Method and system for glomerular super-resolution microscopy imaging

Abstract

This invention relates to a method and system for super-resolution glomerular microscopy. It includes: acquiring target data from renal ultrasound echoes; performing microbubble motion pattern separation based on multiple sets of spatiotemporal filtering thresholds to generate several sets of fast and slow microbubble signals; performing microbubble trajectory generation processing on each set of fast and slow microbubble signals to generate corresponding sets of fast and slow microbubble motion trajectories; filtering the motion trajectories of each set of fast and slow microbubble motion trajectories to generate candidate glomerular cluster sets; subsequently, performing glomerular verification and filtering on each pair of candidate glomerular clusters generated from the sets of fast and slow microbubble motion trajectories to generate corresponding target glomerular sets; extracting the basic imaging features of each target glomerulus within the target glomerular set; and performing super-resolution glomerular microscopy based on the basic imaging features of all target glomeruli. This invention can effectively achieve super-resolution glomerular microscopy and improve the stability and accuracy of super-resolution glomerular microscopy.

Description

Technical Field

[0001] This invention relates to an imaging method and system, and more particularly to a glomerular super-resolution microscopic imaging method and system. Background Technology

[0002] The glomerulus is a key microstructure of the kidney that enables core physiological functions such as filtration and detoxification. Essentially, it is a network of coiled capillaries located between the afferent and efferent arterioles. The difference in diameter between the afferent and efferent arterioles creates a high filtration pressure. Under this high filtration pressure, metabolic toxins such as urea and creatinine, as well as excess water in the blood, are filtered into the Bowman's capsule and form primary urine. At the same time, blood cells and large protein molecules in the blood are effectively retained, thus enabling the kidney to play a crucial role in purifying the blood and excreting metabolic waste.

[0003] The glomerulus is affected by a variety of chronic diseases, such as hypertension, diabetes, autoimmune diseases, and cancer. These diseases can damage the glomerulus and cause it to lose its filtration function. Therefore, imaging the glomerulus and studying the mechanisms of glomerular function based on glomerular imaging images is crucial for the detection, diagnosis, and treatment of these related diseases.

[0004] The diameter of a human glomerulus is approximately 200 micrometers, a size far below the resolution limit of most medical imaging techniques. Therefore, clinical imaging techniques cannot observe individual glomeruli. Currently, glomerular function can only be indirectly assessed through blood or urine tests, but these tests only reflect the overall glomerular filtration rate.

[0005] Currently, glomerular functional imaging technology is at a critical juncture, transitioning from indirect inference to direct visualization. On the one hand, routine non-invasive imaging techniques in clinical practice (such as ultrasound and functional MRI) can indirectly reflect the population effect of glomerular function by assessing overall renal blood flow, perfusion, oxygenation, and fibrosis. However, these techniques are limited by spatial resolution (millimeter level), making it impossible to directly observe individual glomeruli with a diameter of approximately 200 micrometers, nor can they specifically distinguish the fine pathological changes of different cells and filtration barriers within the glomerulus. On the other hand, pathological imaging based on renal biopsy (light microscopy, immunofluorescence, electron microscopy), while providing the "gold standard" information on glomerular structure and function, is an invasive, static, and sampling-based ex vivo examination that cannot achieve long-term dynamic monitoring and carries the risk of sampling errors.

[0006] Emerging technologies, such as confocal microscopy and ultra-high resolution ultrasound, have shown initial potential for imaging at the glomerular level in vivo, but they are still in the clinical exploration stage and face challenges such as equipment miniaturization, operation standardization, and extensive validation.

[0007] With the continuous advancement of technology, the academic community has proposed a sensor-based super-resolution ultrasound imaging technology. This technology uses microbubbles as sensors for their surrounding microenvironment. By separating different microbubble motion modes that correspond to specific microstructures, and locating and tracking the "centroid" of microbubbles in different modes, it achieves ultra-high resolution imaging of the microstructure and function of the kidney glomeruli.

[0008] In the current clinical setting, sensor-based super-resolution ultrasound imaging technology uses clinically certified color Doppler ultrasound diagnostic instruments and matching contrast probes. Specifically, during standard ultrasound contrast examinations, the operator administers microbubble contrast agents intravenously according to a predetermined protocol and continuously stores the dynamic image sequence of the entire contrast process in DICOM format. These dynamic image data are essentially two-dimensional intensity images formed after real-time beamforming, filtering, and data compression within the device. The frame rate is limited by the clinical imaging depth and mode, and is typically 15 to 60 Hz.

[0009] After acquiring the dynamic image sequence in DICOM format, the analysis of the sensor-based super-resolution ultrasound imaging is performed on a separate offline workstation, in a time-sharing manner. The specific process may include:

[0010] First, the long sequence is divided into manageable short data blocks. For each short data block, the key step is to perform dual filtering separation: first, singular value decomposition (SVD) spatiotemporal filtering is used to suppress the organizational background, and then temporal bandpass filtering is used to initially separate the microbubble signal into two subsets, "fast" and "slow". Subsequently, in each subset, an image intensity-based localization algorithm is used to detect individual microbubbles, and the Hungarian algorithm is used in combination with two different sets of motion parameters suitable for fast and slow signals for trajectory tracking.

[0011] Finally, by calculating the normalized distance, residence time, and dispersion of the trajectory and combining them with the spatial information of the renal cortex, motion clusters exhibiting the characteristics of "low speed, high entanglement, and long residence" are identified. These clusters are then determined and reconstructed into super-resolution microscopic images of the glomerular capillary network, while outputting quantitative parameters such as density and distribution.

[0012] As can be seen from the above description, the transformation of sensor-based super-resolution ultrasound imaging technology from experimental research to clinical application still faces several key challenges: First, motion pattern separation strategies are rigid and lack robustness. Specifically, when separating fast and slow microbubbles, existing methods mostly rely on fixed spatiotemporal bandpass filter thresholds, failing to fully consider the continuous spectrum characteristics and dynamic variations of hemodynamics between individuals, regions, and pathological states. This single set of fixed thresholds not only lacks cross-scenario adaptability but may also incorrectly segment microbubble signals originating from the same physiological structure (such as a single glomerulus) but exhibiting velocity fluctuations due to local flow field pulsation into multiple discrete velocity categories. This disrupts the physiological consistency and integrity of microbubble trajectories and ultimately leads to a decrease in the accuracy of downstream glomerular identification and counting.

[0013] Second, in glomerular identification, the model features are limited and lack multi-source cross-validation. Specifically, current identification methods mostly rely on limited kinematic or morphological indicators such as normalized distance and residence time, resulting in insufficient feature dimensions and difficulty in achieving high specificity in complex backgrounds. Furthermore, existing schemes generally lack cross-validation mechanisms based on multiple channels, multiple parameters, or multiple sampling, making the identification results susceptible to noise, artifacts, and individual differences. This poses challenges to stability and repeatability, limiting the potential of this technology for accurate quantitative diagnosis.

[0014] As can be seen from the above description, the key issues mentioned above restrict the accuracy, stability, and scalability of sensor-based super-resolution ultrasound imaging technology in clinical in vivo glomerular structure and function imaging applications. Summary of the Invention

[0015] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method and system for super-resolution microscopy of the glomerulus, which can effectively realize super-resolution microscopy of the glomerulus and improve the stability and accuracy of super-resolution microscopy of the glomerulus.

[0016] According to the technical solution provided by the present invention, a method for super-resolution microscopic imaging of glomeruli is provided, the super-resolution microscopic imaging method comprising: Acquire kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, wherein the kidney ultrasound echo target data includes several frames of kidney ultrasound echo in-phase orthogonal signal groups that are time-correlated; The target data of renal ultrasound echo were separated into microbubble motion patterns based on multiple sets of spatiotemporal filtering thresholds to generate several sets of fast and slow microbubble signals. Each set of fast and slow microbubble signals corresponds to a set of spatiotemporal filtering thresholds, and each set of fast and slow microbubble signals includes a subset of fast microbubble signals and a subset of slow microbubble signals. Microbubble trajectory generation processing is performed on all fast and slow microbubble signal sets to generate fast and slow microbubble motion trajectory sets corresponding to each fast and slow microbubble signal set. The fast and slow microbubble motion trajectory sets include fast microbubble motion trajectory sets corresponding to fast microbubble signal subsets and slow microbubble motion trajectory sets corresponding to slow microbubble signal subsets. For each set of fast and slow microbubble motion trajectories, motion trajectory screening is performed to generate several candidate glomerular cluster sets that initially match the glomerular trajectory features; thereafter, at least based on spatial overlap, all candidate glomerular cluster sets are screened for glomerular body verification to generate the corresponding target glomerular set. The basic imaging features of each target glomerulus within the target glomerulus set are extracted, and glomerular super-resolution microscopy is performed based on the basic imaging features of all target glomerulus.

