EIT region segmentation and quantitative analysis method and system based on ultrasonic dissection calibration

Abstract

The invention discloses an EIT region segmentation and quantitative analysis method and system based on ultrasonic anatomy calibration, and belongs to the field of medical image and respiratory physiology monitoring. Processing the ultrasound image to identify anatomical markers such as a rib line, a pleural line and a diaphragm line; performing multi-modal image registration based on the physical coordinates and the common anatomical anchor points; mapping the anatomical landmark to an EIT grid; performing hierarchical constraint region segmentation on the EIT grid; and calculating an impedance average value and a tidal impedance variation based on a segmentation result. According to the method, individualized accurate region segmentation of the EIT image is achieved through ultrasonic anatomical structure calibration, data pollution is avoided, the EIT can provide a quantitative report with a clear anatomical identifier, and the application value and reliability of the EIT in clinical scenes such as pulmonary ventilation efficiency evaluation, pulmonary edema monitoring and pleural effusion quantification are greatly improved.

Description

Technical Field

[0001] This application relates to the field of medical imaging and respiratory physiological monitoring, specifically to a method and system for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration. Background Technology

[0002] Electrical Impedance Tomography (EIT), a non-invasive, real-time lung function monitoring technology, has been widely used in respiratory physiology monitoring, lung ventilation distribution assessment, and dynamic tracking of lung diseases due to its advantages such as being radiation-free, bedside operation, and high temporal resolution. EIT measures changes in thoracic impedance using a surface electrode array to reconstruct images of lung ventilation function, providing continuous and dynamic lung function information for clinical practice. However, a core limitation of EIT technology lies in its significantly low image resolution, typically consisting of only tens to hundreds of pixels. This results in images that cannot clearly represent key anatomical boundaries of the lungs, such as interlobar fissures, the outline of the heart, the position of the diaphragm, and the pleural line. This physical characteristic poses a fundamental challenge to EIT in clinical quantitative analysis: physicians cannot accurately delineate regions of interest (ROIs) with well-defined anatomical locations based on EIT images, thus hindering the precise quantification of lung function.

[0003] In existing technologies, region segmentation and quantitative analysis of EIT images mainly rely on two methods: one is the simple geometric partitioning method, such as mechanically dividing the thoracic cavity into four quadrants or upper and lower parts. This method completely ignores individual anatomical differences and changes in body position, resulting in a significant deviation between the segmentation results and the actual lung structure. The other is the operator's manual delineation method, where doctors subjectively delineate ROIs on EIT images. This method not only highly relies on the operator's experience, easily introducing subjective errors and repeatability differences, but also cannot adapt to changes in patient position (such as changing from supine to lateral) or pathological changes (such as changes in lung morphology due to pleural effusion). More importantly, both of these methods completely ignore the patient's individual anatomical characteristics, such as rib morphology, diaphragm position, heart size, and other anatomical variations, leading to systematic biases in the quantitative analysis results. Specifically, impedance signals from non-lung tissues (such as the heart, ribs, and abdominal tissues) are incorrectly included in the lung function area, or some lung tissues (such as the lower lobe) are omitted, causing data contamination problems. Clinical studies have shown that this imprecise regional definition significantly reduces the correlation between quantitative parameters of EIT (such as tidal impedance change ΔZ) and the gold standard (such as CT) (the correlation coefficient of the traditional method is approximately -0.63), which severely restricts the credibility and application value of EIT in key clinical scenarios such as lung ventilation efficiency assessment, pulmonary edema monitoring, and pleural effusion quantification.

[0004] Therefore, there is an urgent need for a quantitative analysis method for EIT that can integrate individualized anatomical information, adapt to changes in body position, and achieve precise ROI definition, in order to overcome the current technical bottlenecks and improve the practicality and reliability of EIT in clinical pulmonary function assessment. Summary of the Invention

[0005] This application provides a method and system for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration, which can solve the problems of low EIT image resolution in the prior art, which makes it impossible to accurately define the anatomical region of the lung and achieve individualized quantitative analysis.

[0006] In a first aspect, embodiments of this application provide a method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration, wherein the method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration includes: Simultaneous acquisition of electrical impedance tomography data and lung ultrasound images; The ultrasound images were processed to identify anatomical landmarks; Initial affine registration is performed based on physical coordinates, and elastic registration is implemented through shared anatomical anchor points; The anatomical landmarks are mapped onto an electrical impedance tomography (EIT) grid; hierarchical constrained region segmentation is performed on the grid; and quantitative parameters are calculated based on the segmentation results.

