Method and device for segmenting ultrasonic image sequence

Abstract

The invention provides an ultrasonic image sequence segmentation method and device, and the method is characterized in that the method comprises the following steps: an original template obtaining step: obtaining an original contour template of a target organ, and the original contour template shows the contour form of the target organ in a natural state; an image acquisition step: acquiring an ultrasonic image sequence of a plurality of image frames of the target organ in a deformation state due to extrusion of an instrument, and positioning an iteration starting frame in the ultrasonic image sequence; a deformation field construction step: based on the original contour template and the iteration start frame, establishing a deformation field representing a space mapping relation between a natural state and a deformation state of the target organ; and a contour segmentation step: based on the original contour template and the deformation field, obtaining contour information of the target organ in each image frame of the ultrasonic image sequence in an inter-frame iteration mode, and obtaining a contour sequence of a deformation state of the target organ.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and specifically to a segmentation method and apparatus for extracting the contour of a target organ based on two-dimensional ultrasound image sequence segmentation, which is required in procedures such as prostate biopsy or treatment. Background Technology

[0002] In existing technologies, intraoperative ultrasound images generally suffer from low resolution, high noise, and numerous artifacts, resulting in blurred tissue edges and making accurate segmentation difficult using conventional edge detection algorithms. Furthermore, to obtain complete images of the target organ, ultrasound probes often need to scan within narrow cavities such as the rectum. The physical insertion of the probe or surgical instruments inevitably compresses soft tissues such as the prostate, causing severe nonlinear deformation of the target organ's morphology in the acquired images, deviating from its natural state. Most existing segmentation methods do not specifically model and compensate for this real-time tissue deformation caused by instrument intervention, leading to decreased segmentation accuracy. In addition, deep learning methods rely on large amounts of precisely labeled training data, but the labeling cost of medical images is extremely high; methods based on preoperative CT / MRI and intraoperative ultrasound registration typically require the introduction of external positioning devices such as optical or electromagnetic sensors, increasing the complexity and cost of the surgical system.

[0003] Therefore, in existing technologies, how to efficiently and accurately segment ultrasound image sequences has become a technical challenge. Summary of the Invention

[0004] The purpose of this application is to provide a method and apparatus for segmenting ultrasound image sequences that can perform efficient and accurate segmentation of ultrasound image sequences.

[0005] To achieve the above objectives, one solution of this application is a segmentation method for an ultrasound image sequence, comprising the following steps: an original template acquisition step: acquiring an original contour template of a target organ, wherein the original contour template shows the contour shape of the target organ in its natural state; an image acquisition step: acquiring an ultrasound image sequence of multiple image frames of the target organ in a deformed state due to instrument compression, and locating an iteration start frame therein; a deformation field construction step: based on the original contour template and the iteration start frame, establishing a deformation field representing the spatial mapping relationship between the target organ in its natural state and deformed state; and a contour segmentation step: based on the original contour template and the deformation field, acquiring the contour information of the target organ in each image frame of the ultrasound image sequence in an inter-frame iteration manner to obtain a contour sequence of the deformed state of the target organ.

[0006] According to this invention, by establishing a spatial association between the original contour template in its natural state and the ultrasound image in its deformed state, and combining inter-frame iteration to achieve contour extraction of the ultrasound image sequence, the prior template effectively eliminates ultrasound image noise and artifact interference, ensuring the matching of the segmentation result with the intraoperative deformed state, thus improving the efficiency and accuracy of contour segmentation of the ultrasound image sequence. Simultaneously, it eliminates the need for separately configured external equipment and a large amount of labeled training data, reducing the cost and barrier to clinical application.

[0007] Preferably, in the contour segmentation step, based on the original contour template and the deformation field, the contour information of the target organ in the iteration start frame is obtained, and then... Based on the contour information of the target organ in the iteration start frame, the contour information of the target organ in the remaining image frames of the ultrasound image sequence excluding the iteration start frame is obtained by inter-frame iteration, so as to obtain the contour sequence of the deformation state of the target organ.

[0008] According to this invention, after contour extraction is completed starting from the initial frame of the iteration, iterative extraction is then performed on the remaining image frames. The contour information of the initial frame is derived from the original contour template and the deformation field, resulting in high accuracy. Furthermore, by using this as a basis to pass segmentation constraints to subsequent frames, the accumulation of segmentation errors can be effectively suppressed, ensuring the continuity and accuracy of contour extraction from the initial frame to the remaining frames. This avoids the error problems caused by independent segmentation frame by frame and improves the consistency of contour segmentation in ultrasound image sequences.

[0009] Preferably, in the contour segmentation step, the inter-frame iteration is performed using the ultrasound image sequence as the iteration carrier or the restored ultrasound image sequence as the iteration carrier. The restored ultrasound image sequence is based on the deformation field to deform the ultrasound image sequence, thereby restoring it to a natural state image sequence.

[0010] Preferably, when the ultrasound image sequence is used as the iterative carrier for inter-frame iteration, the original contour template is deformed based on the deformation field to obtain a deformed contour template corresponding to the iteration start frame. Starting from the deformed contour template, the contour information of the target organ in each image frame is obtained in the ultrasound image sequence in an inter-frame iterative manner to obtain the contour sequence of the deformed state of the target organ.

[0011] Preferably, when using the restored ultrasound image sequence as an iterative carrier for inter-frame iteration, the ultrasound image sequence is deformed based on the deformation field to obtain the restored ultrasound image sequence. The contour information of the target organ in each image frame of the restored ultrasound image sequence is obtained by inter-frame iteration to obtain the restored contour sequence. Based on the inverse transformation of the deformation field, the restored contour sequence is converted into the contour sequence of the deformed state of the target organ.

[0012] According to the present invention, two carrier selection schemes for inter-frame iteration are provided: iteration can be based directly on ultrasound image sequences in their deformed state, or iteration can be based on ultrasound image sequences restored to their natural state. This allows the segmentation method to adapt to different clinical application scenarios. For example, for intraoperative navigation scenarios with high real-time requirements, ultrasound image sequences can be selected as the iteration carrier; for preoperative planning and postoperative assessment scenarios with higher accuracy requirements, restored ultrasound image sequences can be selected as the iteration carrier. This significantly improves the flexibility and scenario adaptability of the segmentation method, achieving optimal segmentation results under different clinical needs.

