Artificial intelligence-assisted gallbladder image automatic detection method and system, and electronic equipment

Abstract

The invention relates to the technical field of artificial intelligence, and provides an artificial intelligence assisted gallbladder image automatic detection method and system and electronic equipment, and the method comprises the steps: generating a corresponding generative image for an original abdomen CT image, and reconstructing the CT image of a gallbladder in a contraction state into a gallbladder-filled CT image; performing part identification on the generative image to determine whether a gallbladder part exists in the generative image; if the gall bladder part exists in the generative image, carrying out part identification on the original abdomen CT image to obtain a contour area of the gall bladder part; calculating the gallbladder volume according to the contour area of the gallbladder part, and detecting whether the gallbladder size is abnormal or not according to the gallbladder volume to obtain a gallbladder detection result. According to the invention, the automatic detection of the gall bladder image is realized, the rapid analysis of the CT image can be assisted, the influence of the current state of the gall bladder on the recognition accuracy of the gall bladder part is effectively avoided, and a more accurate gall bladder image detection result is rapidly obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical artificial intelligence, and in particular to an artificial intelligence assisted gallbladder image automatic detection method and system and electronic equipment. BACKGROUND

[0002] The gallbladder is an important part of the human digestive system, and gallbladder cancer is a common gastrointestinal malignancy. Therefore, accurately predicting and analyzing the morphological condition of the gallbladder is very important for preventing gallbladder diseases. In the prior art, Computed Tomography (CT) is widely used in the diagnosis and treatment of the gallbladder part of the digestive system by performing one after another cross-sectional scanning around the part to be measured, which has the advantages of large scanning range, fast scanning time, and clear image. At present, the identification and analysis of the morphological condition of the gallbladder based on CT images mainly rely on the manual analysis of CT images by doctors, which is not only inefficient, but also subjective and has low detection accuracy. SUMMARY

[0003] In view of the above problems, the present application is proposed to provide an artificial intelligence assisted gallbladder image automatic detection method, system and electronic equipment to overcome the above problems.

[0004] In one aspect of the present application, an artificial intelligence assisted gallbladder image automatic detection method is provided, which comprises:

[0005] generating a corresponding generative image for the original abdominal CT image to reconstruct the CT image with the gallbladder in the filling or contraction state into a CT image with the gallbladder filled;

[0006] performing part recognition on the generative image to determine whether there is a gallbladder part in the generative image;

[0007] if there is a gallbladder part in the generative image, performing part recognition on the original abdominal CT image to obtain the contour area of the gallbladder part;

[0008] calculating the volume of the gallbladder according to the contour area of the gallbladder part, and detecting whether the size of the gallbladder is abnormal according to the volume of the gallbladder to obtain a gallbladder detection result.

[0009] Further, after performing part recognition on the original abdominal CT image to obtain the contour area of the gallbladder part, the method further comprises:

[0010] performing abnormal detection on the CT value in the contour area of the gallbladder part, and detecting whether there is an abnormal density shadow in the gallbladder according to whether the CT value is higher than a set CT threshold to obtain a gallbladder detection result.

[0011] Further, generating a corresponding generative image for the original abdominal CT image comprises:

[0012] A pre-trained GAN generative adversarial network model is used to generate corresponding generative images for the original abdominal CT images. The GAN generative adversarial network model includes a generator and a discriminator. The generator is used to generate generative images corresponding to the abdominal CT images, and the discriminator is used to determine whether the input generative image is a real CT image of the gallbladder filling state or a generative image generated by the generator.

[0013] The training steps of the GAN generative adversarial network model include: constructing a data set, the data set containing image pairs of different patients in different conditions, wherein the image pairs are constructed in such a manner that a CT image of gallbladder contraction corresponds to a CT image of gallbladder filling of the same patient; a CT image of cholecystectomy corresponds to a CT image of cholecystectomy; and a CT image of gallbladder filling corresponds to a CT image of gallbladder filling of the same patient; and using the data set to train the generative adversarial network to generate a CT image of the gallbladder filling by inputting an abdominal CT image of gallbladder contraction.

