Image information processing method, electronic device, and readable storage medium

Abstract

The present application provides an image information processing method, an electronic device, and a readable storage medium. The image information processing method comprises: identifying a target object in a real-time image; registering the target object in the real-time image with a target object in a pre-annotated image; and displaying the real-time image annotated with the target object. In the image information processing method provided in the present application, when the target object in the real-time image is identified, the target object in the pre-annotated image is automatically registered with the target object in the real-time image, thereby significantly improving the accuracy of target object identification in the real-time image. Subsequently, the target object is annotated in the real-time image to assist an operator in performing an operation, thereby reducing the workload of the operator and the risk of negligence.

Description

Image information processing methods, electronic devices, and readable storage media Technical Field

[0001] This application relates to the field of video surveillance, specifically to an image information processing method, an electronic device, and a readable storage medium. Background Technology

[0002] Ultrasound (B-scan) can be used for real-time monitoring. While ultrasound examinations have advantages such as good real-time performance, low cost, and ease of integration, they also have disadvantages such as insufficient image resolution, limited spatial resolution, and the ability to display only localized images. Therefore, focused ultrasound surgery requires not only ultrasound experience from the surgeon but also excellent spatial reasoning skills, which limits the application and promotion of this technology.

[0003] Focused Ultrasound Surgery (FUS) is a typical minimally invasive surgery that relies on non-invasive imaging monitoring technology to locate the target area and assess safety and effectiveness. Ultrasound imaging is a mainstream and economical monitoring method, but the resolution of ultrasound images is insufficient, and its application in the special environment of FUS further degrades the quality of real-time images, adversely affecting the use of the FUS system by physicians.

[0004] MRI (Magnetic Resonance Imaging) images have high resolution and complete cross-sectional information. They can not only show complete cross-sections, but also have three-dimensional information. After manual segmentation, MRI images can be intuitively expressed using three-dimensional visualization technology, which has many advantages that ultrasound imaging does not have.

[0005] Doctors often need to combine MRI with real-time ultrasound, which not only relies on the doctor's experience but also requires additional time to adjust the registration relationship, increasing the doctor's workload.

[0006] Therefore, how to reduce the workload of doctors is a technical problem that urgently needs to be solved by those in the field. Summary of the Invention

[0007] This application aims to solve at least one of the technical problems existing in the prior art, and proposes an image information processing method, electronic device and readable storage medium, which reduces the requirements of real-time monitoring on doctors' experience and reduces the workload of doctors when using real-time monitoring.

[0008] To achieve the purpose of this application, an image information processing method is provided, comprising:

[0009] Identify target objects in real-time images;

[0010] Register the target object in the real-time image with the target object in the pre-labeled image;

[0011] The real-time image labeled with the target object is displayed.

[0012] In some embodiments, after identifying the target object in the real-time image, the method further includes:

[0013] Compare the identified confidence level with the preset confidence level;

[0014] If the recognition confidence level is greater than or equal to the preset confidence level, then the step of registering the target object in the real-time image with the target object in the pre-labeled image is performed;

[0015] If the recognition confidence level is less than the preset confidence level, then the real-time image without the target object is displayed.

[0016] In some embodiments, after identifying the target object in the real-time image, the method further includes:

[0017] Extract the boundaries of the target object and related objects from the real-time image;

[0018] Determine whether the boundary of the target object intersects with the boundary of the related object;

[0019] If so, then perform the step of registering the target object in the real-time image with the target object in the pre-labeled image;

[0020] If not, the real-time image without the target object labeled is displayed.

[0021] In some embodiments, displaying the real-time image labeled with the target object includes:

[0022] Display a stereoscopic image of the target object based on the pre-annotated image;

[0023] The 3D image is cut according to the position of the real-time image to form a cut image;

[0024] The cut image is then fused with the real-time image for display.

[0025] In some embodiments, registering the target object in the real-time image with the target object in the pre-labeled image further includes:

[0026] Determine the coordinates of the initial real-time image and the pre-annotated image corresponding to the initial real-time image;

[0027] The detection probe is used to acquire the real-time image to determine whether the trend of change of the target object in the real-time image is reasonable based on the direction of movement of the detection probe.