[0013] Preferably, in the contour segmentation step, the image frame with extracted contour information is used as a reference frame, the extracted contour information is passed to the next frame to be processed, and the contour information in the frame to be processed is extracted based on the passed contour information; this process is iterated until the contour information of all image frames is extracted.

[0014] According to the present invention, the high-precision contour information extracted with reference to the original contour template is used as a priori constraint and passed to subsequent frames to be processed. The contours of the subsequent frames to be processed are extracted based on this constraint. Thus, the high similarity and continuity between frames of the ultrasound image sequence are fully utilized.

[0015] Preferably, after the contour information extracted from the reference frame is passed to the next frame to be processed, a local search region is defined in the frame to be processed based on the contour information; within the local search region, the contour information in the frame to be processed is obtained at least based on the image grayscale distribution features.

[0016] According to the present invention, instead of performing a global search on the entire image, the contour extraction range of the frame to be processed is limited to a local region surrounding a reference contour. This operation significantly reduces the computational load of contour extraction and improves the segmentation speed. Simultaneously, it avoids interference from irrelevant regions in the image, allowing contour extraction to focus on the edge regions of the target organ, further improving the accuracy and stability of contour segmentation.

[0017] Preferably, after the contour information extracted from the reference frame is passed to the next frame to be processed, a scaling factor is determined based at least on the relative positional relationship between the iteration start frame, the reference frame, and the frame to be processed; then, the contour information passed from the reference frame is scaled based on the scaling factor to obtain scaled contour information, and the local search area is delineated with the scaled contour information as a reference.

[0018] Preferably, the scaling factor S is calculated according to the following formula:

[0019] Among them, Z current Z reference Z0 and Z0 are the positions of the frame to be processed, the corresponding reference frame, and the iteration start frame in the restored ultrasound image sequence, respectively; L is the normalized length, which is the distance from the iteration start frame to the end frame in the direction of the frame to be processed.

[0020] According to the present invention, a scaling mechanism based on inter-frame relative position is introduced to pre-adjust the size of the contour information transmitted to the frame to be processed, so that the reference contour can adapt to the actual size changes of the target organ in the frame to be processed. This mechanism simulates the morphological changes of the organ naturally contracting / expanding from the center to both ends, making the defined local search area fit the edge of the target organ in the frame to be processed more closely, further improving the coherence and accuracy of inter-frame iterative segmentation, and reducing segmentation errors caused by size deviations.

[0021] Preferably, in the deformation field construction step, on the iteration start frame, a first feature point set corresponding to the contour of the target organ is extracted; on the original contour template, a second feature point set corresponding to the first feature point set is extracted; and the deformation field is obtained based at least on the spatial coordinate transformation between the first feature point set and the second feature point set.

[0022] According to the present invention, a spatial mapping relationship between the natural state and the deformed state is established through spatial coordinate transformation of the feature point set, enabling precise quantification of the deformation field. The feature point set selects key anatomical sites on the contour of the target organ. The deformation field constructed based on the coordinate transformation of the feature points can accurately describe the local deformation caused by instrument compression, which is the technical basis for the contour segmentation method of the present invention to achieve deformation compensation using the original template.

[0023] Preferably, the iteration start frame is any one of the key image frame groups, which includes at least one of the following: a leading frame, which is a frame selected from several frames in the starting region of the ultrasound image sequence where the target organ's features are continuously greater than a preset threshold; an ending frame, which is a frame selected from several frames in the ending region of the ultrasound image sequence where the target organ's features are continuously greater than a preset threshold; and a maximum area frame, which is the frame in the ultrasound image sequence where the target organ has the largest area.

[0024] According to the present invention, a leading frame, a trailing frame, and a frame with the largest area are selected from an ultrasound image sequence, preferably with the frame with the largest area selected as the iteration starting frame. This ensures that the contour information of the iteration starting frame has high recognizability and accuracy.

[0025] Preferably, the original template acquisition step includes at least one of the following as the original contour template: segmenting the target organ and extracting its maximum cross-sectional contour based on the preoperative MRI image of the same patient as the original contour template; obtaining the original contour template by statistically analyzing the maximum cross-sectional contours of multiple samples based on an open-source ultrasound image dataset of the target organ; and obtaining the original contour template by converting the typical size and shape parameters of the target organ into a geometric model based on the doctor's experience data.

[0026] Another aspect of the present invention is a segmentation device for an ultrasound image sequence, comprising: an original template acquisition unit for acquiring an original contour template of a target organ, the original contour template showing the contour shape of the target organ in its natural state; an image acquisition unit for acquiring an ultrasound image sequence of multiple image frames of the target organ in a deformed state due to instrument compression, and locating an iteration start frame therein; a deformation field construction unit for establishing a deformation field representing the spatial mapping relationship between the target organ in its natural state and deformed state based on the original contour template and the iteration start frame; and a contour segmentation unit for acquiring the contour information of the target organ in each image frame of the ultrasound image sequence in an inter-frame iteration manner based on the original contour template and the deformation field, thereby obtaining a contour sequence of the deformed state of the target organ.

[0027] According to this invention, by establishing a spatial association between the original contour template in its natural state and the ultrasound image sequence in its deformed state, and combining this with inter-frame iteration to extract the contour of the ultrasound image sequence, the prior template effectively eliminates ultrasound image noise and artifact interference, ensuring the matching of the segmentation result with the intraoperative deformed state, thus improving the efficiency and accuracy of contour segmentation of the ultrasound image sequence. Simultaneously, it eliminates the need for additional external equipment and a large amount of labeled training data, reducing the cost and barrier to clinical application. Attached Figure Description

[0028] To more clearly illustrate this application, the accompanying drawings will be described and explained below. Obviously, the drawings described below only illustrate certain aspects of some exemplary embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0029] Figure 1 This is a schematic diagram illustrating the principle of using the original contour template of the target organ.

[0030] Figure 2 This is a schematic diagram of the ultrasound image sequence of the target organ after deformation and its corresponding contour information. Detailed Implementation

[0031] Various exemplary embodiments of this application are described in detail below with reference to the accompanying drawings. The descriptions of the exemplary embodiments are merely illustrative and are in no way intended to limit the application or its application or use. This application can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the application thorough and complete, and to fully express the scope of the application to those skilled in the art. It should be noted that, unless otherwise stated, the relative arrangement of components and steps, numerical expressions, and values ​​set forth in these embodiments should be interpreted as merely exemplary and not as limiting.