[0014] Furthermore, the gallbladder volume is calculated based on the contour area of ​​the gallbladder, including:

[0015] The pixel values ​​of the gallbladder area in the original abdominal CT image are marked according to the contour area of ​​the gallbladder area, and the number of voxels in the gallbladder area is counted according to the marking results;

[0016] Obtaining the pixel spacing and slice thickness of the original abdominal CT image according to tag information in the DICOM file corresponding to the original abdominal CT image, and calculating the voxel volume of each voxel in the original abdominal CT image according to the pixel spacing and slice thickness;

[0017] The gallbladder volume was calculated based on the voxel volume of each voxel and the number of voxels in the gallbladder area in the original abdominal CT images.

[0018] Furthermore, the method further comprises:

[0019] The quality of the imaging diagnosis report corresponding to the original abdominal CT image is checked according to the gallbladder detection result obtained from the original abdominal CT image.

[0020] Furthermore, performing quality inspection on the imaging diagnostic report corresponding to the original abdominal CT image based on the gallbladder detection result obtained from the original abdominal CT image includes:

[0021] Performing a text description on the gallbladder detection result obtained from the original abdominal CT image, performing numerical processing on the text description to identify different text description states of the gallbladder part using different numerical values, and representing the obtained numerical values ​​as a first one-dimensional array;

[0022] The text information of the imaging findings in the imaging diagnosis report is split into clauses, and the entity description of the gallbladder location is extracted from each clause to obtain the entity description of the gallbladder location;

[0023] The entity description is digitized using the same digitization method, so that different entity description states of the gallbladder part are marked with different digits, and the obtained digits are expressed as a second one-dimensional array;

[0024] Consistency verification is performed on the elements at the same position in the first one-dimensional array and the second one-dimensional array in sequence.

[0025] Another aspect of the present invention further provides an artificial intelligence-assisted gallbladder imaging automatic detection system, the system comprising a functional module for implementing any of the artificial intelligence-assisted gallbladder imaging automatic detection methods described above, specifically comprising:

[0026] An image generation module is used to generate a corresponding generative image for the original abdominal CT image, so as to reconstruct the CT image of the gallbladder in a filled or contracted state into a CT image of the gallbladder in a filled state;

[0027] a part recognition module, configured to perform part recognition on the generated image to determine whether a gallbladder part exists in the generated image;

[0028] The part recognition module is further configured to perform part recognition on the original abdominal CT image to obtain a contour area of ​​the gallbladder part when it is determined that the gallbladder part exists in the generated image;

[0029] The abnormality detection module is used to calculate the gallbladder volume according to the contour area of ​​the gallbladder part, and to detect whether the gallbladder size is abnormal according to the gallbladder volume to obtain the gallbladder detection result.

[0030] Furthermore, the system further comprises:

[0031] The quality detection module is used to perform quality detection on the imaging diagnosis report corresponding to the original abdominal CT image according to the gallbladder detection result obtained from the original abdominal CT image.

[0032] Another aspect of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor;

[0033] When the computer program is executed by the processor, the steps of the artificial intelligence-assisted automatic gallbladder imaging detection method as described in any one of the above items are implemented.

[0034] Another aspect of the present application also provides a computer program product, which stores an artificial intelligence assisted gallbladder image automatic detection program on the computer program product, and the artificial intelligence assisted gallbladder image automatic detection program, when executed by a processor, implements the steps of the artificial intelligence assisted gallbladder image automatic detection method according to any one of the above.

[0035] The artificial intelligence assisted gallbladder image automatic detection method, system and electronic device provided by the embodiments of the present application can realize automatic detection of gallbladder images, assist in rapid analysis of CT images, avoid subjectivity of pure human analysis, improve reading efficiency of CT images, and effectively avoid the influence of the current state of the gallbladder on the accuracy of gallbladder part recognition based on the generated image to quickly obtain more accurate gallbladder image detection results, which has important clinical application value.

[0036] The above description is only a summary of the technical solutions of the present application. In order to enable the technical means of the present application to be more clearly understood, the present application can be implemented according to the content of the specification, and in order to enable the above and other purposes, characteristics and advantages of the present application to be more apparent and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0037] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments and are not meant to limit the present application. In the drawings:

[0038] Figure 1 A flowchart of an artificial intelligence assisted gallbladder image automatic detection method according to an embodiment of the present application;

[0039] Figure 2 A flowchart of an artificial intelligence assisted gallbladder image automatic detection method according to another embodiment of the present application;

[0040] Figure 3 A flowchart of an artificial intelligence assisted gallbladder image automatic detection method according to another embodiment of the present application;

[0041] Figure 4 A structural schematic diagram of an artificial intelligence assisted gallbladder image automatic detection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0043] Those skilled in the art will understand that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by those skilled in the art in the art to which the present invention pertains. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with those in the context of the prior art and, unless specifically defined, will not be interpreted in an idealized or overly formal sense.

