Image analysis method, device, equipment and storage medium
Through the timing image analysis method, the abnormal areas are determined in reverse using the attribute data of the normal area, which solves the problem of disease changes and the poor applicability of new lesions in the prior art, and improves the reliability and accuracy of medical imaging analysis.
Patent Information
- Application Number
- CN202010500297.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-06-04
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2040-06-04
AI Technical Summary
The existing medical imaging analysis methods rely on initial lesion data and are difficult to adapt to changes in disease types and new lesions, resulting in reduced reliability and accuracy of analysis results, especially in the case of dynamic changes in lung diseases.
By acquiring the timing image, the normal area in the region of interest is determined, and the information of the abnormal area is reversely determined based on the attribute data of the normal area, and the quantitative analysis of image analysis results is realized, which is suitable for the analysis of different stages and new lesions.
It improves the reliability and accuracy of image analysis results, is suitable for medical image analysis of special stages and new lesions, reduces dependence on initial lesions data, and enhances the applicability of medical analysis.
Smart Images

Figure CN111815557B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing technology, and in particular to an image analysis method, device, equipment and storage medium. Background Art
[0002] With the development of artificial intelligence technology, using traditional machine learning or deep learning to analyze images has gradually become a research trend in the field of image processing.
[0003] In the related technologies, in the medical field, when using artificial intelligence technology to analyze medical images, it is often necessary to collect enough data containing lesions, and then build a corresponding lesion processing model based on the collected lesion data, and then perform image analysis tasks such as segmentation and detection based on the constructed lesion processing model.
[0004] It can be seen that the above-mentioned medical image analysis needs to rely on the initial lesion data. However, for some parts with more types of diseases (such as the lungs, etc.), it not only takes a lot of manpower and material resources to collect data that includes most of the diseases; but also can only perform image analysis on a single disease or a few specific diseases. If the disease corresponding to the lesion site changes, or the source of the lesion changes, or some new diseases appear, the lesion processing model originally constructed based on the initial lesion data will not be applicable, and thus will not be able to effectively and accurately perform the corresponding image analysis tasks. For example, there are usually many types of lung diseases, and the lesions corresponding to lung diseases are dynamically changing. For example, for pneumonia caused by viruses or bacteria, the image manifestations of lung lesions are different at different stages. Therefore, the lesion processing model constructed only by the initial lung lesion data will not be applicable to other stages of lung detection and analysis, which will reduce the reliability and accuracy of the medical analysis results and even delay the patient's condition. Summary of the invention
[0005] The present application provides an image analysis method, apparatus, device and storage medium to solve at least one of the above technical problems.
[0006] In one aspect, the present application provides an image analysis method, comprising:
[0007] Acquire a set of images to be analyzed, wherein the set of images to be analyzed includes at least one time-series image;
[0008] Determine a region of interest image corresponding to the target object in the at least one time-series image; the region of interest image includes a normal region and an abnormal region;
[0009] Determine a normal area in the region of interest image corresponding to the at least one time-series image;
[0010] Based on the attribute data of the normal area corresponding to the at least one time-series image, an image analysis result representing abnormal information of the abnormal area in the image set to be analyzed is determined.
[0011] On the other hand, an image analysis device is also provided, comprising:
[0012] An acquisition module, used for acquiring a set of images to be analyzed, wherein the set of images to be analyzed includes at least one time-series image;
[0013] A first determination module is used to determine a region of interest image corresponding to the target object in the at least one time-series image; the region of interest image includes a normal region and an abnormal region;
[0014] A second determination module, configured to determine a normal area in the region of interest image corresponding to the at least one time-series image;
[0015] The analysis module is used to determine an image analysis result representing abnormal information of an abnormal area in the image set to be analyzed based on the attribute data of a normal area corresponding to the at least one time-series image.
[0016] In some embodiments, the second determining module may include:
[0017] a determining unit, configured to segment the normal region in the region of interest image corresponding to the at least one time-series image using a trained image segmentation model, to obtain the normal region in the region of interest image corresponding to the at least one time-series image;
[0018] The image segmentation model is trained by using the analysis sample image set and the segmentation results of the corresponding normal areas of the target object as a training set.
[0019] In some embodiments, when the number of the time-series images is multiple, the analysis module may include:
[0020] A first attribute change determination unit is used to determine attribute change data of the corresponding normal area based on the attribute data of the normal area corresponding to the multiple time-series images; the attribute change data of the normal area represents the change of regional attributes of the normal area corresponding to the multiple time-series images at different time nodes;
[0021] a second attribute change determining unit, configured to determine attribute change data of abnormal areas corresponding to the plurality of time-series images based on the attribute change data of normal areas corresponding to the plurality of time-series images and the attribute data of the image of the region of interest;
[0022] An analysis unit, configured to perform quantitative analysis on the image set to be analyzed based on the attribute change data of the abnormal regions corresponding to the multiple time-series images, and determine an image analysis result representing abnormal information of the abnormal regions in the image set to be analyzed;
[0023] Wherein, the attribute change data includes regional volume change data and / or regional volume proportion change data.
[0024] In some embodiments, when the number of the time-series images is multiple, the analysis module may include:
[0025] A preprocessing unit, used to preprocess the region of interest image corresponding to the first time-series image and the region of interest image corresponding to the second time-series image, respectively, to obtain a first time-series processed image and a second time-series processed image;
[0026] A first attribute determination unit, configured to determine first attribute data of a corresponding abnormal region based on attribute data of a normal region corresponding to the first time-series processed image, wherein the first attribute data includes first region position data and first region size data;
[0027] A second attribute determination unit, configured to determine second attribute data of a corresponding abnormal region based on the attribute data of a normal region corresponding to the second time-series processed image, wherein the second attribute data includes second region position data and second region size data;
[0028] Based on the first region position data and the second region position data, position matching is performed on the abnormal region in the first time series processed image and the abnormal region in the second time series processed image to obtain a region matching mapping relationship;
[0029] A quantitative analysis unit, configured to perform quantitative analysis on the first time series processed image and the second time series processed image based on the region matching mapping relationship, the first region size data and the second region size data, and determine an image analysis result representing abnormal change information of abnormal regions in the image set to be analyzed;
[0030] The acquisition time node of the first time-series image is earlier than the acquisition time node of the second time-series image, and the number of the first time-series image and the number of the second time-series image are at least one.
[0031] In some embodiments, the image analysis device may further include:
[0032] A detection module, used to detect pixel values of each region in the region of interest image corresponding to the at least one time-series image;
[0033] An abnormal region determination module, used to determine an abnormal region in the region of interest image corresponding to the at least one time-series image based on a comparison result between a pixel value of each region and a preset pixel threshold;
[0034] A first region attribute determination module, used to determine the region attribute of the abnormal region in the region of interest image corresponding to the at least one time-series image, wherein the region attribute includes the region size and / or the region quantity;
[0035] A first marking type determination module, configured to determine a marking type corresponding to each time-series image according to a comparison result between a region attribute of an abnormal region in a region-of-interest image corresponding to the at least one time-series image and a preset attribute threshold;
[0036] The first marking module is used to mark the at least one time-series image based on the determined marking type corresponding to the at least one time-series image.
[0037] In some embodiments, the image analysis device may further include:
[0038] A second region attribute determination module is used to determine the region attribute of the abnormal region in the region of interest image corresponding to the at least one time-series image based on the normal region in the region of interest image corresponding to the at least one time-series image; the region attribute includes region size and / or region number;
[0039] A second marking type determination module, configured to determine a marking type corresponding to each time-series image according to a comparison result between a region attribute of an abnormal region in a region-of-interest image corresponding to the at least one time-series image and a preset attribute threshold;
[0040] The second marking module is used to mark the at least one time-series image based on the determined marking type corresponding to the at least one time-series image.
