Liver tumor real-time tracking method based on ultrasonic image

Abstract

The invention provides a real-time liver tumor tracking method based on an ultrasonic image, which comprises the following steps: developing an invisible liver tumor through ultrasonic contrast, determining an initial coordinate position, and selecting and defining anatomical markers around the liver tumor; determining a spatial position relationship between the anatomical marker and the liver tumor according to the coordinate data of the anatomical marker and the liver tumor, and constructing a template data set; performing real-time tracking on a plurality of anatomical markers in the conventional ultrasonic image by using a discriminant correlation filter network, obtaining dynamic coordinate information of each anatomical marker, and calculating geometric features of the anatomical markers in the conventional ultrasonic image; finding a frame which is most matched with the current frame of the conventional ultrasonic image in the template data set through geometric features to obtain optimal relative position coordinates of the liver tumor; and performing coordinate affine transformation on the anatomical marker coordinates of the current frame and the matching frame, and performing fine adjustment on the initial coordinate position of the liver tumor through the transformation matrix to obtain a real-time positioning result. According to the method, stable and real-time tumor positioning and tracking can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a real-time liver tumor tracking method based on ultrasonic imaging. Background Art

[0002] Image-guided percutaneous thermal ablation is one of the three radical treatments for early-stage liver cancer. Conventional ultrasound has become the most commonly used guidance and monitoring tool in thermal ablation of liver cancer due to its advantages such as real-time, cost-effectiveness, easy access and no radiation exposure. However, enhanced computed tomography (CECT) and enhanced magnetic resonance imaging (CEMRI) show that 14.6% to 55.0% of focal liver nodules may not be displayed in ultrasound images due to isoechoicity, cirrhosis and other reasons. These inconspicuous lesions may not only lead to missed diagnosis in ultrasound examinations, but also bring great difficulties to patients undergoing thermal ablation under ultrasound guidance.

[0003] Contrast-enhanced ultrasound (CEUS) can be used to locate these inconspicuous lesions before thermal ablation and guide puncture operations. However, CEUS is easily affected by aerosol during ablation, making it difficult to achieve continuous lesion tracking. In addition, CEUS imaging time is short and requires repeated injections of ultrasound contrast agents, which increases the complexity of the operation. Therefore, CEUS is not an ideal method for real-time tracking of lesions. Ultrasound fusion imaging provides a means for real-time and stable tracking of invisible lesions. It uses an electromagnetic positioning system to register and fuse real-time ultrasound images with CT or MR images to achieve synchronous display of multimodal images. However, the registration accuracy of US-CT / MR fusion imaging is highly dependent on the operator's experience and is affected by factors such as breathing and liver displacement, which may cause image errors. For this reason, frequent fine-tuning and alignment of images are required during surgery, limiting the widespread application of this technology.

[0004] In summary, in the prior art, the real-time positioning method for invisible liver tumors during ultrasound-guided thermal ablation has poor accuracy. Summary of the invention

[0005] In view of this, the present invention provides a real-time liver tumor tracking method based on ultrasound imaging to solve the above problems.

[0006] The present invention provides a real-time tracking method for liver tumors based on ultrasound imaging, comprising: determining the initial coordinate position of a liver tumor that is invisible through ultrasound contrast imaging, selecting and defining anatomical markers around the liver tumor; determining the spatial position relationship between the anatomical markers and the liver tumor based on the coordinate data of the two, and constructing a template data set based on the spatial position relationship; using a discriminant correlation filter network to track multiple anatomical markers in a conventional ultrasound image in real time to obtain dynamic coordinate information of each anatomical marker; based on the dynamic coordinate information, calculating the geometric features of the anatomical markers in the conventional ultrasound image; matching the geometric features to find a frame in the template data set that best matches the current frame in the conventional ultrasound image to obtain the best relative position coordinates of the liver tumor; performing coordinate affine transformation on the anatomical marker coordinates of the current frame and the matching frame, and fine-tuning the initial coordinate position of the liver tumor through a transformation matrix to obtain a real-time positioning result.

[0007] In another implementation of the present invention, the selecting and defining the anatomical markers around the liver tumor includes: selecting at least three non-collinear anatomical markers through DCFNet; the anatomical markers are obvious structures around the liver tumor.