[0028] If reasonable, then proceed with the step of displaying the real-time image labeled with the target object;

[0029] If this is not reasonable, the real-time image without the target object labeled will be displayed.

[0030] In some embodiments, determining whether the change trend of the target object in the real-time image is reasonable based on the movement direction of the detection probe includes:

[0031] The theoretical coordinates of the current real-time image are determined based on the displacement of the detection probe;

[0032] Determine whether the coordinates of the target object in the real-time image obtained by identification match the theoretical coordinates. If they match, the change trend of the target object in the real-time image is reasonable; otherwise, it is unreasonable.

[0033] In some embodiments, determining the theoretical coordinates of the current real-time image based on the displacement of the detection probe further includes:

[0034] The theoretical position of the target object is corrected according to the preset deformation.

[0035] In some embodiments, before identifying the target object in the real-time image, the method further includes:

[0036] Identify the target object in the image sequence;

[0037] The target object in each image of the image sequence is labeled to form any of the above-mentioned pre-labeled images.

[0038] On the other hand, this application also provides an image information processing method, including:

[0039] Receive image sequences;

[0040] Identify the target object in the image sequence;

[0041] The target object in each image of the image sequence is labeled to form any of the above-mentioned pre-labeled images.

[0042] On the other hand, this application also provides an image information processing method, including:

[0043] Receive image sequences;

[0044] Identify target objects in an image sequence;

[0045] The target objects in each image of the image sequence are labeled to form any of the above-mentioned pre-labeled images.

[0046] Furthermore, this application also provides an electronic device, comprising:

[0047] One or more processors;

[0048] A memory having stored one or more programs that, when executed by one or more processors, cause the one or more processors to implement any of the image information processing methods described above.

[0049] In another aspect, this application also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described image information processing methods.

[0050] This application has the following beneficial effects:

[0051] The image information processing method provided in this application automatically registers the target object in the pre-annotated image with the target object in the real-time image after the target object is identified in the real-time image, significantly improving the accuracy of target object identification in the real-time image. Subsequently, the target object is annotated in the real-time image to assist operators in their operations, reducing their workload and the risk of negligence. Attached Figure Description

[0052] Figure 1 is a flowchart of an image information processing method provided in a specific embodiment of this application;

[0053] Figure 2 is a schematic diagram of a real-time image after recognition according to a specific embodiment of this application;

[0054] Figure 3 is a schematic diagram of a pre-annotated image in one specific embodiment of this application;

[0055] Figure 4 is a schematic diagram of a pre-annotated image and an implementation image shown in a specific embodiment of this application;

[0056] Figure 5 is a flowchart of an information processing method provided in another specific embodiment of this application;

[0057] Figure 6 is a schematic diagram showing the bladder and uterus simultaneously.

[0058] Figure 7 is a schematic diagram showing the bladder when it is obscured by the uterus.

[0059] Figure 8 is a flowchart of the tracking process;

[0060] Figure 9 shows a comparison of artifacts generated in real-time images;

[0061] Figure 10 is a schematic diagram of artifacts generated in real-time images. Detailed Implementation

[0062] To enable those skilled in the art to better understand the technical solutions of this application, the image information processing method, electronic device, and readable storage medium provided in this application will be described in detail below with reference to the accompanying drawings.

[0063] The image information processing method provided in this application, as shown in Figure 1, includes:

[0064] S100: Identify target objects in real-time images;

[0065] Real-time images can be acquired using devices such as focused ultrasound transducers and B-ultrasound. The identification of target objects in the real-time images can be accomplished by a first identification model. Specifically, the target object can be a primary target tissue; in the specific embodiment shown in Figure 2, the target object is the uterus. The target object can also be other target tissues, such as the gallbladder, stomach, or intestines. The identification process requires identifying points on the contour boundary of the target object and converting these points into spatial coordinates. Since real-time identification requires relatively little computing power, the first identification model can be deployed at the application end.