[0044] Example 1

[0045] The embodiment of the present invention provides an artificial intelligence-assisted gallbladder imaging automatic detection method, such as Figure 1 As shown, the artificial intelligence-assisted gallbladder imaging automatic detection method proposed in the present invention includes the following steps:

[0046] S11. Generate a corresponding generative image for the original abdominal CT image, so as to reconstruct the CT image of the gallbladder in a filled or contracted state into a CT image of the gallbladder in a filled state.

[0047] S12: Perform part recognition on the generated image to determine whether a gallbladder part exists in the generated image.

[0048] S13. If the gallbladder part exists in the generated image, perform part recognition on the original abdominal CT image to obtain the contour area of ​​the gallbladder part.

[0049] In this embodiment, if the gallbladder portion does not exist in the generated image, it is determined that the gallbladder has been removed.

[0050] S14. Calculate the gallbladder volume according to the contour area of ​​the gallbladder, and detect whether the gallbladder size is abnormal according to the gallbladder volume to obtain a gallbladder detection result.

[0051] The artificial intelligence-assisted automatic gallbladder image detection method provided by an embodiment of the present invention generates a corresponding generative image for the original abdominal CT image to reconstruct a CT image of the gallbladder in a filled or contracted state into a CT image of a filled gallbladder, and performs part recognition based on the generative image. If it is determined that the gallbladder part exists in the generative image, the part recognition is performed on the original abdominal CT image to obtain the contour area of ​​the gallbladder part and then calculate the gallbladder volume to achieve abnormal gallbladder size detection. The present invention can realize automatic gallbladder image detection, assist in rapid analysis of CT images, avoid the subjectivity of simple manual analysis, and improve the reading efficiency of CT images. Moreover, by determining whether the gallbladder part exists in the image based on the generative image, it can effectively avoid the influence of the current state of the gallbladder on the gallbladder part recognition accuracy, thereby quickly obtaining more accurate gallbladder image detection results, and has important clinical application value.

[0052] In an embodiment of the present invention, generating a corresponding generative image for the original abdominal CT image in step S11 specifically includes: generating a corresponding generative image for the original abdominal CT image using a pretrained GAN generative adversarial network model. The GAN generative adversarial network model includes a generator and a discriminator. The generator is configured to generate a generative image corresponding to the abdominal CT image, and the discriminator is configured to determine whether the input generative image is a true CT image of a filled gallbladder or a generative image generated by the generator. In this embodiment, the gallbladder has both filled and contracted states. When bile is released, the gallbladder contracts, reducing the gallbladder area and affecting the accuracy of gallbladder location recognition. To avoid the impact of the current gallbladder state on gallbladder location recognition accuracy, the present invention generates a corresponding generative image for the original abdominal CT image to reconstruct a CT image of a filled or contracted gallbladder into a CT image of a filled gallbladder. Location recognition is pre-performed based on the generative image. When the gallbladder location is present in the generative image, location recognition is performed on the original abdominal CT image.

[0053] Furthermore, the training steps of the GAN generative adversarial network model include: constructing a dataset containing image pairs of different patients under different conditions, wherein the image pairs are constructed in the following manner: a CT image of gallbladder contraction corresponds to a CT image of the same patient with a full gallbladder; a CT image of cholecystectomy corresponds to a CT image of cholecystectomy; and a CT image of a full gallbladder corresponds to a CT image of the same patient with a full gallbladder; and using the dataset to train the generative adversarial network to generate a CT image of the gallbladder with a full gallbladder when inputted as an abdominal CT image with a contracted gallbladder. Specifically, abdominal CT images are first collected to ensure that the dataset contains a rich collection of images from different patients and under different conditions, including abdominal CT images of the same patient taken with a full gallbladder and with a contracted gallbladder. Image pairs are then constructed: an image of gallbladder contraction corresponds to an image of a full gallbladder; an image of cholecystectomy corresponds to an image of a cholecystectomy; and an image of a full gallbladder corresponds to an image of a full gallbladder. The generative adversarial network is trained based on the dataset to generate a CT image of the gallbladder with a contracted gallbladder when inputted as an abdominal CT image with a contracted gallbladder. The constructed GAN model consists of a generator and a discriminator. The generator is responsible for generating corresponding images in image pairs, such as the filled state of the gallbladder. It receives an image of the contracted state as input and learns to generate the corresponding image. The discriminator is responsible for determining whether the input image is a real filled state image or a generated image. Its task is to improve the realism of the images produced by the generator. In each training iteration, the discriminator is trained first, using real and generated images to update its parameters. Then, the generator is trained to generate images that can deceive the discriminator. This cycle continues until the model converges.