[0008] In another implementation of the present invention, determining the spatial position relationship between the anatomical marker and the liver tumor based on the coordinate data of the anatomical marker and the liver tumor includes: calculating the geometric features of the anatomical marker based on the coordinate data of the anatomical marker and the liver tumor; constructing a feature vector based on the geometric features, and determining the spatial position relationship between the anatomical marker and the liver tumor based on the feature vector.

[0009] In another implementation of the present invention, the geometric features include the modulus of a vector, the distance from a point to a vector, the angle between vectors, and the area of ​​a polygon formed by the coordinate points.

[0010] In another implementation of the present invention, matching through the geometric features, finding a frame in the template data set that best matches the current frame in the conventional ultrasound image, and obtaining the optimal relative position coordinates of the liver tumor, includes: evaluating the similarity between the feature vector of the current frame in the conventional ultrasound image and the feature vector in the template data set through a composite similarity coefficient; taking the template frame with the highest composite similarity coefficient value in the template data set as a matching frame; and determining the optimal relative position coordinates of the liver tumor based on the matching frame.

[0011] In another implementation of the present invention, the composite similarity coefficient is expressed as:

[0012] csc=α·ρ(I,F)+(1-α)·J(I,F)

[0013] Wherein, ρ and J represent the correlation coefficient and Jaccard similarity coefficient respectively; α is a preset parameter, I is the feature vector of the current frame in the conventional ultrasound image, and F is the feature vector in the template data set.

[0014] The real-time liver tumor tracking method based on ultrasound imaging of the present invention utilizes multiple visible targets in conventional ultrasound to indirectly locate invisible tumors. The method is simple and direct, requiring only ultrasound equipment without additional operation and auxiliary equipment. The operator only needs to find the largest cross-section of the tumor based on the guidance information during conventional ultrasound scanning, and has low requirements for the use of equipment and operating experience, which has significant advantages in terms of puncture operation time and efficiency. The synchronization and high precision of DCFNet tracking of multiple targets, as well as the relative stability of anatomical markers and tumor position structures, can effectively resist displacement caused by breathing, and achieve stable and real-time tumor positioning and tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. By reading the detailed description of the following implementations, the advantages and benefits of the solutions become clear to those skilled in the art. The drawings are only used to illustrate the preferred implementations and are not to be considered as limiting the present invention.

[0016] In the attached picture:

[0017] Figure 1 The figure is a flow chart of a method for real-time tracking of liver tumors based on ultrasound imaging according to an embodiment of the present invention.

[0018] Figure 2 This is a flow chart of a method for real-time tracking of liver tumors based on ultrasound images, taking the extraction of three anatomical markers as an example, according to an embodiment of the present invention.

[0019] Figure 3 The figure is a schematic diagram of the clinical application process of a real-time liver tumor tracking method based on ultrasound imaging according to an embodiment of the present invention.

[0020] Figure 4 FIG. 1 is a schematic diagram of the processing process of conventional ultrasound and CEUS according to an embodiment of the present invention.

[0021] Figure 5 Schematic diagram of the DCFNet online tracking process according to an embodiment of the present invention.

[0022] Figure 6 A schematic diagram of equipment deployment according to an embodiment of the present invention.

[0023] Figure 7This is a schematic diagram of invisible liver tumor localization results of 3 clinical data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and in detail below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in the field based on the embodiments in the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0025] Figure 1 A flowchart of a method for real-time tracking of liver tumors based on ultrasound imaging provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, this embodiment mainly includes:

[0026] S101, using ultrasound contrast imaging to develop an invisible liver tumor, determine the initial coordinate position of the liver tumor, and select and define anatomical landmarks around the liver tumor.

[0027] Exemplarily, the initial coordinate position of a liver tumor that is not visible through ultrasound contrast imaging is determined. Specifically, the tumor is not visible under conventional ultrasound, and some invisible liver tumors are visualized in CEUS by using contrast agents such as SonoVue.

[0028] S102: Determine the spatial position relationship between the anatomical marker and the liver tumor according to the coordinate data thereof, and construct a template data set according to the spatial position relationship.

[0029] S103, using a discriminant correlation filter network (DCFNet) to track multiple anatomical landmarks in a conventional ultrasound image in real time to obtain dynamic coordinate information of each anatomical landmark.

