Medical image-based organ segmentation modeling method and terminal

Abstract

The application discloses an organ segmentation modeling method and terminal based on medical images, receives inputted continuous medical images, obtains a first image containing a target organ from the continuous medical images, receives positive and negative control points marked on the first image, identifies the target organ on the first image according to the marking result, continuously tracks the organ in the continuous sequence images in the section based on the identified organ region, and further completes automatic identification of the whole organ, generates a three-dimensional model of the organ based on the original section continuous images, establishes at least two other section continuous medical images according to the inputted continuous medical images, and identifies and models the organ in the other section continuous medical images. Finally, the three-dimensional segmentation model of the target organ is obtained by calibrating all the section models, and accurate organ segmentation results are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, in particular to an organ segmentation modeling method based on medical images and a terminal. BACKGROUND

[0002] With the rapid development of artificial intelligence, software for automatic segmentation of medical images is emerging. The widely used software includes 3D slicer, ITK-SNAP, etc. The automatic segmentation algorithms relied on by these software mainly include traditional algorithms such as threshold-based, region growing, level set method, active contour model, etc., and deep learning algorithms such as U-Net, CNN, etc.

[0003] The prior art discloses a method for multi-organ segmentation modeling of abdominal computed tomography images using edge perception algorithm. In this method, an abdominal multi-organ segmentation network is constructed, which mainly includes a voxel segmentation module, an edge perception module and a fusion module. The abdominal multi-organ segmentation model is obtained by training the network with abdominal CT images in the training set. However, the above segmentation method is only suitable for abdominal organs of CT images, and has poor applicability to other organs or lesions. For those automatic segmentation algorithms that can be widely adapted to various organ tissues, the actual segmentation effect in most scenes is not ideal, which limits their effectiveness in practical applications.

[0004] In addition, the traditional automatic segmentation algorithm is very sensitive to the selection of seed points, boundaries and thresholds. In order to obtain more accurate results, it may be necessary to manually set multiple boundary points or make a lot of corrections after automatic segmentation, which may even exceed the complexity of direct manual segmentation in operation. As for deep learning algorithms, they require a large amount of labeled training set data, which not only increases the workload, but also requires high computing power.

[0005] At the same time, the existing traditional automatic segmentation algorithm is difficult to fully meet the needs of medical segmentation. These algorithms are easily affected by image noise, and their segmentation effect is usually not satisfactory when the intensity distribution of the target and the background is overlapped. Due to the characteristics of medical images, the density between organs is often similar or overlapped, which makes it difficult to achieve ideal segmentation. Although deep learning algorithms perform better than traditional algorithms in specific scenarios, they still face great challenges when dealing with physiological variations or rare lesions and abnormal tissues. SUMMARY

[0006] The technical problem to be solved by the present application is to provide an organ segmentation modeling method based on medical images and a terminal, which can segment and model organs in various medical images while ensuring the accuracy of segmentation.

[0007] To solve the above technical problems, the technical scheme adopted by the present application is:

[0008] An organ segmentation modeling method based on medical images, comprising the steps of:

[0009] S1, receiving continuous medical images, obtaining a first image containing a target organ from the continuous medical images, receiving positive and negative control points marked on the first image, and identifying the target organ on the first image according to the marking result;

[0010] S2, identifying and tracking the target organ in the continuous medical images, calculating the overlap ratio of the continuous tracking results, and if the overlap ratio is greater than a preset ratio, obtaining the tracking result of the target organ, and generating a three-dimensional model of the target organ;

[0011] S3, establishing continuous medical images of at least two sections other than the section where the original continuous medical images are located according to the continuous medical images, and returning to execute steps S1 and S2 in turn using the established continuous medical images, to obtain three-dimensional models of the target organ in at least two sections;

[0012] S4, calibrating the three-dimensional models of the target organ in all sections to obtain a three-dimensional segmentation model of the target organ.