[0032] As used in this application, the words “including” or “comprising” or similar terms mean that the element preceding the word covers the element listed after the word, and do not exclude the possibility that it may also cover other elements.

[0033] All terms used in this application (including technical or scientific terms) have the same meaning as understood by one of ordinary skill in the art to which this application pertains, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant art, and not as being interpreted with idealized or highly formalized meanings, unless explicitly defined herein.

[0034] For components, specific model numbers and other parameters of components not described in detail in this section, the interrelationships between components and control circuits, these may be considered as techniques, methods and devices known to those skilled in the art, but where appropriate, such techniques, methods and devices should be considered part of the specification.

[0035] It should be noted that although the operations of the method described in this application are given a specific order, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. On the contrary, the steps described in this application may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0036] The following is for reference Figure 1 and Figure 2 The segmentation method of the ultrasound image sequence of this application is described. Figure 1 This is a schematic diagram illustrating the principle of using the original contour template of the target organ. Figure 2 This is a schematic diagram of an ultrasound image sequence showing the deformed state of a target organ and its corresponding contour information. The segmentation described in this application refers to extracting information such as contour shape.

[0037] For ease of explanation, the prostate is used as the target organ and the ultrasound probe as the aforementioned instrument, but it is not limited to these. (Reference) Figure 1 , Figure 2 The Z-axis is used as the depth direction of the ultrasound probe, and the plane perpendicular to the Z-axis is the cross-section. After the ultrasound probe is inserted into the human cavity along the Z-axis, a set of cross-sectional images of the target organ are acquired, as shown in Z1 to Z2 in the figure. n As shown. However, inserting an ultrasound probe into the cavity can cause the target organ to be compressed and deformed, therefore Z1 to Z n This is a sequence of ultrasound images showing the real-time deformation of the target organ.

[0038] Accurate segmentation of the target organ's contour in ultrasound image sequences is crucial for surgical planning. Furthermore, since surgery is performed with instruments inserted, the contour information of the target organ extracted from the ultrasound image sequence for surgical planning should be a contour sequence deformed by the instruments. However, intraoperative ultrasound images generally suffer from low resolution, high noise, and numerous artifacts, resulting in blurred tissue edges that are difficult to segment accurately using conventional edge detection algorithms.

[0039] Preoperatively, a raw template of the target organ can be obtained, showing its contour information in its natural state. However, this raw template cannot be directly used to assist in contour segmentation. This is because the contour information of the target organ extracted from ultrasound image sequences is a contour sequence deformed by instrument compression, while the raw template of the target organ obtainable preoperatively is its contour shape in its natural state without instrument compression.

[0040] In this invention, the applicants, through in-depth research and repeated experiments, have innovatively proposed an inter-frame iteration method based on the original template, which enables accurate segmentation of the contour of the target organ in an ultrasound image sequence.

[0041] The ultrasound image sequence segmentation method of this application includes the following steps: Original template acquisition steps: Obtain the original contour template of the target organ, which shows the contour shape of the target organ in its natural state; Image acquisition steps: Acquire an ultrasound image sequence of multiple image frames of the target organ in a deformed state due to instrument compression, and locate the iterative starting frame within it; Deformation field construction steps: Based on the original contour template and the iteration start frame, establish a deformation field representing the spatial mapping relationship between the target organ in its natural state and deformed state; Contour segmentation step: Based on the original contour template and the deformation field, the contour information of the target organ in each image frame of the ultrasound image sequence is obtained by inter-frame iteration, so as to obtain the contour sequence of the deformation state of the target organ.

[0042] As a preferred approach, the original template acquisition steps include at least one of the following: segmenting the target organ and extracting its maximum cross-sectional contour based on the preoperative MRI images of the same patient as the original contour template; obtaining the original contour template by statistically analyzing the maximum cross-sectional contours of multiple samples from an open-source ultrasound image dataset of the target organ; and obtaining the original contour template by converting the typical size and shape parameters of the target organ into a geometric model based on the doctor's experience data.

[0043] Specifically, the original contour template acquisition step aims to obtain the contour shape of the target organ at a specific level in its natural state. In this embodiment, the contour shape of the largest cross-section, i.e., the prostate, with the largest area on the cross-sectional image, is taken as the original contour template as a reference constraint for subsequent segmentation. In fact, it is not limited to this. The contour shape of other specific cross-sectional images can also be taken as the original contour template, which will not be elaborated here.

[0044] For example, this application provides three complementary approaches to obtaining the original contour template, which can be preferentially used according to accuracy requirements and degree of individualization: 1. Individualized Template: Obtain preoperative prostate MRI, CT, or abdominal ultrasound images from the same patient. Use segmentation algorithms (such as level set, graph cut, etc.) or manual delineation to segment the three-dimensional shape of the prostate. Extract the layer with the largest cross-sectional area and use its two-dimensional contour as the original contour template. Preoperative prostate MRI images are preferred as they provide the highest template accuracy.

[0045] 2. Statistical General Template: Utilizing an open-source dataset containing a large number of prostate ultrasound images and annotations, the maximum cross-sectional contours of multiple samples in the dataset are extracted. By calculating the average shape of these contours or performing principal component analysis, a statistical general contour template is generated as the original contour template. This template represents common morphologies in the population.

[0046] 3. Empirical Geometric Template: Based on experience, doctors know that the typical shape of the prostate resembles a chestnut, with its largest cross-section approximating an ellipse or convex polygon, and it has a common range of dimensions (such as the length of the major and minor axes). Based on this empirical knowledge, a parameterized geometric contour template is obtained as the original contour template, such as a standard ellipse.

[0047] The original contour template obtained in the above manner is stored as an ordered set of points or in the form of a parametric equation, denoted as T. prior .

[0048] Next, the image acquisition steps will be explained.

[0049] As a preferred method, in the image acquisition step, the operator inserts an ultrasound probe through the rectum to scan the entire prostate, obtaining a series of continuous two-dimensional cross-sectional images, forming an ultrasound image sequence Z={Z1,Z2,...,Z...} n},like Figure 2 As shown.

[0050] In the ultrasound image sequence Z={Z1,Z2,...,Z n In the process, several image frames are selected as key image frame groups. The key image frame group includes at least one of the following: a leading frame, which is a frame selected from several frames in the starting region of the ultrasound image sequence where the target organ's features are continuously greater than a preset threshold; an ending frame, which is a frame selected from several frames in the ending region of the ultrasound image sequence where the target organ's features are continuously greater than a preset threshold; and a maximum area frame, which is the frame with the largest area of ​​the target organ in the ultrasound image sequence.