[0054] In an embodiment of the present invention, performing part recognition on the generative image to determine whether a gallbladder part exists in the generative image in step S12 includes: performing part recognition on the generative image using a pre-trained part recognition network model to determine whether a gallbladder part exists in the generative image.

[0055] The training steps for the region recognition network model include collecting historical abdominal CT image data, annotating the gallbladder and liver regions within the images, and training a semantic segmentation model based on the Unet2D network. Anatomically, the liver is primarily located in the right upper abdomen, while the gallbladder is attached to the underside of the liver. These two regions are closely physiologically connected. By identifying these two regions, the present invention further identifies the gallbladder location, effectively preventing model misidentification.

[0056] In the embodiment of the present invention, calculating the gallbladder volume according to the contour area of ​​the gallbladder in step S14 specifically includes the following steps (not shown in the drawings):

[0057] S141 , marking pixel values ​​of the gallbladder region in the original abdominal CT image according to the contour region of the gallbladder region, and counting the number of voxels in the gallbladder region according to the marking result.

[0058] In this embodiment, a generative image is first reconstructed for each CT image in the acquired abdominal CT image data. After the part recognition network model is sequentially called for each generative image, it is verified whether each image contains the gallbladder region and the liver region. There is a close physiological connection between the liver and gallbladder. Based on this correspondence, recognition results that only recognize the gallbladder region but not the liver region are eliminated. A determination is made as to whether the generative image recognizes the gallbladder region. If not, it is determined that the gallbladder has been removed. If so, the part recognition network model is called on the original CT image data to obtain the contour region of the gallbladder region in the original CT image data and mark the pixel values ​​of the gallbladder region. In one specific embodiment, the gallbladder region can be marked as 1, and other regions can be marked as 0. By counting the voxels marked as 1 in the gallbladder region, that is, summing the voxels marked as 1 in the gallbladder region, the number of voxels in the gallbladder region, gallbladder_voxels, is obtained.

[0059] S142. Obtain the pixel spacing and slice thickness of the original abdominal CT image according to the tag information in the DICOM file corresponding to the original abdominal CT image, and calculate the voxel volume of each voxel in the original abdominal CT image according to the pixel spacing and slice thickness.

[0060] The purpose of obtaining the voxel volume voxel_volume is to determine the actual physical size of each voxel. By reading the tag information of the dicom file, the pixel spacing (Pixel spacing) and slice thickness (Slicethickness) of the image are obtained. The voxel volume calculation formula is as follows:

[0061] voxel_volume=voxel_size_x*voxel_size_y*voxel_size_z, unit: mm 3 ;

[0062] Among them, voxel_size_x represents the first number of pixel spacing, voxel_size_y represents the second number of pixel spacing, and voxel_size_z represents the slice thickness.

[0063] S143. Calculate the gallbladder volume gallbladder_volume based on the voxel volume of each voxel and the number of voxels in the gallbladder area in the original abdominal CT image. The gallbladder volume calculation formula is:

[0064] gallbladder_volume=voxel_volume*gallbladder_voxels, unit: mm 3 When the gallbladder area cannot be identified, the volume is 0mm 3 .

