Dynamic image guided intracranial hematoma puncture positioning method and system
By performing feature component analysis and clustering on consecutive frames of intraoperative ultrasound images, the problem of blurred edema boundaries was solved, enabling precise hematoma localization and improving the accuracy and safety of puncture.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- THE FIRST AFFILIATED HOSPITAL OF GUIZHOU UNIV OF TRADITIONAL CHINESE MEDICINE
- Filing Date
- 2025-11-29
- Publication Date
- 2026-07-24
AI Technical Summary
In existing technologies, the blurring of hematoma boundaries caused by edema makes it difficult for traditional image segmentation methods to accurately capture the blurred boundaries, thus affecting the accuracy of hematoma tissue localization analysis.
By acquiring continuous frames of ultrasound images within the intraoperative bone window cavity, target regions are screened, feature components are divided and clustered based on the distribution characteristics of pixels, and hematoma regions are screened by combining morphological distribution direction and feature parameter indicators. The segmentation accuracy is improved by clustering and localization adjustment.
It enables precise delineation of the boundaries between hematoma and edema, improves the accuracy of puncture localization, and reduces damage to surrounding brain tissue.
Smart Images
Figure CN121400940B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of region segmentation technology, specifically to a method and system for puncturing and locating intracranial hematoma under dynamic image guidance. Background Technology
[0002] Intracerebral hematoma is a common critical condition in neurosurgery, with a high incidence and rapid progression, resulting in high rates of disability and mortality. Clinically, minimally invasive therapies combined with medical methods are often used to treat intracerebral hematomas. By removing the hematoma as early as possible, the impact on surrounding brain tissue and functional areas can be reduced, thereby halting the further development of neurological deficits.
[0003] The key to minimally invasive surgery for removing intracranial hematomas is accurate puncture of the hematoma to avoid damage to brain tissue. Although minimally invasive puncture for removing hematomas in hypertensive intracerebral hemorrhage is less traumatic and shorter in operation time, the puncture must be very accurate during the procedure. Otherwise, the minimally invasive puncture will not achieve the purpose of removing the hematoma and effectively decompressing the brain. Therefore, ultrasound technology is usually used to dynamically observe the hematoma removal process during the operation and to improve accuracy by dynamically locating the hematoma area.
[0004] However, during the procedure, when the hematoma is aspirated under ultrasound monitoring until its severity decreases, the hematoma cavity may collapse. Due to the space-occupying effect of the hematoma, it can cause edema in the surrounding brain tissue, which is closely connected to the hematoma cavity and appears as a strong echo during ultrasound. The edema creates a gradual transition area between the hematoma and normal tissue, resulting in blurred boundaries of the hematoma tissue. Traditional image segmentation and enhancement methods are unable to accurately capture the blurred boundary details, which can easily lead to missegmentation or omission of regions and affect the accuracy of the localization analysis of the hematoma tissue. Summary of the Invention
[0005] To address the technical problem in existing technologies where edema blurs the boundaries of hematoma tissue, easily leading to misclassification or omission of areas and affecting the accuracy of hematoma localization analysis, the present invention aims to provide a dynamic image-guided intracranial hematoma puncture localization method and system. The specific technical solution adopted is as follows: This invention provides a dynamic image-guided method for intracranial hematoma puncture and localization, the method comprising: Acquire continuous frames of ultrasound images within the intraoperative bone window cavity, and filter out target areas with hematoma based on the complexity of the distribution within each frame of ultrasound image. In each ultrasound image, the morphological distribution direction is determined based on the distribution of pixels in the target region; feature components are divided and their distribution directions are determined by the similarity of pixel characteristics in the target region; the distribution trend of each feature component is obtained by the distribution of each feature component in the target region and the deviation between the component distribution direction and the morphological distribution direction; and the feature parameter index of each feature component is obtained based on the distribution trend of each feature component in consecutive frames in time, the approximation of its area distribution, and the pixel value. Clustering is performed based on the feature parameters between feature components to select the current feature region; localization is then performed based on the location and area changes of the current feature region.