[0066] Optionally, the first recognition model can be trained using an image recognition algorithm, such as the yolaCT (You Only Look At CoefficienTs) algorithm. During training, it is first necessary to collect ultrasound image samples and label the features of the target object in the samples to form a sample library. Then, the yolaCT algorithm is used for training, optimizing and continuously supplementing the samples with countersamples to improve the model's accuracy and generalization ability, ultimately obtaining a usable and stable model. Of course, users can also choose other artificial intelligence image recognition algorithms; this is not a limitation.

[0067] S200, Register the target object in the real-time image with the target object in the pre-annotated image;

[0068] Pre-labeled images are images that correspond to real-time images. For example, real-time images can be images captured in real time by focused ultrasound transducers, B-mode ultrasound, etc., while pre-labeled images can be images obtained by magnetic resonance imaging (MRI). MRI images typically include multiple cross-sectional tomographic images, which can be arranged to form a three-dimensional image. The imaging quality of MRI is generally higher than that of real-time images acquired by focused ultrasound transducers, B-mode ultrasound, etc., thus enabling better identification of target objects.

[0069] The pre-labeled image can be identified by a second recognition model in each body layer image, and the target object can be marked in the body layer image to form a pre-labeled image. In the specific embodiment shown in Figure 3, the body layer image not only marks the uterus, but also marks fibroids, bladder, sacrum and coccyx, urethral orifice, endometrium, and pubis with multiple colors. Users can select the labels to be marked as needed, which is not limited here. The second recognition model can be trained by algorithms such as nnUnet (Neural Network U-Net). The pre-labeled image is usually pre-labeled and used directly in the real-time monitoring process. Of course, users can also choose other artificial intelligence image recognition algorithms, which are not limited here.

[0070] The registration process can be achieved using the ICP (Iterative Closest Point) algorithm. The registration module can extract key features from the recognition results of pre-labeled images and real-time images, and then use the ICP algorithm for registration.

[0071] Furthermore, because real-time ultrasound images often only show a partial view, the complete boundaries of the target object are not visible in the real-time image. For example, when the target object is the uterus, the real-time image often only shows the lower edge of the uterus. The first recognition model can only identify the reliable lower edge. During the registration process, the registration module can extract the local boundaries of the target object in the real-time image as features. The ICP algorithm can then establish a registration matrix based on these features and the boundaries of the target object in the pre-labeled image, thereby registering the real-time image with the pre-labeled image.

[0072] S300: Displays a real-time image with the target object marked.

[0073] After identification and registration, the method provided in this embodiment can accurately determine the location of the target object. The target object is then labeled on the real-time image for easy viewing by doctors and other operators. Labeling the target object on the real-time image assists doctors and other operators in completing their tasks, reducing the demands on them. In the specific embodiment shown in Figure 4, the pre-labeled image and the real-time image are displayed side-by-side. This embodiment, based on identification on the real-time image, automatically registers the identification results of the real-time image using the pre-labeled image, improving the accuracy of real-time image labeling. Furthermore, manual registration is no longer required from the operator, reducing their workload.

[0074] It should be noted that the real-time image with labeled target objects displayed in this embodiment can be either a real-time image with labels, or a pre-labeled image fused with a real-time image, then labeled and displayed. Fusing the pre-labeled image with the real-time image improves image clarity and facilitates identification and reference by operators.

[0075] In practical applications, real-time ultrasound images often suffer from poor image quality and numerous artifacts in surgical settings, leading to errors in artificial intelligence (AI) recognition. If incorrect features are used for ICP registration, the registration result will inevitably be erroneous. Therefore, this application incorporates additional judgment criteria to reduce the occurrence of misidentification.

[0076] In the specific embodiment shown in Figure 5, after identifying the target object in the real-time image, the method further includes:

[0077] Compare the identified confidence level with the preset confidence level;

[0078] When real-time images are recognized using artificial intelligence image recognition algorithms, the first recognition model can output not only the recognition result but also the recognition confidence level. The preset confidence level value can be set according to the user's needs and is not limited here. In one specific embodiment of this application, the preset confidence level is 90%.