[0007] In conjunction with the first aspect, in one implementation, the rib line is identified by Hough transform or U-Net, the pleural line is identified by phase symmetry filter or linear structure deep learning model, and the diaphragm line is identified by dynamic programming.

[0008] In conjunction with the first aspect, in one embodiment, the common anatomical anchor points include the center point of the second intercostal space and the apex of the diaphragm; the center point of the second intercostal space is extracted based on the parasternal line, and the apex of the diaphragm is extracted based on the intersection of the diaphragm and the chest wall.

[0009] In conjunction with the first aspect, in one embodiment, the layered constraints include hard constraints and strong soft constraints; the hard constraints are set at the grid boundaries corresponding to the rib and heart regions; the strong soft constraints are set at the grid boundaries corresponding to the pleural line and diaphragm line.

[0010] In conjunction with the first aspect, in one embodiment, the quantitative parameters include the average impedance and the change in humidity impedance.

[0011] In conjunction with the first aspect, in one implementation, the cost of the strong soft constraint is 100 times the cost of the ordinary boundary.

[0012] Secondly, embodiments of this application provide an EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration, the system comprising: The multimodal data synchronous acquisition module is used to simultaneously acquire electrical impedance tomography data and lung ultrasound images; An ultrasound image processing and anatomical landmark recognition module is used to process the ultrasound images to identify the rib line, pleural line and diaphragmatic line; The multimodal image registration and mapping module is used to perform initial affine registration based on physical coordinates, implement elastic registration through shared anatomical anchor points, and map anatomical landmarks to electrical impedance tomography grids. The EIT image anatomical region segmentation module is used to perform hierarchical constrained region segmentation on the grid. The precise quantitative analysis module is used to calculate the average impedance and the change in humid impedance based on the segmentation results.

[0013] In conjunction with the second aspect, in one embodiment, the ultrasound image processing and anatomical landmark recognition module includes: a submodule for recognizing rib lines using Hough transform or U-Net; a submodule for recognizing pleural lines using phase symmetry filters or linear structure deep learning models; and a submodule for recognizing diaphragmatic lines using dynamic programming.

[0014] In conjunction with the second aspect, in one embodiment, the common anatomical anchor points include the center point of the second intercostal space and the center point of the fourth intercostal space; the center point of the second intercostal space is extracted based on the parasternal line, and the apex of the diaphragm is extracted based on the intersection of the diaphragm and the chest wall.

[0015] In conjunction with the second aspect, in one embodiment, the layered constraints include hard constraints and strong soft constraints; the hard constraints are set at the grid boundaries corresponding to the rib and heart regions; the strong soft constraints are set at the grid boundaries corresponding to the pleural line and diaphragm line.

[0016] The beneficial effects of the technical solutions provided in this application include: By simultaneously acquiring EIT impedance data and lung ultrasound images, anatomical landmarks such as rib lines, pleural lines, and diaphragmatic lines are identified. Based on physical coordinates and common anatomical anchor points, elastic registration is performed (using a thin-plate spline TPS model). Ultrasound anatomical structure information is transformed into spatial constraints for EIT image segmentation, realizing automatic ROI segmentation and reliable quantitative analysis. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the EIT region segmentation and quantitative analysis method based on ultrasound anatomical calibration of this application. Figure 2 This is a schematic diagram of the functional modules of an embodiment of the EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0020] In a first aspect, embodiments of this application provide a method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration.

[0021] In one embodiment, reference is made to Figure 1 , Figure 1 This is a schematic flowchart of an embodiment of the EIT region segmentation and quantitative analysis method based on ultrasound anatomical calibration according to this application. Figure 1 As shown, the methods for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration include: Step S1: Simultaneously acquire electrical impedance tomography (EIT) data and lung ultrasound images.

[0022] Step S2: Process ultrasound images to identify anatomical landmarks.

[0023] Step S3: Perform initial affine registration based on physical coordinates, and then perform elastic registration through shared anatomical anchor points.

[0024] Step S4: Map the anatomical landmarks onto the electrical impedance tomography grid.

[0025] Step S5: Perform hierarchical constraint region segmentation on the mesh.