[0041] In some embodiments, when the region attributes include region size and region quantity, the first marker type determination module or the second marker type determination module may include:
[0042] A first marking type determining unit, configured to determine that the corresponding time series image is of a first marking type if the size of the abnormal area is greater than or equal to a preset size threshold; the first marking type is used to indicate a normal area in a marking region of interest image;
[0043] A second marking type determination unit, configured to determine that the corresponding time series image is of a second marking type if the size of the abnormal region is smaller than a preset size threshold and the number of regions of the abnormal region is greater than or equal to a preset number threshold; the second marking type is used to indicate marking a normal region or an abnormal region in the region of interest image according to a comparison result of a ratio of a region attribute of the abnormal region in the region of interest and a preset ratio threshold;
[0044] The third marking type determining unit is used to determine that the corresponding time series image is of the third marking type if the number of abnormal regions is less than a preset number threshold; the third marking type is used to indicate abnormal regions in the marking region of interest image.
[0045] On the other hand, a computer device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the image analysis method as described in any one of the above items.
[0046] On the other hand, a computer storage medium is provided, wherein at least one instruction or at least one program is stored in the computer storage medium, and the at least one instruction or at least one program is loaded and executed by a processor to implement the image analysis method as described in any one of the above items.
[0047] The image analysis method, device, equipment and storage medium provided by this application have the following technical effects:
[0048] The embodiment of the present application obtains a set of images to be analyzed, which includes at least one time-series image; determines a region of interest image corresponding to a target object in at least one time-series image; the region of interest image includes a normal region and an abnormal region; determines a normal region in the region of interest image corresponding to at least one time-series image; and determines an image analysis result representing abnormal information of the abnormal region in the image set to be analyzed based on attribute data of the normal region corresponding to at least one time-series image. The present application does not directly use the feature data of the abnormal region to segment the abnormal region in the time-series image, but reversely determines the corresponding abnormal region through the normal region in the region of interest image, so that the image analysis method provided is not limited to the feature data of the existing abnormal region, and can be applied to image analysis of other stages, thereby improving the reliability and accuracy of the image analysis results. In particular, in the medical field, the image analysis method can be applied to the image analysis of special stage images or new lesions, thereby improving the reliability and accuracy of medical analysis results. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0050] Figure 1 It is a flowchart of an image analysis method provided in an embodiment of the present application;
[0051] Figure 2 is a flow chart of another image analysis method provided in an embodiment of the present application;
[0052] Figure 3 is a flow chart of step S107 in an image analysis method provided in an embodiment of the present application;
[0053] Figure 4 It is a flowchart of another image analysis method provided in an embodiment of the present application;
[0054] Figure 5 It is a flowchart of another image analysis method provided in an embodiment of the present application;
[0055] Figure 6 It is a flowchart of another image analysis method provided in an embodiment of the present application;
[0056] Figure 7 is a schematic diagram of the process of an image analysis method provided in an embodiment of the present application;
[0057] Figure 8 is a schematic diagram of the process of another image analysis method provided in an embodiment of the present application;
[0058] Fig. 9 is a structural block diagram of an image analysis device provided in an embodiment of the present application;
[0059] Fig.10 It is a hardware structure diagram of a computer device for implementing the method provided in an embodiment of the present application. DETAILED DESCRIPTION
[0060] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.
[0061] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.
[0062] The following describes a specific embodiment of an image analysis method of the present application. The method can be applied to an image analysis device, which can be implemented in software and / or hardware, and the device can be independent of an electronic device or integrated in an electronic device with an image analysis function, such as a server, a terminal, an image imaging system, etc.
[0063] Among them, the server may include but is not limited to at least one of an independent server, a distributed server, a cloud server, and a server cluster. The distributed server may specifically be a blockchain structure, and any node in the blockchain structure may execute or participate in the execution of the image analysis method. The terminal may include but is not limited to software running in a physical device, such as an application or client installed on the device, and may also include but is not limited to at least one of the types of physical devices such as a desktop computer, a tablet computer, a laptop computer, a smart phone, etc. with an application installed. Taking the medical field as an example, the image imaging system may include but is not limited to a medical imaging system of at least one modality such as an X-ray imaging system, a computed tomography (CT) system, a magnetic resonance imaging (MRI) system, a positron emission tomography (PET) system, etc.
[0064] Figure 1 It is a flowchart of an image analysis method provided in an embodiment of the present application. The present application provides method operation steps as described in the embodiment or flowchart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many orders, and does not represent the only order of execution. When the actual system or product is executed, it can be executed in sequence or in parallel (for example, in a parallel processor or multi-threaded processing environment) according to the method shown in the embodiment or the drawings. Specifically, Figure 1 As shown, the method may include:
[0065] S101, obtaining a set of images to be analyzed, where the set of images to be analyzed includes at least one time-series image.
[0066] In this embodiment, the terminal can obtain at least one time-series image of the subject acquired by the image acquisition device to form an image set to be analyzed. The time-series image may be a two-dimensional image or a three-dimensional image obtained at different time points. If the number of time-series images is multiple, the model of the image acquisition device may be the same or different, and the modality of the image acquisition device may also be the same or different, and the present application does not make specific restrictions on this. Of course, the image set to be analyzed may also be pre-acquired by the image acquisition device and pre-stored in a server or cloud, for example. When the terminal needs to analyze the image set to be analyzed, it can directly obtain it from the corresponding storage server or cloud, and the present application does not make specific restrictions on this.
[0067] As an example only, in the medical field, the subject may be a biological body such as a human body or an animal; accordingly, the image set to be analyzed is a medical image set constructed by at least one time-series image acquired by using a medical acquisition device to acquire a specific part of the subject, and the medical image set includes but is not limited to a CT image set, an X-ray image set, an MRI image set, etc. In other fields, the subject may be an object to be detected, a component, etc., and accordingly, the image set to be analyzed may be acquired by acquiring the subject through an image acquisition device such as a camera, a video camera, a scanning electron microscope, a transmission electron microscope, a microscope, etc.
[0068] S103, determining a region of interest image corresponding to the target object in at least one time-series image; the region of interest image includes a normal region and an abnormal region.
[0069] The target object may be a detection target for image detection and analysis, and the number of the target object may be one or more. Taking the medical field as an example, the target object may be an organ or tissue of a living organism, including but not limited to at least one of the lungs, liver, heart, kidneys, spleen, stomach, gallbladder, reproductive organs, etc. The embodiments of the present application are mainly described by taking the lung as an example of the target object, but this should not be construed as a limitation on the present application.
[0070] In this embodiment, the region of interest image includes a normal region and an abnormal region. The normal region is used to reflect the region of the target object in the region of interest image in a normal state; the abnormal region is used to reflect the region of the target object in the region of interest image in an abnormal state. Exemplarily, in the medical field, the abnormal region is the region corresponding to the lesion organ or lesion tissue, such as a nodule region, a cyst region, a tumor region, etc.; the normal region can be the region corresponding to the organ or tissue in the target object except the abnormal region.
[0071] Typically, each time-series image in the acquired image set to be analyzed includes not only the region of interest corresponding to the target object, but also the background region other than the region where the target object is located. After acquiring the image set to be analyzed, the terminal may perform background region removal processing on the image set to be analyzed to obtain an image corresponding to the region of interest corresponding to the target object.