[0079] If the recognition confidence level is greater than or equal to the preset confidence level, then the step of registering the target object in the real-time image with the target object in the pre-labeled image is performed;

[0080] If the recognition confidence level is lower than the preset confidence level, then a real-time image without the target object will be displayed.

[0081] For example, after step S100, the first recognition model outputs a recognition confidence level. If the confidence level is greater than or equal to 90%, the recognition result is considered reliable and can be applied to subsequent steps, such as registration. If the confidence level is less than a preset confidence level, the recognition result is considered unreliable. In this case, the recognition result may contain errors, and continuing registration will produce incorrect registration results, which cannot be used to assist the operator. Therefore, only real-time images of unlabeled target objects are displayed.

[0082] The target object is usually adjacent to other related objects, and these related objects are also captured in real-time images. Therefore, the reliability of the first recognition model's recognition of the target object can be determined by whether the boundary of the target object intersects with the boundaries of related objects.

[0083] In some embodiments, after identifying the target object in the real-time image, the method further includes:

[0084] Extract the boundaries of the target object and related objects in a real-time image;

[0085] The boundaries of the target object and related objects can be identified using a first recognition model. Related objects can specifically include the bladder, skin, etc. If the fibroid is clearly displayed, the related objects may also include the fibroid's boundary. Real-time images often only show a portion of the target object and related objects; therefore, the local boundaries of the target object and related objects can be extracted.

[0086] The state of the target object and some related objects can also affect their display in the real-time image. For example, as shown in Figure 6, the bladder is not visible in the real-time image when it is small. As shown in Figure 7, it is visible in the real-time image when the bladder is full.

[0087] Determine whether the boundary of the target object intersects with the boundaries of related objects;

[0088] Based on the extracted boundaries of the target object and related objects, it can be determined whether they intersect. Specifically, as shown in Figure 6, since the bladder is not visible, the lower boundary of the uterus is close to the upper boundary of the skin, meaning that the boundaries of the uterus and the skin intersect. As shown in Figure 7, the bladder is full and visible in the real-time image. The lower boundary of the uterus intersects with the upper right boundary of the bladder.

[0089] If so, then perform the step of registering the target object in the real-time image with the target object in the pre-labeled image;

[0090] Since the boundary of the target object usually intersects with the boundary of related objects, if the boundary of the target object intersects with the boundary of related objects, the recognition result of the first recognition model is considered reliable. Therefore, registration and other operations can continue to be performed.

[0091] If not, display a live image of the target object that is not labeled.

[0092] If the boundary of the extracted target object does not intersect with the boundary of related objects, the recognition result of the first recognition model is considered unreliable and cannot be used. In this case, registration often results in errors. Therefore, only real-time images without labeled target objects are displayed.

[0093] It should be noted that judging the accuracy of a recognition result by its credibility and by whether the boundary of the target object intersects with the boundaries of related objects are not mutually exclusive methods in image processing. As shown in Figure 5, during application, image processing methods can use both methods to judge the credibility of the recognition result, reducing the probability of false recognition and avoiding misleading operators.

[0094] In some embodiments, the real-time image may be fused with a pre-annotated image for display to improve image quality. Therefore, displaying a real-time image labeled with the target object includes:

[0095] Display a 3D image of the target object based on a pre-annotated image;

[0096] Pre-annotated images are formed by annotating images captured by magnetic resonance imaging (MRI). Multiple pre-annotated images can be arranged to form a stereoscopic image. A stereoscopic image contains the target object and related objects.

[0097] The 3D image is cut according to the position of the real-time image to form a cut image;

[0098] When a detection probe acquires real-time images, the angle often differs from that of the pre-annotated image. To better correspond with the real-time image, the stereo image can be re-cut based on the position of the real-time image. The resulting cut image corresponds better to the real-time image, containing the target object and related objects that correspond to those in the real-time image.

[0099] The cut image is merged and displayed with the real-time image.