[0026] Step S6: Calculate quantitative parameters based on the segmentation results.

[0027] In this embodiment, an EIT image region segmentation and quantitative analysis method based on ultrasound anatomical structure calibration solves the technical problem in related technologies where low EIT image resolution leads to the inability to accurately define lung anatomical regions and achieve individualized quantitative analysis. This method simultaneously acquires EIT impedance data and lung ultrasound images, identifies anatomical landmarks such as rib lines, pleural lines, and diaphragmatic lines, and performs elastic registration based on physical coordinates and common anatomical anchor points (using a thin-plate spline TPS model). This transforms ultrasound anatomical structure information into spatial constraints for EIT image segmentation, achieving automatic ROI segmentation and reliable quantitative analysis.

[0028] Specifically, by using hard and soft constraints along the rib line, pleural line, and diaphragm line, the segmentation boundary is ensured to strictly conform to the anatomical structure, avoiding signal contamination from non-lung tissues, including the heart and rib areas. This allows EIT images to accurately define anatomical regions such as the left and right lungs and the upper and lower lobes of the lungs, thus achieving anatomical localization.

[0029] Based on real-time ultrasound, the system acquires individualized anatomical features, including the location of the center point of the second intercostal space and the apex of the diaphragm. The system automatically adapts to the patient's anatomical differences and positional changes, including changing from supine to lateral position, eliminating the shortcomings of the traditional geometric zoning method that ignores individual differences and achieving individualized self-adaptation.

[0030] Without the need for manual delineation, the ROI is automatically segmented through a layered map cutting model, and the average impedance and tidal impedance changes are calculated. This provides a quantitative report with clear anatomical markers (e.g., a 30% decrease in ventilation in the posterior segment of the left lower lobe), achieving fully automated quantitative analysis.

[0031] In a preclinical trial involving 20 patients with pleural effusion, the correlation coefficient between the effusion index calculated by EIT and the effusion volume measured by CT was -0.91, which was significantly better than the correlation coefficient of -0.63 of the traditional geometric partitioning method. This significantly improved the application value and reliability of EIT technology in key clinical scenarios such as lung ventilation efficiency assessment, pulmonary edema monitoring, and pleural effusion quantification.

[0032] In one specific embodiment, the synchronous acquisition step includes using a flexible patch integrating an EIT electrode array and an ultrasound transducer array to simultaneously acquire EIT impedance data and lung ultrasound image data of the target lung region. The EIT electrode array comprises no fewer than 16 electrodes for exciting and measuring body surface potentials. The ultrasound transducer array operates in B-mode with a frequency range of 2-5 MHz to generate two-dimensional ultrasound images displaying the ribs, pleural line, and diaphragm line.

[0033] The ultrasound anatomical recognition step includes processing the acquired ultrasound images and automatically identifying and extracting key anatomical landmarks, including the rib line, pleural line, and diaphragmatic line. Specifically, the rib line is detected using a Hough transform or a convolutional neural network model with a U-Net structure; the pleural line is tracked within the intercostal region using a phase-symmetric filter; and the diaphragmatic line is identified in the basal region of the lung using a dynamic programming algorithm.

[0034] The high-precision elastic registration and mapping steps include performing initial affine registration based on the preset physical coordinates of the EIT electrodes and the ultrasound probe on the patch. The center point of the second intercostal space and the apex of the diaphragm are extracted from the ultrasound image as the first anchor point set, and the corresponding functional points are inferred from the EIT impedance image based on impedance distribution characteristics as the second anchor point set. Using the first and second anchor point sets as control points, the thin-plate spline TPS model is solved to obtain the nonlinear spatial mapping relationship. Using the solved elastic transformation, the anatomical landmarks such as the rib lines, pleural lines, and diaphragm lines are accurately mapped to the EIT finite element mesh space.

[0035] The hierarchical constraint segmentation step involves constructing a graph model on the EIT finite element mesh, where each mesh element serves as a node and the connections between adjacent mesh elements serve as edges. Edges corresponding to the mapped ribs and heart region are set as hard constraints, with a segmentation cost set to infinity. Edges corresponding to the mapped pleural line and diaphragm line are set as strong soft constraints, with a segmentation cost set to 100 times the ordinary boundary cost. Region cost is set based on the impedance value of each mesh element. A maximum flow / minimum cut algorithm is used for optimization, automatically generating multiple regions of interest (ROIs) conforming to individualized anatomical structures, including the left lung, right lung, upper lobe, and lower lobe.