[0051] In this embodiment, the Z1 side is taken as the starting side, and the region near the starting side is the starting region of the ultrasound image sequence. n On the lateral termination side, the region near the termination side is the termination region of the ultrasound image sequence. Automatic thresholding (such as the Otsu algorithm) or fast edge detection is used to calculate the approximate area of ​​the prostate region, i.e., the contour area, for each frame of the image. Figure 2 Z1 on the right ’ To Z n ’ As shown. It should be noted that the automatic thresholding segmentation (such as the Otsu algorithm) or fast edge detection here are just examples, and the specific methods and algorithms are not limited.

[0052] In the initial region of the ultrasound image sequence, several frames will have a contour area smaller than a preset threshold; these frames are discarded because their contour areas are too small. When frames with contour areas exceeding the preset threshold begin to appear, it signifies the emergence of frames with practical value. However, due to inherent image and algorithmic errors, fluctuations may occur after the initial appearance of frames with contour areas exceeding the preset threshold, and the next frame may again have a contour area smaller than the threshold. Therefore, only when multiple consecutive frames have contour areas exceeding the preset threshold are the results considered reliable. From these frames, the first frame with a contour area exceeding the preset threshold is selected as the leading frame, labeled Z. start .

[0053] Similarly, in the termination region of an ultrasound image sequence, there will be multiple frames with contour areas smaller than a preset threshold. These frames are discarded because their contour areas are too small. When frames with contour areas larger than the preset threshold begin to appear, it means that frames with practical value have begun to appear. However, due to errors in the image itself and the algorithm, there may be fluctuations after the first frame with a contour area larger than the preset threshold appears, and the next frame may again have a contour area smaller than the preset threshold. Therefore, only when several consecutive frames have contour areas larger than the preset threshold is the result considered reliable. From these frames, one frame, for example, the last frame with a contour area larger than the preset threshold, is selected as the ending frame and labeled Z. end .

[0054] The shape of the prostate gland dictates that the frame with the largest outline area is usually located near its center. The frame with the largest area is naturally more accurate when segmenting the outline of the target organ. Therefore, in an ultrasound image sequence, the frame with the largest outline area of ​​the prostate gland is selected as the maximum area frame, labeled Z. max It is also added to the key image frame group. Thus, as an example, Z... start Z end Z max The key image frame group constitutes this embodiment. Of course, other frames can be selected to enter the key image frame group; here, only Z is used. start Z end Z max This is a preferred example. The system can provide an interactive interface that allows doctors to manually fine-tune or directly specify these three levels, ensuring reliability.

[0055] Next, the steps for constructing the deformation field and the contour segmentation will be explained.

[0056] As a preferred approach, a first set of feature points corresponding to the contour of the target organ is extracted from a specific frame of the key image frame group; a second set of feature points corresponding to the first set of feature points is extracted from the original contour template; and a deformation field is obtained based at least on the spatial coordinate transformation between the first and second set of feature points. Here, the "specific frame of the key image frame group" is also the starting frame for subsequent inter-frame iterations.

[0057] Here, the largest area frame Z is used. max For example, in Z max On the image, key feature points on the prostate contour are marked. For example... Figure 2 As shown in the lower part of the figure, four feature points are illustrated: A1 at the top, A2 at the leftmost point, A3 at the bottom where the contour is deformed (due to pressure from the probe), and A4 at the rightmost point. However, the number and location of these feature points are not limited to these. For example, other feature points could include A5 (midpoint of the A1-A2 arc), A6 (midpoint of the A1-A4 arc), A7 (lowest point on the left), and A8 (lowest point on the right). The specific number and location of these feature points are not limited; this illustration only uses A1-A4 as an example. These feature points constitute the first feature point set, denoted as P. patient In practical applications, this can be achieved by manually selecting points, or by first using the Canny operator to detect edges and then automatically locating feature points. The specific method or algorithm is not limited.

[0058] Next, from the original contour template T prior Found on P patient The point set P corresponding to the anatomical location of each feature point in the middle template That is, in the original contour template, locate the topmost, leftmost, bottom contour deformation (upward concavity), and rightmost feature points, as well as other features contained in P. patient The feature points in the space are transformed to make P patient The feature points in P template Align the feature points in the data.

[0059] For example, this application employs the Thin Plate Spline Transform (TPS) nonlinear registration method to establish a deformation field that precisely matches the corresponding feature points while minimizing the overall bending energy. By solving the TPS coefficients, the deformation field Φ, representing the spatial mapping relationship between the deformation state corresponding to the current ultrasound image sequence and the natural state corresponding to the original contour template, is obtained. In other words, the deformation field Φ is the mapping relationship between the deformation state space corresponding to the real-time ultrasound image sequence and the natural state space corresponding to the original contour template.

[0060] The core of this step lies in constructing a spatial mapping relationship that can quantitatively describe the target organ's "deformation state" and "natural state" under pressure. The principle is not a simple rigid translation, but a nonlinear spatial transformation based on the correspondence of anatomical structures.

[0061] Specifically, this application uses specific frames in a key image frame group (such as the maximum area frame Z) max Extract a set of feature points P with clear anatomical significance from the data. patient and the original contour template T representing the natural state prior The corresponding feature point set P on template Establishing spatial correspondence. Since the deformation of the target organ caused by compression is usually nonlinear and local (e.g., indentation caused by ultrasound probe compression), this application preferably uses a nonlinear registration algorithm such as Thin Plate Spline Interpolation (TPS) to construct the deformation field Φ. This algorithm can accurately match all corresponding feature points while minimizing the bending energy function, ensuring that the entire deformation field is spatially smooth and continuous, thereby simulating the real physical deformation process of the target organ from a locally compressed state to its natural state. Thin Plate Spline Transformation (TPS) is only one preferred algorithm; this application does not limit the specific algorithm used.

[0062] In this way, the deformation field Φ precisely records the displacement vector that each pixel in the ultrasound image sequence should have during the mapping process. It establishes a unified and reversible spatial transformation benchmark, thus laying a spatially consistent foundation for accurately segmenting the contour shape of the prostate.