[0006] Furthermore, the method for obtaining the target region includes: Each frame of ultrasound image is divided into candidate regions through region growing; for any candidate region, the average pixel value of all pixels in the candidate region is taken as the pixel mean of the candidate region. After calculating the difference between the pixel value and the mean pixel value of each pixel in the candidate region, the average value of all differences is calculated to obtain the feature complexity of the candidate region; the selection factor of the candidate region is obtained by combining the mean pixel value and the feature complexity of the candidate region. The candidate region corresponding to the maximum screening factor in each frame of ultrasound image is taken as the target region of each frame of ultrasound image.
[0007] Furthermore, the method for obtaining the morphological distribution direction includes: A three-dimensional spatial coordinate system is constructed with the lower left corner of each ultrasound image as the origin, the two side lengths of the ultrasound image as the horizontal and vertical axes, and the pixel value of the pixels in the ultrasound image as the vertical axis. The pixels of the target area in each ultrasound image are mapped to obtain the target three-dimensional space. Principal component analysis is performed on points in the target three-dimensional space to obtain the principal component directions; the projection of the principal component directions onto the two-dimensional planes of the horizontal and vertical axes is used as the morphological distribution direction of the target region.
[0008] Furthermore, the step of dividing feature components and determining the component distribution direction based on the similarity of pixel characteristics in the target region includes: In any target region, the feature components of the target region are obtained. The pixel values of the pixels in each feature component are the same and the pixel positions are connected. Obtain the centroid position of each feature component; in each feature component, take the direction from each pixel position to the centroid position as the pointing direction of each pixel; calculate the sum of the pointing directions of all pixels in each feature component as the component distribution direction of each feature component.
[0009] Furthermore, the method for obtaining the distribution trend performance includes: For any target region, calculate the similarity between the morphological distribution direction of the target region and the component distribution direction of each feature component to obtain the trend consistency of each feature component. The proportion of each feature component in the number of pixels in the target region is used as the distribution weight of each feature component; the product of the distribution weight of each feature component and the trend consistency is used as the distribution trend performance of each feature component.
[0010] Furthermore, the method for obtaining the feature parameter index includes: For any feature component, between any two adjacent ultrasound images, the ratio of the area of the feature component in the later ultrasound image to the area in the earlier ultrasound image is used as the contraction intensity index of the feature component between the two ultrasound images. The difference between the distribution trend of the feature component in the subsequent ultrasound image and the distribution trend in the previous ultrasound image is used as the directional stability index of the feature component between the two ultrasound images. By combining the contraction intensity index and directional stability index of this feature component between the two ultrasound images, the development consistency index of this feature component between the two ultrasound images is obtained. By negatively correlating the sum of the developmental consistency indices of this feature component between all adjacent ultrasound images, a continuous feature performance index of this feature component is obtained. By combining the pixel values in this feature component with the continuous feature performance index, the feature parameter index of this feature component is obtained.
[0011] Furthermore, the method for obtaining the feature region includes: The similarity of the feature parameter indices between feature components is used as a weight to weight the positional distance between feature components, thus obtaining the feature similarity between feature components. In the current target region, using feature similarity as the clustering metric, all feature components are clustered to obtain two clusters; Clusters with higher mean values of feature parameters for all feature components in a cluster are designated as target clusters; regions composed of all feature components in the target cluster are designated as feature regions.
[0012] Furthermore, the localization based on the current feature region location and area changes includes: Obtain the location and area of the hematoma region in the preoperative ultrasound image; calculate the area difference between the preoperative area and the characteristic region as the area change rate. The transmission, display, and positioning adjustments are performed by registering the current feature region with its preoperative location and adjusting the area change.
