和残差块(resblock)，残差块的数量可以根据实际情况进行设置，通过设置残差块可以提升网络深度，从而可以提高提取得到的内容特征表示的深度；或者，内容编码子网络还可以采用顺序连接的卷积层和池化层，卷积层和池化层的组数可以根据实际情况进行设置，例如，可以采用一组顺序连接的卷积层和池化层、两组顺序连接的卷积层和池化层、三组顺序连接的卷积层和池化层等等，在此不做限定。
在另一个实施场景中，为了提高病灶识别的便利性，识别网络中还可以包括一区域分割子网络，从而利用识别网络的区域分割子网络对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。示例性地，区域分割子网络可以采用Unet、Vnet等，在此不做限定。此外，病灶区域可以包括包围病灶的区域，例如，病灶的轮廓等等，在此不做限定。
在又一个实施场景中，内容特征表示可以由一张量来表示，例如，可以通过一低分辨率的张量来表示内容特征表示，在此不做限定。
在又一个实施场景中，为了提高病灶识别的准确性，还可以将多个待识别医学图像 的内容特征表示进行第二融合处理，得到最终内容特征表示，并对最终内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，故能够有利于使最终内容特征表示补偿单一待识别医学图像中可能存在的病灶不明显或运动干扰产生的伪影等问题，从而在利用最终内容特征表示进行病灶识别时，能够提高病灶识别的准确性。示例性地，第二融合处理具体可以是将多个待识别医学图像的内容特征表示进行拼接(concatenate)处理以实现对内容特征表示的融合处理，在一些实施例中，在进行拼接处理后，可以经过几个简单的没有池化的卷积层得到最终内容特征表示；这里，对内容特征进行拼接处理可以视为张量的串联运算；或者，也可以是将多个待识别医学图像的内容特征表示进行相加(add)处理以实现对内容特征表示的融合处理，具体可以视为张量的求和运算；或者，还可以将多个待识别医学图像的内容特征表示进行堆叠，再利用诸如1*1的卷积核对堆叠后的内容特征表示进行卷积运算，以实现对内容特征表示的融合处理；或者，还可以将多个待识别医学图像的内容特征表示进行加权处理，在此不做限定。此外，最终内容特征表示和多个待识别医学图像的内容特征表示的维度相同。
在又一个实施场景中，为了提升医生体验，还可以在得到待识别医学图像中的病灶区域之后，提示当前显示的待识别医学图像的病灶区域。例如，可以采用预设线条(如，加粗线、点划线、双实线等)和/或预设颜色(如，黄色、红色、绿色等)表示病灶区域；或者，还可以采用预设符号(例如，指向病灶区域的箭头等)表示病灶区域，具体可以根据实际情况进行设置，在此不做限定。
在又一个实施场景中，可以将上述经训练的识别网络设置于影像后处理工作站、摄片工作站、计算机辅助阅片系统、远程医疗诊断场景，云平台辅助智能诊断场景等，从而能够实现对待识别医学图像的自动识别，提高识别效率。
在一个实施场景中，至少一个待识别医学图像为对同一对象扫描得到的，故为了便于医生阅片，在得到每一待识别医学图像所属的扫描图像类别之后，还可以将至少一个待识别医学图像按照其扫描图像类别进行排序，例如，可以按照T1加权反相成像、T1加权同相成像、造影前平扫、动脉早期、动脉晚期、门脉期、延迟期、T2加权成像、扩散加权成像、表面扩散系数成像的预设顺序进行排序，此外，预设顺序还可以根据医生习惯进行设置，在此不做限定，从而能够提高医生阅片的便捷性。
在另一个实施场景中，为了进一步提高阅片的便捷性，还可以将按照扫描图像类别进行排序后的至少一个待识别医学图像进行同屏显示，例如，待识别医学图像的数量为5个，则可以在5个显示窗口中分别显示待识别医学图像。故此，能够降低医生翻阅多个待识别医学图像来回对照的时间，提升阅片效率。
在又一个实施场景中，至少一个待识别医学图像为对同一对象扫描得到的，故为了在扫描过程中进行质量控制，在得到每一待识别医学图像所属的扫描图像类别之后，还可以判断待识别医学图像的扫描图像类别是否存在重复，并在存在重复时，输出第一预警信息，以提示扫描人员。例如，若存在两张扫描图像类别均为“延迟期”的待识别医学图像，则可以认为扫描过程中存在扫描质量不合规的情况，故为了提示扫描人员，可以输出第一预警信息，示例性地，可以输出预警原因(如，存在扫描图像类别重复的待识别医学图像等)。或者，在得到每一待识别医学图像所属的扫描图像类别之后，还可以判断至少一个待识别医学图像的扫描图像类别中不存在预设扫描图像类别，并在不存在预设扫描图像类别时，输出第二预警信息，以提示扫描人员。例如，预设扫描图像类别为“门脉期”，若至少一个待识别医学图像中不存在扫描图像类别为“门脉期”的图像，则可以认为扫描过程中存在扫描质量不合规的情况，故为了提示扫描人员，可以输出第二预警信息，示例性地，可以输出预警原因(如，待识别医学图像中不存在门脉期图像等)。或者，在得到每一待识别医学图像所属的扫描图像类别之后，还可以判断待识别医学图像的扫描图像类别的分类置信度是否小于预设置信度阈值，并在待识别医学图像的扫描图像类 别的分类置信度小于预设置信度阈值时，输出第三预警信息，以提示扫描人员。示例性地，分类置信度可以是分类处理子网络在进行分类处理时预测得到的。例如，分类处理子网络在进行分类处理时，预测得到每个待识别医学图像的扫描图像类别和对应的分类置信度，若扫描图像类别为“门脉期”的待识别医学图像的分类置信度(如，20％)低于预设置信度阈值(如，90％)，则可以认为扫描图像类别为“门脉期”的待识别医学图像的图像质量不合规，故为了提示扫描人员，可以输出第三预警信息，示例性地，可以输出预警原因(如，待识别医学图像可能存在扫描质量不佳的问题)。故此，能够在扫描过程中实现图像质控，以在与实际相悖时，能够及时纠错，避免病人二次挂号。
上述方案，通过获取多个待识别医学图像，并提取每一待识别医学图像的风格特征表示，从而对多个待识别医学图像的风格特征表示进行分类处理，故能够在分类处理时，考虑多个待识别医学图像在各自风格特征上的差异，进而能够提高识别得到的扫描图像类别的准确性，且由于能够对多个待识别医学图像的风格特征表示进行分类处理，并得到每一待识别医学图像的扫描图像类别，故能够一次得到多个待识别医学图像的扫描图像类别，从而能够提高图像识别的效率，故此，上述方案能够提高图像识别的效率和准确性。