[0063] It is worth noting that in the above description, several key image frames (e.g., the leading frame, the ending frame, and the frame with the largest area) are first extracted from the ultrasound image sequence to form a key image frame group. Then, a key image frame is selected from the key image frame group as the iteration start frame. This is because defining a key image frame group including the leading and ending frames is helpful for both the inter-frame iteration described later in this invention and the contour segmentation of each image frame that is already being performed. Therefore, in this invention, a frame with the largest area is added to the key image frame group and used as the iteration start frame for inter-frame iteration. Of course, not defining a key image frame group and directly specifying the iteration start frame does not affect the implementation of this invention. Furthermore, the iteration start frame is not limited to the frame with the largest area.

[0064] The aforementioned contour segmentation steps include a first segmentation mode as in Embodiment 1 and a second segmentation mode as in Embodiment 2, which will be described separately below.

[0065] First, the first segmentation mode of Example 1 will be explained.

[0066] In the first segmentation mode, each frame of the ultrasound image sequence is deformed based on the deformation field to obtain a restored ultrasound image sequence; in the restored ultrasound image sequence, the contour information of the target organ in each frame is obtained by inter-frame iteration to obtain a restored contour sequence; based on the inverse transformation of the deformation field, the restored contour sequence is converted into contour information corresponding to the natural state of the target organ.

[0067] As a preferred method, in deformation correction, deformation correction is performed on each frame of the ultrasound image sequence based on the deformation field to obtain a restored ultrasound image sequence. Specifically, the obtained deformation field Φ is applied to the entire ultrasound image sequence Z. For each frame of image Z in the sequence... k Resampling is performed using the deformation field Φ (e.g., bilinear interpolation). This involves generating a completely new, corrected image based on the mapping relationship given by the deformation field Φ, ultimately resulting in the corrected image sequence Z′={Z1′,Z2′,...,Z...}. n At this point, the prostate morphology of all images in the sequence has been "corrected" to a shape consistent with the original contour template T. prior Similar shapes and spaces that are closer to their natural state greatly facilitate subsequent unified segmentation.

[0068] In other words, this method uses a deformation field to uniformly "pull back" all images with different degrees of deformation to the original contour template T. prior The corresponding natural state space solves the problem of "inconsistent deformation" at different locations, reduces the morphological uncertainty caused by mechanical compression, and paves the way for processing all frames using a unified method.

[0069] Furthermore, in the corrected ultrasound image sequence, the contour morphology of the target organ is similar to the pre-acquired original contour template, which allows the original contour template to be directly used as a strong shape constraint for subsequent iterative segmentation, thereby greatly improving the accuracy of segmentation.

[0070] As a preferred approach, in the inter-frame iteration, a specific frame in the restored ultrasound image sequence is used as the iteration start frame, and the contour information of the target organ in the iteration start frame is extracted with reference to the original contour template; the frame in the restored ultrasound image sequence with extracted contour information is used as the reference frame, and its extracted contour information is passed to the next frame to be processed, and then the contour information in the frame to be processed is extracted, and the process is iterated until the contour information of all frames is extracted.

[0071] Preferably, the corrected maximum area frame Z max ′ (i.e. Z) maxUsing the corrected image as the starting frame for iteration, and the original contour template as the reference contour, information such as the contour shape of the prostate in the starting frame is extracted. Specifically, the original contour template is overlaid on the Z-axis. max On the original contour template, a preset search distance (e.g., 5 pixels, the specific pixel value is determined based on the image resolution and its features) is extended both inwards and outwards along the normal direction of the contour lines. The area within this search distance is the local search region. Within the narrow band of this local search region, the gray-level gradient of the image is calculated. The points with the largest gradients are connected to form the Z-axis. max The precise contour information of the frame is obtained. This process can be efficiently completed within a narrow band using an active contour model (Snake) or a fast traversal method. Here, only the frame with the largest area after correction is used as an example, but other frames can also be used as the starting frame for iteration; no specific limitation is made here.

[0072] However, when processing frames that are not the maximum area, such as the current frame Z to be processed... k Located in Z max Downstream of the ultrasound probe along the depth direction, its reference profile is taken from a specific frame of the extracted profile information on its upstream side, for example, Z. k-1 ′ represents the frame Z to be processed k A reference frame is used. The outline shape of the reference frame is overlaid on the frame to be processed. A preset search distance (e.g., 5 pixels) is extended inwards and outwards along the normal direction of the reference frame's outline. The area within this search distance is the local search region. Within the narrow band of this local search region, the grayscale gradient of the image is calculated. The points with the largest gradients are connected to form the precise contour information of the frame to be processed. This process can be efficiently completed within a narrow band using an active contour model (Snake) or a fast traversal method.

[0073] More preferably, a size pre-scaling mechanism based on the relative position between frames is introduced here. After the contour information extracted from the reference frame is passed to the next frame to be processed, the scaling factor S is determined at least based on the relative positional relationship between the iteration start frame, the reference frame, and the frame to be processed. Then, the contour information passed from the reference frame is scaled based on the scaling factor S to obtain scaled contour information. The local search region is delineated on the frame to be processed with the scaled contour information as a reference.

[0074] Preferably, the scaling factor S is calculated according to the following formula:

[0075] Among them, Z current Z referenceZ0 and Z1 represent the positions of the frame to be processed, its corresponding reference frame, and the iteration start frame in the ultrasound image sequence, respectively; L is the normalized length, which is the distance from the iteration start frame to the end frame in the direction of the frame to be processed. For example, if the frame to be processed is located downstream of the iteration start frame along the depth direction of the ultrasound probe, and the reference frame is located upstream of the frame to be processed, then the distance from the iteration start frame to the end frame in the downstream direction is L.

[0076] The scaling factor S physically simulates the quadratic curve pattern of the prostate gland's smooth contraction from the center to both ends. Then, with the centroid of the contour as the center, the contour shape of the reference frame is scaled by a factor of S to obtain a contour shape that more closely matches the actual size of the current frame to be processed. Using this contour shape as a reference, local search regions are defined inward and outward along the normal direction of its contour lines, and the gray-level gradient of the image is calculated within these local search regions. The points where the gradient is largest are connected to form the accurate contour information of the frame to be processed.

[0077] Then, the latest contour information obtained in the current frame to be processed is saved and passed as a new reference contour to the next frame to be processed, for example, with Z... k ′ represents the frame Z to be processed k+1 The reference frame is obtained by repeating the steps of pre-scaling, defining the local search area, and obtaining accurate contour information based on the gray-level gradient, and iterating in sequence until all frames are segmented to obtain the restored contour sequence.