[0013] Furthermore, the method for obtaining the candidate region includes: Based on the preset growth point locations in the ultrasound image, the region is grown according to the growth criteria to obtain the growth region; after the growth is completed, the growth regions with pixel values higher than the preset deviation threshold of the adjacent growth regions are screened out, and the remaining growth regions are selected as candidate regions. The growth criterion is: growth can be performed when the difference in pixel value between the pixel to be grown and the growth point is less than a preset growth threshold; the pixel to be grown is a pixel within a preset neighborhood range of the growth point.
[0014] The present invention also provides a dynamic image-guided intracranial hematoma puncture and localization system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.
[0015] The present invention has the following beneficial effects: This invention extracts target regions containing hematoma from each frame of ultrasound images through preliminary region analysis, further analyzes the blurring effect of edema, and performs more refined region segmentation. Then, considering the centripetal contraction of hematoma during the absorption phase, the target region is further component-wise decomposed and analyzed. Based on the distribution trend of the characteristic components and the overall regional directional trend, the consistency between each characteristic component and the overall development process is preliminarily analyzed, reflecting the degree of manifestation of the hematoma region. Furthermore, through real-time changes in consecutive frames over time, and by observing the inter-frame distribution trend and area approximation, the dynamic change trend of hematoma and edema interference during the hematoma thinning process is captured. By observing the uniformity of fluctuations over time and the consistency between the contraction trend and the characteristics of the hematoma tissue, characteristic parameter indicators are obtained, reflecting the likelihood that a component belongs to the hematoma tissue. Finally, characteristic regions are segmented through clustering, and the changes in location and area reflect the changes in positioning, assisting in the adjustment of puncture positioning. This invention preliminarily distinguishes the interference of edema tissue by performing more refined component analysis and overall distribution consistency in the region, and accurately delineates the boundaries of hematoma and edema based on the degree of change of each component in continuous ultrasound image frames, effectively improving the segmentation accuracy and making subsequent puncture positioning more accurate. Attached Figure Description
[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 A flowchart illustrating a dynamic image-guided intracranial hematoma puncture and localization method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of centripetal contraction provided in one embodiment of the present invention. Detailed Implementation
[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a dynamic image-guided intracranial hematoma puncture and localization method and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the dynamic image-guided intracranial hematoma puncture and localization method and system provided by the present invention.
[0021] Please see Figure 1 The diagram illustrates a flowchart of a dynamic image-guided intracranial hematoma puncture and localization method according to an embodiment of the present invention, which includes the following steps: S1: Acquire continuous frames of ultrasound images within the intraoperative bone window cavity, and filter out target areas with hematoma based on the complexity of the distribution within each frame of ultrasound images.
[0022] Preoperatively, brain CT images of the patient are taken, and 3D reconstruction is performed using 3D-Slicer software to initially locate the cerebral hematoma, plan the puncture path, and determine the endoscopic puncture point. During the procedure, a portable ultrasound scanner is used to scan the intracranial hematoma through the bone window cavity, obtaining real-time ultrasound images. The puncture is guided by ultrasound, and changes in the hematoma and surrounding brain tissue are monitored in real time. The monitoring data is transmitted to an image processing platform, and the puncture point is repositioned based on changes in the hematoma. In this embodiment, the acquisition frequency can be set to 15 frames per second; however, the operator can adjust this according to the specific implementation scenario, and it is not specified here.
[0023] During the procedure, as the hematoma is aspirated under ultrasound monitoring until its severity decreases, the hematoma cavity may collapse. The space-occupying effect of the hematoma causes edema in the surrounding brain tissue, which is closely connected to the hematoma cavity and also appears as strong echoes on ultrasound. This results in a mixture of echoes caused by edema and hematoma, leading to blurred boundaries and difficulty in accurately segmenting the hematoma tissue during intraoperative hematoma localization using ultrasound images. However, during continuous scanning, the gradient of the hematoma margin gradually decreases and shrinks centripetally during the absorption phase. Simultaneously, in continuous ultrasound images, the hematoma tissue dynamically changes according to the aspiration process, while the changes in edema interference are relatively static. By analyzing the changes in hematoma morphology and monitoring its evolution in real time, the strong echo interference of static edema on the edge segmentation can be suppressed.