[0065] The size of the gallbladder is affected by its physiological state. For example, when the gallbladder is filled with bile, its size will increase accordingly; and when the bile is emptied, the gallbladder will shrink. Abnormal gallbladder size may indicate gallbladder disease or other health problems. For example, an enlarged gallbladder may be seen in diseases such as acute cholecystitis, gallstones, and gallbladder cancer; a shrinkage of the gallbladder may be seen in gallbladder atrophy. The gallbladder exists in states of filling and contraction. The present invention determines the size of the gallbladder by calculating the gallbladder volume and then setting a fixed threshold to determine whether there is no gallbladder, the gallbladder is shrunken, the gallbladder is of normal size, or the gallbladder is enlarged.

[0066] In another embodiment of the present invention, Figure 2 As shown, after performing part recognition on the original abdominal CT image to obtain the contour area of ​​the gallbladder part, the method further includes step S15:

[0067] S15. Perform abnormality detection on the CT value within the contour area of ​​the gallbladder, and detect whether there is an abnormal density shadow in the gallbladder based on whether the CT value is higher than a set CT threshold, thereby obtaining a gallbladder detection result.

[0068] The CT value of a normal gallbladder is generally between -10 and 30HU. Among them, -10 to 0HU: usually represents the density of bile in the gallbladder; 0 to 30HU: may include the density of the gallbladder wall and other components. Gallstones usually appear as high density on CT images, and the CT value can reach 300HU or higher, depending on the composition of the stones. Cholecystitis may cause thickening of the gallbladder wall, and its CT value will be higher than the normal range. In this embodiment, after performing site recognition on the original abdominal CT image to obtain the contour area of ​​the gallbladder site, the CT value in the contour area of ​​the gallbladder site can also be detected for abnormalities. If the CT value is higher than the set CT threshold, it is determined that the gallbladder density value is abnormal and an abnormal density shadow is present.

[0069] In another embodiment of the present invention, Figure 3 As shown, after obtaining the gallbladder test results, the method further includes the step of performing a quality assessment on the imaging diagnostic report corresponding to the original abdominal CT image. The specific implementation method includes the following steps:

[0070] S16. Performing a quality inspection on the imaging diagnostic report corresponding to the original abdominal CT image based on the gallbladder detection result obtained from the original abdominal CT image. Specifically, the inspection includes: performing a text description on the gallbladder detection result obtained from the original abdominal CT image, performing numerical processing on the text description, using different numerical values ​​to identify different text description states of the gallbladder part, and representing the obtained numerical values ​​as a first one-dimensional array; performing clause splitting on the text information seen in the imaging diagnostic report, and extracting the entity description of the gallbladder part from each clause to obtain an entity description of the gallbladder part; performing numerical processing on the entity description using the same numerical processing method, using different numerical values ​​to identify different entity description states of the gallbladder part, and representing the obtained numerical values ​​as a second one-dimensional array; and sequentially performing consistency verification on the elements at the same position in the first one-dimensional array and the second one-dimensional array.

[0071] In this embodiment, after obtaining the gallbladder detection result based on the abdominal CT image, the radiological imaging report corresponding to the image is obtained, and then the "image findings" text information is extracted from the radiological imaging report. The entire text is split into sentences, and the pre-trained entity extraction model is called for each sentence in turn to obtain the entity description of the gallbladder part. The output entity description of the gallbladder part is digitized to obtain a set of numerical representations. In a specific example: no gallbladder is represented by 0, and a visible gallbladder is represented by 1; a small gallbladder is represented by 0, a normal gallbladder size is represented by 1, and an enlarged gallbladder is represented by 2; no abnormal density shadow is represented by 0, and an abnormal density shadow is represented by 1. For example, [visible gallbladder, normal size, abnormal] is represented by [1, 1, 1]. Among them, the training steps of the entity extraction model specifically include: constructing a training data set, extracting text information seen in the image from the radiology report database, and extracting entities describing the parts of the text: using any annotation rule such as BIO, BIOES and BMES to tokenize the text information; based on the Chinese NLP pre-training model BERT training annotation data, realize the entity extraction of the description of the gallbladder in the text information seen in the image. For example, the text information in the report is input: "The gallbladder is not large, the wall is not thick, and there is no high-density stone shadow inside it." The entity extraction model is called to obtain the output: [Visible gallbladder, normal size, no abnormalities]; for another example, the text information in the report is input: "The gallbladder is not displayed", the entity extraction model is called to obtain the output: [No gallbladder is seen]; for another example, the text information in the report is input: "The gallbladder wall is slightly thickened, and multiple high-density stone shadows with a diameter of less than 12 mm are seen inside." The entity extraction model is called to obtain the output: [Visible gallbladder, normal size, abnormal].