[0100] Since the sharpness of a cut image is usually higher than that of a real-time image, fusing the cut image with the real-time image can improve the sharpness of the real-time image. Furthermore, the cut image's position corresponds to that of the real-time image, allowing for better fusion and improving the quality of the fused image.

[0101] When acquiring real-time images, the detection probe moves as needed, and the corresponding real-time images will also change. Tracking the movement of the detection probe can determine whether the real-time images are reasonable.

[0102] In some embodiments, registering a target object in a real-time image with a target object in a pre-labeled image further includes:

[0103] Determine the coordinates of the initial real-time image and the pre-annotated image corresponding to the initial real-time image;

[0104] After the detection probe acquires the initial real-time image, the position of the corresponding pre-annotated image is determined through a registration step, enabling the tracking of the detection probe's movement on the pre-annotated image. Since the coordinates of the target object's boundary in the pre-annotated image have already been labeled, the coordinates of the initial real-time image are known parameters after registration.

[0105] The direction of movement of the detection probe determines whether the trend of change of the target object in the real-time image is reasonable. The detection probe is used to acquire real-time images.

[0106] The detection probe moves under the control of the application, which tracks the probe's coordinate changes. Based on these coordinate changes, the position of the pre-annotated image corresponding to the real-time image, as well as the changing trends of the target object and related objects, can be determined. The motion coordinates of the real-time image are known parameters. For example, a change in the probe's detection depth will cause displacement of the boundaries of the target object and related objects in the real-time image. The direction of this displacement should typically correspond to the change in the probe's depth. If the displacement direction does not match the probe's movement direction, the changing trend of the target object and related objects is unreasonable; conversely, if they do, the changing trend is reasonable. Therefore, analyzing the direction of coordinate changes can determine whether the changing trend of the target object in the real-time image is reasonable.

[0107] If reasonable, proceed with the step of displaying a real-time image labeled with the target object;

[0108] If the trend of change of the target object in the real-time image is reasonable, the real-time image is considered reliable, and therefore the real-time image with the target object labeled can be displayed. The real-time image can be displayed alone, or it can be merged with a pre-labeled image or a segmented image for display.

[0109] If this is not reasonable, then display a real-time image of the target object that is not labeled.

[0110] If the trend of the target object in the real-time image is unreasonable, it indicates that the real-time image is unreliable and the labeled target object may be incorrect. Therefore, to avoid interfering with the operator, the real-time image without labeled target objects is displayed directly.

[0111] Optionally, the reasonableness of the target object's change trend in the real-time image can be determined based on the movement direction of the detection probe, as shown in Figure 8, including:

[0112] The theoretical coordinates of the current real-time image are determined based on the displacement of the detection probe.

[0113] Since the detection probe moves under the control of the application, the displacement vector of the detection probe is a known parameter. By combining the coordinates of the initial real-time image and the displacement vector of the detection probe, the theoretical coordinates of the current real-time image can be determined.

[0114] The system determines whether the coordinates of the target object in the real-time image match the theoretical coordinates. If they match, the trend of the target object's change in the real-time image is reasonable; otherwise, it is unreasonable.

[0115] As mentioned earlier, the theoretical coordinates of the target object can be determined by combining the initial real-time image and the displacement vector of the detection probe. By identifying the target object in the current real-time image, the current coordinates of the target object can be obtained. Comparing the current coordinates with the theoretical coordinates determines the reliability of the current coordinates. For example, as shown in Figures 9 and 10, skin lines and artifact areas exist in Figure 9A. After the detection probe is moved, the image in Figure 9B is obtained. In Figure 9B, both the artifact area and the skin line have been displaced within the image. Figure 10 displays and compares the curves in Figures 9A and 9B. The skin line in Figures 9A and 9B represents the accurate identification result, showing the combined trend of movement after the detection probe is moved. The artifact area represents an erroneous imaging result; its displacement in the image is often asynchronous with the displacement of the detection probe. Therefore, Figure 10 shows the skin line and artifact area in Figure 9A, as well as their theoretical coordinates after the detection probe is moved. The theoretical coordinates of the skin line after movement can coincide with the skin line in Figure 9B. The theoretical coordinates of the artifact area after movement are a certain distance away from the artifact area in Figure 9B. Therefore, it can be determined that the imaging of the artifact area is inaccurate.