[0024] Ultrasound images inherently contain significant noise, which can easily affect edge detection, interfering with the identification of hematoma edges and leading to inaccurate localization. Because hematoma tissue contains a large amount of blood, it has a high density and appears as a hyperechoic mass in ultrasound images, significantly different from normal brain tissue. Therefore, based on the strong echogenicity of hematoma tissue, the target region in the ultrasound image is initially extracted to facilitate more refined calculations in subsequent steps.
[0025] Preferably, in this embodiment of the invention, the method for obtaining the target region includes: First, each frame of ultrasound image is divided into candidate regions using region growing. Because different tissue structures and lesions within a patient's body have different compositions and shapes, their echo characteristics also differ. In ultrasound images, this manifests as pixel value differences between regions corresponding to different tissue structures and lesions.
[0026] Therefore, in this embodiment of the invention, based on the preset growth point positions in the ultrasound image, region growth is performed according to growth criteria to obtain a growth region. Growth points are uniformly set at the center of every 20×20 region in the image. The growth criterion is: when the pixel value difference between the pixel to be grown and the growth point is less than a preset growth threshold, growth can be performed. The pixel to be grown is a pixel within a preset neighborhood range of the growth point. The preset growth threshold is set to 15, and the preset neighborhood range is an eight-neighbor range. The specific numerical settings of the growth criteria can be adjusted by the implementer and are not limited here.
[0027] Growth stops when all pixels belong to a certain growth region. After growth is complete, growth regions with pixel values higher than the preset deviation threshold of adjacent growth regions are screened out, and the remaining growth regions are selected as candidate regions. Since the skeletal part also exhibits extremely high echo reflection during scanning, this part has clear boundaries with other regions and uniform pixels, so it can be directly screened to reduce the complexity of subsequent analysis. In this embodiment, the preset deviation threshold can be set to 100, which can be adjusted by the implementer, but there are no restrictions here.
[0028] Different regions may correspond to different tissue structures or lesion components. Intracranial hematomas are complex in composition and may contain different components such as fresh blood, blood clots, and serum depending on the time of hematoma formation. These different components have different effects on the reflection of ultrasound signals, resulting in uneven echoes inside the hematoma. Therefore, by analyzing the complexity of grayscale signals in different hyperechoic regions, regions containing hematomas can be screened out.
[0029] For any candidate region, the average pixel value of all pixels in the candidate region is taken as the pixel mean of the candidate region, reflecting the echo situation. After calculating the difference between the pixel value of each pixel in the candidate region and the pixel mean, the average of all differences is calculated to obtain the feature complexity of the candidate region, reflecting the degree of pixel offset.
[0030] Finally, the selection factor of the candidate region is obtained by combining the pixel mean and feature complexity of the candidate region. In this embodiment of the invention, the product of the pixel mean and feature complexity of the candidate region is used as the selection factor of the candidate region. When the pixel mean distribution and feature complexity of the region are higher, it indicates that the high echo features in the region are more significant and the composition is more complex, and the possibility of it belonging to hematoma tissue is higher.
[0031] Therefore, the candidate region corresponding to the maximum screening factor in each frame of ultrasound image is taken as the target region of each frame of ultrasound image for subsequent analysis.
[0032] S2: In each ultrasound image, the morphological distribution direction is determined based on the distribution of pixels in the target region; feature components are divided and their distribution directions are determined by the similarity of pixel characteristics in the target region; the distribution trend of each feature component is obtained by the distribution of each feature component in the target region and the deviation between the component distribution direction and the morphological distribution direction; the feature parameter index of each feature component is obtained based on the distribution trend and area distribution approximation of each feature component in consecutive frames in time sequence, as well as the pixel value.