[0072] Based on the same numerical processing rules, the text descriptions corresponding to the gallbladder test results are numerically processed to obtain a set of numerical representations, which are then compared with the numerical representations corresponding to the radiological imaging reports to verify whether the descriptions are consistent. If they are consistent, the quality inspection is passed. If they are inconsistent, the radiological imaging report is subjected to secondary verification and error correction reminders in the RIS graphic system.

[0073] The embodiment of the present invention uses a neural network model to identify the location of the gallbladder in the CT image and detect whether the gallbladder shape, size, and CT value are abnormal, and verify it with the text description of the gallbladder in the "image findings" in the radiology report. For example, if the CT image identifies that "the gallbladder has been removed", but the radiology report still describes "the gallbladder findings", it means that the radiology report is incorrectly written. For another example, if the CT image identifies that "the gallbladder size is normal", but the radiology report describes that "the gallbladder is not large", it means that the radiology report is incorrectly written. For another example, if the CT image identifies that "an abnormal density shadow is identified in the gallbladder area", but the radiology report describes that "no abnormal density shadow is found in the cystic cavity", it means that the radiology report is incorrectly written. The incorrectly written areas are marked in the RIS graphic system to achieve quality detection of the radiology report.

[0074] The artificial intelligence-assisted automatic gallbladder image detection method provided by the embodiment of the present invention not only realizes automatic gallbladder image detection, but also can assist in rapid analysis of CT images, effectively avoid the impact of the current state of the gallbladder on the accuracy of gallbladder location identification, and obtain more accurate gallbladder image detection results. It can also realize quality inspection of radiological image reports based on the gallbladder image detection results, ensure the accuracy and usability of radiological image reports, thereby reducing the risk of missed diagnosis and misdiagnosis, and improving the detection rate and diagnostic accuracy of diseases.

[0075] For simplicity of description, the method embodiments are described as a series of actions. However, those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because certain steps can be performed in other orders or simultaneously according to the embodiments of the present invention. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0076] Example 2

[0077] An embodiment of the present invention provides an artificial intelligence-assisted automatic gallbladder image detection system, which includes a functional module for implementing the artificial intelligence-assisted automatic gallbladder image detection method as described in any of the above items. Figure 4 The present invention schematically shows a schematic diagram of an artificial intelligence-assisted gallbladder imaging automatic detection system. Figure 4The system of the embodiment of the present invention specifically includes an image generation module 401, a part recognition module 402 and an abnormality detection module 403, wherein:

[0078] An image generation module 401 is configured to generate a corresponding generative image for the original abdominal CT image, so as to reconstruct a CT image of the gallbladder in a filled or contracted state into a CT image of the gallbladder in a filled state;

[0079] a part recognition module 402, configured to perform part recognition on the generated image to determine whether a gallbladder part exists in the generated image;

[0080] The part recognition module 402 is further configured to perform part recognition on the original abdominal CT image to obtain a contour area of ​​the gallbladder part when it is determined that the gallbladder part exists in the generated image;

[0081] The abnormality detection module 403 is used to calculate the gallbladder volume according to the contour area of ​​the gallbladder part, and detect whether the gallbladder size is abnormal according to the gallbladder volume to obtain a gallbladder detection result.

[0082] In an embodiment of the present invention, the system further includes a CT value detection module (not shown in the accompanying drawings). The CT value detection module is configured to, after the part recognition module 402 performs part recognition on the original abdominal CT image to obtain a contour area of ​​the gallbladder part, perform abnormality detection on the CT value within the contour area of ​​the gallbladder part, and detect whether the gallbladder has an abnormal density shadow based on whether the CT value is higher than a set CT threshold, thereby obtaining a gallbladder detection result.