[0013] To solve the above technical problems, another technical scheme adopted by the present application is:

[0014] An organ segmentation modeling terminal based on medical images, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the following steps:

[0015] S1, receiving continuous medical images, obtaining a first image containing a target organ from the continuous medical images, receiving positive and negative control points marked on the first image, and identifying the target organ on the first image according to the marking result;

[0016] S2, identifying and tracking the target organ in the continuous medical images, calculating the overlap ratio of the continuous tracking results, and if the overlap ratio is greater than a preset ratio, obtaining the tracking result of the target organ, and generating a three-dimensional model of the target organ;

[0017] S3, establishing continuous medical images of at least two sections other than the section where the original continuous medical images are located according to the continuous medical images, and returning to execute steps S1 and S2 in turn using the established continuous medical images, to obtain three-dimensional models of the target organ in at least two sections;

[0018] S4, calibrating the three-dimensional model of the target organ in all sections to obtain a three-dimensional segmentation model of the target organ.

[0019] The present application has the advantages that: receiving inputted continuous medical images, obtaining a first image containing a target organ from the continuous medical images, receiving positive and negative control points marked on the first image, performing automatic identification of the target organ on the first image according to the marking result, performing continuous tracking of the organ on the continuous sequence images based on the identified organ region, and further completing automatic identification of the whole organ, generating a three-dimensional model of the organ based on the original section continuous images, establishing at least two continuous medical images of other sections based on the inputted continuous medical images, and performing identification and modeling of the organ on the continuous medical images of the other sections. Finally, calibration is performed based on the models of all sections to obtain a three-dimensional segmentation model of the target organ, so that accurate organ segmentation results are ensured. Therefore, the present application can be adapted to various medical images and is suitable for various scanning sites, and the delineation of the target organ only needs to mark a small number of control points on the corresponding site to achieve the purpose, and the identification and tracking are combined with the continuous medical images to further ensure the accuracy of the segmentation. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of an organ segmentation and modeling method based on medical images according to an embodiment of the present application;

[0021] Figure 2 A schematic diagram of an organ segmentation and modeling terminal based on medical images according to an embodiment of the present application;

[0022] Figure 3 A flowchart of a three-dimensional model construction process of a target organ of a medical image according to an embodiment of the present application.

[0023] Label explanation:

[0024] 1. An organ segmentation and modeling terminal based on medical images; 2. a memory; 3. a processor. DETAILED DESCRIPTION

[0025] To explain the technical content, purposes and effects of the present application in detail, the following will be described in conjunction with the embodiments and the accompanying drawings.

[0026] Please refer to Figure 1 The embodiment of the present application provides an organ segmentation and modeling method based on medical images, which comprises the following steps:

[0027] S1, receiving continuous medical images, obtaining a first image containing a target organ from the continuous medical images, receiving positive and negative control points marked on the first image, and identifying the target organ on the first image according to the marking result;

[0028] S2, recognizing and tracking a target organ in continuous medical images, calculating an overlap ratio of continuous tracking results, and obtaining a tracking result of the target organ if the overlap ratio is greater than a preset ratio, and generating a three-dimensional model of the target organ;

[0029] S3, establishing continuous medical images of at least two sections other than a section of the original continuous medical images according to the continuous medical images respectively, and returning to execute steps S1 and S2 using the established continuous medical images in sequence to obtain three-dimensional models of the target organ in the at least two sections;

[0030] S4, calibrating the three-dimensional models of the target organ in all sections to obtain a three-dimensional segmentation model of the target organ.

[0031] From the above description, the beneficial effects of the present application are that: the input continuous medical images are received, a first image containing a target organ is obtained from the continuous medical images, positive and negative control points marked on the first image are received, automatic recognition of the target organ is performed on the first image according to the marking result, continuous tracking of the organ in the continuous sequence images in the section is performed based on the recognized organ region, and then automatic recognition of the entire organ is completed, a three-dimensional model of the organ on the continuous images based on the original section is generated, at least two continuous medical images of other sections are established according to the input continuous medical images, and recognition and modeling of the organ in the continuous medical images of the other sections are performed. Finally, calibration is performed based on the models of all sections to obtain a three-dimensional segmentation model of the target organ, and accurate organ segmentation results are ensured. Therefore, the present application can adapt to various medical images and be suitable for various scanning sites, and the delineation of the target organ only needs to mark a small number of control points in the corresponding part to achieve, and recognition tracking is performed in combination with the continuous medical images, and the accuracy of segmentation is further ensured.