[0030] Exemplarily, DCFNet is used to track multiple anatomical markers in conventional ultrasound images in real time to obtain the dynamic coordinate information of each marker. The operation is to select a predefined batch of markers and input them into DCFNet. Taking three anatomical markers as an example, the coordinate information of the three anatomical markers are (x1, y1)(x2, y2)(x3, y3), and DCFNet processes and outputs the corresponding coordinate information frame by frame (x1 ′ ,y1 ′ )(x2 ′ ,y2 ′ )(x3 ′ ,y3 ′ ), in which case the position and angle of the ultrasound scan are required to be set according to the scanning format used when collecting the template data set.

[0031] S104. Calculate geometric features of anatomical landmarks in the conventional ultrasound image based on the dynamic coordinate information.

[0032] S105 , matching is performed through the geometric features to find a frame in the template data set that best matches the current frame in the conventional ultrasound image, and obtain the best relative position coordinates of the liver tumor.

[0033] S106 , performing coordinate affine transformation on the coordinates of the anatomical landmarks of the current frame and the matching frame, and fine-tuning the initial coordinate position of the liver tumor through a transformation matrix to obtain a real-time positioning result.

[0034] For example, since the computational geometry features represent relative position relationships rather than absolute positions, position alignment is required. The affine transformation matrix is ​​calculated using the coordinates of the markers in the matching frame and the current frame, and the obtained transformation matrix is ​​used to adjust the tumor coordinates (x t ,y t ) to obtain accurate positioning results (x t ′ ,y t ′ ), the tumor radius is set according to the recorded value of the template dataset. The visual guidance information presented on the conventional ultrasound image ultimately includes the center of the landmark, the center of the invisible tumor, and the tumor area.

[0035] The real-time liver tumor tracking method based on ultrasound imaging of the present invention utilizes multiple visible targets in conventional ultrasound to indirectly locate invisible tumors. The method is simple and direct, requiring only ultrasound equipment without additional operation and auxiliary equipment. The operator only needs to find the largest cross-section of the tumor based on the guidance information during conventional ultrasound scanning, and has low requirements for the use of equipment and operating experience, which has significant advantages in terms of puncture operation time and efficiency. The synchronization and high precision of DCFNet tracking of multiple targets, as well as the relative stability of anatomical markers and tumor position structures, can effectively resist displacement caused by breathing, and achieve stable and real-time tumor positioning and tracking.

[0036] In another implementation of the present invention, the selecting and defining of anatomical markers around the liver tumor includes: selecting at least three non-collinear anatomical markers through DCFNet; the anatomical markers are obvious structures around the liver tumor.

[0037] Exemplarily, obvious structures around the liver tumor that can be observed in conventional ultrasound images, such as vascular structures or other anatomical markers, are selected to facilitate stable tracking, and at least three non-collinear markers are selected, otherwise, the scanning position or angle of the ultrasound probe is adjusted.

[0038] In another implementation of the present invention, determining the spatial position relationship between the anatomical marker and the liver tumor based on the coordinate data of the anatomical marker and the liver tumor includes: calculating the geometric features of the anatomical marker based on the coordinate data of the anatomical marker and the liver tumor; constructing a feature vector based on the geometric features, and determining the spatial position relationship between the anatomical marker and the liver tumor based on the feature vector.

[0039] Exemplarily, the coordinate data of the collected markers and the developed tumors are obtained through CEUS examination. CEUS imaging is used to determine the scanning position and area of ​​the ultrasound probe, and then the corresponding conventional ultrasound and CEUS videos are collected for coordinate extraction. In the conventional ultrasound image sequence, DCFNet is used to track multiple markers to obtain coordinate information. In the CEUS image sequence, the presence of the tumor is first manually identified, and then the DCFNet combined with manual annotation method is used to extract the coordinate and radius information of the tumor for the image with the tumor. Conventional ultrasound and CEUS images are scans of the same position section, so it is only necessary to unify the above-mentioned tumor coordinates into the conventional ultrasound image coordinate system through translation transformation.

[0040] The DCFNet in the present invention is an efficient target tracking algorithm that is independently built and trained, combining a deep learning network with a correlation filter. The network includes a feature extractor implemented with a fully convolutional network (FCN) and a tracker with a differentiable DCF layer. The FCN processes the input to generate a rich feature map, thereby capturing the spatial and contextual information that is critical to distinguishing the target from the surrounding environment; the tracker achieves adaptive tracking by learning spatial filters that can be optimized end-to-end, enabling the network to maintain high accuracy and adaptability in different scenes and changes in target appearance. The FCN in DCFNet is jointly trained using the natural image dataset ILSVRC2015 and the ultrasound image dataset CLUST. During online tracking, synchronous tracking of multiple targets can be achieved through batch input.