[0083] In an embodiment of the present invention, the image generation module 401 is specifically configured to generate a corresponding generative image for an original abdominal CT image using a pre-trained GAN generative adversarial network model. The GAN generative adversarial network model includes a generator and a discriminator. The generator is configured to generate a generative image corresponding to the abdominal CT image, and the discriminator is configured to determine whether the input generative image is a true CT image of a filled gallbladder or a generative image generated by the generator. The training steps of the GAN generative adversarial network model include: constructing a dataset containing image pairs of different patients under different conditions, wherein the image pairs are constructed in such a manner that: a CT image of gallbladder contraction corresponds to a CT image of the same patient with a filled gallbladder; a CT image of gallbladder removal corresponds to a CT image of gallbladder removal; and a CT image of a filled gallbladder corresponds to a CT image of the same patient with a filled gallbladder; and the dataset is used to train the generative adversarial network to generate a CT image of the gallbladder filled state by inputting an abdominal CT image of a contracted gallbladder.

[0084] In an embodiment of the present invention, the abnormality detection module 403 is specifically used to mark the pixel values ​​of the gallbladder area in the original abdominal CT image according to the contour area of ​​the gallbladder area, and count the number of voxels in the gallbladder area according to the marking result; obtain the pixel spacing and slice thickness of the original abdominal CT image according to the tag information in the DICOM file corresponding to the original abdominal CT image, and calculate the voxel volume of each voxel in the original abdominal CT image according to the pixel spacing and slice thickness; calculate the gallbladder volume according to the voxel volume of each voxel and the number of voxels in the gallbladder area in the original abdominal CT image.

[0085] In an embodiment of the present invention, the system further includes a quality detection module not shown in the accompanying drawings, and the quality detection module is used to perform quality detection on the imaging diagnostic report corresponding to the original abdominal CT image based on the gallbladder detection result obtained from the original abdominal CT image.

[0086] Furthermore, the quality detection module is specifically used to provide a text description of the gallbladder detection results obtained from the original abdominal CT image, perform numerical processing on the text description, use different numerical values ​​to identify different text description states of the gallbladder part, and represent the obtained numerical values ​​as a first one-dimensional array; split the text information seen in the image in the imaging diagnosis report into clauses, and extract the entity description of the gallbladder part for each clause to obtain the entity description of the gallbladder part, use the same numerical processing method to perform numerical processing on the entity description, use different numerical values ​​to identify different entity description states of the gallbladder part, and represent the obtained numerical values ​​as a second one-dimensional array; and perform consistency verification on the elements at the same position in the first one-dimensional array and the second one-dimensional array in turn.

[0087] In the specific implementation process of the second embodiment, reference may be made to the first embodiment, and the corresponding technical effects are achieved.

[0088] Example 3

[0089] An embodiment of the present invention provides a computer program product having stored thereon an artificial intelligence-assisted automatic gallbladder image detection program. When the artificial intelligence-assisted automatic gallbladder image detection program is executed by a processor, the steps of the above artificial intelligence-assisted automatic gallbladder image detection method embodiment are implemented, for example Figure 1 Steps S11-S14 are shown.

[0090] In the specific implementation process of Example 3, reference may be made to Example 1, and the corresponding technical effects are achieved.

[0091] Example 4

[0092] The embodiment of the present application provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor to implement the steps in the above-mentioned artificial intelligence assisted gallbladder image automatic detection method embodiment, for example Figure 1 The steps S11-S14 are shown.

[0093] In the implementation process, the embodiment four can refer to the embodiment one, and has the corresponding technical effects.

[0094] In addition, those skilled in the art can understand that although some embodiments herein include certain features rather than other features included in other embodiments, the combination of features of different embodiments means to be within the scope of the present application and forms different embodiments. For example, any one of the claimed embodiments can be used in any combination.

[0095] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An artificial intelligence-assisted gallbladder imaging automatic detection method, characterized in that: The method comprises: generating corresponding generative images for the original abdominal CT image to reconstruct the CT image of the gallbladder in a filled or contracted state into a CT image of the gallbladder in a filled state; Performing part recognition on the generated image to determine whether a gallbladder part exists in the generated image; If the gallbladder part exists in the generated image, the original abdominal CT image is subjected to part recognition to obtain the contour area of ​​the gallbladder part; The gallbladder volume is calculated based on the contour area of ​​the gallbladder, and the gallbladder size is detected to determine whether it is abnormal, thereby obtaining the gallbladder test results.

2. The method according to claim 1, characterized in that After performing part recognition on the original abdominal CT image to obtain the contour area of ​​the gallbladder part, the method further includes: The CT value within the contour area of ​​the gallbladder is detected for abnormalities. The gallbladder is detected for abnormal density based on whether the CT value is higher than the set CT threshold, and the gallbladder detection result is obtained.