[0032] Further, the receiving of the positive and negative control points marked on the first image and the recognition of the target organ on the first image in sequence according to the marking result comprises:

[0033] extracting image feature information from the first image, and constructing a predictor of the first image based on the image feature information;

[0034] receiving the positive and negative control points marked on the target organ in the first image, inputting the positive and negative control points into the predictor, and identifying an image segmentation result of the target organ in the first image.

[0035] From the above description, for the operator, the delineation of the target organ only needs to mark a small number of control points in the corresponding part, and the non-target tissue is removed by only adjusting the positive marking to negative marking and performing corresponding control point operation.

[0036] Further, the recognizing and tracking the target organ in the continuous medical images comprises:

[0037] constructing a first image sequence containing the first image and images after the first image, and a second image sequence containing the first image and images before the first image;

[0038] recognizing and tracking the target organ in the first image sequence and the second image sequence according to the image segmentation result of the target organ in the first image, to obtain an initial tracking result.

[0039] As can be seen from the above description, the continuous tracking of the organ on the target organ region of the upper and lower sequence images can complete the automatic recognition of the entire organ.

[0040] Further, the calculating the overlap ratio of the continuous tracking results, if the overlap ratio is greater than a preset ratio, obtaining the tracking result of the target organ comprises:

[0041] calculating the overlap ratio of the tracking result of the current image and the initial tracking result of the next image in the image sequence from the first image, if the overlap ratio is greater than or equal to a preset ratio, storing the initial tracking result of the next image, and taking the next image as the current image to calculate the overlap ratio, until all images in the image sequence are completed for the calculation of the overlap ratio;

[0042] if the overlap ratio is less than the preset ratio, considering that the target organ does not exist in the next image, ending the organ segmentation.

[0043] As can be seen from the above description, the continuous tracking results are checked, the operation logic is simple and easy to understand, the computing power requirement is not high, and the drawing efficiency is significantly improved, and the organ "drifting" situation of multiple non-continuous regions is excluded.

[0044] Further, the calibrating the three-dimensional model of the target organ in all sections to obtain a three-dimensional segmentation model of the target organ comprises:

[0045] converting the three-dimensional model of the target organ in all sections to the same coordinate system, calculating the intersection of the three-dimensional models of all sections, and obtaining the three-dimensional segmentation model of the target organ according to the intersection result.

[0046] As can be seen from the above description, the organ boundary recognition is combined with the transverse section, the sagittal section and the coronal section, which can complement each other and improve the accuracy of organ segmentation.

[0047] Please refer to Figure 2Another embodiment of the present application provides a medical image-based organ segmentation modeling terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements each step of the medical image-based organ segmentation modeling method described above when executing the computer program.

[0048] The medical image-based organ segmentation modeling method and terminal described above are suitable for various medical images and various scanning parts, and can ensure the accuracy of segmentation.

[0049] Embodiment one

[0050] Please refer to Figure 1 and Figure 3 A medical image-based organ segmentation modeling method, characterized in that it comprises the following steps:

[0051] S1, receiving continuous medical images, obtaining a first image containing a target organ from the continuous medical images, receiving positive and negative control points marked on the first image, and identifying the target organ on the first image according to the marking result.

[0052] S11, receiving an input medical image sequence in DICOM format, the medical image sequence containing n medical images, selecting any one containing a desired target organ, and setting it as the ith image, denoted as I i , where 0 < i ≤ n.

[0053] S12, extracting image feature information from the first image, and constructing a predictor of the first image based on the image feature information.