[0033] Because the hematoma compresses the surrounding brain tissue, causing displacement or deformation, it is accompanied by secondary edema. As the hematoma cavity is continuously aspirated, the hematoma volume decreases and the blood accumulated in the cavity is gradually drained. This results in the echo intensity of the target area gradually weakening in continuous ultrasound images. The area that was originally compressed by the hematoma may have some degree of release. As the edema interference range expands, the boundary between the hematoma and edema merges and becomes difficult to distinguish in the ultrasound image.
[0034] In ultrasound images, the density structure of hematoma tissue changes, while edema interference appears relatively stable. During the absorption phase, the edge gradient of the hematoma gradually decreases and centripetal contraction occurs, while blurred edge edema does not exhibit the aforementioned characteristics. By analyzing the consistency of the distribution of blurred edge areas with the overall target area in ultrasound images, the interference of edema tissue can be preliminarily distinguished.
[0035] Therefore, the overall morphological distribution direction is first determined based on the structural distribution of pixels in the target area. In this embodiment of the invention, a three-dimensional spatial coordinate system is constructed with the lower left corner of each frame of ultrasound image as the origin, the two side lengths of the ultrasound image as the horizontal and vertical axes, and the pixel value of the pixels in the ultrasound image as the vertical axis. The pixels of the target area in each frame of ultrasound image are mapped to obtain the target three-dimensional space. Combining the comprehensive distribution state of position and pixel value, principal component analysis is performed on the points in the target three-dimensional space to obtain the principal component direction, which reflects the structural distribution of the overall area. The projection of the principal component direction onto the two-dimensional plane of the horizontal and vertical axes is taken as the morphological distribution direction of the target area. It should be noted that the construction of the coordinate space and the principal component analysis are formulas and techniques well known to those skilled in the art, and will not be elaborated here.
[0036] In a target region, there are similar component regions. By further refining the subdivision, feature components representing different characteristics are obtained, and the morphological distribution of the feature components is analyzed. In this embodiment of the invention, in any target region, the feature components of that target region are obtained. In each feature component, the pixel values of the pixels are the same and the pixel positions are connected. That is, all pixels that are adjacent in position and have the same pixel value are recorded as one feature component.
[0037] Obtain the centroid position of each feature component. In each feature component, take the direction from each pixel position to the centroid position as the pointing direction of each pixel. Calculate the sum of the pointing directions of all pixels in each feature component as the component distribution direction of each feature component. By comprehensively analyzing the regional directional distribution degree of the feature components, the overall distribution trend of the feature components can be represented.
[0038] Since edema interference often exists at the boundary of hematoma tissue, away from the center of hematoma, and the growth state of edema tissue differs from that of hematoma structure, the pixel distribution in the edema interference area deviates significantly from the regional morphological parameters of the target area. Therefore, the deviation between the pixels in the target area and the regional morphological parameters is calculated. When there is a large difference between the pixels and the regional morphological parameters and the deviation constitutes a certain scale at the edge of the area, the pixels are more likely to be edema interference.
[0039] Therefore, by combining the deviation analysis to determine the distribution trend performance of each feature component, in this embodiment of the invention, the method for obtaining the distribution trend performance includes: For any target region, the similarity between the morphological distribution direction of the target region and the component distribution direction of each feature component is calculated to obtain the trend consistency of each feature component. In this embodiment of the invention, sampling cosine similarity can be selected to calculate the consistency between the two directions. The larger the cosine similarity, the higher the trend consistency. In other embodiments of the invention, inverse cosine or Manhattan distance methods can also be selected, and no limitation is made here.
[0040] Furthermore, the proportion of each feature component in the target region is used as the distribution weight of each feature component. The larger the scale of the feature component in the region, the better the distribution analysis reflects the consistency trend. Then, the product of the distribution weight of each feature component and the trend consistency is used as the distribution trend performance of each feature component. The greater the distribution trend performance, the more significant the consistency between the component and the whole.