[0017] When filtering the motion trajectories for each set of fast and slow microbubble motion trajectories, the following steps are included: The set of fast and slow microbubble motion trajectories is subjected to primary dynamic coarse screening to generate a set of low-speed microbubble motion trajectories. The trajectory feature information of each low-speed microbubble motion trajectory within the set of low-speed microbubble motion trajectories is extracted, wherein the trajectory feature information includes at least kinematic features, morphological features, and spatiotemporal features; Based on the trajectory feature information of each low-speed microbubble motion trajectory, unsupervised clustering is performed on all low-speed microbubble motion trajectories to generate a low-speed motion trajectory core point cluster containing several low-speed microbubble motion trajectories. When the cluster features of a low-speed motion trajectory core point cluster are identified as glomerular feature clusters, the corresponding low-speed motion trajectory core point clusters are configured as candidate glomerular clusters, and a candidate glomerular cluster set is formed based on all candidate glomerular clusters.

[0018] When performing a preliminary dynamic coarse-grained screening of the set of fast and slow microbubble motion trajectories, the following are included: Determine the set of microbubble velocity information corresponding to the set of fast and slow microbubble motion trajectories, wherein the set of microbubble velocity information includes several microbubble velocities; The velocity distribution of microbubble velocities within the statistical set of microbubble velocity information is analyzed, and coarse velocity threshold information is generated based on the statistical velocity distribution. Calculate the trajectory velocity characteristics of each fast microbubble motion trajectory and slow microbubble motion trajectory within the set of fast and slow microbubble motion trajectories; When a trajectory velocity feature matches the coarse velocity threshold information, the fast microbubble motion trajectory or slow microbubble motion trajectory corresponding to the trajectory velocity feature is identified as a low-speed microbubble motion trajectory. A set of low-speed microbubble motion trajectories is formed based on all the low-speed microbubble motion trajectories.

[0019] When generating coarse velocity threshold information, the following are included: Based on the velocities of all microbubbles within the set of microbubble velocity information, a velocity distribution histogram is constructed for the set of fast and slow microbubble motion trajectories. On the constructed velocity distribution histogram, the corresponding main velocity distribution peak, secondary velocity distribution peak, and velocity distribution trough region located between the main velocity distribution peak and the secondary velocity distribution peak are determined; Velocity clustering is performed on all microbubble velocities within the velocity distribution trough region to generate coarse velocity threshold information after clustering. When performing speed clustering, the number of clusters is set to 2. When the number of clusters is 2, the coarse speed threshold information includes a first coarse speed threshold and a second coarse speed threshold. When the first coarse velocity threshold is less than the second coarse velocity threshold, the trajectory velocity feature is not greater than the first coarse velocity threshold, and the trajectory velocity feature matches the coarse velocity threshold information.

[0020] When generating the core point cluster of low-speed motion trajectory, the following is included: Construct a high-dimensional orthogonal metric space corresponding to trajectory feature information; Based on the trajectory feature information corresponding to each low-speed microbubble motion trajectory, all low-speed microbubble motion trajectories are distributed in the high-dimensional orthogonal metric space. In a high-dimensional orthogonal metric space, density-based unsupervised clustering is performed on all low-speed microbubble motion trajectories to obtain core point clusters at least in the high-dimensional orthogonal metric space. The core point clusters include several low-speed motion trajectory core point clusters, and each low-speed motion trajectory core point cluster includes several low-speed microbubble motion trajectories. For each low-speed motion trajectory core point cluster, the corresponding point cluster trajectory features of all low-speed microbubble motion trajectories within the low-speed motion trajectory core point cluster are statistically analyzed. Based on the statistically analyzed point cluster trajectory features, the cluster type features corresponding to the low-speed motion trajectory core point cluster are identified. When the point cluster trajectory features have the characteristics of high curvature and long dwell time, the cluster features of the core point cluster of the low-speed motion trajectory are identified as glomerular feature clusters.

[0021] When performing density-based unsupervised clustering on all low-velocity microbubble trajectories, including: Configure unsupervised classification conditions, wherein the unsupervised classification conditions include cluster neighborhood radius and cluster point number threshold. Based on the unsupervised classification conditions configured above, each low-speed microbubble motion trajectory is identified as a core point, boundary point, or noise point. Starting from any unvisited core point, based on the rules of density reachability and density connectivity, all core points and boundary points that satisfy density reachability are clustered into the same cluster, forming the corresponding low-speed motion trajectory core point cluster.

[0022] When performing sphericity verification screening on all candidate glomerular cluster sets, the following is included: For each candidate glomerular cluster in the candidate glomerular cluster set, generate the corresponding minimum circumcircle of each candidate glomerular cluster, so as to use the minimum circumcircle to represent a candidate glomerulus, and the center of the minimum circumcircle represents the center position of the candidate glomerulus. For each minimum circumcircle, determine the spatial overlap between the minimum circumcircle and the corresponding minimum circumcircles of the other sets of fast and slow microbubble motion trajectories, where, When the spatial overlap state is spatial correlation overlap, the candidate glomerular cluster corresponding to the current smallest circumcircle is configured with other candidate glomerular clusters that satisfy spatial correlation overlap as glomerular spatial overlap clusters. When each candidate glomerular cluster within the overlapping glomerular space has consistent characteristics, the smallest circumcircle configuration corresponding to the candidate glomerular cluster is configured as the target glomerulus. Based on the target glomeruli determined in the above manner, a set of target glomeruli is generated.

[0023] Determining the spatial overlap status includes: Calculate the positional coincidence and structural overlap rate between a minimum circumcircle and other minimum circumcircles, where, Positional overlap is the distance between the centers of the smallest circumcircles; The structural overlap rate is the proportion of the area overlap between the smallest circumcircles. When the positional overlap degree matches the positional overlap degree threshold, and the structural overlap rate matches the overlap rate threshold, the corresponding spatial overlap degree state is spatial correlation overlap.

[0024] For each candidate glomerular cluster within the spatial overlap cluster, a feature consistency check is performed to determine the corresponding feature consistency status of each candidate glomerular cluster within the spatial overlap cluster. When performing consistency feature verification, the following are included: Determine the glomerular motion characteristics and glomerular morphology characteristics of each candidate glomerular cluster; When the glomerular motion characteristics and glomerular morphology characteristics of all candidate glomerular clusters meet the similarity condition, then each candidate glomerular cluster within the spatially overlapping glomerular cluster has characteristic consistency.

[0025] The basic imaging features of the target glomerulus include local blood flow velocity distribution characteristics, mean transit time, and perfusion intensity; When performing glomerular super-resolution microscopy based on the fundamental imaging features of all target glomeruli, including: The basic imaging features of each target glomerulus are mapped to a spatial location. Subsequently, glomerular density distribution map, mean blood flow velocity map, and perfusion heterogeneity map of the renal cortex are generated by interpolation rendering.

[0026] When acquiring kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, the following steps are included: Acquire renal ultrafast ultrasound contrast imaging data generated based on ultrafast plane wave imaging technology, wherein the renal ultrafast ultrasound contrast imaging data includes several frames of renal ultrasound contrast echo data generated sequentially along the time dimension; Each frame of renal ultrasound contrast imaging data was demodulated to generate a set of renal ultrasound echo in-phase orthogonal signals corresponding to each frame of renal ultrasound contrast imaging data. Kidney ultrasound echo target data is generated based on all in-phase orthogonal signal groups of kidney ultrasound echoes.

[0027] A glomerular super-resolution imaging system includes an imaging processing terminal, wherein, For any kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, the imaging processing terminal uses the method described above to perform glomerular super-resolution microscopy.

[0028] The advantages of this invention are as follows: Microbubble motion pattern separation is performed on renal ultrasound echo target data based on multiple sets of spatiotemporal filtering thresholds to generate several sets of fast and slow microbubble signals; microbubble trajectory generation processing is performed on all sets of fast and slow microbubble signals to generate sets of fast and slow microbubble motion trajectories corresponding to each set; motion trajectory screening is performed on each set of fast and slow microbubble motion trajectories to generate several sets of candidate glomerular clusters that initially match the glomerular trajectory features; subsequently, glomerular clusters are screened based on at least spatial overlap to generate corresponding target glomerular sets; basic imaging features of each target glomerulus within the target glomerular set are extracted; and glomerular super-resolution microscopy is performed based on the basic imaging features of all target glomeruli. This effectively achieves glomerular super-resolution microscopy and improves the stability and accuracy of glomerular super-resolution microscopy. Attached Figure Description

[0029] Figure 1 This is a schematic flowchart of an embodiment of the glomerular super-resolution imaging method of the present invention.

[0030] Figure 2 This is a schematic diagram of one embodiment of the microbubble motion pattern separation of the present invention.

[0031] Figure 3 This is a schematic diagram of an embodiment of the microbubble trajectory generation process of the present invention.

[0032] Figure 4 This is a schematic diagram of one embodiment of the present invention for generating a set of candidate glomerular clusters. Detailed Implementation

[0033] The present invention will be further described below with reference to specific accompanying drawings and embodiments.