[0054] In this embodiment, in order to automatically segment and outline the image, a SAM model (Segment Anything Model) is used for image analysis. In the identification process, first, the SAM model is applied to the image I i for feature encoding (Encode), and the image features are extracted as image feature information T i in the form of a tensor to prepare for identification, and the feature information is stored.

[0055] At the beginning of the identification algorithm, the image I i is loaded, the feature information T i is extracted, and the SAM model is input into the algorithm predictor P i .

[0056] S13, receiving positive and negative control points of a target organ mark in the first image, inputting the positive and negative control points into the predictor, and identifying an image segmentation result of the target organ in the first image.

[0057] Specifically, based on the image I i , the target organ is marked with positive and negative control points to automatically identify and delineate the target organ, wherein the positive control points represent the target organ tissue to be included in the delineation, and the negative control points represent the tissue outside the target organ to be removed. i For the jth control point on the image I j , the information can be represented as: p j = {x j , y j , δ j}, wherein (x j , y j ) represents the coordinates of the control point in the x and y directions on the image I i , and δ j represents the weight of the control point, wherein the positive control point takes a value of 1, and the negative control point takes a value of 0.

[0058] In the prediction stage of the identification algorithm, the control points p j of the image are input into the predictor P i , the predictor performs target segmentation, and outputs an image segmentation result M i . After confirming the target organ boundary range, the segmentation result is set as the region of interest (ROI) of the image I i for subsequent tracking.

[0059] S2, identifying and tracking the target organ in the continuous medical images, calculating the overlap ratio of the continuous tracking results, and obtaining the tracking result of the target organ if the overlap ratio is greater than a preset ratio, and generating a three-dimensional model of the target organ.

[0060] In this embodiment, the organ identification and tracking of the multi-layer continuous image sequence is performed according to the organ segmentation result of a single cross-sectional image, so as to obtain the three-dimensional delineation and segmentation of the same organ in the continuous images. The specific idea is to regard the continuously scrolling medical images as a continuous video, and regard the delineated target organ ROI as a certain object in the video, and track the tissue related to the object in the up and down scrolling images, that is, the concept of "video tracking object" is analogized to achieve the purpose of automatically delineating the three-dimensional structure of the target tissue in the entire sequence of images.

[0061] Specifically, the XMem model (Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model) is used for the identification and tracking of continuous organs. Based on the automatic segmentation of the target organ's ROI in the first step, the organ region is tracked and identified in consecutive images above and below, thereby obtaining a 3D segmentation model of the target organ. This is divided into the following steps:

[0062] S21. Construct a first image sequence containing the first image and images after the first image, and a second image sequence containing the first image and images before the first image.

[0063] In image I i The extracted image segmentation result M i As input to the XMem model, construct S1 = {I i ,I i+1 ,I i+2 ,…} and S2={I i ,I i-1 ,I i-2 The two image sequences, , ..., represent the preceding and following image sequences of the selected images, respectively.

[0064] S22. Based on the image segmentation results of the target organ in the first image, identify and track the target organ in the first image sequence and the second image sequence to obtain the initial tracking result.

[0065] Specifically, taking image sequence S1 as an example for image tracking, where I i M is the first image in the image sequence. i This represents the segmentation result of the organs in the image. First, an XMem model is created, denoted as X1, and I... i With M i Input it into the model. Then, take the next image I. i+1 Input model X1, the model performs target tracking prediction, and outputs result M' i+1 These are preliminary results of the prediction.

[0066] S23. Starting from the first image in the image sequence, calculate the overlap ratio between the tracking result of the current image and the initial tracking result of the next image. If the overlap ratio is greater than or equal to a preset ratio, store the initial tracking result of the next image and use the next image as the current image to calculate the overlap ratio until all images in the image sequence have completed the calculation of the overlap ratio. If the overlap ratio is less than the preset ratio, it is considered that there is no target organ in the next image, and the organ segmentation ends.