请参阅图2，图2是训练识别网络一实施例的流程示意图。具体而言，识别网络包括前述实施例中的内容编码子网络、风格编码子网络、分类处理子网络和区域分割子网络，具体过程如下：
步骤S21：获取多个样本医学图像，其中，多个样本医学图像标注有其真实扫描图像类别和真实病灶区域。
一次训练过程中所使用的多个样本医学图像可以是对同一对象扫描得到的。例如，某次训练所采用的样本医学图像可以是对对象A扫描得到的，另一次训练所采用的样本医学图像可以是对对象B扫描得到的。
样本医学图像也可以包括CT图像、MR图像，在此不做限定。具体可以参阅前述实施例中的待识别医学图像，在此不再赘述。
样本医学图像所标注的真实扫描图像类别和真实病灶区域可以是临床医生、影像科医生等具有医学影像知识的人员标注的。扫描图像类别具体可以根据实际情况进行设置，例如，样本医学图像是对肝脏进行扫描得到的，则扫描图像类别具体可以包括与时序有关的造影前平扫、动脉早期、动脉晚期、门脉期、延迟期；或者，扫描图像类别还可以包括与描参数有关的T1加权反相成像、T1加权同相成像、T2加权成像、扩散加权成像、表面扩散系数成像，具体可以参阅前述实施例中的相关步骤，在此不再赘述。当样本医学图像是对其他脏器扫描得到时，可以以此类推，在此不再一一举例。真实病灶区域可以采用多边形进行标注，例如，可以采用多边形标注病灶轮廓等，在此不做限定。
请结合参阅图3，图3是训练识别网络一实施例的状态示意图。如图3所示，在一次训练过程中，可以获取样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n，其中，n的数值可以根据实际情况进行设置，例如，当扫描图像类别包括与时序有关的造影前平扫、动脉早期、动脉晚期、门脉期、延迟期时，n的数值可以设置为小于或等于5的整数，例如，5、4、3等等；或者，当扫描图像类别包括与描参数有关的T1加权反相成像、T1加权同相成像、T2加权成像、扩散加权成像、表面扩散系数成像时，n的数值可以设置为小于或等于5的整数，例如，5、4、3等等；或者，当扫描图像类别既包括与描参数有关的T1加权反相成像、T1加权同相成像、T2加权成像、扩散加权成像、表面扩散系数成像，也包括与时序有关的造影前平扫、动脉早期、动脉晚期、门脉期、延迟期时，n的数值可以设置为小于或等于10的整数，例如，10、9、8等等，具体可以根据实际情况进行设置，在此不做限定。
步骤S22：利用风格编码子网络分别提取每一样本医学图像的样本风格特征表示， 并利用内容编码子网络分别提取每一样本医学图像的样本内容特征表示。
风格编码子网络和内容编码子网络的网络结构具体可以参阅前述实施例中的相关步骤，在此不再赘述。请结合参阅图3，样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n在经过风格编码子网络进行风格特征提取之后，可以分别得到样本风格特征表示1、样本风格特征表示2、样本风格特征表示3、……、样本风格特征表示n；经过内容编码子网络进行内容特征提取之后，可以分别得到样本内容特征表示1、样本内容特征表示2、样本内容特征表示3、……、样本内容特征表示n。样本风格特征表示和样本内容特征表示具体可以参考前述实施例中的风格特征表示、内容特征表示，在此不再赘述。
步骤S23：利用分类处理子网络对多个样本医学图像的样本风格特征表示进行分类处理，得到每一样本医学图像的预测扫描图像类别，并利用区域分割子网络对多个样本医学图像的样本内容特征表示进行病灶识别，得到每一样本医学图像中的预测病灶区域。
在一个实施场景中，可以对多个样本医学图像的样本风格特征表示进行拼接处理，得到最终样本风格特征表示，并利用分类处理子网络对最终样本风格特征表示进行分类处理，得到每一样本医学图像的预测扫描图像类别，从而最终样本风格特征表示能够表示每一样本医学图像的样本风格特征表示与其他样本医学图像的样本风格特征表示之间的差异，故利用分类处理子网络对最终样本风格特征表示进行分类处理时，能够提高分类处理的准确性。
如图3所示，在利用分类处理子网络对样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n的样本风格特征表示进行分类处理时，可以将上述样本医学图像的样本风格特征进行拼接处理(或堆叠处理等其他处理方式，具体可以参阅前述公开实施例，在此不再赘述)，得到最终样本风格特征表示，从而利用分类处理子网络对最终样本风格特征表示进行分类处理，得到样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n的预测扫描图像类别。
在一个实施场景中，可以将多个样本医学图像的样本内容特征表示进行融合处理，得到最终样本内容特征表示，并利用区域分割子网络对最终样本内容特征表示进行病灶识别，得到每一样本医学图像中的预测病灶区域，从而能够有利于使最终样本内容特征表示补偿单一样本医学图像中可能存在的病灶不明显或运动干扰产生的伪影等问题，从而在利用区域分割子网络对最终样本内容特征表示进行病灶识别时，能够提高病灶识别的准确性。示例性地，上述融合处理的操作可以包括以下任一者：将多个样本医学图像的样本内容特征表示进行拼接处理，利用至少一个卷积层对多个样本医学图像的样本内容特征表示进行特征提取，且最终样本内容特征表示和多个样本医学图像的内容特征表示的维度相同。具体可以参阅前述实施例中的相关步骤，在此不再赘述。
如图3所示，在利用区域分割子网络对样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n的样本内容特征表示进行病灶识别时，可以将上述样本医学图像的样本内容特征表示进行融合处理(如，拼接处理、相加处理等，具体可以参阅前述公开实施例，在此不再赘述)，得到最终样本内容特征表示，从而利用区域分割子网络对最终样本内容特征表示进行病灶识别，得到样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n中的预测病灶区域。
步骤S24：利用真实扫描图像类别和预测扫描图像类别的差异，调整风格编码子网络和分类处理子网络的网络参数，以及利用真实病灶区域和预测病灶区域的差异，调整内容编码子网络和区域分割子网络的网络参数。