[0072] In an optional embodiment, the terminal may use the target object segmentation model to segment the target object in at least one time-series image to obtain the region of interest of the target object, and then remove the background area other than the region of interest in the at least one time-series image to obtain an image containing only the region of interest corresponding to the target object, that is, the region of interest image. At this time, the region of interest image may be a segmentation mask image, and the segmentation mask image may specifically be a binary image. For example, the binary value of the region of interest in the segmentation mask image may be 1, and the binary value of the background region may be 0. Of course, in practical applications, the binary values of different regions may be intuitively displayed using different colors.
[0073] In another optional embodiment, the terminal can also obtain attribute information such as feature attributes, position attributes or contour attributes of the target object based on prior knowledge, and then determine the target object based on the prior knowledge, and then obtain the region of interest corresponding to the target object from at least one time-series image, that is, obtain an image that only contains the region of interest corresponding to the target object, that is, the region of interest image.
[0074] In practical applications, taking the medical field as an example, the region of interest image may be an image corresponding to a substantial area of the target object, such as a lung parenchyma image (i.e., a whole lung image), specifically a left lung image, a right lung image, or a double lung image. The terminal may use the whole lung segmentation model to segment the lungs in the image set to be analyzed, and output a whole lung image, i.e., a region of interest image corresponding to the lungs.
[0075] S105: Determine a normal area in a region of interest image corresponding to at least one time-series image.
[0076] In the field of image analysis, related technologies usually directly use a pre-built model to detect or identify abnormal areas in the region of interest after acquiring the image of the region of interest, and then obtain corresponding image analysis results based on the abnormal areas detected or analyzed. However, in some cases, if the characteristic attributes of the abnormal area have changed, that is, the abnormal area is a new type of abnormal area, the previously built model will not be able to effectively detect or identify the new type of abnormal area, thereby reducing the accuracy and reliability of the image analysis results.
[0077] In this embodiment, after determining the region of interest image corresponding to the target object in at least one time-series image, the terminal does not directly determine the corresponding abnormal region from the region of interest image, but determines the corresponding normal region from the region of interest image, and then obtains the image analysis result corresponding to the abnormal region in reverse based on the determined normal region.
[0078] In an optional embodiment, the terminal may use a preset image segmentation algorithm to segment the normal area in the region of interest image corresponding to at least one time-series image, thereby obtaining the corresponding normal area.
[0079] In another optional embodiment, the terminal may use an image segmentation model to determine a normal region in an image of a region of interest corresponding to at least one time-series image. The image segmentation model is used to segment a normal region from an image of interest corresponding to the time-series image, so as to subsequently determine an image analysis result of a corresponding abnormal region based on the segmented normal region.
[0080] Optionally, the image segmentation model may be a neural network model, such as a deep learning model. The deep learning model includes but is not limited to models such as convolutional neural networks (CNN), deep neural networks (DNN), and recurrent neural networks (RNN). The convolutional neural network model may include but is not limited to one or more of a V-Net model, a DenseNet model, a FPN model, and a U-Net model.
[0081] Optionally, the image segmentation model may be a separate neural network or a cascade network composed of at least two neural networks with the same or different network structures. If the image segmentation model is a cascade network, the input of the next neural network at least includes the output of the previous neural network, and the cascade network is used to perform cascade segmentation processing on the region of interest image respectively, and the output results of each cascade network are combined to obtain the normal area in each region of interest image.
[0082] As just an example, the above S105 determining the normal area in the region of interest image corresponding to the at least one time-series image may include:
[0083] The trained image segmentation model is used to segment the normal area in the region of interest image corresponding to at least one time series image to obtain the normal area in the region of interest image corresponding to at least one time series image. The image segmentation model is trained by using the segmentation results of the normal area in the analysis sample image set and the corresponding target object as a training set.
[0084] In practical applications, taking the target object as the lungs as an example, after the terminal obtains the region of interest image corresponding to the lungs, it can use at least one time series image of the lung region of interest as a model input, input it into a trained image segmentation model, and use the image segmentation model to segment the normal area in the lungs, and output a segmented image corresponding to the normal area of at least one lung region of interest image.
[0085] S107, determining an image analysis result representing abnormal information of an abnormal area in the image set to be analyzed based on attribute data of a normal area corresponding to at least one time-series image.
[0086] After determining the normal area of the target object in at least one time-series image, attribute data corresponding to the normal area of the target object can be obtained, and then an image analysis result representing abnormal information of the abnormal area in the image set to be analyzed is determined based on the attribute data corresponding to the normal area of the target object.
[0087] The attribute data corresponding to the normal area may be a region attribute state used to characterize the normal area, for example, including but not limited to region position data, region size data, region shape data, etc. The region size data may include region area, region volume, etc.
[0088] The image analysis result may include one or more of an image analysis result characterizing abnormal information of an abnormal region in at least one time-series image, and a follow-up analysis result of qualitative and quantitative analysis characterizing abnormal change information of each abnormal region in multiple time-series images.
[0089] For example, in the medical field, if the number of time-series images is one, the attribute data of the abnormal area can be determined based on the attribute data of the normal area of the target object in the time-series image, combined with the attribute data of the area of interest in the time-series image, and then the image analysis result representing the abnormal information of the abnormal area in the time-series image can be determined. If the number of time-series images is multiple, in addition to determining the image analysis result of the abnormal information of the abnormal area in at least one time-series image, the change of the abnormal area in multiple time-series images can also be determined to obtain the follow-up analysis results of the qualitative and quantitative analysis of the abnormal change information.
[0090] The above-mentioned embodiment obtains a set of images to be analyzed, which includes at least one time-series image; determines a region of interest image corresponding to a target object in at least one time-series image; the region of interest image includes a normal region and an abnormal region; determines a normal region in the region of interest image corresponding to at least one time-series image; and determines an image analysis result representing abnormal information of abnormal regions in the set of images to be analyzed based on attribute data of the normal region corresponding to at least one time-series image. The present application does not directly segment abnormal regions in time-series images using feature data of abnormal regions, but reversely determines corresponding abnormal regions through normal regions in the region of interest images, so that the provided image analysis method is not limited to feature data of abnormal regions that are only applicable to the current stage, and can be applied to image analysis at other stages, thereby improving the reliability and accuracy of image analysis results.
[0091] Especially in the medical field, the above image analysis method can be applied to medical images of special stages or medical images with new lesions, thereby improving the reliability and accuracy of the medical image analysis results. In addition, the present application does not require the detection of lesions, and by quantifying and calculating the normal areas of the same patient at different stages, the purpose of quantitative follow-up of abnormal areas can be achieved.
[0092] In an alternative embodiment, if Figure 2 As shown, when there are multiple time-series images, the above S107 determines the image analysis result representing the abnormal information of the abnormal area in the image set to be analyzed based on the attribute data of the normal area corresponding to at least one time-series image, which may specifically include:
[0093] S201, determining attribute change data of the corresponding normal region based on attribute data of the normal region corresponding to the multiple time-series images; the attribute change data of the normal region represents the change of regional attributes of the normal region corresponding to the multiple time-series images at different time nodes.
[0094] S203, determining the attribute change data of the abnormal area corresponding to the multiple time-series images based on the attribute change data of the normal area corresponding to the multiple time-series images and the attribute data of the region of interest image.
[0095] S205 , performing quantitative analysis on the image set to be analyzed based on the attribute change data of the abnormal regions corresponding to the multiple time-series images, and determining an image analysis result representing abnormal information of the abnormal regions in the image set to be analyzed.