[0078] After image correction, this application extracts the prostate contour through "inter-frame iteration". Its effect is that it can efficiently handle common problems in medical image sequences such as uneven grayscale, blurred boundaries and noise interference, avoiding the drawbacks of traditional frame-by-frame segmentation methods that are time-consuming and prone to errors.

[0079] The specific process involves a specific frame in the restored ultrasound image sequence (such as the corrected maximum area frame Z). max (T) is used as the starting frame for iteration. Since this frame has been precisely corrected by the deformation field Φ, it can be stably referenced to the original contour template T. prior High-precision contour information is extracted. Then, taking advantage of the high similarity and continuity between frames in the Z-axis direction of the ultrasound image sequence, the currently accurately segmented frame is used as a reference frame, and its contour information is used as a priori shape constraint and passed to the next frame to be processed.

[0080] Preferably, the next frame to be processed is the adjacent frame of the reference frame, and the adjacent frames have a higher similarity. However, it is not limited to this. It can also be several frames apart from the reference frame. The second segmentation mode of Embodiment 2 is consistent with this, and will not be described again.

[0081] The iterative transfer process does not simply copy the contours, but combines the grayscale gradient of the image in the frame to be processed with the shape constraints from the reference frame to perform local fine-grained search and adjustment. This method makes full use of the continuity information of the sequence images, enabling the contour extraction process to proceed smoothly and stably throughout the entire sequence, thereby quickly generating the restored contour sequence of all frames in the entire sequence. Furthermore, since each iteration is constrained by the accurate result of the previous reference frame, the accumulation of segmentation errors is effectively suppressed.

[0082] As a preferred approach, in the inverse transformation of the deformation field, the restored contour sequence is converted into a contour sequence of the deformation state of the target organ based on the inverse transformation of the deformation field.

[0083] The deformation field Φ is a mapping function from the ultrasound image sequence Z to the corrected image sequence Z′. During surgery, the surgeon sees real-time ultrasound images. If the segmentation results only exist in the corrected images, they cannot be directly superimposed on the real-time ultrasound images, thus losing their clinical guidance value. This step, through inverse transformation, precisely "projects" the segmentation results back onto the real-time ultrasound image sequence containing the deformation state. This allows the surgeon to clearly see the outline boundary of the target organ on a real-time ultrasound image that perfectly corresponds to the real-time position of the ultrasound probe, providing the most intuitive visual basis for precise surgical navigation.

[0084] This step utilizes the mathematical invertibility of the deformation field, ensuring that the segmentation results benefit from the prior knowledge of the original contour template and can accurately regress to the real-time ultrasound image sequence containing deformation, thus achieving a closed loop for the entire method in clinical applications.

[0085] Next, the second segmentation mode of Example 2 will be described. Only the differences from Example 1 will be emphasized; the parts that are the same as in Example 1 will not be repeated.

[0086] The difference between the second segmentation mode and the first segmentation mode is that the second segmentation mode does not require deformation correction of the entire ultrasound image sequence. Instead, it maps the original contour template to the deformation state space and directly completes the inter-frame iterative segmentation of the contour on the original ultrasound image sequence, thereby omitting the inverse transformation step and further improving processing efficiency.

[0087] In the second segmentation mode, after the deformation field construction step, the original contour template T is first segmented based on the established deformation field Φ. prior Perform deformation mapping to generate the iteration start frame corresponding to a specific frame in the key image frame group (e.g., the maximum area frame Z). max The deformable profile template T) deformedThis mapping process utilizes the invertibility of the deformation field Φ, causing each contour point in the original contour template to move to the corresponding position in the deformation state space according to the displacement vector described by the deformation field Φ, thereby obtaining an initial contour that matches the actual deformation state of the target organ in the image of the initial iteration frame.

[0088] Subsequently, using the deformed profile template T deformed Take it as the starting point, that is, take it as the starting frame of the iteration (e.g., the maximum area frame Z). max The initial contour of the image is obtained and directly segmented into frames using inter-frame iteration within the real-time ultrasound image sequence Z. The iteration process is as follows: the obtained initial frame (e.g., the frame with the largest area Z) is used for iteration. max The contour information of ) is the aforementioned deformable contour template T deformed The reference contour is passed to the next frame to be processed, and combined with information such as image grayscale gradient and edge features, the precise contour of the target organ in the frame to be processed is extracted within the local search area. The most recently obtained precise contour in the frame to be processed is used as the new reference contour, and the process continues to iterate towards both ends of the sequence until the segmentation of all frames is completed. The method of delineating the local search area and obtaining precise contour information based on grayscale gradient is the same as in Example 1, and will not be repeated here.

[0089] In this invention, the core idea is the same in both Embodiment 1 and Embodiment 2. That is, with the aid of the original template, the deformation field between the original contour template of the target organ in its natural state and the ultrasound image sequence of the target organ in its deformed state is calculated. Then, using the original contour template, the contour information of the target organ is segmented in the initial frame of the iteration.

[0090] In the case of inter-frame iteration using ultrasound image sequences as the iterative carrier (Example 2), the original contour template is deformed into the space of the ultrasound image sequence, thereby directly segmenting the contour information of the target organ in the initial frame of the iteration. Furthermore, in subsequent inter-frame iterations, the contour information of the target organ is also directly segmented in each image frame.

[0091] In the case of inter-frame iteration using the restored ultrasound image sequence as the iterative carrier (Example 1), it is necessary to deform the ultrasound image frames to the space of the original contour template, i.e., generate the restored ultrasound image sequence. Next, in the restored ultrasound image sequence, the contour information of the target organ is segmented from each image frame. Since the contour information of the target organ segmented at this time is in its natural state, it is necessary to perform an inverse transformation based on the deformation field to transform it into a deformed state under instrument compression. Of course, this inverse transformation can be performed uniformly after the contour segmentation of the target organ has been completed in all image frames, or it can be performed during the inter-frame iteration process, after the contour segmentation of the target organ in each image frame is completed, by inversely transforming the contour information of the target organ in that image frame into a deformed state under instrument compression. Furthermore, the restored ultrasound image sequence can be generated uniformly or generated one by one during the iteration process. In short, the final result should be a contour sequence that represents the deformed state of the target organ.