[0034] To effectively achieve super-resolution microscopic imaging of the glomerulus and improve its stability and accuracy, this invention provides a method for super-resolution microscopic imaging of the glomerulus. Specifically, the super-resolution microscopic imaging method includes: Acquire kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, wherein the kidney ultrasound echo target data includes several frames of kidney ultrasound echo in-phase orthogonal signal groups that are time-correlated; The target data of renal ultrasound echo were separated into microbubble motion patterns based on multiple sets of spatiotemporal filtering thresholds to generate several sets of fast and slow microbubble signals. Each set of fast and slow microbubble signals corresponds to a set of spatiotemporal filtering thresholds, and each set of fast and slow microbubble signals includes a subset of fast microbubble signals and a subset of slow microbubble signals. Microbubble trajectory generation processing is performed on all fast and slow microbubble signal sets to generate fast and slow microbubble motion trajectory sets corresponding to each fast and slow microbubble signal set. The fast and slow microbubble motion trajectory sets include fast microbubble motion trajectory sets corresponding to fast microbubble signal subsets and slow microbubble motion trajectory sets corresponding to slow microbubble signal subsets. For each set of fast and slow microbubble motion trajectories, motion trajectory screening is performed to generate several candidate glomerular cluster sets that initially match the glomerular trajectory features; thereafter, at least based on spatial overlap, all candidate glomerular cluster sets are screened for glomerular body verification to generate the corresponding target glomerular set. The basic imaging features of each target glomerulus within the target glomerulus set are extracted, and glomerular super-resolution microscopy is performed based on the basic imaging features of all target glomerulus.

[0035] Figure 1 This illustration shows an embodiment of the present invention for super-resolution microscopic imaging of the glomerulus. As shown in the figure, when performing super-resolution microscopic imaging, kidney ultrasound echo target data should be acquired first. Kidney ultrasound echo target data can be acquired using existing commonly used techniques. In one embodiment of the present invention, acquiring kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology includes: Acquire renal ultrafast ultrasound contrast imaging data generated based on ultrafast plane wave imaging technology, wherein the renal ultrafast ultrasound contrast imaging data includes several frames of renal ultrasound contrast echo data generated sequentially along the time dimension; Each frame of renal ultrasound contrast imaging data was demodulated to generate a set of renal ultrasound echo in-phase orthogonal signals corresponding to each frame of renal ultrasound contrast imaging data. Kidney ultrasound echo target data is generated based on all in-phase orthogonal signal groups of kidney ultrasound echoes.

[0036] Similar to existing technologies, when acquiring renal ultrafast contrast-enhanced ultrasound (UHE) scan data, the target kidney should first be identified. Then, commonly used plane wave imaging techniques are employed to perform UHE scans on the target kidney to obtain the required UHE scan data. Generally, the UHE scan data should include multiple frames of renal ultrasound echo data. For example, transmitting a plane wave to the target kidney generates one frame of renal ultrasound echo data through the corresponding echo signal. The method for generating renal ultrasound echo data is consistent with existing technologies. Because the transmitted plane wave has temporal characteristics, the generated multiple frames of renal ultrasound echo data also exhibit temporal characteristics.

[0037] It should be noted that the number of frames of renal ultrasound contrast echo data within the renal ultrafast ultrasound contrast imaging scan data can be selected as needed, such as based on the transmission frequency of the plane wave and the scanning time of the target kidney. Since the frame rate of generating renal ultrasound contrast echo data can be in the hundreds of hertz range, it can provide key temporal resolution for capturing the transient motion of microbubbles in the glomerulus, laying the physical basis for glomerular super-resolution imaging.

[0038] After acquiring renal ultrafast contrast-enhanced ultrasound (UHE) scan data, each frame of renal UHE echo data is demodulated to generate corresponding renal ultrasound echo in-phase orthogonal signals. The modulation process is consistent with existing technologies, such as demodulating each renal UHE echo signal within each frame of UHE echo data to generate corresponding in-phase orthogonal signals. Based on all in-phase orthogonal signals, a renal ultrasound echo in-phase orthogonal signal set can be obtained. Based on all renal ultrasound echo in-phase orthogonal signal sets, renal ultrasound echo target data can be generated. It is understood that since there is a one-to-one correspondence between the renal ultrasound echo in-phase orthogonal signals and the renal contrast-enhanced ultrasound echo signals, the renal ultrasound echo in-phase orthogonal signal sets within the renal ultrasound echo data also possess the same temporal characteristics, meaning there is a temporal correlation between the renal ultrasound echo in-phase orthogonal signal sets.

[0039] Depend on Figure 1It is understood that after acquiring the target data of renal ultrasound echo, microbubble motion pattern separation should be performed. This separation allows for the determination of the microbubble's trajectory and occurrence time within the kidney. The method of microbubble motion pattern separation can be consistent with existing technologies, such as using a spatiotemporal filtering threshold to filter the renal ultrasound echo target data to achieve microbubble motion pattern separation. As described in the background art, existing technologies use a set of spatiotemporal filtering thresholds to generate a set of fast and slow microbubble signals during microbubble motion pattern separation. However, in one embodiment of this invention, multiple sets of spatiotemporal filtering thresholds are used, generating multiple sets of fast and slow microbubble signals. Each set of fast and slow microbubble signals corresponds to a set of spatiotemporal filtering thresholds, such as... Figure 2 As shown.

[0040] When multiple sets of spatiotemporal filtering thresholds are used for microbubble motion pattern separation and multiple sets of fast and slow microbubble signals are generated, the limitation of insufficient robustness of a single spatiotemporal filtering threshold in cross-clinical scenarios and glomerular imaging can be overcome, significantly improving the universality of clinical applications and reducing the probability of glomerular identification errors. When using multiple sets of spatiotemporal filtering thresholds, microbubble motion pattern separation can be performed in parallel using each set. However, the method and process of using each set of spatiotemporal filtering thresholds for microbubble motion pattern separation can be consistent with existing technologies, such as... Figure 2 The illustration shows an embodiment of microbubble motion mode separation using the dual filtering separation mentioned in the background art. Specifically, when using dual filtering separation, the dual filtering separation can be a multi-level filtering technique that integrates singular value decomposition filtering and frequency domain filtering.

[0041] When using a dual-filter separation method for microbubble motion pattern separation, singular value decomposition (SVD) filtering should be performed first. Specifically, SVD filtering reconstructs the target data of renal ultrasound echoes into a two-dimensional spatiotemporal matrix. Then, SVD is performed on this two-dimensional spatiotemporal matrix, decomposing it into a series of spatiotemporal patterns (represented by singular vectors) and their energies (singular values). By threshold selection, the singular value components representing specific microbubble motion patterns are retained, while other components are set to zero. Then, an inverse SVD transformation is performed to reconstruct the microbubble signal of the specific motion pattern. For frequency domain filtering, the bandpass filtering method mentioned in the background technique can be used. The object of frequency domain filtering is the microbubble signal reconstructed by SVD filtering. The method and process of frequency domain filtering will not be elaborated here.

[0042] As explained above, each set of spatiotemporal filtering thresholds should include the truncation order of the singular value decomposition filtering state and the bandpass filtering range of the bandpass filtering state. When multiple sets of spatiotemporal filtering thresholds are used in parallel for microbubble motion pattern separation, the multiple sets of spatiotemporal filtering thresholds should exhibit a gradient distribution, meaning that the threshold values ​​of each set of spatiotemporal filtering thresholds should not be completely identical. Specifically, the threshold value should be determined based on satisfying the gradient distribution and achieving the required microbubble motion pattern separation. For example, the range of values ​​for the spatiotemporal filtering thresholds can be determined using existing commonly used techniques. Subsequently, the threshold value of each set of spatiotemporal filtering thresholds can be configured using existing commonly used methods. The specific methods and processes for configuring each set of spatiotemporal filtering thresholds will not be detailed here.

[0043] In addition, to optimize the filtering effect at the time start point, windowed frequency domain filtering is also used to suppress boundary transient effects. When using windowed frequency domain filtering, the window function can be selected as a Hanning window or a Hamming window. The window function is applied to the coefficients of the frequency domain filter to smooth the truncation of edges and suppress spectral leakage, thereby improving the overall filtering effect. Specifically, the method and process of using windowed frequency domain filtering can be consistent with existing technologies, and will not be elaborated here.

[0044] For each set of fast and slow microbubble signals generated by microbubble motion pattern separation, the set includes a fast microbubble signal subset and a slow microbubble signal subset. It is understood that the fast microbubble signal subset may include several frames of fast microbubble signal groups. The number of frames in the fast microbubble signal groups should correspond to the number of frames in the in-phase orthogonal signal groups of the renal ultrasound echo, and there should be a one-to-one correspondence between the fast microbubble signal groups and the in-phase orthogonal signal groups of the renal ultrasound echo. The situation of the slow microbubble signal subset can be found in the corresponding description of the fast microbubble signal subset here.

[0045] Depend on Figure 1 It is known that after obtaining multiple sets of fast and slow microbubble signals, microbubble trajectory generation processing should be performed on all sets of fast and slow microbubble signals. Specifically, after microbubble trajectory generation processing, a corresponding set of fast and slow microbubble motion trajectories can be generated. One set of fast and slow microbubble signals can generate one set of fast and slow microbubble motion trajectories after microbubble trajectory generation processing. The set of fast and slow microbubble motion trajectories should include a set of fast microbubble motion trajectories and a set of slow microbubble motion trajectories. Generally, when performing microbubble trajectory generation, performing microbubble trajectory generation processing on a subset of fast microbubble signals can generate a corresponding set of fast microbubble motion trajectories, while performing microbubble trajectory generation processing on a subset of slow microbubble signals can generate a corresponding set of slow microbubble motion trajectories. Therefore, microbubble trajectory generation processing can be performed on the subsets of fast microbubble signals and the subsets of slow microbubble signals in parallel.