[0067] Since the main idea of continuous video tracking is to track the similarity of objects, the specific object in a frame of image and its consecutive frame may have a large displacement and less overlapping part. In the continuous layers of medical images, the organs in the upper and lower layers of each organ have a large overlapping area. If only the "object similarity" is used, other similar organs that are not continuous may be outlined, resulting in a large error.

[0068] Therefore, in the present embodiment, the result M' i+1 is checked and judged, which is divided into the following cases: if the result M' i+1 does not contain any target, it means that the target does not exist in the image I i+1 , and the algorithm of image tracking ends; if the result M' i+1 contains the target, it needs to be identified and checked. The checking process takes the similarity of the segmentation result as the evaluation index, and the calculation method is as follows:

[0069]

[0070] Wherein, Area(X) represents the area of X.

[0071] In the checking process, firstly, the "drifting" situation of multiple non-continuous regions needs to be excluded, that is, the M' i+1 output by the algorithm may generate a region completely not overlapping with M i outside the range of M i .

[0072] The exclusion process is as follows: assuming that there are a continuous regions in M' i+1 , each continuous region is marked as X a . For each X a , if sim(X a , M i ) = 0, X a is excluded from M' i+1 . Then, M' i+1 is further checked, and a threshold t ∈ (0, 1) is set. If sim(M' i+1 , M i ) ≥ t, it is considered that the identification result of M' i+1 is reliable, and the segmentation result M i+1 of the image I i+1 is set as M' i+1 . The algorithm continues to track the next frame, and the above process is repeated until there is no tracking result in the image, and the algorithm ends. If im(M' i+1 , M i ) < t, it is considered that there is no target on the image I i+1 , and the algorithm is terminated.

[0073] The above algorithm flow realizes the organ continuous tracking process of the image sequence S1, and the output result is the segmentation sequence R1={M i ,M i+1 ,M i+2 ,…}. The same tracking process is applied to the sequence S2, and the segmentation sequence R1={M i ,M i-1 ,M i-2 ,…} is obtained. R1 and R2 are combined to obtain the image tracking result R={…M i-2 ,M i-1 ,M i ,M i+1 ,M i+2 ,…}, which is the organ continuous tracking result output of the sequence image.

[0074] S3, according to the continuous medical image, at least two continuous medical images of sections other than the section of the original continuous medical image are established, and steps S1 and S2 are returned to be executed in turn using the established continuous medical images, to obtain a three-dimensional model of the target organ of at least two sections.

[0075] Specifically, the section of the medical image can include a transverse section, a sagittal section and a coronal section. In this embodiment, the original continuous medical image can be a medical image obtained from the transverse section. Then, the sagittal section and the coronal section are established based on the original continuous medical image, and the three-dimensional delineation is performed by using the processes of steps S1 and S2, and then the three-dimensional model of the same organ under three sections is obtained.

[0076] S4, calibrating the three-dimensional models of the target organ in all sections to obtain a three-dimensional segmentation model of the target organ.

[0077] The step S4 includes:

[0078] Converting the three-dimensional models of the target organ in all sections to the same coordinate system, calculating the intersection of the three-dimensional models of all sections, and obtaining the three-dimensional segmentation model of the target organ according to the intersection result.

[0079] In the embodiment, the images analyzed by the medical images are any one of the transverse plane, the sagittal plane and the coronal plane, which can correspond to the x, y and z planes in the three-dimensional space respectively. Based on the medical images of a single dimension, the images of the other two sections can be constructed for verification. Since the medical images are gray images, organs are arranged closely with each other, and the density / signal of some organs is similar to that of other organs, so there is no obvious interface in some sections, and the boundary is difficult to accurately identify. In the embodiment, it can be found that the interface of the organ and other tissues observed in different directions is different, for example, the boundary of some organs that are difficult to distinguish in the transverse plane is easier to identify in the sagittal plane and the coronal plane, that is, the effect of complementary identification of different sections can be achieved.