在一个实施场景中，可以利用真实扫描图像类别和预测扫描图像类别，计算得到风格编码子网络和分类处理子网络的第一损失值，并利用第一损失值调整风格编码子网络和分类处理子网络的网络参数。在一个实施场景中，可以采用交叉熵损失(cross entropy  loss)或者softmax逻辑损失(logistic softmax loss)等来计算第一损失值，在此不做限定。
在一个实施场景中，可以利用真实病灶区域和预测病灶区域，计算得到内容编码子网络和区域分割子网络的第二损失值，并利用第二损失值调整内容编码子网络和区域分割子网络的网络参数。在一个实施场景中，可以采用二分类交叉熵损失(binary cross-entropy loss)或者dice系数损失来计算第二损失值，在此不做限定。示例性地，dice系数损失是一种集合相似度度量函数，通常用于计算两个样本的相似度(范围为0～1)，具体可以采用公式(1)进行计算：
Figure PCTCN2021106479-appb-000001
上式中，loss dice表示以dice系数损失计算得到的第二损失值，X表示真实病灶区域，Y表示预测病灶区域，X∩Y表示真实病灶区域与预测病灶区域的交集。
在一个实施场景中，为了提高扫描图像类别识别的准确性，还可以在训练过程中，获取每一样本医学图像的样本风格特征表示的样本数据分布情况，并利用样本数据分布情况之间的差异，调整风格编码子网络的网络参数，进而能够有利于使得后续提取到的风格特征表示之间相互独立，故能够有利于提高识别到的扫描图像类别的准确性。示例性地，可以在训练过程中，采用KL散度度量样本数据分布情况之间的差异，并将其作为第三损失值，以此来约束风格特征表示的分布。KL散度(Kullback–Leibler divergence)，又称相对熵(relative entropy)，是描述两个概率分布P和Q差异的一种方法，具体可以采用公式(2)进行计算：
Figure PCTCN2021106479-appb-000002
上式中，μ(X)和λ(X)分别表示两个样本风格特征表示的样本数据分布情况，E μ表示其中一个样本风格特征表示的概率分布期望，μ(x)和λ(x)分别表示两个样本风格特征表示中的元素x在各自样本数据分布情况中的分布概率。在一个实施场景中，可以采用高斯分布函数获取样本风格特征表示的样本数据分布情况，从而通过上述训练，能够使风格编码子网络后续提取得到的风格特征表示为中心相同且各向异性的高斯分布，进而故能够有利于提高识别到的扫描图像类别的准确性。
在一个实施场景中，为了能够使风格编码子网络尽可能地提取到完整准确的风格特征，而内容编码子网络尽可能地提取到完整准确的风格特征，还可以利用一样本风格特征表示和一样本内容特征表示，构建得到与样本风格特征表示对应的重建图像，并利用重建图像与对应的样本风格特征表示所属的样本医学图像之间的差异，调整风格编码子网络和内容编码子网络的网络参数，从而能够在训练过程中使重建图像与对应的样本医学图像尽可能地相同，进而能够使风格编码子网络尽可能地提取到完整准确的风格特征，而内容编码子网络尽可能地提取到完整准确的风格特征，故此能够有利于提高后续扫描图像类别以及病灶识别的准确性，示例性地，可以将重建图像与对应的样本风格特征表示所属的样本医学图像之间的差异作为第四损失值。在一个实施场景中，可以利用每一样本医学图像的样本风格特征表示和样本内容特征表示，得到对应样本医学图像的域内重建图像，并利用每一样本医学图像及其域内重建图像的差异，调整风格编码子网络和内容编码子网络的网络参数，从而确保分解得到的样本内容特征表示和样本风格特征表示能够稳定重建其自身，防止中途变异。在另一个实施场景中，还可以利用每一样本医学图像的样本风格特征表示和任一其他样本医学图像的样本内容特征表示(或最终样本内容特征表示)，得到对应样本医学图像的跨域重建图像，并利用每一样本医学图像及其跨域重建图像的差异，调整风格编码子网络和内容编码子网络的网络参数，从而确保提取得到的内容特征表示是医学图像之间真正彼此重合的基础特征集。示例性地，在训练 过程中，可以利用生成器实现重建，利用鉴别器鉴定重建图像是真实的样本医学图像还是重建得到的图像，故可以采用生成对抗损失(Generative Adversarial Networks loss，GAN loss)来度量上述跨域重建的损失值，以及L1范数损失来度量上述域内重建的损失值，具体在此不再赘述。
如图3所示，在进行图像重建时，可以将样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n分别对应的样本内容特征表示1、样本内容特征表示2、样本内容特征表示3、……、样本内容特征表示n进行融合处理，得到最终样本内容特征表示，并分别将样本医学图像1、样本医学图像2、样本医学图像3、……、样本医学图像n对应的样本风格特征表示1、样本风格特征表示2、样本风格特征表示3、……、样本风格特征表示n与最终样本内容特征表示进行重建，得到重建图像1、重建图像2、重建图像3、……、重建图像n，从而实现域内重建，并采用L1范数损失来度量上述域内重建的损失值。此外，还可以采用样本医学图像1对应的样本风格特征表示1与其他样本医学图像(即样本医学图像2～n)对应的样本内容特征表示进行跨域重建，其他样本医学图像以此类推，并采用生成对抗损失(GAN loss)来度量上述跨域重建的损失值。
在一个实施场景中，还可以同时计算上述第一损失值、第二损失值、第三损失值和第四损失值，并根据这些损失值来对识别网络的网络参数进行调整，以提高内容编码子网络对于病灶相关的内容特征的获取度，使风格编码子网络不响应与病灶相关的特征，提高图像识别的鲁棒性，并使得后续提取到的风格特征表示之间相互独立，且使风格编码子网络能够提取到完整准确的风格特征表示，内容编码子网络能够提取到完整准确的内容特征表示，从而提高图像识别的准确性。
区别于前述实施例，在对风格编码子网络和分类处理子网络进行训练的同时，加入对内容编码子网络和区域分割子网络的训练，从而能够在提高区域分割子网络的病灶识别能力的同时，提高内容编码子网络对于病灶相关的内容特征的获取度，进而能够有利于使得风格编码子网络不响应与病灶相关的特征，使后续分类时不受病灶相关特征的影响，故能够提高图像识别的鲁棒性。
请参阅图4，图4是本申请图像识别装置40一实施例的框架示意图。图像识别装置40包括图像获取模块41、风格提取模块42和分类处理模块43，图像获取模块41配置为获取多个待识别医学图像；风格提取模块42配置为分别提取每一待识别医学图像的风格特征表示；分类处理模块43配置为对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