3. The method according to claim 1, characterized in that The corresponding generative images generated for the original abdominal CT image include: A pre-trained GAN generative adversarial network model is used to generate corresponding generative images for the original abdominal CT images. The GAN generative adversarial network model includes a generator and a discriminator. The generator is used to generate generative images corresponding to the abdominal CT images, and the discriminator is used to determine whether the input generative image is a real CT image of the gallbladder filling state or a generative image generated by the generator. The training steps of the GAN generative adversarial network model include: constructing a data set, the data set containing image pairs of different patients in different conditions, wherein the image pairs are constructed in such a manner that a CT image of gallbladder contraction corresponds to a CT image of gallbladder filling of the same patient; a CT image of cholecystectomy corresponds to a CT image of cholecystectomy; and a CT image of gallbladder filling corresponds to a CT image of gallbladder filling of the same patient; and using the data set to train the generative adversarial network to generate a CT image of the gallbladder filling by inputting an abdominal CT image of gallbladder contraction.

4. The method according to claim 1, wherein Calculation of gallbladder volume based on the contour area of ​​the gallbladder includes: The pixel values ​​of the gallbladder area in the original abdominal CT image are marked according to the contour area of ​​the gallbladder area, and the number of voxels in the gallbladder area is counted according to the marking results; Obtaining the pixel spacing and slice thickness of the original abdominal CT image according to tag information in the DICOM file corresponding to the original abdominal CT image, and calculating the voxel volume of each voxel in the original abdominal CT image according to the pixel spacing and slice thickness; The gallbladder volume was calculated based on the voxel volume of each voxel and the number of voxels in the gallbladder area in the original abdominal CT images.

5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: The quality of the imaging diagnosis report corresponding to the original abdominal CT image is checked according to the gallbladder detection result obtained from the original abdominal CT image.

6. The method according to claim 5, characterized in that The quality inspection of the imaging diagnostic report corresponding to the original abdominal CT image based on the gallbladder test results obtained from the original abdominal CT image includes: Performing a text description on the gallbladder detection result obtained from the original abdominal CT image, performing numerical processing on the text description to identify different text description states of the gallbladder part using different numerical values, and representing the obtained numerical values ​​as a first one-dimensional array; The text information of the imaging findings in the imaging diagnosis report is split into clauses, and the entity description of the gallbladder location is extracted from each clause to obtain the entity description of the gallbladder location; The entity description is digitized using the same digitization method, so that different entity description states of the gallbladder part are marked with different digits, and the obtained digits are expressed as a second one-dimensional array; Consistency verification is performed on the elements at the same position in the first one-dimensional array and the second one-dimensional array in sequence.

7. An artificial intelligence-assisted gallbladder imaging automatic detection system, characterized in that: The system includes a functional module for implementing the artificial intelligence-assisted automatic gallbladder imaging detection method according to any one of claims 1 to 6, specifically including: An image generation module is used to generate a corresponding generative image for the original abdominal CT image, so as to reconstruct the CT image of the gallbladder in a filled or contracted state into a CT image of the gallbladder in a filled state; a part recognition module, configured to perform part recognition on the generated image to determine whether a gallbladder part exists in the generated image; The part recognition module is further configured to perform part recognition on the original abdominal CT image to obtain a contour area of ​​the gallbladder part when it is determined that the gallbladder part exists in the generated image; The abnormality detection module is used to calculate the gallbladder volume according to the contour area of ​​the gallbladder part, and to detect whether the gallbladder size is abnormal according to the gallbladder volume to obtain the gallbladder detection result.

8. The system according to claim 7, characterized in that The system further comprises: The quality detection module is used to perform quality detection on the imaging diagnosis report corresponding to the original abdominal CT image according to the gallbladder detection result obtained from the original abdominal CT image.

9. An electronic device, characterized in that: The electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor; When the computer program is executed by the processor, the steps of the artificial intelligence-assisted automatic gallbladder imaging detection method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product stores an artificial intelligence-assisted automatic gallbladder image detection program. When the artificial intelligence-assisted automatic gallbladder image detection program is executed by the processor, the steps of the artificial intelligence-assisted automatic gallbladder image detection method as described in any one of claims 1 to 6 are implemented.

Images