[0046] The method for generating microbubble trajectories for fast and slow microbubble signal subsets can be consistent with existing technologies. The following example illustrates the microbubble trajectory generation method and process. In one feasible embodiment, the following is an example: For fast-moving and slow-moving microbubble signal subsets, a method based on local gradient distribution weighting should be used for microbubble centroid localization, respectively. The following example illustrates the method for microbubble centroid localization using a fast-moving microbubble signal subset. For the method and process of microbubble centroid localization using a slow-moving microbubble signal subset, please refer to the corresponding documentation.

[0047] It should be understood that, since each subset of fast microbubble signals may include several frames of fast microbubble signal groups, and a two-dimensional image of fast microbubble signals can be generated from each frame of fast microbubble signal groups, multiple corresponding frames of two-dimensional images of fast microbubble signals can be generated for each subset of fast microbubble signals, and the generated two-dimensional images of fast microbubble signals also satisfy the aforementioned temporal relationship. The specific method for generating two-dimensional images of fast microbubble signals can be consistent with existing technologies and will not be elaborated here.

[0048] Within each frame of the fast microbubble signal two-dimensional image, each microbubble candidate region is determined, and the image gradient magnitude and orientation field of each microbubble candidate region are calculated. Then, the centroid of the current microbubble candidate region is estimated using the traditional intensity-based weighted average method. The specific method for estimating the centroid of the microbubble candidate region is consistent with the existing technology, and the specific process will not be described here.

[0049] In practice, the method for determining each microbubble candidate region can be consistent with existing technologies. For example, the approximate center of the microbubble (not the actual microbubble center) can be determined by using local maxima, and then a fixed-size neighborhood of the approximate center coordinates (e.g., 5) is taken as the image region. 5 or 7 7) To represent a single microbubble, the selected image region here is the microbubble candidate region. It can be understood that within each frame of a fast microbubble signal two-dimensional image, the above centroid localization method can obtain one or more microbubble centroids. In addition, the above iterative calculation-optimized composite weighted centroid localization method can significantly improve the localization stability of low-speed, weak signal microbubbles.

[0050] After locating all microbubble centroids within each frame of the fast microbubble signal 2D image, a microbubble centroid set can be generated based on all microbubble centroids, thus obtaining the microbubble centroid set for all frames of the fast microbubble signal 2D image. Subsequently, using an improved Kalman filter tracking method, temporal correlation is performed on the centroid set of each fast microbubble signal subset. The temporal correlation method can be consistent with existing technologies. For example, for a microbubble centroid within a frame of the fast microbubble signal 2D image, the position of the corresponding microbubble centroid in the next temporal sequence is determined through temporal correlation. A fast microbubble motion trajectory can be formed through the corresponding temporally correlated microbubble centroids. For each fast microbubble signal subset, a fast microbubble motion trajectory set can be formed based on all corresponding fast microbubble motion trajectories, such as... Figure 3 As shown.

[0051] Generally, each set of fast microbubble motion trajectories may include multiple fast microbubble motion trajectories. The number of fast microbubble motion trajectories should be related to the number of microbubble centroids determined above and the trajectory tracking process. Specifically, the trajectory tracking mentioned here has the same meaning as the trajectory tracking mentioned in the background art. For details, please refer to the corresponding description above.

[0052] Depend on Figure 1 It is known that after obtaining the set of fast and slow microbubble motion trajectories corresponding to each set of fast and slow microbubble signals, motion trajectory screening should be performed first to obtain the candidate glomerular cluster set for each set of fast and slow microbubble motion trajectories. In one embodiment of the present invention, the motion trajectory screening for each set of fast and slow microbubble motion trajectories includes: The set of fast and slow microbubble motion trajectories is subjected to primary dynamic coarse screening to generate a set of low-speed microbubble motion trajectories. The trajectory feature information of each low-speed microbubble motion trajectory within the set of low-speed microbubble motion trajectories is extracted, wherein the trajectory feature information includes at least kinematic features, morphological features, and spatiotemporal features; Based on the trajectory feature information of each low-speed microbubble motion trajectory, unsupervised clustering is performed on all low-speed microbubble motion trajectories to generate a low-speed motion trajectory core point cluster containing several low-speed microbubble motion trajectories. When the cluster features of a low-speed motion trajectory core point cluster are identified as glomerular feature clusters, the corresponding low-speed motion trajectory core point clusters are configured as candidate glomerular clusters, and a candidate glomerular cluster set is formed based on all candidate glomerular clusters.

[0053] In practice, when screening motion trajectories, a preliminary coarse-grained dynamics screening should be performed on each set of fast and slow microbubble motion trajectories. The purpose of this preliminary coarse-grained dynamics screening is to generate a set of low-speed microbubble motion trajectories, such as... Figure 4As shown, it is understandable that the set of low-velocity microbubble motion trajectories obtained through screening may represent the motion trajectories of the glomeruli. The method for primary kinetic coarse screening will be explained in detail below. Generally, the set of low-velocity microbubble motion trajectories will include multiple low-velocity microbubble motion trajectories.

[0054] After obtaining the set of low-speed microbubble motion trajectories, the trajectory feature information of each low-speed microbubble motion trajectory within the set should be extracted. Figure 4 It is understood that the trajectory feature information includes at least kinematic features, morphological features, and spatiotemporal features. Kinematic features may include average velocity and acceleration; morphological features may include curvature and branching structure; and spatiotemporal features may include normalized distance, spatial relationship parameters of neighboring trajectories, and dwell time. Specifically, For each low-speed microbubble motion trajectory, the average velocity and acceleration within the kinematic features can be extracted using existing techniques. As explained above, each low-speed microbubble motion trajectory may include multiple microbubble centroids. The time interval between adjacent microbubble centroids is the interval of the emitted plane wave, and the distance between the microbubble centroids can be determined accordingly. The specific determination method can be consistent with existing techniques. For example, each microbubble centroid on the low-speed microbubble motion trajectory is located on a corresponding image frame. The distance between the microbubble centroids can be determined by calculating the distance between the corresponding coordinates of the microbubble centroids. After determining the distance between adjacent microbubbles and the corresponding time, the velocity between adjacent microbubble centroids can be determined. Then, the corresponding average velocity and acceleration can be calculated using existing methods. Of course, other methods can also be used to calculate the required kinematic features, which will not be illustrated here. The curvature of the morphological features specifically refers to the curvature of the trajectory of low-speed microbubble motion. For a given low-speed microbubble motion trajectory, the corresponding curvature can be calculated using existing calculation methods. For the branching structure of the low-speed microbubble motion trajectory, generally, the number of neighboring pixels of each pixel in the binary image of the low-speed microbubble motion trajectory can be calculated. If there is a pixel in the binary image with a number of neighboring pixels greater than 2, it is considered that the trajectory has a branching point, thereby determining the corresponding branching structure.

[0055] The normalized distance within the spatiotemporal characteristics specifically refers to the ratio of the true distance between the starting points of the trajectory to the Euclidean distance. The residence time within the spatiotemporal characteristics specifically refers to the time from the trajectory's appearance to its disappearance. The spatial relationship parameters of neighboring trajectories generally refer to the number of adjacent low-speed microbubble motion trajectories around the low-speed microbubble motion trajectory, as described above. The normalized distance and residence time can be calculated using existing commonly used methods; the specific calculation process will not be elaborated here.

[0056] The spatial relationship parameters of neighboring trajectories can be statistically determined using existing common methods. For example, a spatial neighborhood distance can be set, and all low-speed microbubble motion trajectories within this distance can be considered as adjacent low-speed microbubble motion trajectories. It is understood that the size of the spatial neighborhood distance can be selected and determined according to actual needs; for example, it can be set to 10 μm. Of course, other values ​​can also be used, such as those selected based on imaging accuracy and experience. When statistically analyzing neighboring low-speed microbubble motion trajectories, one feasible approach is to calculate the distance between the centroid of each microbubble on the current low-speed microbubble motion trajectory and other low-speed microbubble motion trajectories, taking the centroid of that microbubble as the center. When the calculated distance is within the set spatial neighborhood distance, the corresponding low-speed microbubble motion trajectory is considered as an estimate of the motion of the adjacent low-speed microbubble. Of course, other calculation methods can also be used to determine whether a low-speed microbubble motion trajectory can be used as an estimate of the motion of the adjacent low-speed microbubble; specific methods will not be illustrated here.

[0057] Depend on Figure 4 It can be seen that after obtaining the trajectory feature information of each low-speed microbubble motion trajectory, unsupervised clustering should be performed on all low-speed microbubble motion trajectories based on the trajectory feature information of each low-speed microbubble motion trajectory. After unsupervised clustering, a low-speed motion trajectory core point cluster containing several low-speed microbubble motion trajectories can be generated. The low-speed motion trajectory core point cluster generally includes some low-speed microbubble motion trajectories obtained by clustering.