[0036] The precise quantitative calculation steps include calculating the average impedance, impedance change curve over time, and tidal impedance change ΔZ within each ROI based on the ROI obtained from the above hierarchical constraint segmentation steps.

[0037] The results presentation and reporting steps include visualizing the quantitative analysis results and generating a structured clinical report containing anatomical region location information. The report associates the quantitative parameters with the corresponding anatomical region names.

[0038] Furthermore, in one embodiment, the rib line is identified by Hough transform or U-Net, the pleural line is identified by phase symmetry filter or linear structure deep learning model, and the diaphragm line is identified by dynamic programming.

[0039] In this embodiment, the rib line is identified by Hough transform, the pleural line by phase symmetry tracking, and the diaphragm line by dynamic programming. This solves the technical problem of large region segmentation errors in EIT images due to inaccurate identification of anatomical landmarks in related technologies. This technical solution provides anatomical boundary information based on elastic registration for subsequent region segmentation by accurately identifying key anatomical landmarks such as the rib line, pleural line, and diaphragm line. This allows EIT images to define lung anatomical regions (such as the left lung, right lung, upper lobe, and lower lobe), avoiding signal contamination from non-lung tissues. This significantly improves the clinical reliability of lung ventilation distribution assessment (correlation with CT reaches r=-0.91), enabling EIT technology to provide quantitative reports with clear anatomical landmarks, such as a 30% decrease in ventilation in the posterior segment of the left lower lobe. Furthermore, in one embodiment, the common anatomical anchor points include the center point of the second intercostal space and the apex of the diaphragm. The center point of the second intercostal space is extracted based on the parasternal line, and the apex of the diaphragm is extracted based on the intersection of the diaphragm and the chest wall.

[0040] In this embodiment, the technical problem of insufficient EIT image registration accuracy in related technologies, which cannot adapt to individualized anatomical differences and changes in body position, is solved by using the center point of the second intercostal space (extracted based on the parasternal line) and the apex of the diaphragm (extracted based on the intersection of the diaphragm and the chest wall). This technical solution establishes an elastic registration basis (using a thin-plate spline TPS model) by accurately extracting the center point of the second intercostal space and the apex of the diaphragm, enabling EIT images to accurately reflect individualized anatomical structures, automatically adapt to changes in body position, and significantly improve clinical reliability (correlation with CT reaches r = -0.91). This allows EIT technology to provide quantitative reports with clear anatomical markers, such as a 30% decrease in ventilation in the posterior segment of the left lower lobe, and other clinically practical information. Furthermore, in one embodiment, the above-mentioned layered constraints include hard constraints and strong soft constraints. The hard constraints are set at the grid boundaries corresponding to the rib and cardiac regions. The strong soft constraints are set at the grid boundaries corresponding to the pleural line and the diaphragm line.

[0041] In this embodiment, by setting hard constraints in the rib and heart regions and strong soft constraints in the pleural and diaphragmatic lines, the technical problems of data contamination and unreliable quantitative analysis caused by inaccurate segmentation boundaries in EIT images in related technologies are solved. This technical solution ensures that the segmentation boundaries strictly conform to the anatomical structure through a layered constraint mechanism, avoiding signal contamination from non-lung tissues. This allows EIT images to accurately define the anatomical regions of the lungs, significantly improving the accuracy and reliability of lung ventilation distribution assessment. It enables EIT technology to provide quantitative reports with clear anatomical markers, such as a 30% decrease in ventilation in the posterior segment of the left lower lobe, providing clinically useful information.

[0042] Furthermore, in one embodiment, the quantitative parameters mentioned above include the average impedance and the change in moisture impedance.

[0043] In this embodiment, by calculating the average impedance and the change in tidal impedance, the technical problem of low clinical application value caused by the lack of clear anatomical landmarks in quantitative analysis of EIT images in related technologies is solved. This technical solution, by providing quantitative parameters with clear anatomical landmarks, enables EIT technology to provide precise clinical information such as a 30% decrease in ventilation in the posterior segment of the left lower lobe, significantly improving its application value and reliability in key clinical scenarios such as lung ventilation efficiency assessment, pulmonary edema monitoring, and pleural effusion quantification.