[0041] In another implementation of the present invention, the geometric features include the modulus of a vector, the distance from a point to a vector, the angle between vectors, and the area of ​​a polygon formed by the coordinate points.

[0042] For example, Figure 4 As shown, taking three anatomical landmarks as examples, the structural relationship between them is described by computational geometry features; to ensure the rotation and translation invariance of the features, the following geometric measurements are taken: the magnitude of the vector, the distance from point to vector, the angle between vectors, and the area of ​​the triangle formed by the three points.

[0043] Assuming the coordinates of the three markers are (x1, y1) (x2, y2) (x3, y3), the vector membrane L is calculated as follows:

[0044]

[0045] The distance VL from a point to a vector is calculated as follows:

[0046]

[0047] The angle A between the vectors is calculated as follows:

[0048]

[0049] The area S of a triangle is calculated as follows:

[0050]

[0051] Finally, the feature vector composed of the computational geometric features is:

[0052] F n =[L 12 ,L 13 ,L 23 ,VL 1-23 ,VL 2-13 ,VL 3-12 ,A1,A2,A3,S]

[0053] For each frame of the template dataset, the corresponding F is calculated n The extracted tumor information includes whether there is a tumor (positive = 1, negative = 0), coordinates (x t ,y t ), and radius r. At this point, the establishment of the template data set containing the feature vector F and tumor information is completed.

[0054] In another implementation of the present invention, matching through the geometric features, finding a frame in the template data set that best matches the current frame in the conventional ultrasound image, and obtaining the optimal relative position coordinates of the liver tumor, includes: evaluating the similarity between the feature vector of the current frame in the conventional ultrasound image and the feature vector in the template data set through a composite similarity coefficient; taking the template frame with the highest composite similarity coefficient value in the template data set as a matching frame; and determining the optimal relative position coordinates of the liver tumor based on the matching frame.

[0055] Exemplarily, the coordinates of the landmarks in the current frame are used to calculate the structural feature vector I containing the eigenvalues, wherein the feature vector I corresponds to the F corresponding to each frame of the template data set. n have the same number of eigenvalues ​​as those contained in .

[0056] In another implementation of the present invention, the composite similarity coefficient is expressed as:

[0057] csc=α·ρ(I,F)+(1-α)·J(I,F)

[0058] Wherein, ρ and J represent the correlation coefficient and Jaccard similarity coefficient respectively; α is a preset parameter, I is the feature vector of the current frame in the conventional ultrasound image, and F is the feature vector in the template data set.

[0059] In another implementation of the present invention, Figure 2 As shown, taking the selection of three anatomical landmarks as an example, a real-time tracking method for ultrasound-invisible liver tumors is introduced, involving preoperative and intraoperative modules.

[0060] (1) Preoperative module: used to establish a template data set. First, invisible tumors are visualized by CEUS. The position and angle of the ultrasound probe are determined based on the tumor location and the distribution of markers. The probe position is recorded, and conventional ultrasound images and CEUS are acquired simultaneously. The acquired images are then processed, including using DCFNet to continuously track three selected markers in conventional ultrasound for multiple frames, extracting the coordinate information of the markers, and using DCFNet combined with manual annotation to extract tumor information in CEUS, such as whether a tumor exists, coordinates (x t ,y t ), and tumor radius r; for each frame of the template image, the feature vector F is calculated using the coordinates of the three markers (x1, y1) (x2, y2) (x3, y3) n ; The tumor coordinates are unified into the coordinate system of conventional ultrasound through translation transformation; the feature vector and tumor information of each frame image constitute the template data set.

[0061] (2) Intraoperative module: used to locate and track invisible tumors in real time through conventional ultrasound images. First, scan according to the probe posture recorded before surgery to obtain conventional ultrasound images in real time; after observing the three predefined anatomical landmarks, manually mark the landmark positions and batch input them into the DCFNet network for synchronous tracking; the dynamic coordinates of the output landmarks are (x1 ′ ,y1 ′ )(x2 ′ ,y2 ′ )(x3 ′ ,y3 ′ ), use these coordinates to calculate the feature vector I of the current frame; perform a similarity evaluation between I and the feature vector set F of the template data set, and calculate the composite similarity coefficient csc. If the csc value of a template frame is the highest and greater than 0.998, then the template frame is the best reference frame, and its corresponding marker coordinates and tumor information will be output; if a tumor exists, the coordinates of the marker in the best template frame (x1, y1)(x2, y2)(x3, y3) and the coordinates in the current frame (x1 ′ ,y1′ )(x2 ′ ,y2 ′ )(x3 ′ ,y3 ′ ) to perform an affine transformation and apply the resulting transformation matrix to the tumor coordinates (x t ,y t ), and obtain the tumor localization result after position alignment (x t ′ ,y t ′ ); Marker Center (x1 ′ ,y1 ′ )(x2 ′ ,y2 ′ )(x3 ′ ,y3 ′ ), invisible tumor center (x t ′ ,y t ′ ) and the tumor area with radius r are marked on the ultrasound image for visualization purposes.