[0058] Understandably, after unsupervised clustering, one or more clusters of low-velocity motion trajectory core points can be obtained. Subsequently, the cluster class characteristics of each low-velocity motion trajectory core point cluster should be determined. These cluster class characteristics generally represent the corresponding performance features of all low-velocity microbubble motion trajectories within the corresponding low-velocity motion trajectory core point cluster. When the cluster class characteristics of a low-velocity motion trajectory core point cluster are identified as glomerular feature clusters, then the corresponding low-velocity motion trajectory core point cluster is configured as a candidate glomerular cluster. Figure 4 The figure illustrates one embodiment of cluster features, where cluster features can be linear clusters, branching clusters, and feature clusters, i.e. Figure 4 The feature clusters in the model are the glomerular feature clusters. When the cluster feature is a glomerular feature cluster, that is, the low-speed microbubble motion trajectory within the core cluster of the entire low-speed motion trajectory is similar to the motion trajectory of the glomerulus. Generally, each set of fast and slow microbubble motion trajectories, after the above motion trajectory screening, can yield one or more candidate glomerular clusters. Based on all the candidate glomerular clusters, a candidate glomerular cluster set can be formed.

[0059] In one embodiment of the present invention, the primary dynamic coarse screening of the set of fast and slow microbubble motion trajectories includes: Determine the set of microbubble velocity information corresponding to the set of fast and slow microbubble motion trajectories, wherein the set of microbubble velocity information includes several microbubble velocities; The velocity distribution of microbubble velocities within the statistical set of microbubble velocity information is analyzed, and coarse velocity threshold information is generated based on the statistical velocity distribution. Calculate the trajectory velocity characteristics of each fast microbubble motion trajectory and slow microbubble motion trajectory within the set of fast and slow microbubble motion trajectories; When a trajectory velocity feature matches the coarse velocity threshold information, the fast microbubble motion trajectory or slow microbubble motion trajectory corresponding to the trajectory velocity feature is identified as a low-speed microbubble motion trajectory. A set of low-speed microbubble motion trajectories is formed based on all the low-speed microbubble motion trajectories.

[0060] In practice, when performing a preliminary dynamic coarse screening of each set of fast and slow microbubble motion trajectories, the set of microbubble velocity information corresponding to the set of fast and slow microbubble motion trajectories should be determined first. The set of microbubble velocity information specifically refers to the microbubble velocity contained in the entire set of fast and slow microbubble motion trajectories, that is, it should include all microbubble velocities contained in the set of slow microbubble motion trajectories and the set of fast microbubble motion trajectories. Therefore, the set of microbubble velocity information should include several microbubble velocities. Here, the microbubble velocity generally refers to the corresponding velocity between two adjacent microbubble centroids on a motion trajectory.

[0061] After determining the aggregated microbubble velocity information, the velocity distribution of all microbubble velocities within the aggregated microbubble velocity information can be statistically analyzed. Subsequently, coarse velocity threshold information can be generated based on the velocity distribution. In one embodiment of the present invention, generating the coarse velocity threshold information includes: Based on the velocities of all microbubbles within the set of microbubble velocity information, a velocity distribution histogram is constructed for the set of fast and slow microbubble motion trajectories. On the constructed velocity distribution histogram, the corresponding main velocity distribution peak, secondary velocity distribution peak, and velocity distribution trough region located between the main velocity distribution peak and the secondary velocity distribution peak are determined; Velocity clustering is performed on all microbubble velocities within the velocity distribution trough region to generate coarse velocity threshold information after clustering. When performing speed clustering, the number of clusters is set to 2. When the number of clusters is 2, the coarse speed threshold information includes a first coarse speed threshold and a second coarse speed threshold. When the first coarse velocity threshold is less than the second coarse velocity threshold, the trajectory velocity feature is not greater than the first coarse velocity threshold, and the trajectory velocity feature matches the coarse velocity threshold information.

[0062] In one embodiment of the present invention, constructing a velocity distribution histogram includes: Based on the velocities of all microbubbles within the set of microbubble velocity information, a microbubble velocity distribution interval is generated. The determined microbubble velocity distribution interval is divided into several microbubble velocity sub-intervals. The probability distribution of microbubble velocity is statistically analyzed for each microbubble velocity sub-interval, and a velocity distribution histogram is constructed based on the statistically analyzed probability distribution of microbubble velocity. Specifically, generating the microbubble velocity distribution interval involves determining the maximum and minimum values ​​of all microbubble velocities. The minimum and maximum microbubble velocities constitute the microbubble velocity distribution interval. To improve the accuracy of generating coarse velocity threshold information, the microbubble velocity distribution interval can be divided. For example, it can be divided into intervals based on the difference between the minimum and maximum microbubble velocities, or other methods can be used. The number of microbubble velocity sub-intervals and the corresponding microbubble velocity ranges can be selected as needed. Generally, more microbubble velocity sub-intervals result in more accurate coarse velocity threshold information, but reduce the generation speed. A trade-off between generation accuracy and generation speed is generally necessary.

[0063] After dividing the area into microbubble velocity sub-intervals, the number of all microbubble velocities in each sub-interval is counted, thus obtaining the microbubble velocity distribution probability for each sub-interval. Subsequently, a velocity distribution histogram can be constructed, where the horizontal axis represents the microbubble velocity sub-interval, and the vertical axis represents the corresponding microbubble velocity distribution probability for each sub-interval.

[0064] After constructing the velocity distribution histogram, the main peak and secondary peak of the velocity distribution can be obtained within the histogram. Specifically, the main peak and secondary peak refer to the two largest microbubble velocity distribution probabilities within the velocity distribution histogram. The velocity distribution trough region specifically refers to the sub-interval range of microbubble velocity distribution between the main peak and the secondary peak.

[0065] After determining the velocity distribution trough region, velocity clustering can be performed on all microbubble velocities within the trough region. Commonly used clustering methods can be employed, such as K-means clustering. After selecting the clustering method, coarse velocity threshold information can be generated after clustering. Figure 4As explained above, the coarse velocity threshold information can be used to identify fast or slow microbubble motion trajectories as low-speed microbubble motion trajectories. Therefore, the number of clusters can be 2. After clustering, the first coarse velocity threshold and the second coarse velocity threshold can be obtained. The specific method and process of obtaining the first and second coarse velocity thresholds through clustering can be consistent with the existing technology and will not be elaborated here.

[0066] For each fast-moving microbubble trajectory and each slow-moving microbubble trajectory, the corresponding trajectory velocity feature can be the average velocity or median velocity of the corresponding trajectory. Of course, other velocity information can also be used as the trajectory velocity feature, which can be selected as needed. After determining the velocity trajectory feature, if the first coarse velocity threshold is less than the second coarse velocity threshold, and the trajectory velocity feature is not greater than the first coarse velocity threshold, then the trajectory velocity feature is considered to match the coarse velocity threshold information. At this time, the microbubble trajectory corresponding to the trajectory velocity feature can be identified as a low-speed microbubble trajectory.

[0067] In specific implementation, when the trajectory velocity feature is greater than the first coarse velocity threshold but less than the second coarse velocity threshold, the microbubble motion trajectory corresponding to the trajectory velocity threshold can be identified as a medium-speed microbubble motion trajectory. When the trajectory velocity feature is not less than the second coarse velocity threshold, the microbubble motion trajectory corresponding to the trajectory velocity feature can be identified as a high-speed microbubble motion trajectory. Figure 4 In the context of high-speed microbubble motion trajectory set, the set of high-speed microbubble motion trajectories specifically refers to the set generated by all identified high-speed microbubble motion trajectories. Similarly, the meaning of the set of medium-speed microbubble motion trajectory set can be obtained.

[0068] In one embodiment of the present invention, generating a cluster of core points for a low-speed motion trajectory includes: Construct a high-dimensional orthogonal metric space corresponding to trajectory feature information; Based on the trajectory feature information corresponding to each low-speed microbubble motion trajectory, all low-speed microbubble motion trajectories are distributed in the high-dimensional orthogonal metric space. In a high-dimensional orthogonal metric space, density-based unsupervised clustering is performed on all low-speed microbubble motion trajectories to obtain core point clusters at least in the high-dimensional orthogonal metric space. The core point clusters include several low-speed motion trajectory core point clusters, and each low-speed motion trajectory core point cluster includes several low-speed microbubble motion trajectories. For each low-speed motion trajectory core point cluster, the corresponding point cluster trajectory features of all low-speed microbubble motion trajectories within the low-speed motion trajectory core point cluster are statistically analyzed. Based on the statistically analyzed point cluster trajectory features, the cluster type features corresponding to the low-speed motion trajectory core point cluster are identified. When the point cluster trajectory features have the characteristics of high curvature and long dwell time, the cluster features of the core point cluster of the low-speed motion trajectory are identified as glomerular feature clusters.