上述方案，通过获取多个待识别医学图像，并提取每一待识别医学图像的风格特征表示，从而对多个待识别医学图像的风格特征表示进行分类处理，故能够在分类处理时，考虑多个待识别医学图像在各自风格特征上的差异，进而能够提高识别得到的扫描图像类别的准确性，且由于能够对多个待识别医学图像的风格特征表示进行分类处理，并得到每一待识别医学图像的扫描图像类别，故能够一次得到多个待识别医学图像的扫描图像类别，从而能够提高图像识别的效率，故此，上述方案能够提高图像识别的效率和准确性。
在一些实施例中，分类处理模块43包括第一融合处理子模块，配置为将多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示；分类处理模块43包括分类处理子模块，配置为对最终风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
区别于前述实施例，在对多个待识别医学图像的风格特征表示进行分类处理时，将多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示，故最终风格特征表示能够表示每一待识别医学图像的风格特征表示与其他待识别医学图像的风格特征表示之间的差异，故利用最终风格特征表示进行分类处理能够提高识别得到的 扫描图像类别的准确性。
在一些实施例中，图像识别装置40还包括图像排除模块、图像显示模块、第一预警模块、第二预警模块和第三预警模块中的至少一个模块；
所述图像排除模块，配置为将多个待识别医学图像按照其扫描图像类别进行排序；所述图像显示模块，配置为将按照扫描图像类别进行排序后的至少一个待识别医学图像进行同屏显示；所述第一预警模块，配置为在待识别医学图像的扫描图像类别存在重复时，输出第一预警信息，以提示扫描人员；所述第二预警模块，配置为在多个待识别医学图像的扫描图像类别中不存在预设扫描图像类别时，输出第二预警信息，以提示扫描人员；所述第三预警模块，配置为在所述待识别医学图像的扫描图像类别的分类置信度小于预设置信度阈值时，输出第三预警信息，提示扫描人员。
区别于前述实施例，在确定得到每一待识别医学图像所属的扫描图像类别之后，执行将至少一个待识别医学图像按照其扫描图像类别进行排序，能够提高医生阅片的便捷性；将按照扫描图像类别进行排序后的至少一个待识别医学图像进行同屏显示，能够免去医生翻阅待识别医学图像时来回对照，从而能够提高医生阅片的效率；在待识别医学图像的扫描图像类别存在重复时，输出第一预警信息，以提示扫描人员，在至少一个待识别医学图像的扫描图像类别中不存在预设扫描图像类别时，输出第二预警信息，以提示扫描人员，能够在扫描过程中实现图像质控，以在与实际相悖时，能够及时纠错，避免病人二次挂号。
在一些实施例中，图像识别装置40还包括预处理模块，配置为对每一待识别医学图像进行预处理，其中，预处理包括以下至少一种：将待识别医学图像的图像尺寸调整至预设尺寸、将待识别医学图像的图像强度归一化至预设范围。
区别于前述实施例，在提取风格特征表示之前，对每一目标区域的图像数据进行预处理，且预处理包括以下至少一种：将目标区域的图像尺寸调整至预设尺寸，将目标区域的图像强度归一化至预设范围，故能够有利于提高后续图像识别的准确性。
在一些实施例中，图像识别装置40还包括内容提取模块，配置为分别提取每一待识别医学图像的内容特征表示；图像识别装置40还包括病灶识别模块，配置为对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
区别于前述实施例，通过提取每一待识别医学图像的内容特征表示，并对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，能够在得到每一待识别医学图像的扫描图像类别的同时，确定其中的病灶区域，故能够有利于提高整体阅片效能，同时能够有利于消除病灶对扫描图像类别识别带来的干扰，从而能够提高图像识别的准确性。
在一些实施例中，病灶识别模块包括第二融合处理子模块，配置为将多个待识别医学图像的内容特征表示进行第二融合处理，得到最终内容特征表示；病灶识别模块包括病灶识别子模块，配置为对最终内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
区别于前述实施例，将多个待识别医学图像的内容特征表示进行第二融合处理，得到最终内容特征表示，能够有利于使最终内容特征表示补偿单一待识别医学图像中可能存在的病灶不明显或运动干扰产生的伪影等问题，从而在利用最终内容特征表示进行病灶识别时，能够提高病灶识别的准确性。
在一些实施例中，病灶识别模块还包括病灶提示子模块，配置为提示当前显示的待识别医学图像的病灶区域。
区别于前述实施例，通过提示当前显示的待识别医学图像的病灶区域，能够提升医生阅片体验。
在一些实施例中，第二融合处理子模块具体配置为执行以下任一者：将多个待识别 医学图像的内容特征表示进行拼接处理；将多个待识别医学图像的内容特征表示进行相加处理；其中，最终内容特征表示和多个待识别医学图像的内容特征表示的维度相同。
区别于前述实施例，通过将多个待识别医学图像的内容特征表示进行拼接处理，或者将多个待识别医学图像的内容特征表示进行相加处理中的任一者，得到最终内容特征表示，且最终内容特征表示和多个待识别医学图像的内容特征表示的维度相同，能够通过多种方式得到最终内容特征表示，从而能够提高图像识别的鲁棒性。
在一些实施例中，风格提取模块42具体配置为利用识别网络的风格编码子网络分别提取每一待识别医学图像的风格特征表示；分类处理模块43具体配置为利用识别网络的分类处理子网络对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别；内容提取模块具体配置为利用识别网络的内容编码子网络分别提取每一待识别医学图像的内容特征表示；病灶识别模块具体配置为利用识别网络的区域分割子网络对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