[0062] In another implementation of the present invention, Figure 3 As shown in the figure, the clinical application process of the real-time liver tumor tracking method based on ultrasound imaging is as follows:

[0063] (1) In the preoperative stage, the patient is examined using any ultrasound device, and 1.0-1.5 mL of SonoVue contrast agent is rapidly injected into the antecubital vein, followed by 5 mL of normal saline to visualize liver tumors that are not visible on CEUS.

[0064] (2) Find and determine the section where at least three significant anatomical landmarks and the tumor can be clearly observed, and record the probe position; use the workstation to synchronously acquire conventional ultrasound and CEUS at this position, and slowly translate the probe for scanning to ensure the data integrity in time and space.

[0065] (3) The data collected before surgery were processed on the workstation, and DCFNet was used to track the markers and output their coordinates. DCFNet was combined with manual annotation to extract tumor information.

[0066] (4) The coordinate information of all frames is input into the calculation program to generate a template data set. The storage format is not limited to Excel table.

[0067] (5) During the intraoperative stage, the workstation is connected to any ultrasound machine to read ultrasound images in real time and perform scans according to the recorded ultrasound probe position.

[0068] (6) When the predefined anatomical landmarks are observed, the ultrasound probe position is fixed and the algorithm is started at the workstation.

[0069] (7) After the algorithm is started, first manually select the marker on the currently frozen ultrasound image, and then start tracking after confirming that it is correct (DCFNet online tracking process is as follows Figure 5 shown).

[0070] (8) Slowly move the ultrasound probe to find the section containing the tumor according to the visual prompts of the predicted results, and fine-tune the probe according to the guidance information during the scan until the largest section of the tumor is located;

[0071] (9) Based on the results and visualization of the algorithm output, invisible tumors can be tracked stably and in real time on conventional ultrasound, and subsequent surgical operations such as puncture can then be performed.

[0072] Refer to the clinical use process and Figure 6 The equipment required for this method includes an ultrasound machine, a workstation, and an external display screen. The ultrasound machine is responsible for real-time image acquisition as the acquisition end; the workstation is connected to the ultrasound machine through a graphics acquisition card to read the ultrasound image and run the algorithm; the output results are visualized on the ultrasound image and displayed on the external display screen for the doctor's reference.

[0073] As shown in Table 1 and Figure 7 As shown in the table, the data volume of the template dataset, the tumor localization results using the template dataset as intraoperative data for self-verification, and the tumor localization results using an independent test dataset are listed. Accuracy and precision are used to quantify the tumor recognition effect, and the positioning error is the Euclidean distance between the coordinates output by the algorithm and the true label. The results show that this method can better judge whether a tumor exists based on the location information of the marker, proving that the marker tracking and feature matching effects are good; the average value of the positioning error is within 2mm, which meets clinical requirements.

[0074] Table 1. Quantitative results of 3 clinical review data

[0075]

[0076] This method simplifies the clinical puncture process of liver tumors. In the preoperative stage, a template set is established by collecting conventional ultrasound and CEUS data, and the probe position is marked on the patient's body surface. In the intraoperative stage, only the markers in the initial ultrasound image need to be manually defined. As long as the markers remain visible in the conventional ultrasound scan image, the markers and target tumors can be automatically and continuously tracked throughout the process, avoiding the need to rely on CEUS for tumor localization multiple times during surgery.

[0077] In another aspect of the present invention, an electronic device includes: a processor, a memory, a communication bus, and a communication interface.

[0078] in:

[0079] The processor, memory and communication interface communicate with each other through a communication bus.

[0080] Communication interface, used to communicate with other electronic devices or servers.

[0081] The processor is used to execute the program, and specifically can execute the steps of any one of the methods for real-time tracking of liver tumors based on ultrasound images in the above embodiments.