[0096] The attribute change data may include regional volume change data and / or regional volume percentage change data. That is, the attribute change data of the normal region includes regional volume change data and / or regional volume percentage change data of the normal region; the attribute change data of the abnormal region includes regional volume change data and / or regional volume percentage change data of the abnormal region. Of course, the attribute change data is not limited to this, and may also include, for example, regional weight change data, and other regional size change data.
[0097] Specifically, after determining the normal area in the region of interest image corresponding to multiple time-series images, the terminal can use an image detection tool or an image detection algorithm to obtain the attribute data of the normal area corresponding to the multiple time-series images, and then extract the time-series images corresponding to the same target object belonging to the same subject based on the attribute data of the normal area corresponding to the multiple time-series images, calculate the attribute change data of the normal area of the same target object of the extracted multiple time-series images, and then determine the attribute change data of the abnormal area corresponding to the multiple time-series images in combination with the attribute data of the region of interest image; then, perform quantitative analysis on the image set to be analyzed based on the determined attribute change data of the abnormal area, and determine the image analysis results that characterize the abnormal information of the abnormal area in the image set to be analyzed.
[0098] It should be understood that when determining the image analysis result of the abnormal area, it can be achieved by first extracting multiple abnormal areas from the area of interest based on the attribute change data of the normal area of the area of interest, and then obtaining the corresponding image analysis results based on the attribute change data of the multiple abnormal areas. Alternatively, it can also be achieved by directly obtaining the attribute change data of the abnormal area based on the attribute change data of the normal area of the area of interest without extracting the abnormal area, and then obtaining the corresponding image analysis result based on the attribute change data of the abnormal area, and this application does not make specific limitations on this.
[0099] In practical applications, taking the target object as the lung as an example, the terminal obtains multiple lung CT images of the same patient at different times, extracts the lung parenchymal area in each lung CT image to obtain a lung parenchymal area image, determines the normal area representing normal lung development from the lung parenchymal area image, and obtains attribute data of normal areas corresponding to the multiple lung CT images based on an image detection tool or an image detection algorithm, such as area size, area volume, etc.; then, calculates the change data of the attribute data of the normal areas corresponding to the multiple lung CT images to obtain the attribute change data of the corresponding normal areas, such as area size change data, area volume change data, etc. Then, the attribute data of the lung parenchymal area image, such as the total area size and total area volume of the lung parenchymal area, is combined with the calculated attribute change data of the normal area to perform a subtraction operation or a complement operation to obtain the attribute change data of the lesion area in multiple lung parenchymal area images, and then the image analysis results characterizing the abnormal information of the lesion area in the image set to be analyzed are determined based on the attribute change data of the lesion area, so as to determine the lesion changes of the patient at different periods, achieve the purpose of tracking and following up the lesion area, and then assist the doctor to understand the patient's condition in time according to the lesion changes, which is conducive to improving the accuracy and reliability of the image analysis results.
[0100] In the above embodiment, the property changes of the normal area in the region of interest image can be reversely determined to achieve the purpose of quantitative follow-up.
[0101] In another alternative embodiment, if Figure 3 As shown, when there are multiple time-series images, the above S107 determines the image analysis result representing the abnormal information of the abnormal area in the image set to be analyzed based on the attribute data of the normal area corresponding to at least one time-series image, which may specifically include:
[0102] S301 , preprocessing the region of interest image corresponding to the first time-series image and the region of interest image corresponding to the second time-series image respectively to obtain a first time-series processed image and a second time-series processed image.
[0103] The acquisition time node of the first time sequence image is earlier than the acquisition time node of the second time sequence image, and the first time sequence image and the second time sequence image are images of the same subject acquired at different acquisition time nodes. The number of the first time sequence image and the second time sequence image can be at least one.
[0104] Optionally, the preprocessing includes but is not limited to one or more of denoising, data enhancement, resampling, normalization, etc. In practical applications, the specific process of preprocessing can be adaptively adjusted according to actual conditions, as long as the preprocessed images can be effectively compared and analyzed, and this application does not make specific limitations on this.
[0105] S303, based on the attribute data of the normal area corresponding to the first time-series processed image, determine first attribute data of the corresponding abnormal area, where the first attribute data includes first area position data and first area size data.
[0106] S305 , based on the attribute data of the normal area corresponding to the second time-series processed image, determine second attribute data of the corresponding abnormal area, where the second attribute data includes second area position data and second area size data.
[0107] The number of normal regions may be one or more. Adjacent normal regions may be connected to each other or may not be connected to each other. For example, taking the target object as the lung, if the number of normal regions is multiple, the multiple normal regions may not be connected to each other. Correspondingly, the number of abnormal regions may be one or more. If the number of abnormal regions is multiple, the multiple abnormal regions may not be connected to each other.
[0108] The first attribute data of the abnormal area is used to reflect the regional attributes of the abnormal area in the first time-series processed image, and the first attribute data may include first area position data and first area size data. Among them, the first area position data is used to reflect the position of each abnormal area in the first time-series processed image, which may include the coordinate data of the key points, the position coordinates of the detection box or outline of the abnormal area, etc. The key point includes but is not limited to the center point, centroid, etc. of the abnormal area. The first area size data is used to reflect the size of each abnormal area in the first time-series processed image, such as the area of the abnormal area, the volume of the abnormal area, etc. The second attribute data of the abnormal area is similar to the first attribute data of the abnormal area, and will not be repeated here.
[0109] Specifically, after obtaining the normal area corresponding to the first time-series processed image, the attribute data of the normal area corresponding to the first time-series processed image can be extracted using an image detection tool or an image detection algorithm, and then combined with the attribute data of the first time-series processed image, the first attribute data of each abnormal area except the normal area in the first time-series processed image can be obtained, and then the first area position data and first area size data of each abnormal area in the first time-series processed image can be obtained.
[0110] Similarly, after obtaining the normal area corresponding to the second time-series processed image, the attribute data of the normal area corresponding to the second time-series processed image can be extracted using an image detection tool or an image detection algorithm, and then combined with the attribute data of the second time-series processed image to obtain the second attribute data of each abnormal area except the normal area in the second time-series processed image, and then obtain the second area position data and second area size data of each abnormal area in the second time-series processed image.
[0111] S307 , based on the first region position data and the second region position data, position matching is performed on the abnormal region in the first time series processed image and the abnormal region in the second time series processed image to obtain a region matching mapping relationship.
[0112] Optionally, after obtaining the first region position data and the second region position data, a position matching algorithm may be used to match the coordinate positions of key points in the abnormal region in the first time series processed image and the abnormal region in the second time series processed image to obtain a region matching mapping relationship. That is, for each first key point in the abnormal region in the first time series processed image, a corresponding matching key point is found in the second key point in the abnormal region in the second time series processed image.
[0113] If the position correlation between the first key point and the second key point is greater than or equal to the preset correlation threshold, the first key point and the second key point are determined to be a matching key point pair, thereby constructing a key point mapping relationship of the matching key point pair. The position correlation can be measured according to the coordinate similarity of the key points. On the contrary, if the position correlation between the first key point and the second key point is less than the preset correlation threshold, it is determined that there is no corresponding matching key point for the first key point, and the matching key point corresponding to the first key point can be set to a null value. After traversing all the first key points, there are still extra unmatched second key points in the second time series processing image, and there is no corresponding matching key point for the second key point, and the matching key point corresponding to the second key point can be set to a null value.
[0114] Since each key point corresponds to an abnormal region one-to-one, the region matching mapping relationship between abnormal regions can be determined based on the key point mapping relationship. The region matching mapping relationship can be embodied in the form of a mapping table, etc., which is not specifically limited in this application.