区别于前述实施例，利用识别网络的风格编码子网络分别提取每一待识别医学图像的风格特征表示，利用识别网络的分类处理子网络对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别，利用识别网络的内容编码子网络分别提取每一待识别医学图像的内容特征表示，利用识别网络的区域分割子网络对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域，能够利用识别网络执行风格特征表示的提取、分类处理、内容特征表示的提取以及病灶识别等任务，故能够有利于提高图像识别的效率。
在一些实施例中，图像识别装置40还包括样本获取模块，配置为获取多个样本医学图像，其中，多个样本医学图像标注有其真实扫描图像类别和真实病灶区域；图像识别装置40还包括特征提取模块，配置为利用风格编码子网络分别提取每一样本医学图像的样本风格特征表示，并利用内容编码子网络分别提取每一样本医学图像的样本内容特征表示；图像识别装置40还包括识别处理模块，配置为利用分类处理子网络对多个样本医学图像的样本风格特征表示进行分类处理，得到每一样本医学图像的预测扫描图像类别，并利用区域分割子网络对多个样本医学图像的样本内容特征表示进行病灶识别，得到每一样本医学图像中的预测病灶区域；图像识别装置40还包括第一调整模块，配置为利用真实扫描类别和预测扫描类别的差异，调整风格编码子网络和分类处理子网络的网络参数，以及利用真实病灶区域和预测病灶区域的差异，调整内容编码子网络和区域分割子网络的网络参数。
区别于前述实施例，在对风格编码子网络和分类处理子网络进行训练的同时，加入对内容编码子网络和区域分割子网络的训练，从而能够在提高区域分割子网络的病灶识别能力的同时，提高内容编码子网络对于病灶相关的内容特征的获取度，进而能够有利于使得风格编码子网络不响应与病灶相关的特征，使后续分类时不受病灶相关特征的影响，故能够提高图像识别的鲁棒性。
在一些实施例中，图像识别装置40还包括分布获取模块，配置为获取每一样本医学图像的样本风格特征表示的样本数据分布情况；图像识别装置40还包括第二调整模块，配置为利用样本数据分布情况之间的差异，调整风格编码子网络的网络参数。
区别于前述实施例，在训练过程中，同时获取样本数据分布情况，并利用样本数据分布情况之间的差异，调整风格编码子网络的网络参数，故能够有利于使后续提取到风格特征表示之间相互独立，从而能够有利于提高识别到的扫描图像类别的准确性。
在一些实施例中，图像识别装置40还包括图像重建模块，配置为利用一样本风格特征表示和一内容特征表示，构建得到与样本风格特征表示对应的重建图像；图像识别装置40还包括第三调整模块，配置为利用重建图像与对应的样本风格特征表示所属的样本 医学图像之间的差异，调整风格编码子网络和内容编码子网络的网络参数。
区别于前述实施例，在训练过程中，同时利用一样本风格特征表示和一内容特征表示，构建得到与样本风格特征表示对应的重建图像，并利用重建图像与对应的样本风格特征表示所属的样本医学图像之间的差异，调整风格编码子网络和内容编码子网络的网络参数，从而能够使风格编码子网络尽可能地提取到完整准确的风格特征，而内容编码子网络尽可能地提取到完整准确的风格特征，进而能够有利于提高后续扫描图像类别以及病灶识别的准确性。
在一些实施例中，风格编码子网络包括：顺序连接的下采样层和全局池化层；和/或，内容编码子网络包括以下任一者：顺序连接的下采样层和残差块、顺序连接的卷积层和池化层。
区别于前述实施例，通过将风格编码子网络设置为包括顺序连接的下采样层和全局池化层，能够有利于在简化网络结构的同时便于网络训练；通过将内容编码子网络设置为包括以下任一者：顺序连接的下采样层和残差块、顺序连接的卷积层和池化层，能够有利于在简化网络结构的同时便于网络训练。
请参阅图5，图5是本申请电子设备50一实施例的框架示意图。电子设备50包括相互耦接的存储器51和处理器52，处理器52配置为执行存储器51中存储的程序指令，以实现上述任一图像识别方法实施例的步骤。在一个实施场景中，电子设备50可以包括但不限于：微型计算机、服务器，此外，电子设备50还可以包括笔记本电脑、平板电脑等移动设备，在此不做限定。
具体而言，处理器52用于控制其自身以及存储器51以实现上述任一图像识别方法实施例的步骤。处理器52还可以称为中央处理单元(Central Processing Unit，CPU)。处理器52可能是一种集成电路芯片，具有信号的处理能力。处理器52还可以是通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。另外，处理器52可以由集成电路芯片共同实现。
上述方案，能够提高图像识别的效率和准确性。
请参阅图6，图6为本申请计算机可读存储介质60一实施例的框架示意图。计算机可读存储介质60存储有能够被处理器运行的程序指令601，程序指令601用于实现上述任一图像识别方法。
上述方案，能够提高图像识别的效率和准确性。
相应地，本申请实施例还提供了一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现上述任意一种图像识别方法。
在本申请所提供的几个实施例中，应该理解到，所揭露的方法和装置，可以通过其它的方式实现。例如，以上所描述的装置实施方式仅仅是示意性的，例如，模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性、机械或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施方式方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是 各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本申请各个实施方式方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
工业实用性
本申请实施例公开了一种图像识别方法及装置、电子设备、计算机存储介质和计算机程序，图像识别方法包括：获取多个待识别医学图像；分别提取每一所述待识别医学图像的风格特征表示；对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别。上述方案，能够提高图像识别的效率和准确性。