[0044] Furthermore, in one embodiment, the cost of the aforementioned strong soft constraint is 100 times the cost of a normal boundary. Secondly, embodiments of this application also provide an EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration.

[0045] In one embodiment, reference is made to Figure 2 , Figure 2 This is a functional module diagram of an embodiment of the EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration according to this application. Figure 2 As shown, the EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration includes: The multimodal data synchronous acquisition module 1 is used to synchronously acquire electrical impedance tomography data and lung ultrasound images.

[0046] The ultrasound image processing and anatomical landmark recognition module 2 is used to process the ultrasound images to identify the rib line, pleural line and diaphragm line.

[0047] The multimodal image registration and mapping module 3 is used to perform initial affine registration based on physical coordinates, implement elastic registration through common anatomical anchor points, and map anatomical landmarks to electrical impedance tomography grids. EIT image anatomy region segmentation module 4 is used to perform hierarchical constrained region segmentation on the above grid.

[0048] The precise quantitative analysis module 5 is used to calculate the average impedance and the change in moisture impedance based on the segmentation results.

[0049] The aforementioned ultrasound image processing and anatomical landmark recognition module 2 includes: a submodule for recognizing rib lines using Hough transform; a submodule for recognizing pleural lines using phase symmetry tracing; and a submodule for recognizing diaphragmatic lines using dynamic programming.

[0050] In this embodiment, an EIT image region segmentation and quantitative analysis method based on ultrasound anatomical structure calibration solves the technical problem in related technologies where low EIT image resolution leads to the inability to accurately define lung anatomical regions and achieve individualized quantitative analysis. This method simultaneously acquires EIT impedance data and lung ultrasound images, identifies anatomical landmarks such as rib lines, pleural lines, and diaphragmatic lines, and performs high-precision multimodal registration based on physical coordinates and shared anatomical anchor points. This transforms ultrasound anatomical structure information into high-precision spatial constraints for EIT image segmentation, achieving automatic and accurate ROI segmentation and reliable quantitative analysis.

[0051] Furthermore, in one embodiment, the multimodal data synchronous acquisition module 1 is used to simultaneously acquire EIT impedance data and lung ultrasound image data of the same patient's thoracic region via a flexible patch integrating an EIT electrode array and an ultrasound transducer array. The aforementioned EIT electrode array includes no fewer than 16 electrodes for exciting and measuring body surface potentials. The aforementioned ultrasound transducer array operates in B-mode with a frequency range of 2-5 MHz, used to generate two-dimensional ultrasound images displaying the ribs, pleural line, and diaphragm line.

[0052] Furthermore, in one embodiment, the ultrasound image processing and anatomical landmark recognition module 2 includes: The rib recognition unit is used to automatically detect and locate the arc-shaped upper edge of the rib shadow after preprocessing the ultrasound image, using a convolutional neural network model with Hough transform or U-Net structure.

[0053] The pleural line recognition and tracking unit is used to enhance and identify horizontal pleural line features in the intercostal region between adjacent ribs using a phase symmetry filter, and to continuously track them to determine the upper boundary of the lung.

[0054] The diaphragm line recognition unit is used to identify the strong echo interface formed by lung tissue and liver / spleen tissue in the base region of the lung, and to use a dynamic programming algorithm to find the globally optimal smooth curve and identify the curve as the diaphragm line, so as to determine the lower boundary of the lung and the separation line between the left and right lungs.

[0055] In this embodiment, the rib recognition unit accurately and automatically identifies the location of ribs, providing crucial anatomical boundary references for EIT image segmentation. This avoids misidentifying rib areas as lung tissue, significantly reducing data contamination and enabling EIT images to precisely distinguish between lung and rib regions, thus improving the accuracy of lung function assessment. The pleural line recognition and tracking unit accurately identifies the upper boundary of the lungs. By continuously tracking the pleural line, it provides a reliable basis for segmenting the upper and lower lobes of the lungs, ensuring that EIT images accurately reflect the anatomical structure of the lungs and avoiding misidentification of other areas of the thoracic cavity as lung tissue, greatly improving the reliability of lung ventilation distribution assessment. The diaphragm line recognition unit accurately determines the lower boundary of the lungs and the separation line between the left and right lungs. By identifying the strong echogenic interface between the lungs and the liver / spleen, it provides crucial anatomical references for EIT images, enabling the system to automatically distinguish between the left and right lungs and the upper and lower lobes, achieving precise segmentation of the lung regions and providing clinical quantitative analysis results with clear anatomical localization.