[0115] S309, based on the region matching mapping relationship, the first region size data and the second region size data, quantitatively analyze the first time series processed image and the second time series processed image to determine an image analysis result representing abnormal change information of the abnormal region in the image set to be analyzed.
[0116] After determining the regional matching mapping relationship, the regional size change data of the abnormal area in the first time series processing image that matches the abnormal area in the second time series processing image can be determined, and then the change range of the abnormal area proportion can be determined based on the regional size change data, so as to achieve the purpose of follow-up quantitative analysis of the progress of the abnormal area.
[0117] In practical applications, taking the target part as the lung as an example, the segmentation mask of the whole lung is performed for the time series images collected at different stages of the same patient, and its minimum three-dimensional external rectangular frame is obtained. The rectangular frames corresponding to different time series images are named sample1 and sample2, and sample1 and sample2 are resampled to the same size. After that, the lesion area in the resampled rectangular frame image is determined, the key points and area size of each lesion area are obtained, and the position association information of the key points of each lesion area of the resampled sample1 and sample2 is calculated. Assume that the two lesion areas in the resampled sample1 are box1_a and box1_b, and the two lesion areas in sample2 are box2_a and box2_b. When the position correlation of the key points in box1_a and box2_a is greater than a certain preset correlation threshold, it is considered that box1_a and box2_a have a regional matching relationship, and box2_a is developed from box1_a. If the key point correlation of box2_a to all lesion areas in sample1 is lower than a certain preset correlation threshold, box2_a is considered to be a newly added lesion area. If the correlation of box1_b with respect to the key points of all lesion areas in sample2 is lower than a certain preset correlation threshold, box1_b is considered to be the lesion area to be absorbed. After determining the lesion type of each lesion area in different stages, the regional size change data of the lesion area can be determined according to the regional size of the lesion area and the regional matching mapping relationship, and then the change range of the regional proportion of the same patient at different stages can be determined according to the regional size change data, so as to achieve the purpose of follow-up quantitative analysis of the patient's disease progression.
[0118] In the above-mentioned embodiment, a region matching mapping relationship is obtained by positionally matching the abnormal region in the first time series processed image with the abnormal region in the second time series processed image. Then, based on the region matching mapping relationship, the first region size data and the second region size data, a more accurate quantitative analysis is performed on the first time series processed image and the second time series processed image, thereby realizing quantitative follow-up analysis and improving the reliability and accuracy of the follow-up analysis results.
[0119] In some embodiments, the above-mentioned image analysis method may further include an image marking step.
[0120] In an alternative embodiment, if Figure 4 and 7 As shown, the image marking step may include:
[0121] S401, detecting pixel values of each region in a region-of-interest image corresponding to at least one time-series image.
[0122] After obtaining the region of interest image corresponding to at least one time-series image, the pixel value of each pixel in the region of interest image can be identified, and then each pixel is traversed and cluster analysis is performed, and the neighboring pixel points whose pixel difference between the pixel points meets the preset pixel difference value and the pixel point are divided into a set to obtain multiple pixel point sets, each of which can correspond to a region, and then the region of interest image is divided into multiple regions. Since the pixel value range of the pixel points in each region may be different, the pixel values of each pixel point in each region can be averaged to obtain the pixel mean corresponding to each region as the pixel value of each region in the region of interest image.
[0123] S403: Determine an abnormal region in a region of interest image corresponding to at least one time series image based on a comparison result between a pixel value of each region and a preset pixel threshold.
[0124] After determining the pixel value of each region in the region of interest image, the pixel value of each region may be compared with a preset pixel threshold to determine an abnormal region in the region of interest image corresponding to at least one time series image based on the comparison result.
[0125] Optionally, if the comparison result is that the pixel value of the region is greater than or equal to a preset pixel threshold, the region is determined to be an abnormal region in the region of interest image. Otherwise, the region is determined to be a normal region in the region of interest image.
[0126] It should be noted that, in practical applications, the pixel value can also be converted into a gray value or a CT value for calculation, so as to determine the abnormal area in the image of the region of interest corresponding to at least one time series image based on the comparison result of the gray value or CT value of each region with the preset gray threshold or the preset CT value threshold. The preset pixel threshold, the preset gray threshold, the preset CT value threshold and the comparison rule can be determined according to the actual situation and prior knowledge, and this application does not make specific limitations on this.
[0127] Taking the target object as the lung as an example, the region of interest image of the lung is usually a CT image, which is represented by different grayscales to reflect the degree of absorption of radiation by different tissues or parts. Normal areas in the lungs are usually low-density areas, and their CT values are usually low; while abnormal areas in the lungs (i.e., lesion areas) are high-density areas, and their CT values are usually high. Then, the abnormal area in the region of interest image of the lungs can be determined based on the CT value of the comparison area being greater than or equal to the preset CT threshold.
[0128] S405, determining a region attribute of an abnormal region in a region of interest image corresponding to at least one time-series image, where the region attribute includes a region size and / or a region quantity.
[0129] After determining the abnormal region in the region of interest image corresponding to at least one time-series image, the region attribute of each abnormal region can be counted, and the region attribute can include but is not limited to one or two sub-attributes of region size and region number. Of course, the region attribute can also include other suitable sub-attributes. Among them, the region size can be determined according to the image dimension, which can specifically be the region area, region volume, etc.
[0130] S407: Determine a marking type corresponding to each time-series image according to a comparison result of a region attribute of an abnormal region in a region-of-interest image corresponding to at least one time-series image with a preset attribute threshold.
[0131] After determining the regional attributes of the abnormal region in the region of interest image, each sub-attribute in the regional attributes is compared with the size of the corresponding preset attribute threshold, and the regional type of the abnormal region in each time series image is determined based on the comparison results of all sub-attributes, and then the marking type corresponding to each time series image is determined based on the regional type corresponding to the abnormal region. The regional type of the abnormal region may include a large area abnormal region, a small area abnormal region, a relatively small area abnormal region (such as a nodular lesion region), etc. The marking type may include, for example, an abnormal region marking type, a normal region marking type, a mixed region marking type, a personalized marking type, etc.
[0132] Among them, the specific value of the preset attribute threshold can be set and adjusted according to actual conditions, and this application does not make any specific limitations on this.
[0133] S409: Mark at least one time-series image based on the determined mark type corresponding to the at least one time-series image.
[0134] After determining the mark type corresponding to at least one time-series image, each time-series image may be differentially marked to present a differentiated image analysis result to the user.
[0135] It should be noted that the above steps S401-S403 can be performed after step S103, or after step S105 and before step S107. The above steps S405-S409 can be performed immediately after step S403, or can be separated and performed after step S105 or step S107. Of course, the above steps S401-S409 can also be performed in other appropriate orders, and this application does not specifically limit this.
[0136] In the above-mentioned embodiment, all abnormal areas in the area of interest image are determined by detecting the pixel values of each area in the area of interest image corresponding to at least one time series image, and then the area type to which the abnormal area belongs is determined based on the comparison result of the area attributes of all abnormal areas and the preset attribute threshold value, and then the marking type of the corresponding time series image is determined according to the area type, and then the time series image is differentially segmented and marked according to the marking type, so as to further improve the accuracy of the image analysis results.
[0137] In another alternative embodiment, if Figure 5 and 8 As shown, in order to perform more accurate image marking, the image marking step may specifically include:
[0138] S501, based on a normal area in a region of interest image corresponding to at least one time-series image, determine a region attribute of an abnormal area in a region of interest image corresponding to at least one time-series image; the region attribute includes a region size and / or a region quantity.