[0082] Specifically, the program may include program codes including computer operation instructions.

[0083] The processor may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the smart device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0084] The memory is used to store programs. The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0085] The program can be specifically used to enable the processor to execute to implement the steps of any one of the methods for real-time tracking of liver tumors based on ultrasound images described in the embodiments. The specific implementation of each step in the program can refer to the corresponding descriptions in the steps and units executed by any one of the methods for real-time tracking of liver tumors based on ultrasound images in the above steps, which will not be repeated here. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described devices and modules can refer to the corresponding process description in the aforementioned method embodiments.

[0086] The above method according to the embodiment of the present invention can be implemented in a server equipped with a central processing unit CPU and an image processing unit COU.

[0087] Thus far, specific embodiments of the present invention have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results.

[0088] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, back, etc.) are only used to explain the relative position relationship between the components in a certain specific order (as shown in the accompanying drawings). If the specific order changes, the directional indication will also change accordingly.

[0089] In the description of the present invention, the terms "first" and "second" are only used to facilitate the description of different components or names, and cannot be understood as indicating or implying a sequential relationship, relative importance, or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0090] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0091] It should be noted that although the specific embodiments of the present invention are described in detail in conjunction with the accompanying drawings, it should not be understood as limiting the scope of protection of the present invention. Within the scope described in the claims, various modifications and variations that can be made by those skilled in the art without creative work still belong to the scope of protection of the present invention.

[0092] The examples of the embodiments of the present invention are intended to concisely illustrate the technical features of the embodiments of the present invention so that those skilled in the art can intuitively understand the technical features of the embodiments of the present invention, and are not intended to be improper limitations of the embodiments of the present invention.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time liver tumor tracking method based on ultrasound imaging, characterized in that: include: Determine the initial coordinate position of the liver tumor by developing an invisible liver tumor through ultrasound contrast imaging, and select and define anatomical landmarks around the liver tumor; Determining a spatial position relationship between the anatomical marker and the liver tumor according to the coordinate data thereof, and constructing a template data set according to the spatial position relationship; The discriminant correlation filter network is used to track multiple anatomical landmarks in conventional ultrasound images in real time to obtain the dynamic coordinate information of each anatomical landmark; Calculating geometric features of anatomical landmarks in the conventional ultrasound image based on the dynamic coordinate information; Matching the geometric features is performed to find a frame in the template data set that best matches the current frame in the conventional ultrasound image, and obtain the best relative position coordinates of the liver tumor; The coordinates of the anatomical landmarks of the current frame and the matching frame are subjected to coordinate affine transformation, and the initial coordinate position of the liver tumor is fine-tuned through the transformation matrix to obtain a real-time positioning result.

2. The method according to claim 1, characterized in that The selecting and defining anatomical markers around the liver tumor comprises: At least three non-collinear anatomical landmarks were selected through DCFNet; The anatomical landmarks are distinct structures surrounding the liver tumor.

3. The method according to claim 1, characterized in that Determining the spatial positional relationship between the anatomical marker and the liver tumor according to the coordinate data thereof includes: Calculating geometric features of the anatomical marker based on the coordinate data of the anatomical marker and the liver tumor; A feature vector is constructed according to the geometric feature, and a spatial position relationship between the anatomical marker and the liver tumor is determined according to the feature vector.

4. The method according to claim 3, characterized in that The geometric features include the magnitude of a vector, the distance from a point to a vector, the angle between vectors, and the area of ​​a polygon formed by the coordinate points.

5. The method according to claim 3, characterized in that: The matching by the geometric features, finding a frame in the template data set that best matches the current frame in the conventional ultrasound image, and obtaining the best relative position coordinates of the liver tumor, includes: Performing similarity evaluation on the feature vector of the current frame in the conventional ultrasound image and the feature vector in the template data set by using a composite similarity coefficient; The template frame with the highest composite similarity coefficient value in the template data set is taken as the matching frame; The optimal relative position coordinates of the liver tumor are determined based on the matching frames.

6. The method according to claim 5, characterized in that The composite similarity coefficient is expressed as: csc=α·ρ(I,F)+(1-α)·J(I,F) Wherein, ρ and J represent the correlation coefficient and Jaccard similarity coefficient respectively; α is a preset parameter, I is the feature vector of the current frame in the conventional ultrasound image, and F is the feature vector in the template data set.

Images