Claims (27)

  1. 一种图像识别方法，包括：
    获取多个待识别医学图像；
    分别提取每一所述待识别医学图像的风格特征表示；
    对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别。
  2. 根据权利要求1所述的图像识别方法，其中，所述对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别包括：
    将所述多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示；
    对所述最终风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别。
  3. 根据权利要求1或2所述的图像识别方法，其中，所述对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别之后，所述图像识别方法还包括以下至少一者：
    将所述多个待识别医学图像按照其扫描图像类别进行排序；
    将按照扫描图像类别进行排序后的至少一个所述待识别医学图像进行同屏显示；
    若所述待识别医学图像的扫描图像类别存在重复，则输出第一预警信息，以提示扫描人员；
    若所述多个待识别医学图像的扫描图像类别中不存在预设扫描图像类别，则输出第二预警信息，以提示扫描人员；
    若所述待识别医学图像的扫描图像类别的分类置信度小于预设置信度阈值，则输出第三预警信息，以提示扫描人员。
  4. 根据权利要求1至3任一项所述的图像识别方法，其中，所述分别提取每一所述待识别医学图像的风格特征表示之前，所述方法还包括：
    对每一所述待识别医学图像进行预处理，其中，所述预处理包括以下至少一种：将所述待识别医学图像的图像尺寸调整至预设尺寸、将所述待识别医学图像的图像强度归一化至预设范围。
  5. 根据权利要求1至4任一项所述的图像识别方法，其中，所述方法还包括：
    分别提取每一所述待识别医学图像的内容特征表示；
    对所述多个待识别医学图像的内容特征表示进行病灶识别，得到每一所述待识别医学图像中的病灶区域。
  6. 根据权利要求5所述的图像识别方法，其中，所述对所述多个待识别医学图像的内容特征表示进行病灶识别，得到每一所述待识别医学图像中的病灶区域包括：
    将所述多个待识别医学图像的内容特征表示进行第二融合处理，得到最终内容特征表示；
    对所述最终内容特征表示进行病灶识别，得到每一所述待识别医学图像中的病灶区域；
    和/或，所述方法还包括：
    提示当前显示的所述待识别医学图像的病灶区域。
  7. 根据权利要求6所述的图像识别方法，其中，所述将所述多个待识别医学图像的内容特征表示进行第二融合处理，包括以下任一者：
    将所述多个待识别医学图像的内容特征表示进行拼接处理；
    将所述多个待识别医学图像的内容特征表示进行相加处理；
    其中，所述最终内容特征表示和所述多个待识别医学图像的内容特征表示的维度相同。
  8. 根据权利要求5至7任一项所述的图像识别方法，其中，所述分别提取每一所述待识别医学图像的风格特征表示，包括：
    利用识别网络的风格编码子网络分别提取每一所述待识别医学图像的风格特征表示；
    所述对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别，包括：
    利用所述识别网络的分类处理子网络对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别；
    所述分别提取每一所述待识别医学图像的内容特征表示，包括：
    利用所述识别网络的内容编码子网络分别提取每一所述待识别医学图像的内容特征表示；
    所述对所述多个待识别医学图像的内容特征表示进行病灶识别，得到每一所述待识别医学图像中的病灶区域，包括：
    利用所述识别网络的区域分割子网络对所述多个待识别医学图像的内容特征表示进行病灶识别，得到每一所述待识别医学图像中的病灶区域。
  9. 根据权利要求8所述的图像识别方法，其中，所述分别提取每一所述待识别医学图像的风格特征表示之前，所述图像识别方法还包括：
    获取多个样本医学图像，其中，所述多个样本医学图像标注有其真实扫描图像类别和真实病灶区域；
    利用所述风格编码子网络分别提取每一所述样本医学图像的样本风格特征表示，并利用所述内容编码子网络分别提取每一所述样本医学图像的样本内容特征表示；
    利用所述分类处理子网络对所述多个样本医学图像的样本风格特征表示进行分类处理，得到每一所述样本医学图像的预测扫描图像类别，并利用所述区域分割子网络对所述多个样本医学图像的样本内容特征表示进行病灶识别，得到每一所述样本医学图像中的预测病灶区域；
    利用所述真实扫描图像类别和所述预测扫描图像类别的差异，调整所述风格编码子网络和所述分类处理子网络的网络参数，以及利用所述真实病灶区域和所述预测病灶区域的差异，调整所述内容编码子网络和所述区域分割子网络的网络参数。
  10. 根据权利要求9所述的图像识别方法，其中，所述图像识别方法还包括：
    获取每一所述样本医学图像的样本风格特征表示的样本数据分布情况；
    利用所述样本数据分布情况之间的差异，调整所述风格编码子网络的网络参数。
  11. 根据权利要求9所述的图像识别方法，其中，所述图像识别方法还包括：
    利用一所述样本风格特征表示和一所述内容特征表示，构建得到与所述样本风格特征表示对应的重建图像；
    利用所述重建图像与对应的所述样本风格特征表示所属的所述样本医学图像之间的差异，调整所述风格编码子网络和所述内容编码子网络的网络参数。
  12. 根据权利要求8所述的图像识别方法，其中，所述风格编码子网络包括：顺序连接的下采样层和全局池化层；
    和/或，所述内容编码子网络包括以下任一者：顺序连接的下采样层和残差块、顺序连接的卷积层和池化层。
  13. 一种图像识别装置，包括：
    图像获取模块，配置为获取多个待识别医学图像；
    风格提取模块，配置为分别提取每一所述待识别医学图像的风格特征表示；