[0080] Therefore, in the embodiment, the images of the other two dimensions are reconstructed from the images of a single dimension, and the medical images of the transverse plane, the sagittal plane and the coronal plane are obtained. In the three planes, the automatic identification of the target organ is performed according to steps S1 and S2 respectively, and the organ is outlined to obtain the image tracking result. The tracking results of the transverse plane, the sagittal plane and the coronal plane are represented as R A ,R S ,R C . Before the organ is outlined, R A ,R S ,R C needs to be converted to the same coordinate system, and the final outline result of the organ is R F =R A ∩R S ∩R C . Thus, the algorithm completes the segmentation of the organ, and the three-dimensional segmentation model of the target organ is obtained.

[0081] Therefore, the organ segmentation and modeling method in the embodiment has strong generalization performance but low accuracy or high accuracy but limited generalization performance. In the embodiment, the SAM algorithm and the video continuous tracking XMem algorithm are introduced, and the video tracking is used for the automatic segmentation of the medical continuous images in combination with the medical image framework. Based on the deep learning model of big data, the generalization performance of the model is significantly enhanced, so that the model has excellent generalization and accuracy.

[0082] In addition, the input conditions required for outlining the target organ by using the method are simple, and the target region can be quickly and accurately identified and outlined in the form of control point marking after the image is analyzed. The method avoids manual determination of the organ boundary or setting of a threshold to prevent region overflow, and does not need to rely on a large amount of training set data for labeling and learning.

[0083] Meanwhile, a multi-dimensional correction system is further proposed in combination with the overlap ratio of the recognition result. This not only maintains the original high accuracy, but also significantly improves the accuracy of automatic segmentation of organs with boundary overlap problems through further boundary correction.

[0084] Embodiment two

[0085] Please refer to Figure 2 A medical image-based organ segmentation modeling terminal 1 comprises a memory 2, a processor 3, and a computer program stored in the memory 2 and executable on the processor 3, wherein the processor 3 implements each step of the medical image-based organ segmentation modeling method of embodiment one when executing the computer program.

[0086] In summary, the medical image-based organ segmentation modeling method and terminal provided by the present application receive inputted continuous medical images, obtain a first image containing a target organ from the continuous medical images, receive positive and negative control points marked on the first image, perform automatic recognition of the target organ on the first image according to the marking result, perform continuous tracking of the organ in the continuous sequence images in the section based on the recognized organ region, further complete automatic recognition of the entire organ, generate a three-dimensional model of the organ on the continuous images based on the original section, and further perform organ recognition and modeling on the continuous medical images of other sections. Finally, the three-dimensional segmentation model of the target organ is obtained by calibrating all the models of the sections, and accurate organ segmentation results are ensured. Therefore, the present application can adapt to various medical images and is suitable for various scanning sites. Moreover, the delineation of the target organ can be realized by marking a small number of control points on the corresponding site, and the recognition and tracking in combination with the continuous medical images further ensure the accuracy of the segmentation.

[0087] The above description is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent transformation or direct or indirect application in the related technical field based on the content of the specification and drawings is also included in the patent protection scope of the present application.

Claims

1. A method for organ segmentation and modeling based on medical images, characterized in that, Including the following steps: S1. Receive a series of medical images, acquire a first image containing the target organ from the series of medical images, receive positive and negative control points marked on the first image, and identify the target organ on the first image based on the marking results, including: Image feature information is extracted from the first image, and a predictor for the first image is constructed based on the image feature information. The predictor for the first image is constructed using the SAM model. The system receives positive and negative control points labeled with target organs in the first image, inputs the positive and negative control points into the predictor, and identifies the image segmentation result of the target organs in the first image. S2. Identify and track target organs in continuous medical images, calculate the overlap ratio of continuous tracking results, and if the overlap ratio is greater than a preset ratio, obtain the tracking result of the target organ and generate a three-dimensional model of the target organ. S3. Based on the continuous medical images, establish continuous medical images of at least two sections other than the section where the original continuous medical image is located. Use the established continuous medical images to return to the execution steps S1 and S2 in sequence to obtain a three-dimensional model of the target organ of at least two sections. S4. Calibrate the three-dimensional model of the target organ in all cross-sections to obtain a three-dimensional segmentation model of the target organ, including: The three-dimensional models of the target organ in all cross sections are transformed to the same coordinate system, and the intersection of the three-dimensional models of all cross sections is calculated. The three-dimensional segmentation model of the target organ is obtained based on the intersection result.