[0041] During continuous aspiration of hematoma tissue, the hematoma cavity gradually collapses, while the edematous tissue shows no significant structural changes. As the aspiration process continues, the hematoma cavity exhibits centripetal contraction. Please refer to [link to relevant documentation]. Figure 2 The diagram illustrates a centripetal contraction according to an embodiment of the present invention, representing the trend of contraction. Since edema tissue does not exhibit the aforementioned centripetal change, the consistency of the development process of each component can be determined by analyzing the trend of change in continuous frames, further enabling precise differentiation between hematoma tissue and edema interference.
[0042] Preferably, in this embodiment of the invention, the method for obtaining the feature parameter index includes: For any given feature component, the ratio of the area of the feature component in the later ultrasound image to the area in the earlier ultrasound image between any two adjacent frames is used as the contraction intensity index of the feature component between the two ultrasound images. First, the degree of contraction consistency of the component is reflected by the area change between consecutive frames. The smaller the contraction intensity index, the more significant the hematoma tissue characteristics are reflected.
[0043] Furthermore, the difference between the distribution trend of this feature component in the subsequent ultrasound image and the distribution trend in the previous ultrasound image is used as an indicator of the directional stability of this feature component between the two ultrasound images. From the difference in the distribution direction features of consecutive frames, the smaller the difference, the higher the consistency of the directional trend in the process, indicating that the hematoma tissue features are more significant.
[0044] Furthermore, by combining the contraction intensity index and directional stability index of the feature component between the two ultrasound images, a development consistency index of the feature component between the two ultrasound images is obtained. In this embodiment of the invention, the product of the contraction intensity index and the directional stability index is used as the development consistency index, which integrates the area change and directional trend of the contraction process. The smaller the index, the more significant the hematoma tissue characteristics.
[0045] Therefore, a negative correlation mapping is performed on the sum of the developmental consistency indices of this feature component across all adjacent ultrasound images to obtain a continuous feature performance index for this feature component. This index, considering all inter-frame conditions, reflects the overall performance across consecutive frames. A larger continuous feature performance index indicates more uniform temporal fluctuations, a contraction trend similar to hematoma tissue, and a more significant reflection of hematoma tissue characteristics. It should be noted that negative correlation mapping is a technique well-known to those skilled in the art and can be performed using inverse proportional or negative exponential forms; no restrictions or further details are provided here.
[0046] Considering the high echo characteristics, the feature parameter index of the feature component is obtained by combining the pixel value and the continuous feature performance index in the feature component. In this embodiment of the invention, the product of the pixel value and the continuous feature performance index in the feature component is used as the feature parameter index of the feature component. The higher the pixel value and the larger the continuous feature performance index, the higher the probability that the component corresponds to hematoma tissue.
[0047] S3: Cluster the feature components based on their feature parameters and filter out the current feature regions; locate the current feature regions based on their location and area changes.
[0048] Hematoma tissue and edema are distinguished based on the characteristic parameter indices of characteristic components. In this embodiment of the invention, the similarity of the characteristic parameter indices between characteristic components is used as a weight to weight the positional distance between characteristic components to obtain the characteristic similarity between characteristic components. In one embodiment of the invention, the difference in the characteristic parameter indices between two characteristic components is used as a weight, and the product of the positional distance between two characteristic components and the weight is used as the characteristic similarity. The smaller the characteristic similarity, the more similar the characteristics of the characteristic components are, the closer their positions are, and the more likely they are to belong to the same class.
[0049] Therefore, in the current target region, feature similarity is used as the clustering metric to cluster all feature components, resulting in two clusters. This divides the target region into hematoma tissue and edema, thus eliminating the interference of edema on region localization. It should be noted that the clustering method used is a technique well-known to those skilled in the art, such as K-means clustering with a cluster size of 2. Further details and limitations will not be elaborated upon here.
[0050] Finally, the cluster with the highest mean value of all feature parameters of all feature components in the cluster is selected as the target cluster, which represents the hematoma portion to be located. The region formed by all feature components in the target cluster is then selected as the feature region, and the region corresponding to the other cluster is the edema interference region, thus achieving region differentiation.