[0116] In some embodiments, determining the theoretical coordinates of the current real-time image based on the displacement of the detection probe includes:

[0117] The theoretical position of the target object is corrected according to the preset deformation.

[0118] The movement of the detection probe causes a certain degree of compression to the target object and related objects, resulting in deformation. This application can experimentally determine the deformation caused by the detection probe in advance, i.e., preset deformation, and store the preset deformation in the application terminal, thereby establishing the relationship between motion coordinates and tissue changes. The application terminal can correct the theoretical position of the target object based on the preset deformation, improving the accuracy of the theoretical position calculation of the target object.

[0119] Pre-annotated images need to be pre-annotated before operation. The pre-annotation process can be performed on the server side or on the application side.

[0120] In some embodiments, before identifying the target object in a real-time image, the method further includes:

[0121] Identify target objects in an image sequence;

[0122] The image sequence may include multiple tomographic images, and at least some of the tomographic images may contain the target object. For example, the image sequence may be obtained by magnetic resonance imaging (MRI), which typically includes multiple cross-sectional tomographic images that can be arranged to form a three-dimensional image. The second recognition model can identify the target object in each tomographic image. In this embodiment, the tomographic image recognition process is performed at the application end.

[0123] Optionally, the second recognition model can be trained using algorithms such as nnUnet (Neural Network U-Net). nnUnet identifies the target object as an image, and the second recognition model can extract the boundary of the target object and convert each point on the boundary into 3D spatial coordinates. For convenient transmission and 3D visualization, the conversion method can employ the contour extraction algorithm provided by OpenCV. The resulting coordinates can be saved in JSON format for easy transmission.

[0124] Optionally, during the training of the second recognition model, stereoscopic image sequences from magnetic resonance imaging are first collected and labeled to form a sample library. The nnUnet deep learning algorithm is then used for training, optimizing and continuously supplementing the model with countersamples to improve its generalization ability and ultimately obtain a usable stable model. Of course, users can also use other algorithms for training; this is not limited here.

[0125] The target objects in each image of the image sequence are labeled to form a pre-labeled image in any of the above embodiments.

[0126] The target objects in the identified tomographic images can be labeled. Related objects can also be labeled within the tomographic images and distinguished using different labels. When the application receives the JSON information, it extracts points on the contour boundaries of the target object and related objects using the label names, thus reconstructing the three-dimensional surface information of these tissues for 3D visualization. The application can transmit the 3D sequence of the tomographic images, the labels of the target objects, and their corresponding 3D spatial coordinates via a JSON file. Of course, users can also choose the specific content of the recognition results as needed; this is not limited here.

[0127] This application also provides an image information processing method, including:

[0128] Receive image sequences;

[0129] In this embodiment, the pre-annotation process is performed on the server side. Optionally, the image sequence can be obtained from magnetic resonance imaging (MRI). MRI images typically include multiple cross-sectional tomographic images, which can be arranged to form a stereoscopic image. MRI is performed preoperatively and does not require high real-time performance, so it can be deployed on a server. The server can connect to multiple applications and receive stereoscopic sequences of tomographic images sent from multiple applications. The server queues these sequences according to their order, first performing recognition on the sequence at the front of the queue and then transmitting the recognition result back to the application.

[0130] Identify target objects in an image sequence;

[0131] The second recognition model can identify target objects in images at various volumetric layers. This model can be trained using algorithms such as nnUnet (Neural Network U-Net). Since the target objects identified by nnUnet are images, for ease of transmission and 3D visualization, the server can convert the boundary points of the resulting image into 3D spatial coordinates. This conversion can be achieved using the contour extraction algorithm provided by OpenCV, and the resulting coordinates can be saved in JSON format for transmission. The training method for the second recognition model can be referred to the previous section and will not be repeated here.