[0139] S503: Determine a marking type corresponding to each time-series image according to a comparison result of a region attribute of an abnormal region in a region-of-interest image corresponding to at least one time-series image with a preset attribute threshold.
[0140] S505: Mark at least one time-series image based on the determined mark type corresponding to the at least one time-series image.
[0141] Specifically, after executing the above step S105 and obtaining the normal area in the region of interest image corresponding to at least one time-series image, the abnormal area in the region of interest image can be directly determined in combination with the region of interest image, and then the region attributes of the abnormal area in the region of interest image are counted. The region attribute includes but is not limited to one or two sub-attributes of region size and region number. Of course, the region attribute may also include other suitable sub-attributes. Among them, the region size can be determined according to the image dimension, which can specifically be the region area, region volume, etc.
[0142] After determining the regional attributes of the abnormal area in the image of the region of interest, each sub-attribute in the regional attributes is compared with the size of the corresponding preset attribute threshold, and the marking type corresponding to each time-series image is determined based on the comparison results of all sub-attributes. The marking type may include, for example, an abnormal area marking type, a normal area marking type, a mixed area marking type, a personalized marking type, etc. Among them, the specific value of the preset attribute threshold can be set and adjusted according to actual conditions, and this application does not make specific restrictions on this. After determining the marking type corresponding to at least one time-series image, each time-series image can be differentially marked to present differentiated image marking results to the user.
[0143] In the above embodiment, the abnormal region is directly determined by the normal region in the region of interest image, and then the label type corresponding to the time series image is determined according to the region attribute of the abnormal region, thereby realizing differential labeling. Compared with the method of using the pixel value of the region in the above embodiment, the normal region is used to reversely determine the abnormal region in this embodiment, so that a more accurate and reliable abnormal region can be obtained, further improving the accuracy of the image analysis result.
[0144] This application does not directly use the feature data of the abnormal area to segment and mark the abnormal area in the time series image, but reversely marks the corresponding abnormal area through the normal area in the region of interest image, so that the image analysis method provided is not limited to the feature data of the abnormal area only applicable to the current stage, and can be applied to image analysis at other stages, thereby improving the reliability and accuracy of the image analysis results. In particular, in the medical field, the above-mentioned image analysis method can be applied to medical images at special stages or medical images with new lesions, thereby improving the reliability and accuracy of medical image analysis results.
[0145] In an alternative embodiment, if Figure 6 , 7 As shown in FIG. 8 , in the case where the region attribute includes the region size and the region number, the marking type corresponding to each time series image is determined based on the comparison result of the region attribute of the abnormal region in the region of interest image corresponding to at least one time series image with the preset attribute threshold, that is, the above S407 or S503 may specifically include:
[0146] S601, if the size of the abnormal area is greater than or equal to a preset size threshold, determine that the corresponding time series image is a first marking type; the first marking type is used to indicate a normal area in a marking region of interest image.
[0147] S603, if the area size of the abnormal area is smaller than the preset size threshold, and the area number of the abnormal area is greater than or equal to the preset number threshold, the corresponding time series image is determined to be a second marking type; the second marking type is used to indicate marking of a normal area or an abnormal area in the area of interest image based on a comparison result of a proportion of the area attribute of the abnormal area in the area of interest and a preset proportion threshold.
[0148] S605: If the number of abnormal regions is less than a preset number threshold, determine that the corresponding time series image is a third marking type; the third marking type is used to indicate abnormal regions in the region of interest image.
[0149] Specifically, after detecting the abnormal area of each time series image, the size and number of each abnormal area are counted, which can be divided into the following three cases:
[0150] 1) When the size of the abnormal area is greater than or equal to the preset size threshold a, the abnormal area of the time series image is determined to be a large-area abnormal area, and the corresponding time series image is set to a first marking type, which is used to indicate a normal area in the image of the region of interest.
[0151] 2) When the size of the abnormal area is smaller than the preset size threshold a and greater than 0, and the number of abnormal areas is greater than or equal to the preset number threshold b, the abnormal area of the time series image is determined to be a small area abnormal area, and the corresponding time series image is the second marking type.
[0152] The specific marking area in the second marking type can be determined according to the comparison result of the ratio of the area attribute of the abnormal area in the region of interest and the preset ratio threshold. Specifically, after determining that the abnormal area is a small-area abnormal area, the ratio of the area size of the small-area abnormal area in the region of interest is calculated, and then the ratio is compared with the preset ratio threshold t. If the comparison result is that the ratio is greater than or equal to the preset ratio threshold t, the normal area in the region of interest image is marked, otherwise, the abnormal area in the region of interest image is marked.
[0153] 3) When the size of the abnormal region is less than or equal to the preset size threshold a, and the number of the abnormal region is less than or equal to the preset number threshold b, the abnormal region of the time series image is determined to be a smaller abnormal region (such as a nodule or mass lesion in the medical field), and the corresponding time series image is of the third marking type. The third marking type is used to indicate the abnormal region in the marking region of interest image.
[0154] In practical applications, taking the medical field as an example, the preset size threshold a may be a preset volume size threshold, and its value may be set but not limited to 180-250 cm 3 Any value, for example 200cm 3 、220cm 3Etc. The value of the preset quantity threshold b can be set to, but not limited to, any value in the range of 6-12, such as 6, 8, 10, etc. The preset proportion threshold t can be a preset volume proportion threshold, and its value can be set to, but not limited to, any value in the range of 0.06-0.12, such as 0.08, 0.09, etc. Specifically, when the abnormal lesion area is large, the image segmentation model is used to segment the healthy area to obtain a normal area segmentation mask, and the segmented normal area segmentation mask is marked. Because the abnormal lesion area is large enough, it is easier to locate the corresponding lesion area, and there is no need to mark it with a mask. At this time, masks that mark too many lesion areas will also block the line of sight. At this time, the doctor's attention is more attracted to the area without mask markings, which can allow the doctor to focus more on observing subtle changes in the lesion area, improve the doctor's user experience, and effectively help the doctor read the film to determine whether the condition has worsened.
[0155] When the abnormal lesion area is not large, the image segmentation model is first used to segment the healthy area, and then the region of interest is used to perform a subtraction operation with the segmented healthy area to obtain the abnormal area segmentation mask, and the obtained abnormal area segmentation mask is marked.
[0156] Taking the medical field as an example, assume that the image of the lung region of interest is lung mask image A, which is a binary image, in which the mask label of the lung region of interest is 1, and the mask label of the background region is 0. Assume that the segmented image corresponding to the normal lung region is lung mask image B, which is also a binary image, in which the mask label of the healthy lung region is 1, and the mask label of other regions is 0. Subtract the lung mask image A from the lung mask image B, and numerically subtract the mask values of the corresponding regions in the two mask images to obtain the abnormal region mask image C. If the position with a label value of 1 in the abnormal region mask image C is the abnormal lung region; and the position with a label value of -1 or 0 in the abnormal region mask image C can be considered as the region other than the abnormal region, which is not the abnormal lung region.