[0051] Based on the currently segmented feature regions, localization adjustments can be made according to changes in location and area. Since preoperative CT images can clearly distinguish between hematoma and edema, preoperative CT imaging data can be fused with intraoperative real-time ultrasound images. Elastic registration can be used to correct brain tissue displacement and improve the accuracy of hematoma boundary segmentation and localization.
[0052] In this embodiment of the invention, the location and area of the hematoma region in the preoperative ultrasound image are acquired. The area difference between the preoperative area and the feature region is calculated as the area change rate, reflecting the magnitude of change of the hematoma tissue in the intraoperative image. The transmission, display, and positioning adjustment are performed based on the registration of the current feature region with the preoperative location and the area change rate. In one embodiment of the invention, a correlation coefficient method based on feature points can be used to calculate the relative positions in two images, perform elastic registration, and align the image space. The registered image can display the positional contraction changes of the hematoma region before and during the operation, and combined with the area change rate, provides a more accurate real-time data basis for positioning adjustment.
[0053] In summary, this invention extracts the target region containing hematoma from each frame of ultrasound image through preliminary region analysis, further analyzes the blurring effect of edema, and performs more refined region segmentation. Then, considering the centripetal contraction of hematoma during the absorption phase, the target region is further component-wise decomposed and analyzed. Based on the distribution trend of the characteristic components and the overall regional directional trend, the consistency between each characteristic component and the overall development process is preliminarily analyzed, reflecting the degree of manifestation of the hematoma region. Furthermore, through real-time changes in consecutive frames over time, and by observing the inter-frame distribution trend and area approximation, the dynamic change trend of hematoma and edema interference during the hematoma thinning process is captured. By observing the uniformity of temporal fluctuations and the consistency between the contraction trend and the characteristics of the hematoma tissue, characteristic parameter indicators are obtained, reflecting the likelihood that a component belongs to the hematoma tissue. Finally, characteristic regions are segmented through clustering, and the changes in location and area reflect the changes in positioning, assisting in the adjustment of puncture positioning. This invention preliminarily distinguishes the interference of edema tissue by performing more refined component analysis and overall distribution consistency in the region, and accurately delineates the boundaries of hematoma and edema based on the degree of change of each component in continuous ultrasound image frames, effectively improving the segmentation accuracy and making subsequent puncture positioning more accurate.
[0054] The present invention also provides a dynamic image-guided intracranial hematoma puncture and localization system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.
[0055] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0056] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
Claims
1. A method for puncture and localization of intracranial hematoma guided by dynamic image, characterized in that, The method includes: Acquire continuous frames of ultrasound images within the intraoperative bone window cavity, and filter out target areas with hematoma based on the complexity of the distribution within each frame of ultrasound image. In each ultrasound image, the morphological distribution direction is determined based on the distribution of pixels in the target region; feature components are divided and their distribution directions are determined by the similarity of pixel characteristics in the target region; the distribution trend of each feature component is obtained by the distribution of each feature component in the target region and the deviation between the component distribution direction and the morphological distribution direction; and the feature parameter index of each feature component is obtained based on the distribution trend of each feature component in consecutive frames in time, the approximation of its area distribution, and the pixel value. Clustering is performed based on the feature parameters between feature components to select the current feature region; localization is then performed based on the location and area changes of the current feature region. The method for obtaining the morphological distribution direction includes: A three-dimensional spatial coordinate system is constructed with the lower left corner of each ultrasound image frame as the origin. The two side lengths of the ultrasound image are used as the horizontal and vertical axes, and the pixel value of the pixels in the ultrasound image is used as the vertical axis. The pixels of the target region in each ultrasound image frame are mapped to obtain the target three-dimensional space. Principal component analysis is performed on the points in the target three-dimensional space to obtain the principal component directions. The projection of the principal component directions onto the two-dimensional plane of the horizontal and vertical axes is used as the morphological distribution direction of the target region. The step of dividing feature components and determining the component distribution direction based on the similarity