    分类处理模块，配置为对所述多个待识别医学图像的风格特征表示进行分类处理，得到每一所述待识别医学图像的扫描图像类别。
  14. 根据权利要求13所述的装置，其中，所述分类处理模块包括第一融合处理子模块，配置为将多个待识别医学图像的风格特征表示进行第一融合处理，得到最终风格特征表示；所述分类处理模块还包括分类处理子模块，配置为对最终风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别。
  15. 根据权利要求13或14所述的装置，其中，所述装置还包括图像排除模块、图像显示模块、第一预警模块、第二预警模块和第三预警模块中的至少一个模块；
    所述图像排除模块，配置为将多个待识别医学图像按照其扫描图像类别进行排序；
    所述图像显示模块，配置为将按照扫描图像类别进行排序后的至少一个待识别医学图像进行同屏显示；
    所述第一预警模块，配置为在待识别医学图像的扫描图像类别存在重复时，输出第一预警信息，以提示扫描人员；
    所述第二预警模块，配置为在多个待识别医学图像的扫描图像类别中不存在预设扫描图像类别时，输出第二预警信息，以提示扫描人员；
    所述第三预警模块，配置为在所述待识别医学图像的扫描图像类别的分类置信度小于预设置信度阈值时，输出第三预警信息，提示扫描人员。
  16. 根据权利要求13至15任一项所述的装置，其中，所述装置还包括预处理模块，配置为对每一待识别医学图像进行预处理，其中，预处理包括以下至少一种：将待识别医学图像的图像尺寸调整至预设尺寸、将待识别医学图像的图像强度归一化至预设范围。
  17. 根据权利要求13至16任一项所述的装置，其中，所述装置还包括内容提取模块，配置为分别提取每一待识别医学图像的内容特征表示；
    所述装置还包括病灶识别模块，配置为对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
  18. 根据权利要求17所述的装置，其中，所述病灶识别模块包括第二融合处理子模块和病灶识别子模块，配置为将多个待识别医学图像的内容特征表示进行第二融合处理，得到最终内容特征表示；所述病灶识别模块包括病灶识别子模块，配置为对最终内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域；
    和/或，所述病灶识别模块包括病灶提示子模块，配置为提示当前显示的待识别医学图像的病灶区域。
  19. 根据权利要求18所述的装置，其中，所述第二融合处理子模块具体配置为执行以下任一者：将多个待识别医学图像的内容特征表示进行拼接处理；将多个待识别医学图像的内容特征表示进行相加处理；其中，最终内容特征表示和多个待识别医学图像的内容特征表示的维度相同。
  20. 根据权利要求17至19任一项所述的装置，其中，所述风格提取模块具体配置为利用识别网络的风格编码子网络分别提取每一待识别医学图像的风格特征表示；
    所述分类处理模块具体配置为利用识别网络的分类处理子网络对多个待识别医学图像的风格特征表示进行分类处理，得到每一待识别医学图像的扫描图像类别；
    所述内容提取模块具体配置为利用识别网络的内容编码子网络分别提取每一待识别医学图像的内容特征表示；
    所述病灶识别模块具体配置为利用识别网络的区域分割子网络对多个待识别医学图像的内容特征表示进行病灶识别，得到每一待识别医学图像中的病灶区域。
  21. 根据权利要求20所述的装置，其中，所述装置还包括样本获取模块，配置为获 取多个样本医学图像，其中，多个样本医学图像标注有其真实扫描图像类别和真实病灶区域；
    所述装置还包括特征提取模块，配置为利用风格编码子网络分别提取每一样本医学图像的样本风格特征表示，并利用内容编码子网络分别提取每一样本医学图像的样本内容特征表示；
    所述装置还包括识别处理模块，配置为利用分类处理子网络对多个样本医学图像的样本风格特征表示进行分类处理，得到每一样本医学图像的预测扫描图像类别，并利用区域分割子网络对多个样本医学图像的样本内容特征表示进行病灶识别，得到每一样本医学图像中的预测病灶区域；
    所述装置还包括第一调整模块，配置为利用真实扫描类别和预测扫描类别的差异，调整风格编码子网络和分类处理子网络的网络参数，以及利用真实病灶区域和预测病灶区域的差异，调整内容编码子网络和区域分割子网络的网络参数。
  22. 根据权利要求21所述的装置，其中，所述装置还包括分布获取模块，配置为获取每一样本医学图像的样本风格特征表示的样本数据分布情况；
    所述装置还包括第二调整模块，配置为利用样本数据分布情况之间的差异，调整风格编码子网络的网络参数。
  23. 根据权利要求21所述的装置，其中，所述装置还包括图像重建模块，配置为利用一样本风格特征表示和一内容特征表示，构建得到与样本风格特征表示对应的重建图像；
    所述装置还包括第三调整模块，配置为利用重建图像与对应的样本风格特征表示所属的样本医学图像之间的差异，调整风格编码子网络和内容编码子网络的网络参数。
  24. 根据权利要求20所述的装置，其中，所述风格编码子网络包括：顺序连接的下采样层和全局池化层；和/或，内容编码子网络包括以下任一者：顺序连接的下采样层和残差块、顺序连接的卷积层和池化层。
  25. 一种电子设备，包括相互耦接的存储器和处理器，所述处理器配置为执行所述存储器中存储的程序指令，以实现权利要求1至12任一项所述的图像识别方法。
  26. 一种计算机可读存储介质，其上存储有程序指令，所述程序指令被处理器执行时实现权利要求1至12任一项所述的图像识别方法。
  27. 一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现权利要求1至12任一所述的图像识别方法。
PCT/CN2021/106479 2020-09-24 2021-07-15 图像识别方法及装置、设备、存储介质和程序 WO2022062590A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011018760.7 2020-09-24
CN202011018760.7A CN112036506A (zh) 2020-09-24 2020-09-24 图像识别方法及相关装置、设备