[0069] It should be understood that the high-dimensional orthogonal metric space should correspond to the trajectory features contained in the trajectory feature information. For example, if the trajectory feature information mentioned above includes average velocity, acceleration, curvature, minute structure, normalized distance, dwell time, and spatial parameter relationships of neighboring trajectories, then the constructed high-dimensional orthogonal metric space should have 7 dimensions, and the 7 dimensions should be mutually perpendicular. Each dimension corresponds to a trajectory feature. The method of constructing the high-dimensional orthogonal metric space can be consistent with existing technologies, and will not be elaborated here.

[0070] After constructing a high-dimensional orthogonal metric space, since the trajectory feature information of each low-speed microbubble motion trajectory is known, the position of each low-speed microbubble motion trajectory within the high-dimensional orthogonal metric space is determined based on the distribution of the trajectory feature information in the high-dimensional orthogonal metric space. Figure 4 After the low-speed microbubble motion trajectories in the low-speed microbubble motion trajectory set are distributed in a high-dimensional orthogonal metric space, density-based unsupervised clustering is performed. After density-based unsupervised clustering, core point clusters are obtained at least in the high-dimensional orthogonal metric space.

[0071] In one embodiment of the present invention, when performing density-based unsupervised clustering on all low-speed microbubble motion trajectories, the following steps are included: Configure unsupervised classification conditions, wherein the unsupervised classification conditions include cluster neighborhood radius and cluster point number threshold. Based on the unsupervised classification conditions configured above, each low-speed microbubble motion trajectory is identified as a core point, boundary point, or noise point. Starting from any unvisited core point, based on the rules of density reachability and density connectivity, all core points and boundary points that satisfy density reachability are clustered into the same cluster, forming the corresponding low-speed motion trajectory core point cluster.

[0072] It should be noted that during unsupervised clustering, the low-speed microbubble motion trajectory within the high-dimensional orthogonal metric space is considered an independent data point. Subsequently, based on the unsupervised classification criteria, the low-speed microbubble motion trajectory of each independent data point is initially classified. For example, a low-speed microbubble motion trajectory can be classified as a core point, boundary point, or noise point. Specifically, a core point refers to a data point within the cluster neighborhood radius of each data point that has at least a threshold number of cluster points; core points represent high-density regions of behavioral patterns. Boundary points refer to data points within the cluster neighborhood radius of each data point but not exceeding the threshold number of cluster points. When a data point does not meet the criteria of being a core point or boundary point, it is identified as a noise point, which is generally a sparse point.

[0073] When identifying core points, for each data point, the high-dimensional spatial distance between each data point and other data points should be calculated. The specific method for calculating the high-dimensional spatial distance is consistent with existing technologies. For example, after determining the coordinates of each data point in the high-dimensional orthogonal metric space, the corresponding distance is calculated using the distance calculation method between the corresponding coordinates in the high-dimensional space. Based on the calculated distance, the number of data points within the clustering neighborhood radius that are not less than the clustering point threshold for each data point can be statistically determined. It should be noted that the clustering neighborhood radius and the clustering point threshold can be selected as needed. For example, the clustering neighborhood radius and the clustering point threshold can be determined using statistical methods based on the number of dimensions of the high-dimensional orthogonal metric space and the movement of the glomeruli. Specific settings will not be detailed here.

[0074] In unsupervised clustering, we should start from each unvisited core point. Unvisited core points are those that haven't been judged by the rules of density reachability and density connectivity. Initially, all core points are unvisited. It should be noted that in density-based unsupervised clustering, the meanings of density reachability and density connectivity are consistent with existing technologies. For density reachability, a feasible implementation is: given a sequence of points p1.p2.p3...pn, where for the i-th point (i from 1 to n-1), point pi+1 is within the neighborhood of point pi, then p1 to pn are considered density reachable. For density connectivity, if there exists a point o such that points p and q both originate from point o and are density reachable, then points p and q are said to be density connected.

[0075] Based on the density reachability and density adjacency rules mentioned above, for each core point, core points and boundary points that satisfy density reachability can be determined, and the density reachable core points and boundary points can be clustered into the same cluster, thereby forming a low-speed motion trajectory core point cluster. It can be seen that each low-speed motion trajectory core point cluster should include several low-speed microbubble motion trajectories.

[0076] After obtaining the core point clusters of low-speed motion trajectories, the corresponding point cluster trajectory features can be statistically analyzed. These point cluster trajectory features specifically refer to the shape and other characteristics exhibited by the core point clusters of low-speed motion. Based on the statistically analyzed point cluster trajectory features, the cluster type characteristics corresponding to the core point clusters of the low-speed motion trajectory can be determined. Specifically: If the trajectory of all low-speed microbubble motions within the low-speed motion core cluster is approximately straight or an arc with a large radius of curvature, with high directional consistency and relatively low but stable speed, then the cluster characteristics of the low-speed motion core cluster should be identified as a linear cluster. The organ tissue corresponding to the linear cluster can be a linear cluster corresponding to small arcuate arteriovenous branches or relatively straight interlobular arteriovenous branches.

[0077] If the trajectory shapes of all low-speed microbubble movements within the low-speed motion core cluster have obvious "Y" or "T" bifurcation or confluence structures, and the velocity may change in direction and magnitude at the bifurcation point, then the cluster characteristics of the low-speed motion core cluster should be identified as a branch cluster, and the organ tissue corresponding to the branch cluster can be the branch cluster (track) corresponding to the afferent arteriole and the efferent arteriole.

[0078] If the trajectory shapes of all low-velocity microbubble movements within the low-velocity core cluster exhibit high curvature and long residence time, such as highly entangled and rotated trajectories within a very small spatial range, forming clusters or ring structures with extremely low speeds and microbubble residence times significantly longer than in other blood vessels, then the cluster characteristics of the low-velocity core cluster should be identified as glomerular characteristic clusters. It is understood that for each set of fast and slow microbubble movements, after the above trajectory screening, one or more candidate glomerular clusters can be obtained, and a candidate glomerular cluster set can be formed based on all candidate glomerular clusters. After trajectory screening, as explained above, glomerular verification screening should also be performed.

[0079] As can be seen from the unsupervised clustering described above, after clustering to obtain the core point clusters of low-speed motion trajectories, each core point cluster should ensure the trajectory feature information contained in the constructed high-dimensional orthogonal metric space, and can correspondingly determine a main trajectory feature. For example, if the corresponding curvature feature is mainly manifested as an approximately straight line or an arc with a large radius of curvature, then the cluster feature of the core point cluster of low-speed motion is identified as a linear cluster. For other cases identified as branch clusters or feature clusters, please refer to the corresponding explanations here, that is, determine the cluster feature corresponding to the core point cluster of low-speed motion trajectories based on the main manifested features.

[0080] In one embodiment of the present invention, when performing sphericity verification screening on all candidate glomerular cluster sets, the process includes: For each candidate glomerular cluster in the candidate glomerular cluster set, generate the corresponding minimum circumcircle of each candidate glomerular cluster, so as to use the minimum circumcircle to represent a candidate glomerulus, and the center of the minimum circumcircle represents the center position of the candidate glomerulus. For each minimum circumcircle, determine the spatial overlap between the minimum circumcircle and the corresponding minimum circumcircles of the other sets of fast and slow microbubble motion trajectories, where, When the spatial overlap state is spatial correlation overlap, the candidate glomerular cluster corresponding to the current smallest circumcircle is configured with other candidate glomerular clusters that satisfy spatial correlation overlap as glomerular spatial overlap clusters. When each candidate glomerular cluster within the overlapping glomerular space has consistent characteristics, the smallest circumcircle configuration corresponding to the candidate glomerular cluster is configured as the target glomerulus. Based on the target glomeruli determined in the above manner, a set of target glomeruli is generated.

[0081] It should be noted that the above-mentioned motion trajectory screening should be performed for each set of fast and slow microbubble motion trajectories. When performing spherical verification screening, the corresponding candidate glomerular cluster sets for all fast and slow microbubble motion trajectory sets should be utilized. Specifically, for each candidate glomerular cluster set, there are several corresponding low-speed microbubble motion trajectories. Subsequently, based on the low-speed microbubble motion trajectories within the candidate glomerular cluster sets, corresponding candidate trajectory images can be generated. Then, within the generated candidate trajectory images, circumcircle fitting should be performed on each candidate glomerular cluster to generate a minimum circumcircle of each candidate glomerulus. The generated minimum circumcircle represents a candidate glomerulus, and the center of the minimum circumcircle represents the center position of the candidate glomerulus. It should be understood that the candidate trajectory image should include multiple candidate trajectory sub-images, and the number of frames of the candidate trajectory sub-images should be consistent with the number of frames in the aforementioned renal ultrasound echo in-phase orthogonal signal group. Furthermore, the circumcircle fitting method can be consistent with existing technologies; the specific fitting method will not be elaborated here.