of pixel characteristics in the target region includes: In any target region, obtain the feature components of the target region. The pixel values of the pixels in each feature component are the same and the pixel positions are connected. Obtain the centroid position of each feature component. In each feature component, take the direction from each pixel position to the centroid position as the pointing direction of each pixel. Calculate the sum of the pointing directions of all pixels in each feature component as the component distribution direction of each feature component. The methods for obtaining the distribution trend performance include: For any target region, calculate the similarity between the morphological distribution direction of the target region and the component distribution direction of each feature component to obtain the trend consistency of each feature component; take the proportion of the number of pixels of each feature component in the target region as the distribution weight of each feature component; take the product of the distribution weight of each feature component and the trend consistency as the distribution trend performance of each feature component. The method for obtaining the feature parameter index includes: For any feature component, between any two adjacent ultrasound images, the ratio of the area of the feature component in the later ultrasound image to the area in the earlier ultrasound image is used as the contraction intensity index of the feature component between the two ultrasound images. The difference between the distribution trend of the feature component in the subsequent ultrasound image and the distribution trend in the previous ultrasound image is used as the directional stability index of the feature component between the two ultrasound images; the product of the contraction intensity index and the directional stability index of the feature component between the two ultrasound images is used as the development consistency index of the feature component between the two ultrasound images. By negatively correlating the sum of the developmental consistency indices of this feature component between all adjacent ultrasound images, a continuous feature performance index of this feature component is obtained. The product of the pixel value in the feature component and the continuous feature performance index is used as the feature parameter index of the feature component.
2. The method for dynamic image-guided intracranial hematoma puncture and localization according to claim 1, characterized in that, The method for obtaining the target region includes: Each frame of ultrasound image is divided into candidate regions through region growing; for any candidate region, the average pixel value of all pixels in the candidate region is taken as the pixel mean of the candidate region. After calculating the difference between the pixel value and the pixel mean of each pixel in the candidate region, the average of all differences is calculated to obtain the feature complexity of the candidate region; the product of the pixel mean and feature complexity of the candidate region is used as the selection factor for the candidate region. The candidate region corresponding to the maximum screening factor in each frame of ultrasound image is taken as the target region of each frame of ultrasound image.
3. The method for dynamic image-guided intracranial hematoma puncture and localization according to claim 1, characterized in that, The method for obtaining the feature region includes: The similarity of the feature parameter indices between feature components is used as a weight to weight the positional distance between feature components, thus obtaining the feature similarity between feature components. In the current target region, using feature similarity as the clustering metric, all feature components are clustered to obtain two clusters; Clusters with higher mean values of feature parameters for all feature components in a cluster are designated as target clusters; regions composed of all feature components in the target cluster are designated as feature regions.
4. The method for dynamic image-guided intracranial hematoma puncture and localization according to claim 1, characterized in that, The localization based on the current feature region location and area changes includes: Obtain the location and area of the hematoma region in the preoperative ultrasound image; calculate the area difference between the preoperative area and the characteristic region as the area change rate. The transmission, display, and positioning adjustments are performed by registering the current feature region with its preoperative location and adjusting the area change.
5. The method for dynamic image-guided intracranial hematoma puncture and localization according to claim 2, characterized in that, The method for obtaining the candidate region includes: Based on the preset growth point locations in the ultrasound image, the region is grown according to the growth criteria to obtain the growth region; after the growth is completed, the growth regions with pixel values higher than the preset deviation threshold of the adjacent growth regions are screened out, and the remaining growth regions are selected as candidate regions. The growth criterion is: growth can be performed when the difference in pixel value between the pixel to be grown and the growth point is less than a preset growth threshold; the pixel to be grown is a pixel within a preset neighborhood range of the growth point.
6. A dynamic image-guided intracranial hematoma puncture and localization system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the dynamic image-guided intracranial hematoma puncture and localization method as described in any one of claims 1 to 5.