[0132] The target objects in each image of the image sequence are labeled to form a pre-labeled image in any of the above embodiments.

[0133] Optionally, target objects in the identified tomographic images can be labeled. Related objects can also be labeled in the tomographic images and distinguished using different labels. When the application receives the JSON information, it can extract points on the contour boundaries of each tissue using the label names, thus reconstructing the three-dimensional surface information of these tissues for 3D visualization. Therefore, the recognition results returned from the server to the application can include the 3D sequence of the tomographic images and the labels and corresponding 3D spatial coordinates of the target objects. Of course, users can also choose the specific content of the recognition results as needed; this is not limited here.

[0134] This application also provides an electronic device, including:

[0135] One or more processors;

[0136] A memory that stores one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the image information processing method in any of the above embodiments.

[0137] This application also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the image information processing method in any of the above embodiments.

[0138] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of this application, and this application is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of this application, and these modifications and improvements are also considered to be within the scope of protection of this application.

Claims

1. An information processing method of an image, characterized by, comprising: identifying a target object in a real-time image; registering the target object in the real-time image with a target object in a pre-labeled image; displaying the real-time image labeled with the target object.

2. The information processing method according to claim 1, characterized by, after the identifying a target object in a real-time image, further comprising: comparing an identification confidence with a preset confidence; if the identification confidence is greater than or equal to the preset confidence, performing the registering the target object in the real-time image with the target object in the pre-labeled image; if the identification confidence is less than the preset confidence, displaying the real-time image unlabeled with the target object.

3. The information processing method according to claim 1, characterized by, after the identifying a target object in a real-time image, further comprising: extracting a boundary of the target object and a boundary of a related object in the real-time image; judging whether the boundary of the target object intersects with the boundary of the related object; if yes, performing the registering the target object in the real-time image with the target object in the pre-labeled image; if no, displaying the real-time image unlabeled with the target object.

4. The information processing method according to claim 1, characterized by, the displaying the real-time image labeled with the target object, comprising: displaying a stereoscopic image of the target object according to the pre-labeled image; cutting the stereoscopic image according to a position of the real-time image to form a cut image; fusing and displaying the cut image with the real-time image.

5. The information processing method according to claim 1, characterized by, the registering the target object in the real-time image with the target object in the pre-labeled image, further comprising: determining a coordinate of an initial real-time image and a pre-labeled image corresponding to the initial real-time image; judging whether a change trend of the target object in the real-time image is reasonable according to a moving direction of a detection probe, the detection probe being used to acquire the real-time image; if reasonable, performing the displaying the real-time image labeled with the target object; if not reasonable, displaying the real-time image unlabeled with the target object.

6. The information processing method according to claim 5, characterized by, the judging whether a change trend of the target object in the real-time image is reasonable according to a moving direction of a detection probe, comprising: determining a theoretical coordinate of a current real-time image according to a displacement of the detection probe; judging whether a coordinate of the target object in the real-time image obtained by identification and the theoretical coordinate are consistent, if consistent, the change trend of the target object in the real-time image is reasonable, otherwise, not reasonable.

7. The information processing method according to claim 6, characterized by, the determining a theoretical coordinate of a current real-time image according to a displacement of the detection probe, further comprising: correcting a theoretical position of the target object according to a preset deformation variable.

8. The information processing method according to claim 7, characterized by, before the identifying a target object in a real-time image, further comprising: identifying the target object in an image sequence; labeling the target object in each image of the image sequence to form the pre-labeled image of any one of claims 1 to 7.

9. An information processing method of an image, characterized by, comprising: receiving an image sequence; identifying the target object in the image sequence; labeling the target object in each image of the image sequence to form the pre-labeled image of any one of claims 1 to 7. 10.An electronic device, comprising: one or more processors; a memory having stored thereon one or more programs, which, when executed by the one or more processors, cause the one or more processors to carry out the information processing method of an image according to any one of claims 1 to 8.

11. A computer readable medium having stored thereon a computer program, which, when executed by a processor, carries out the information processing method of an image according to any one of claims 1 to 8.

Images