Publications (1)

Publication Number Publication Date
WO2022062590A1 true WO2022062590A1 (zh) 2022-03-31

Family

ID=73574303

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/106479 WO2022062590A1 (zh) 2020-09-24 2021-07-15 图像识别方法及装置、设备、存储介质和程序

Country Status (3)

Country Link
CN (1) CN112036506A (zh)
TW (1) TW202221568A (zh)
WO (1) WO2022062590A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112036506A (zh) * 2020-09-24 2020-12-04 上海商汤智能科技有限公司 图像识别方法及相关装置、设备
CN113516757A (zh) * 2021-07-07 2021-10-19 上海商汤智能科技有限公司 图像显示方法及相关装置和电子设备、存储介质
CN113516758B (zh) * 2021-07-07 2024-10-29 上海商汤善萃医疗科技有限公司 图像显示方法及相关装置和电子设备、存储介质
CN114663715B (zh) * 2022-05-26 2022-08-26 浙江太美医疗科技股份有限公司 医学图像质控、分类模型训练方法、装置及计算机设备
CN115294110B (zh) * 2022-09-30 2023-01-06 杭州太美星程医药科技有限公司 扫描期的识别方法、装置、电子设备和存储介质
CN117351197A (zh) * 2023-12-04 2024-01-05 北京联影智能影像技术研究院 图像分割方法、装置、计算机设备和存储介质
CN118429734B (zh) * 2024-07-05 2024-10-01 之江实验室 磁共振数据分类系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109447966A (zh) * 2018-10-26 2019-03-08 科大讯飞股份有限公司 医学图像的病灶定位识别方法、装置、设备及存储介质
CN110504029A (zh) * 2019-08-29 2019-11-26 腾讯医疗健康(深圳)有限公司 一种医学图像处理方法、医学图像识别方法及装置
CN111340083A (zh) * 2020-02-20 2020-06-26 京东方科技集团股份有限公司 医学图像的处理方法、装置、设备及存储介质
CN112036506A (zh) * 2020-09-24 2020-12-04 上海商汤智能科技有限公司 图像识别方法及相关装置、设备

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109447966A (zh) * 2018-10-26 2019-03-08 科大讯飞股份有限公司 医学图像的病灶定位识别方法、装置、设备及存储介质
CN110504029A (zh) * 2019-08-29 2019-11-26 腾讯医疗健康(深圳)有限公司 一种医学图像处理方法、医学图像识别方法及装置
CN111340083A (zh) * 2020-02-20 2020-06-26 京东方科技集团股份有限公司 医学图像的处理方法、装置、设备及存储介质
CN112036506A (zh) * 2020-09-24 2020-12-04 上海商汤智能科技有限公司 图像识别方法及相关装置、设备

Also Published As

Publication number Publication date
CN112036506A (zh) 2020-12-04
TW202221568A (zh) 2022-06-01

Similar Documents

Publication Publication Date Title
WO2022062590A1 (zh) 图像识别方法及装置、设备、存储介质和程序
TWI776426B (zh) 圖像識別方法、電子設備和電腦可讀儲存介質
Singh et al. Shallow 3D CNN for detecting acute brain hemorrhage from medical imaging sensors
CN110506278B (zh) 隐空间中的目标检测
Wang et al. Robust content-adaptive global registration for multimodal retinal images using weakly supervised deep-learning framework
CN115496771A (zh) 一种基于脑部三维mri图像设计的脑肿瘤分割方法
CN114494296A (zh) 一种基于Unet和Transformer相融合的脑部胶质瘤分割方法与系统
Wang et al. Sk-unet: An improved u-net model with selective kernel for the segmentation of lge cardiac mr images
WO2022095258A1 (zh) 图像目标分类方法、装置、设备、存储介质及程序
Tang et al. Automatic lumbar spinal CT image segmentation with a dual densely connected U-Net
CN110751629A (zh) 一种心肌影像分析装置及设备
CN118247284B (zh) 图像处理模型的训练方法、图像处理方法
Yu et al. Cardiac LGE MRI segmentation with cross-modality image augmentation and improved U-Net
Kascenas et al. Anomaly detection via context and local feature matching
Rajeshkumar et al. Convolutional Neural Networks (CNN) based Brain Tumor Detection in MRI Images
Dikici et al. Constrained generative adversarial network ensembles for sharable synthetic data generation
Pavarut et al. Improving Kidney Tumor Classification With Multi-Modal Medical Images Recovered Partially by Conditional CycleGAN
Balagalla et al. Automated segmentation of standard scanning planes to measure biometric parameters in foetal ultrasound images–a survey
CN111598870A (zh) 基于卷积神经网络端对端推理计算冠状动脉钙化比的方法
CN115294110B (zh) 扫描期的识别方法、装置、电子设备和存储介质
Murthy et al. Deep Learning and MRI Improve Carotid Arterial Tree Reconstruction
US20240153089A1 (en) Systems and methods for processing real-time cardiac mri images
US20240144469A1 (en) Systems and methods for automatic cardiac image analysis
Gia et al. A Computer-Aided Detection to Intracranial Hemorrhage by Using Deep Learning: A Case Study
Tahir et al. A Methodical Review on the Segmentation Types and Techniques of Medical Images

Legal Events

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

Ref document number: 21870968

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: 21870968

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 25/09/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21870968

Country of ref document: EP

Kind code of ref document: A1