[0082] As explained above, each minimum circumcircle belongs to a corresponding set of fast and slow microbubble motion trajectories. During sphere verification and screening, the minimum circumcircle belonging to one set of fast and slow microbubble motion trajectories is spatially overlapped with the minimum circumcircles belonging to other sets of fast and slow microbubble motion trajectories. If there are n sets of fast and slow microbubble motion trajectories, during sphere verification and screening, the minimum circumcircle belonging to one set of fast and slow microbubble motion trajectories should be spatially overlapped with all the minimum circumcircles belonging to the other n-1 sets of fast and slow microbubble motion trajectories. Specifically, when there exists a spatially related minimum circumcircle in each of the minimum circumcircles belonging to the other n-1 sets of fast and slow microbubble motion trajectories, the spatial overlap state of the current minimum circumcircle is spatially related and coincident. Otherwise, the spatial overlap state should be non-spatially related and coincident. Here, "there exists a spatially related minimum circumcircle in each of the other n-1 sets of fast and slow microbubble motion trajectories" specifically means that there should be n-1 minimum circumcircles, and each minimum circumcircle satisfies spatial association with the current minimum circumcircle, and the n-1 minimum circumcircles belong to the n-1 sets of fast and slow microbubble motion trajectories respectively.

[0083] In one embodiment of the present invention, determining the spatial overlap state includes: Calculate the positional overlap and structural overlap rate between a minimum circumcircle and the minimum circumcircles belonging to the trajectories of other fast and slow pairs of microbubbles. Positional overlap is the distance between the centers of the smallest circumcircles; The structural overlap rate is the proportion of the area overlap between the smallest circumcircles. When there exists a minimum circumcircle within each of the sets of trajectories of other fast and slow microbubbles that is spatially correlated with the current minimum circumcircle, then the spatial overlap state of the current minimum circumcircle is spatially correlated and coincident. When two smallest circumcircles satisfy spatial correlation, then at least the following conditions must be met: the degree of positional overlap matches the degree of positional overlap threshold, and the structural overlap rate matches the overlap rate threshold.

[0084] It should be noted that when determining the spatial overlap state, it should also be within the space where the aforementioned candidate trajectory images are located. The aforementioned "one smallest circumscribed circle" is the current smallest circumscribed circle, and the "other circumscribed circles" are the smallest circumscribed circles that belong to different sets of fast and slow bubble motion trajectories from the current smallest circumscribed circle. In addition, the "smallest circumscribed circles belonging to other fast and slow bubble motion trajectories" should include all the smallest circumscribed circles corresponding to other sets of fast and slow bubble motion trajectories, as mentioned above, all the smallest circumscribed circles belonging to the other n-1 sets of fast and slow bubble motion trajectories.

[0085] In practice, after selecting other minimum circumcircles corresponding to the current minimum circumcircle, the positional overlap and structural overlap rate between the two minimum circumcircles should be calculated. When calculating the positional overlap, the distance between the corresponding centers of the two minimum circumcircles should be calculated first. When calculating the structural overlap rate, the sum of the areas of the two minimum circumcircles is calculated, and the area of ​​the overlapping portion is divided by the sum of the areas of the two minimum circumcircles; the quotient is taken as the structural overlap rate. When the positional overlap matches a positional overlap threshold, and the structural overlap rate matches an overlap rate threshold, then the two minimum circumcircles satisfy spatial association. The meaning of spatial association can be found in the above explanation.

[0086] In practice, when determining whether spatial correlation is satisfied, a position overlap threshold and an overlap rate threshold should be set. Generally, the position overlap threshold and overlap rate threshold can be selected as needed. If higher accuracy is required, a smaller position overlap threshold (100 micrometers) and a higher overlap rate threshold (80%) can be set. If the accuracy requirement is not so high, a larger position overlap threshold (200 micrometers) and a lower overlap rate threshold (40%) can be set.

[0087] When the spatial overlap state corresponding to a current minimum circumcircle is spatially correlated overlap, the candidate glomerular cluster corresponding to the current minimum circumcircle should be configured with other candidate glomerular clusters that satisfy spatial correlation overlap as a spatially overlapped glomerular cluster. To further improve the recognition accuracy of glomeruli, feature consistency verification should also be performed.

[0088] In one embodiment of the present invention, feature consistency verification is performed on each candidate glomerular cluster within the spatial overlap cluster to determine the corresponding feature consistency state of each candidate glomerular cluster within the spatial overlap cluster, wherein, When performing consistency feature verification, the following are included: Determine the glomerular motion characteristics and glomerular morphology characteristics of each candidate glomerular cluster; When the glomerular motion characteristics and glomerular morphology characteristics of all candidate glomerular clusters meet the similarity condition, then each candidate glomerular cluster within the spatially overlapping glomerular cluster has characteristic consistency.

[0089] Specifically, the spherical motion characteristics of each candidate glomerular cluster may include the spherical velocity characteristics corresponding to each candidate glomerulus, which may include the average velocity or the median velocity; the spherical morphological characteristics of each candidate glomerular cluster may include the spherical radius corresponding to each candidate glomerulus. When the spherical motion characteristics and spherical morphological characteristics of all candidate glomerular clusters meet the similarity condition, specifically meaning that the spherical motion characteristics and spherical morphological characteristics meet the similarity condition, specifically, the spherical velocity characteristics are within the same velocity range, such as 10-30 micrometers per second; and the spherical morphological characteristics meet the similarity condition, specifically meaning that the spherical radii are within the same value range, such as 20-40 micrometers per second. It should be understood that the spherical radius of each candidate glomerulus can be calculated and determined using existing commonly used methods, the specific calculation methods of which will not be elaborated here.

[0090] Only candidate glomeruli that pass the feature consistency check are considered target glomeruli. Based on all identified target glomeruli, a target glomeruli set can be generated. Subsequently, super-resolution microscopic imaging can be performed based on this target glomeruli set. As can be seen from the above description of the method and process for identifying target glomeruli, this invention possesses strong anti-interference capabilities and significant artifact suppression, thereby significantly improving the robustness, specificity, and quantification ability of glomeruli identification.

[0091] In one embodiment of the present invention, when performing super-resolution microscopy, the basic imaging features of each glomerulus should first be extracted. These basic imaging features of the target glomerulus include local blood flow velocity distribution characteristics, mean transit time, and perfusion intensity. These characteristics can be extracted using existing commonly used techniques. Specifically, the local blood flow velocity distribution characteristics include the maximum, minimum, average, and standard deviation of the velocity distribution. The mean transit time specifically refers to the average residence time of multiple trajectories constituting the glomerulus. The perfusion intensity specifically refers to the number of microvessels entering the glomerulus, generally the product of the number of low-speed trajectories constituting the target glomerulus and the average velocity.

[0092] In one embodiment of the present invention, when performing super-resolution microscopy of glomeruli based on the basic imaging features of all target glomeruli, the basic imaging features of each target glomeruli should be mapped to a spatial location. Subsequently, glomerular density distribution map, mean blood flow velocity map and perfusion heterogeneity map of the renal cortex are generated by interpolation rendering.

[0093] It should be noted that the mapped spatial location refers to the distribution location within the corresponding image after imaging based on the target data of renal ultrasound echo. After mapping to the spatial location, an average blood flow velocity map can be generated after interpolation rendering based on the local blood flow velocity distribution characteristics of all target glomeruli. Similarly, a perfusion heterogeneity map can be generated after interpolation rendering based on the perfusion intensity of all target glomeruli. Based on the positional distribution information of all target glomeruli, a glomerular density distribution map of the renal cortex can be generated after interpolation rendering. The specific interpolation rendering method used during imaging can be consistent with existing technologies, and the specific interpolation rendering method will not be detailed here. By using the generated glomerular density distribution map, average blood flow velocity map, and perfusion heterogeneity map, the functional gradient and spatial variation of the microcirculation in the cortex can be intuitively displayed, achieving precise fusion of structural and functional information, and providing key imaging evidence for the early identification and quantitative assessment of renal pathological states.

[0094] Based on the above description, the present invention can also provide a glomerular super-resolution imaging system, specifically including an imaging processing terminal, wherein the imaging processing terminal performs glomerular super-resolution microscopy imaging on any kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology using the method described above.

[0095] In practice, the imaging processing terminal can use existing commonly used terminal equipment, with the specific type of terminal equipment being determined by its ability to process ultrasound data. After acquiring the target data of the kidney ultrasound echo, the target data of the kidney ultrasound echo is loaded into the imaging processing terminal. The imaging processing terminal performs glomerular super-resolution microscopy imaging using the method described above. The specific method and process of super-resolution microscopy imaging can be found in the corresponding descriptions above, and will not be repeated here.

Claims

1. A method of glomerular super-resolution microscopic imaging, characterized by, The super-resolution microscopy method includes: Acquire kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, wherein the kidney ultrasound echo target data includes several frames of kidney ultrasound echo in-phase orthogonal signal groups that are time-correlated; The target data of renal ultrasound echo were separated into microbubble motion patterns based on multiple sets of spatiotemporal filtering thresholds to generate several sets of fast and slow microbubble signals. Each set of fast and slow microbubble signals corresponds to a set of spatiotemporal filtering thresholds, and each set of fast and slow microbubble signals includes a subset of fast microbubble signals and a subset of slow microbubble signals. Microbubble trajectory generation processing is performed on all fast and slow microbubble signal sets to generate fast and slow microbubble motion trajectory sets corresponding to each fast and slow microbubble signal set. The fast and slow microbubble motion trajectory sets include fast microbubble motion trajectory sets corresponding to fast microbubble signal subsets and slow microbubble motion trajectory sets corresponding to slow microbubble signal subsets. For each set of fast and slow microbubble motion trajectories, motion trajectory screening is performed to generate several candidate glomerular cluster sets that initially match the glomerular trajectory features; thereafter, at least based on spatial overlap, all candidate glomerular cluster sets are screened for glomerular body verification to generate the corresponding target glomerular set. The basic imaging features of each target glomerulus within the target glomerulus set are extracted, and glomerular super-resolution microscopy is performed based on the basic imaging features of all target glomerulus.