[0092] The phrase "obtaining the contour information of the target organ in the iteration start frame based on the original contour template and the deformation field" in this invention broadly includes the two scenarios mentioned above. That is, in Embodiment 1, after restoring the iteration start frame to its natural state based on the deformation field, the contour information of the target organ segmented in the restored iteration start frame can also be broadly interpreted as "the contour information of the target organ in the iteration start frame" since the contour information of the target organ at this time only needs to undergo one inverse transformation to become the contour information of the deformed state.

[0093] During inter-frame iteration, a size pre-scaling mechanism based on the relative position between frames can also be introduced to simulate the physiological morphological change of the target organ gradually shrinking from the center to both ends. Specifically, the size of the contour transmitted to the frame to be processed can be adjusted according to the aforementioned formula for calculating the scaling factor S, so that it better matches the actual size of the target organ in the current frame to be processed, thereby improving the accuracy of segmentation. The size scaling mechanism here is the same as in Example 1 and will not be described again.

[0094] Using the above method, Example 2 directly obtains the deformed contour sequence corresponding to each frame in the original ultrasound image sequence. The contour information in this sequence completely corresponds to the images acquired in real time during surgery and can be directly used for surgical navigation or 3D reconstruction without further inverse transformation processing. Compared to Example 1, this example reduces the calculation of image resampling and inverse transformation, making it particularly suitable for clinical scenarios with higher real-time requirements.

[0095] It should be noted that the two embodiments mentioned above are not mutually exclusive or isolated technical solutions, but rather two optimization paths based on the same core technical framework. They are organically complementary in terms of technical logic, application scenarios, and effect verification, and together they construct a complete technical solution system.

[0096] First, they share a unified technical logic. Both embodiments share the same core technical elements: both use the original contour template as prior knowledge, both construct the deformation field Φ based on key image frame groups, and both employ an inter-frame iterative segmentation strategy. The only difference lies in the application direction of the deformation field—the first segmentation mode applies the deformation field to image deformation correction (mapping the ultrasound image sequence to the natural state space), while the second segmentation mode directly uses the deformation field for template mapping (mapping the original contour template to the deformation state space). This "two-way application" design makes the technical framework of this application highly flexible, enabling the selection of the optimal path according to different needs.

[0097] Secondly, they offer complementary application scenarios. The first segmentation mode is suitable for scenarios with higher accuracy requirements but less stringent real-time demands, such as preoperative planning and postoperative evaluation (and can also be used for intraoperative real-time navigation). By correcting the image to its natural state space, the morphology of the target organ can be obtained that is comparable to standard data such as preoperative MRI, facilitating fusion analysis. The second segmentation mode is more suited to intraoperative real-time navigation scenarios, such as puncture biopsy and local ablation, as its lower latency and higher fidelity easily meet the needs of physicians for real-time operation. The two implementations cover the complete clinical chain from preoperative to intraoperative, forming full-process technical support.

[0098] Third, the two methods allow for mutual verification of results. In clinical applications, both methods can be used simultaneously to process the same set of ultrasound image sequences. The inversely transformed result output from the first segmentation mode is compared with the deformed contour sequence directly output from the second segmentation mode. If the two are highly consistent, the reliability of the segmentation results is verified; if there are significant differences, it suggests that the deformation field construction may be inaccurate or there may be iterative segmentation errors, requiring manual intervention or parameter adjustment. This dual verification mechanism significantly improves the system's credibility.

[0099] In other words, these two embodiments complement each other within the overall framework of this application through two paths: "deformation correction" and "template mapping". On the one hand, they can be flexibly adapted to different clinical scenarios, and on the other hand, through mutual verification and synergy between the two, the overall technical effect is further improved.

[0100] Next, the beneficial effects of this application will be explained.

[0101] This application proposes a complete method for segmenting ultrasound image sequences, which solves the technical problem of target organ deformation caused by instrument compression, thereby affecting segmentation accuracy.

[0102] First, deformation interference is eliminated. Traditional ultrasound segmentation methods operate directly on images subjected to pressure deformation. Organ morphology undergoes non-rigid distortion due to the compression of the ultrasound probe (e.g., the prostate is flattened or elongated), limiting the accuracy of the segmentation algorithm. This application establishes a precise mapping between the deformed state space and the natural state space by constructing a nonlinear deformation field Φ. Based on this, deformation interference can be effectively overcome through a first segmentation mode of "deformation correction" and / or a second segmentation mode of "template mapping".

[0103] Secondly, inter-frame iteration improves the accuracy and efficiency of sequence segmentation. Ultrasound images often suffer from uneven grayscale, blurred boundaries, and noise interference. This application abandons the inefficient method of independent frame-by-frame segmentation and adopts a contour iteration and transfer method. Starting from the initial frame of the iteration, it utilizes the high similarity between frames to transfer the accurate contour of the acquired reference frame as a priori constraint to the next frame to be processed. This ensures that the segmentation of each frame benefits from the contour information of the previous frame, effectively suppressing interference caused by local noise and weak boundaries, and ensuring the continuity and accuracy of the entire sequence segmentation result.

[0104] Third, dynamic size adjustment is achieved during iterative transmission. In the process of iterative transmission of contour information between frames, this application introduces a dynamic size adjustment mechanism based on relative positional relationships to simulate the physiological morphology of the target organ naturally contracting from the center to both ends. This is different from simple contour copying, making the contour information transmitted between frames more consistent with human anatomy, reducing errors caused by abrupt changes in the shape of the target organ, and improving the coherence and accuracy of long sequence image segmentation.

[0105] Fourth, it does not require a large number of data samples. This application proposes a multi-source original contour template construction mechanism, which provides strong shape prior constraints for the segmentation process through individualized preoperative MRI data (high precision), statistical models of open-source datasets (generality), and physician-experienced geometric models (basic anatomical constraints). This strategy enables the system to eliminate noise and artifact interference in ultrasound images using anatomical structural information even without deep learning training data, achieving low-cost and highly adaptive segmentation.

[0106] Fifth, it lowers the threshold and cost for clinical application. This application is entirely based on image processing algorithms, eliminating the need for expensive external hardware. The operation process is simple; doctors only need to mark a few points on keyframes or directly use automatic algorithms to automatically complete the segmentation of the entire ultrasound image sequence, reducing surgical costs and operational burden.

[0107] In summary, the technical approach of this application significantly improves the accuracy, practicality, and efficiency of ultrasound image sequence segmentation, and systematically solves the segmentation problem in a specific scenario where the target organ is compressed and deformed due to the physical insertion of instruments.