[0157] For ease of understanding, combined Figure 7 A detailed description is given. For the same patient, the lung CT time series images of three stages, on the one hand, the label type of the lung CT time series images of each stage is obtained by detecting the lung abnormal area, that is, the CT time series images of stage 1 are of the third label type, the CT time series images of stage 2 are of the second label type, and the CT time series images of stage 3 are of the first label type. On the other hand, the lung segmentation model is used to segment the lung CT time series images of the three stages to obtain the whole lung mask of the three stages (such as Figure 7The three whole lung masks are then segmented into normal areas through the image segmentation model to obtain the normal lung area. According to the normal lung area in the lung CT time series images of each stage, the segmentation mask of each stage is differentially marked according to the label type corresponding to the lung CT time series images of each stage. Figure 7 For the marked part in the figure, the segmentation mask of stage 1 is marked according to the third marking type, that is, the segmentation mask of the abnormal lung area is marked (such as the small lesion area pointed by the white arrow in the figure). For the segmentation mask of stage 2, it is marked according to the second marking type. At this time, the normal lung area is marked according to the preset proportion threshold t, that is, the segmentation mask of the normal lung area is marked (such as the normal area pointed by the white arrow in the figure). For the segmentation mask of stage 3, it is marked according to the third marking type, that is, the segmentation mask of the normal lung area is marked (such as the normal area pointed by the white arrow in the figure).
[0158] Since the abnormal lesion area is small, the abnormal area segmentation mask will not block the doctor's vision too much. Moreover, because the abnormal area segmentation mask is marked, it will attract more attention. The doctor can locate the very small lesion area more easily and quickly, which helps the doctor read the film.
[0159] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.
[0160] Please refer to Fig. 9 , which shows a structural block diagram of an image analysis device provided in an embodiment of the present application. The device has the function of implementing the device in the above method example, and the function can be implemented by hardware or by hardware executing corresponding software. The image analysis device 90 may include:
[0161] An acquisition module 901 is used to acquire a set of images to be analyzed, wherein the set of images to be analyzed includes at least one time-series image;
[0162] A first determination module 902 is used to determine a region of interest image corresponding to the target object in the at least one time-series image; the region of interest image includes a normal region and an abnormal region;
[0163] A second determination module 903, configured to determine a normal area in the region of interest image corresponding to the at least one time-series image;
[0164] The analysis module 904 is used to determine an image analysis result representing abnormal information of an abnormal area in the image set to be analyzed based on the attribute data of the normal area corresponding to the at least one time-series image.
[0165] In some embodiments, the second determining module may include:
[0166] a determining unit, configured to segment the normal region in the region of interest image corresponding to the at least one time-series image using a trained image segmentation model, to obtain the normal region in the region of interest image corresponding to the at least one time-series image;
[0167] The image segmentation model is trained by using the analysis sample image set and the segmentation results of the corresponding normal areas of the target object as a training set.
[0168] In some embodiments, when the number of the time-series images is multiple, the analysis module may include:
[0169] A first attribute change determination unit is used to determine attribute change data of the corresponding normal area based on the attribute data of the normal area corresponding to the multiple time-series images; the attribute change data of the normal area represents the change of regional attributes of the normal area corresponding to the multiple time-series images at different time nodes;
[0170] a second attribute change determining unit, configured to determine attribute change data of abnormal areas corresponding to the plurality of time-series images based on the attribute change data of normal areas corresponding to the plurality of time-series images and the attribute data of the image of the region of interest;
[0171] An analysis unit, configured to perform quantitative analysis on the image set to be analyzed based on the attribute change data of the abnormal regions corresponding to the multiple time-series images, and determine an image analysis result representing abnormal information of the abnormal regions in the image set to be analyzed;
[0172] Wherein, the attribute change data includes regional volume change data and / or regional volume proportion change data.
[0173] In some embodiments, when the number of the time-series images is multiple, the analysis module may include:
[0174] A preprocessing unit, used to preprocess the region of interest image corresponding to the first time-series image and the region of interest image corresponding to the second time-series image, respectively, to obtain a first time-series processed image and a second time-series processed image;
[0175] A first attribute determination unit, configured to determine first attribute data of a corresponding abnormal region based on attribute data of a normal region corresponding to the first time-series processed image, wherein the first attribute data includes first region position data and first region size data;
[0176] A second attribute determination unit, configured to determine second attribute data of a corresponding abnormal region based on the attribute data of a normal region corresponding to the second time-series processed image, wherein the second attribute data includes second region position data and second region size data;
[0177] Based on the first region position data and the second region position data, position matching is performed on the abnormal region in the first time series processed image and the abnormal region in the second time series processed image to obtain a region matching mapping relationship;
[0178] A quantitative analysis unit, configured to perform quantitative analysis on the first time series processed image and the second time series processed image based on the region matching mapping relationship, the first region size data and the second region size data, and determine an image analysis result representing abnormal change information of abnormal regions in the image set to be analyzed;
[0179] The acquisition time node of the first time-series image is earlier than the acquisition time node of the second time-series image, and the number of the first time-series image and the number of the second time-series image are at least one.
[0180] In some embodiments, the image analysis device may further include:
[0181] A detection module, used to detect pixel values of each region in the region of interest image corresponding to the at least one time-series image;
[0182] An abnormal region determination module, used to determine an abnormal region in the region of interest image corresponding to the at least one time-series image based on a comparison result between a pixel value of each region and a preset pixel threshold;
[0183] A first region attribute determination module, used to determine the region attribute of the abnormal region in the region of interest image corresponding to the at least one time-series image, wherein the region attribute includes the region size and / or the region quantity;
[0184] A first marking type determination module, configured to determine a marking type corresponding to each time-series image according to a comparison result between a region attribute of an abnormal region in a region-of-interest image corresponding to the at least one time-series image and a preset attribute threshold;
[0185] The first marking module is used to mark the at least one time-series image based on the determined marking type corresponding to the at least one time-series image.
[0186] In some embodiments, the image analysis device may further include:
[0187] A second region attribute determination module is used to determine the region attribute of the abnormal region in the region of interest image corresponding to the at least one time-series image based on the normal region in the region of interest image corresponding to the at least one time-series image; the region attribute includes region size and / or region number;
[0188] A second marking type determination module, configured to determine a marking type corresponding to each time-series image according to a comparison result between a region attribute of an abnormal region in a region-of-interest image corresponding to the at least one time-series image and a preset attribute threshold;
[0189] The second marking module is used to mark the at least one time-series image based on the determined marking type corresponding to the at least one time-series image.
[0190] In some embodiments, when the region attributes include region size and region quantity, the first marker type determination module or the second marker type determination module may include:
[0191] A first marking type determining unit, configured to determine that the corresponding time series image is of a first marking type if the size of the abnormal area is greater than or equal to a preset size threshold; the first marking type is used to indicate a normal area in a marking region of interest image;
[0192] A second marking type determination unit, configured to determine that the corresponding time series image is of a second marking type if the size of the abnormal region is smaller than a preset size threshold and the number of regions of the abnormal region is greater than or equal to a preset number threshold; the second marking type is used to indicate marking a normal region or an abnormal region in the region of interest image according to a comparison result of a ratio of a region attribute of the abnormal region in the region of interest and a preset ratio threshold;
[0193] The third marking type determining unit is used to determine that the corresponding time series image is of the third marking type if the number of abnormal regions is less than a preset number threshold; the third marking type is used to indicate abnormal regions in the marking region of interest image.
[0194] It should be noted that the specific contents in the device embodiment of the present application can be found in the above method embodiment, which will not be repeated here.
[0195] An embodiment of the present application also provides a computer storage medium, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the image analysis method as described in any of the above embodiments.
[0196] The storage medium may be located in at least one of the multiple network devices of the computer network. Optionally, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.
[0197] An embodiment of the present application provides a computer device, which may include a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the image analysis method described in any of the above embodiments.
[0198] Further, Fig.10 A hardware structure diagram of a computer device for implementing the method provided in the embodiment of the present application is shown. The computer device may be a server, a computer terminal, a mobile terminal or other device. The computer device may also participate in or include the apparatus provided in the embodiment of the present application. Figure 6 As shown, the computer device 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 6 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 6 More or fewer components as shown, or with Figure 6 Different configurations are shown.