2. The glomerular super-resolution microscopy method of claim 1, wherein, When filtering the motion trajectories for each set of fast and slow microbubble motion trajectories, the following steps are included: The set of fast and slow microbubble motion trajectories is subjected to primary dynamic coarse screening to generate a set of low-speed microbubble motion trajectories. The trajectory feature information of each low-speed microbubble motion trajectory within the set of low-speed microbubble motion trajectories is extracted, wherein the trajectory feature information includes at least kinematic features, morphological features, and spatiotemporal features; Based on the trajectory feature information of each low-speed microbubble motion trajectory, unsupervised clustering is performed on all low-speed microbubble motion trajectories to generate a low-speed motion trajectory core point cluster containing several low-speed microbubble motion trajectories. When the cluster features of a low-speed motion trajectory core point cluster are identified as glomerular feature clusters, the corresponding low-speed motion trajectory core point clusters are configured as candidate glomerular clusters, and a candidate glomerular cluster set is formed based on all candidate glomerular clusters.

3. The glomerular super-resolution microscopy method of claim 2, wherein, When performing a preliminary dynamic coarse-grained screening of the set of fast and slow microbubble motion trajectories, the following are included: Determine the set of microbubble velocity information corresponding to the set of fast and slow microbubble motion trajectories, wherein the set of microbubble velocity information includes several microbubble velocities; The velocity distribution of microbubble velocities within the statistical set of microbubble velocity information is analyzed, and coarse velocity threshold information is generated based on the statistical velocity distribution. Calculate the trajectory velocity characteristics of each fast microbubble motion trajectory and slow microbubble motion trajectory within the set of fast and slow microbubble motion trajectories; When a trajectory velocity feature matches the coarse velocity threshold information, the fast microbubble motion trajectory or slow microbubble motion trajectory corresponding to the trajectory velocity feature is identified as a low-speed microbubble motion trajectory. A set of low-speed microbubble motion trajectories is formed based on all the low-speed microbubble motion trajectories.

4. The glomerular super-resolution microscopy method of claim 3, wherein, When generating coarse velocity threshold information, the following are included: Based on the velocities of all microbubbles within the set of microbubble velocity information, a velocity distribution histogram is constructed for the set of fast and slow microbubble motion trajectories. On the constructed velocity distribution histogram, the corresponding main velocity distribution peak, secondary velocity distribution peak, and velocity distribution trough region located between the main velocity distribution peak and the secondary velocity distribution peak are determined; Velocity clustering is performed on all microbubble velocities within the velocity distribution trough region to generate coarse velocity threshold information after clustering. When performing speed clustering, the number of clusters is set to 2. When the number of clusters is 2, the coarse speed threshold information includes a first coarse speed threshold and a second coarse speed threshold. When the first coarse velocity threshold is less than the second coarse velocity threshold, the trajectory velocity feature is not greater than the first coarse velocity threshold, and the trajectory velocity feature matches the coarse velocity threshold information.

5. The glomerular super-resolution microscopy method of claim 2, wherein, When generating the core point cluster of low-speed motion trajectory, the following is included: Construct a high-dimensional orthogonal metric space corresponding to trajectory feature information; Based on the trajectory feature information corresponding to each low-speed microbubble motion trajectory, all low-speed microbubble motion trajectories are distributed in the high-dimensional orthogonal metric space. In a high-dimensional orthogonal metric space, density-based unsupervised clustering is performed on all low-speed microbubble motion trajectories to obtain core point clusters at least in the high-dimensional orthogonal metric space. The core point clusters include several low-speed motion trajectory core point clusters, and each low-speed motion trajectory core point cluster includes several low-speed microbubble motion trajectories. For each low-speed motion trajectory core point cluster, the corresponding point cluster trajectory features of all low-speed microbubble motion trajectories within the low-speed motion trajectory core point cluster are statistically analyzed. Based on the statistically analyzed point cluster trajectory features, the cluster type features corresponding to the low-speed motion trajectory core point cluster are identified. When the point cluster trajectory features have the characteristics of high curvature and long dwell time, the cluster features of the core point cluster of the low-speed motion trajectory are identified as glomerular feature clusters.

6. The glomerular super-resolution microscopic imaging method according to claim 5, characterized in that, When performing density-based unsupervised clustering on all low-velocity microbubble trajectories, including: Configure unsupervised classification conditions, wherein the unsupervised classification conditions include cluster neighborhood radius and cluster point number threshold. Based on the unsupervised classification conditions configured above, each low-speed microbubble motion trajectory is identified as a core point, boundary point, or noise point. Starting from any unvisited core point, based on the rules of density reachability and density connectivity, all core points and boundary points that satisfy density reachability are clustered into the same cluster, forming the corresponding low-speed motion trajectory core point cluster.

7. The glomerular super-resolution microscopic imaging method according to any one of claims 1 to 6, characterized in that, When performing sphericity verification screening on all candidate glomerular cluster sets, the following is included: For each candidate glomerular cluster in the candidate glomerular cluster set, generate the corresponding minimum circumcircle of each candidate glomerular cluster, so as to use the minimum circumcircle to represent a candidate glomerulus, and the center of the minimum circumcircle represents the center position of the candidate glomerulus. For each minimum circumcircle, determine the spatial overlap between the minimum circumcircle and the corresponding minimum circumcircles of the other sets of fast and slow microbubble motion trajectories, where, When the spatial overlap state is spatial correlation overlap, the candidate glomerular cluster corresponding to the current smallest circumcircle is configured with other candidate glomerular clusters that satisfy spatial correlation overlap as glomerular spatial overlap clusters. When each candidate glomerular cluster within the overlapping glomerular space has consistent characteristics, the minimum circumcircle configuration corresponding to the candidate glomerular cluster is used as the target glomerulus. Based on the target glomeruli determined in the above manner, a set of target glomeruli is generated.

8. The glomerular super-resolution microscopic imaging method according to claim 7, characterized in that, Determining the spatial overlap status includes: Calculate the positional coincidence and structural overlap rate between a minimum circumcircle and other minimum circumcircles, where, Positional overlap is the distance between the centers of the smallest circumcircles; The structural overlap rate is the proportion of the area overlap between the smallest circumcircles. When the positional overlap degree matches the positional overlap degree threshold, and the structural overlap rate matches the overlap rate threshold, the corresponding spatial overlap degree state is spatial correlation overlap.

9. The glomerular super-resolution microscopic imaging method according to claim 7, characterized in that, For each candidate glomerular cluster within the spatial overlap of glomerular clusters, a feature consistency check is performed to determine the corresponding feature consistency state of each candidate glomerular cluster within the spatial overlap of glomerular clusters. When performing consistency feature verification, the following are included: Determine the glomerular motion characteristics and glomerular morphology characteristics of each candidate glomerular cluster; When the glomerular motion characteristics and glomerular morphology characteristics of all candidate glomerular clusters meet the similarity condition, then each candidate glomerular cluster within the spatially overlapping glomerular cluster has characteristic consistency.

10. The glomerular super-resolution microscopic imaging method according to any one of claims 1 to 6, characterized in that, The basic imaging features of the target glomerulus include local blood flow velocity distribution characteristics, mean transit time, and perfusion intensity; When performing glomerular super-resolution microscopy based on the fundamental imaging features of all target glomeruli, including: The basic imaging features of each target glomerulus are mapped to a spatial location. Subsequently, glomerular density distribution map, mean blood flow velocity map, and perfusion heterogeneity map of the renal cortex are generated by interpolation rendering.

11. The glomerular super-resolution microscopic imaging method according to any one of claims 1 to 6, characterized in that, When acquiring kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, the following steps are included: Acquire renal ultrafast ultrasound contrast imaging data generated based on ultrafast plane wave imaging technology, wherein the renal ultrafast ultrasound contrast imaging data includes several frames of renal ultrasound contrast echo data generated sequentially along the time dimension; Each frame of renal ultrasound contrast imaging data was demodulated to generate a set of renal ultrasound echo in-phase orthogonal signals corresponding to each frame of renal ultrasound contrast imaging data. Kidney ultrasound echo target data is generated based on all in-phase orthogonal signal groups of kidney ultrasound echoes.

12. A glomerular super-resolution imaging system, characterized in that, Including the imaging processing terminal, among which, For any kidney ultrasound echo target data generated based on ultrafast plane wave imaging technology, the imaging processing terminal performs glomerular super-resolution microscopy imaging using the method described in any one of claims 1 to 11.

Images