[0108] It should be understood that the specific embodiments described above are only used to explain this application, and the scope of protection of this application is not limited thereto. Any changes, substitutions, or combinations made by those skilled in the art within the scope of the technology disclosed in this application, based on the technical solution and inventive concept of this application, should be covered within the scope of protection of this application.

Claims

1. A method for segmenting an ultrasound image sequence, characterized in that, Includes the following steps, Original template acquisition steps: Obtain the original contour template of the target organ, which shows the contour shape of the target organ in its natural state; Image acquisition steps: Acquire an ultrasound image sequence of multiple image frames of the target organ in a deformed state due to instrument compression, and locate the iterative starting frame within it; Deformation field construction steps: Based on the original contour template and the iteration start frame, establish a deformation field representing the spatial mapping relationship between the target organ in its natural state and deformed state; Contour segmentation step: Based on the original contour template and the deformation field, the contour information of the target organ in each image frame of the ultrasound image sequence is obtained by inter-frame iteration, so as to obtain the contour sequence of the deformation state of the target organ.

2. The method for segmenting an ultrasound image sequence according to claim 1, characterized in that, In the contour segmentation step, based on the original contour template and the deformation field, the contour information of the target organ in the iteration start frame is obtained, and then... Based on the contour information of the target organ in the iteration start frame, the contour information of the target organ in the remaining image frames of the ultrasound image sequence excluding the iteration start frame is obtained by inter-frame iteration, so as to obtain the contour sequence of the deformation state of the target organ.

3. The method for segmenting an ultrasound image sequence according to claim 2, characterized in that, In the contour segmentation step, the inter-frame iteration is performed using the ultrasound image sequence as the iteration carrier or the restored ultrasound image sequence as the iteration carrier. The restored ultrasound image sequence is based on the deformation field to deform the ultrasound image sequence, thereby restoring it to a natural state image sequence.

4. The method for segmenting an ultrasound image sequence according to claim 3, characterized in that, When performing inter-frame iteration using the ultrasound image sequence as the iterative carrier, the original contour template is deformed based on the deformation field to obtain a deformed contour template corresponding to the iteration start frame. Starting from the deformed contour template, the contour information of the target organ in each image frame is obtained in the ultrasound image sequence in an inter-frame iterative manner to obtain the contour sequence of the deformed state of the target organ.

5. The method for segmenting an ultrasound image sequence according to claim 3, characterized in that, When performing inter-frame iteration using the restored ultrasound image sequence as the iterative carrier, the ultrasound image sequence is deformed based on the deformation field to obtain the restored ultrasound image sequence. The contour information of the target organ in each image frame of the restored ultrasound image sequence is obtained by inter-frame iteration to obtain the restored contour sequence. Based on the inverse transformation of the deformation field, the restored contour sequence is converted into the contour sequence of the deformed state of the target organ.

6. The method for segmenting an ultrasound image sequence according to any one of claims 2-5, characterized in that, In the contour segmentation step. The image frame from which the contour information has been extracted is used as a reference frame. The extracted contour information is passed to the next frame to be processed, and the contour information in the frame to be processed is extracted based on the passed contour information. The process is repeated until the contour information of all image frames has been extracted.

7. The method for segmenting an ultrasound image sequence according to claim 6, characterized in that, After the contour information extracted from the reference frame is passed to the next frame to be processed, a local search region is defined in the frame to be processed based on the contour information; within the local search region, the contour information in the frame to be processed is obtained at least based on the image grayscale distribution features.

8. The method for segmenting an ultrasound image sequence according to claim 7, characterized in that, After passing the contour information extracted from the reference frame to the next frame to be processed, the scaling factor is determined at least based on the relative positional relationship between the iteration start frame, the reference frame, and the frame to be processed. Then, the contour information transmitted from the reference frame is scaled based on the scaling factor to obtain scaled contour information, and the local search area is delineated with the scaled contour information as a reference.

9. The method for segmenting an ultrasound image sequence according to claim 8, characterized in that, The scaling factor S is calculated according to the following formula: ， Among them, Z current Z reference Z0 and Z0 are the positions of the frame to be processed, the corresponding reference frame, and the iteration start frame in the restored ultrasound image sequence, respectively; L is the normalized length, which is the distance from the iteration start frame to the end frame in the direction of the frame to be processed.

10. The method for segmenting an ultrasound image sequence according to any one of claims 1-5, characterized in that, In the deformation field construction step, On the initial frame of the iteration, a first set of feature points corresponding to the contour of the target organ is extracted; On the original contour template, extract the second feature point set corresponding to the first feature point set; The deformation field is obtained based at least on the spatial coordinate transformation between the first feature point set and the second feature point set.

11. The method for segmenting an ultrasound image sequence according to any one of claims 1-5, characterized in that, The iteration start frame is any one of the key image frame groups, which includes at least one of the following: The first frame is a frame selected from several frames in the starting region of the ultrasound image sequence where the features are continuously greater than a preset threshold. The final frame is a frame selected from several frames in the termination region of the ultrasound image sequence where the target organ's features are continuously greater than a preset threshold. The maximum area frame is the frame in which the target organ has the largest area in the ultrasound image sequence.

12. The method for segmenting an ultrasound image sequence according to any one of claims 1-5, characterized in that, The original template acquisition step includes at least one of the following as the original contour template: Based on the preoperative MRI images of the same patient, the target organ was segmented and its maximum cross-sectional contour was extracted as the original contour template. Based on the open-source ultrasound image dataset of the target organ, the original contour template is obtained by statistically analyzing the maximum cross-sectional contour of multiple samples. Based on the doctor's experience data, the typical size and shape parameters of the target organ are converted into a geometric model to obtain the original contour template.

13. A segmentation device for an ultrasound image sequence, characterized in that, include: Original template acquisition unit: acquires the original contour template of the target organ, the original contour template showing the contour shape of the target organ in its natural state; Image acquisition unit: acquires an ultrasound image sequence of multiple image frames of the target organ in a deformed state due to instrument compression, and locates the iterative starting frame in the sequence; Deformation field construction unit: Based on the original contour template and the iteration start frame, a deformation field is established to represent the spatial mapping relationship between the target organ in its natural state and deformed state; Contour segmentation unit: Based on the original contour template and the deformation field, the contour information of the target organ in each image frame of the ultrasound image sequence is obtained in an inter-frame iteration manner to obtain the contour sequence of the deformation state of the target organ.

Images