[0199] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer device 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).
[0200] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the methods described in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, to implement the above-mentioned neural network processing method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories may be connected to the computer device 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0201] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer device 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0202] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer device 10 .
[0203] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0204] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and server embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0205] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.
[0206] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.
Claims
1. An image analysis method, characterized in that: include: Acquire a set of images to be analyzed, wherein the set of images to be analyzed includes at least one time-series image; Determine a region of interest image corresponding to the target object in the at least one time-series image; The region of interest image includes a normal region and an abnormal region; Determine a normal area in the region of interest image corresponding to the at least one time-series image; Determine, based on the attribute data of the normal area corresponding to the at least one time-series image, an image analysis result representing abnormal information of the abnormal area in the image set to be analyzed; Wherein, when the number of the time-series images is multiple, the image analysis result of determining the abnormal information of the abnormal area in the image set to be analyzed based on the attribute data of the normal area corresponding to the at least one time-series image includes: Based on the attribute data of the normal area corresponding to the multiple time-series images, determining the attribute change data of the corresponding normal area; the attribute change data of the normal area represents the change of the regional attributes of the normal area corresponding to the multiple time-series images at different time nodes; The image analysis result is determined based on the attribute change data of the normal area corresponding to the multiple time-series images and the attribute data of the area of interest image.
2. The method according to claim 1, characterized in that The determining of a normal area in the region of interest image corresponding to the at least one time-series image comprises: Using the trained image segmentation model, segmenting the normal area in the region of interest image corresponding to the at least one time-series image to obtain the normal area in the region of interest image corresponding to the at least one time-series image; The image segmentation model is trained by using the analysis sample image set and the segmentation results of the corresponding normal areas of the target object as a training set.
3. The method according to claim 1 or 2, characterized in that: The image analysis result is determined based on the attribute change data of the normal area corresponding to the multiple time-series images and the attribute data of the image of the region of interest: Determining the attribute change data of the abnormal area corresponding to the multiple time-series images based on the attribute change data of the normal area corresponding to the multiple time-series images and the attribute data of the image of the region of interest; Performing quantitative analysis on the image set to be analyzed based on the attribute change data of the abnormal regions corresponding to the multiple time-series images, and determining an image analysis result representing abnormal information of the abnormal regions in the image set to be analyzed; Wherein, the attribute change data includes regional volume change data and / or regional volume proportion change data.
4. The method according to claim 1 or 2, characterized in that: When there are multiple time-series images, the image analysis result of determining abnormal information representing abnormal areas in the image set to be analyzed based on the attribute data of the normal area corresponding to the at least one time-series image includes: Preprocessing the region of interest image corresponding to the first time-series image and the region of interest image corresponding to the second time-series image respectively to obtain a first time-series processed image and a second time-series processed image; Based on the attribute data of the normal area corresponding to the first time-series processed image, determine first attribute data of the corresponding abnormal area, wherein the first attribute data includes first area position data and first area size data; Determine second attribute data of a corresponding abnormal region based on the attribute data of a normal region corresponding to the second time-series processed image, wherein the second attribute data includes second region position data and second region size data; Based on the first region position data and the second region position data, position matching is performed on the abnormal region in the first time series processed image and the abnormal region in the second time series processed image to obtain a region matching mapping relationship; Based on the region matching mapping relationship, the first region size data and the second region size data, quantitatively analyzing the first time series processed image and the second time series processed image to determine an image analysis result representing abnormal change information of the abnormal region in the image set to be analyzed; The acquisition time node of the first time-series image is earlier than the acquisition time node of the second time-series image, and the number of the first time-series image and the number of the second time-series image are at least one.
5. The method according to claim 1, characterized in that The method further comprises: Detecting pixel values of each region in the region of interest image corresponding to the at least one time-series image; Determine an abnormal area in the region of interest image corresponding to the at least one time-series image based on a comparison result of the pixel value of each area with a preset pixel threshold; Determine a region attribute of an abnormal region in a region of interest image corresponding to the at least one time-series image, wherein the region attribute includes a region size and / or a region quantity; Determining a marking type corresponding to each time-series image according to a comparison result of a region attribute of an abnormal region in a region-of-interest image corresponding to the at least one time-series image and a preset attribute threshold; Based on the determined marking type corresponding to the at least one time-series image, the at least one time-series image is marked respectively.
6. The method according to claim 1, characterized in that The method further comprises: Based on a normal area in the region of interest image corresponding to the at least one time-series image, determining a region attribute of an abnormal area in the region of interest image corresponding to the at least one time-series image; the region attribute includes a region size and / or a region quantity; Determining a marking type corresponding to each time-series image according to a comparison result of a region attribute of an abnormal region in a region-of-interest image corresponding to the at least one time-series image and a preset attribute threshold; Based on the determined marking type corresponding to the at least one time-series image, the at least one time-series image is marked respectively.
7. The method according to claim 5 or 6, characterized in that: In the case where the region attribute includes the region size and the region number, the determining the marking type corresponding to each time-series image according to the comparison result of the region attribute of the abnormal region in the region of interest image corresponding to the at least one time-series image with a preset attribute threshold value comprises: If the size of the abnormal area is greater than or equal to a preset size threshold, the corresponding time series image is determined to be a first marking type; the first marking type is used to indicate a normal area in the marking region of interest image; If the size of the abnormal region is smaller than a preset size threshold, and the number of the abnormal regions is greater than or equal to a preset number threshold, the corresponding time series image is determined to be a second marking type; the second marking type is used to indicate marking a normal region or an abnormal region in the region of interest image according to a comparison result of a ratio of the region attribute of the abnormal region in the region of interest and a preset ratio threshold; If the number of abnormal regions is less than a preset number threshold, the corresponding time series image is determined to be a third marking type; the third marking type is used to indicate abnormal regions in the marking region of interest image.
8. An image analysis device, characterized in that: include: An acquisition module, used for acquiring a set of images to be analyzed, wherein the set of images to be analyzed includes at least one time-series image; A first determining module, configured to determine a region of interest image corresponding to the target object in the at least one time-series image; The region of interest image includes a normal region and an abnormal region; A second determination module, configured to determine a normal area in the region of interest image corresponding to the at least one time-series image; An analysis module, configured to determine an image analysis result representing abnormal information of an abnormal area in the image set to be analyzed based on attribute data of a normal area corresponding to the at least one time-series image; Wherein, when the number of the time-series images is multiple, the image analysis result of determining the abnormal information of the abnormal area in the image set to be analyzed based on the attribute data of the normal area corresponding to the at least one time-series image includes: Based on the attribute data of the normal area corresponding to the multiple time-series images, determining the attribute change data of the corresponding normal area; the attribute change data of the normal area represents the change of the regional attributes of the normal area corresponding to the multiple time-series images at different time nodes; The image analysis result is determined based on the attribute change data of the normal area corresponding to the multiple time-series images and the attribute data of the area of interest image.
9. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the image analysis method according to any one of claims 1 to 7.
10. A computer storage medium, characterized in that: The computer storage medium stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the image analysis method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Analysis method for infant brain medical computer scanning images and realization system
CN101520893A
Method of, and apparatus for, processing medical image
CN104224217A
Gastroscope image auxiliary processing system and method based on ensemble learning
CN107564580A
Method of identifying spinal sagittal image anomaly and computing equipment
CN108830835A
A brain white matter high signal detection and positioning method based on multiple maps
CN109886944A