2. The organ segmentation modeling method based on medical images according to claim 1, characterized in that, The identification and tracking of target organs in continuous medical images includes: Construct a first image sequence containing the first image and images after the first image, and a second image sequence containing the first image and images before the first image; Based on the image segmentation results of the target organ in the first image, the target organ in the first image sequence and the second image sequence is identified and tracked to obtain the initial tracking result.

3. The organ segmentation modeling method based on medical images according to claim 2, characterized in that, The calculation of the overlap ratio of continuous tracking results, if the overlap ratio is greater than a preset ratio, then the tracking result of the target organ is obtained as follows: In the image sequence, starting from the first image, the overlap ratio between the tracking result of the current image and the initial tracking result of the next image is calculated. If the overlap ratio is greater than or equal to a preset ratio, the initial tracking result of the next image is stored, and the next image is used as the current image to calculate the overlap ratio until the overlap ratio of all images in the image sequence has been calculated. If the overlap ratio is less than the preset ratio, it is assumed that the target organ does not exist in the next image, and organ segmentation ends.

4. A terminal for organ segmentation and modeling based on medical images, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: S1. Receive a series of medical images, acquire a first image containing the target organ from the series of medical images, receive positive and negative control points marked on the first image, and identify the target organ on the first image based on the marking results, including: Image feature information is extracted from the first image, and a predictor for the first image is constructed based on the image feature information. The predictor for the first image is constructed using the SAM model. The system receives positive and negative control points labeled with target organs in the first image, inputs the positive and negative control points into the predictor, and identifies the image segmentation result of the target organs in the first image. S2. Identify and track target organs in continuous medical images, calculate the overlap ratio of continuous tracking results, and if the overlap ratio is greater than a preset ratio, obtain the tracking result of the target organ and generate a three-dimensional model of the target organ. S3. Based on the continuous medical images, establish continuous medical images of at least two sections other than the section where the original continuous medical image is located. Use the established continuous medical images to return to the execution steps S1 and S2 in sequence to obtain a three-dimensional model of the target organ of at least two sections. S4. Calibrate the three-dimensional model of the target organ in all cross-sections to obtain a three-dimensional segmentation model of the target organ, including: The three-dimensional models of the target organ in all cross sections are transformed to the same coordinate system, and the intersection of the three-dimensional models of all cross sections is calculated. The three-dimensional segmentation model of the target organ is obtained based on the intersection result.

5. The organ segmentation and modeling terminal based on medical images according to claim 4, characterized in that, The identification and tracking of target organs in continuous medical images includes: Construct a first image sequence containing the first image and images after the first image, and a second image sequence containing the first image and images before the first image; Based on the image segmentation results of the target organ in the first image, the target organ in the first image sequence and the second image sequence is identified and tracked to obtain the initial tracking result.

6. The organ segmentation and modeling terminal based on medical images according to claim 5, characterized in that, The calculation of the overlap ratio of continuous tracking results, if the overlap ratio is greater than a preset ratio, then the tracking result of the target organ is obtained as follows: In the image sequence, starting from the first image, the overlap ratio between the tracking result of the current image and the initial tracking result of the next image is calculated. If the overlap ratio is greater than or equal to a preset ratio, the initial tracking result of the next image is stored, and the next image is used as the current image to calculate the overlap ratio until the overlap ratio of all images in the image sequence has been calculated. If the overlap ratio is less than the preset ratio, it is assumed that the target organ does not exist in the next image, and organ segmentation ends.

Images