[0056] Furthermore, in one embodiment, the multimodal image registration and mapping module 3 includes: A high-precision elastic registration unit is used to establish a high-precision spatial mapping relationship from the ultrasound image coordinate system to the EIT finite element mesh coordinate system. The aforementioned high-precision elastic registration unit includes: The initial affine registration sub-unit is used to calculate the global affine transformation matrix based on the preset physical coordinates of the EIT electrode and the ultrasonic probe on the patch.

[0057] There is a subunit for extracting anatomical anchor points, which is used to extract the center point of the second intercostal space and the apex of the diaphragm from the ultrasound image as the first anchor point set, and infer the corresponding functional landmark points from the EIT image based on the impedance distribution characteristics as the second anchor point set.

[0058] The elastic transformation solution sub-unit is used to solve the elastic transformation based on the thin plate spline model, using the first and second anchor point sets mentioned above as control points, to compensate for the nonlinear spatial distortion caused by the curvature of the human chest wall, soft tissue deformation, and differences in patch fit.

[0059] The anatomical constraint mapping unit is used to accurately map anatomical structural information such as ribs, pleural lines, and diaphragm lines from the pixel space of the ultrasound image to the finite element mesh space of the EIT using the final transformation function obtained by the high-precision elastic registration unit mentioned above.

[0060] In this embodiment, the high-precision elastic registration unit accurately compensates for differences in the curvature of the human thoracic cavity, soft tissue deformation, and patch fit, improving the consistency between EIT images and actual anatomical structures. The initial affine registration subunit provides a stable linear mapping basis based on physical coordinates, simplifying elastic registration calculations and ensuring initial registration accuracy. The common anatomical anchor point extraction subunit extracts stable landmarks such as the center point of the second intercostal space and the apex of the diaphragm, achieving individualized functional correspondence and solving the problem that traditional registration cannot adapt to individual differences. The elastic transformation solution subunit corrects nonlinear distortion based on a thin-plate spline model, improving mapping accuracy and making the registration results closer to real anatomy. The anatomical constraint mapping unit accurately maps the ribs, pleural lines, and diaphragm lines to the EIT mesh, providing reliable boundary constraints for segmentation, avoiding non-lung tissue signal contamination, and improving the reliability of quantitative analysis.

[0061] Furthermore, in one embodiment, the EIT image anatomical region segmentation module 4 includes: A hierarchical constraint-based graph cut segmentation element is used to perform region segmentation on an EIT finite element mesh. The aforementioned hierarchical constraint-based graph cut segmentation element includes: Graph construction sub-elements are used to define each element of the EIT finite element mesh as a node of the graph, and the connection between adjacent elements as an edge of the graph.

[0062] Layered constraint settings are used to set sub-elements for hard and strong soft constraints. The hard constraints are set at the mesh edges corresponding to the rib shadows and heart regions, with an infinite segmentation cost. The strong soft constraints are set at the mesh edges corresponding to the pleural lines and diaphragm lines, with a segmentation cost 2-10 times that of ordinary boundary conditions.

[0063] The region item sets sub-cells to calculate the cost of belonging to different region labels based on the impedance value of each grid cell.

[0064] The optimized solution sub-cell is used to minimize the energy of the constructed graph using the maximum flow / minimum cut algorithm, and outputs the anatomical region label of each grid cell.

[0065] In this embodiment, the graph construction sub-unit transforms the EIT mesh into a graph model, providing a structural foundation for region segmentation and achieving a precise correspondence between mesh units and segmentation boundaries. The hierarchical constraint setting sub-unit uses hard constraints to ensure that segmentation boundaries do not cross the rib / heart region, while strong soft constraints guide the boundaries to closely adhere to the pleural line / diaphragm line, achieving accurate anatomical segmentation. The region term setting sub-unit calculates the region label cost based on impedance values, ensuring that the segmentation results are consistent with EIT data characteristics and improving segmentation accuracy. The optimization solution sub-unit automatically solves for the optimal segmentation using the maximum flow / minimum cut algorithm, achieving automatic and accurate ROI segmentation, avoiding manual intervention, and improving analysis efficiency and reliability.

[0066] Furthermore, in one embodiment, the precise quantitative analysis module 5 includes: The region parameter calculation unit is used to calculate the average impedance and tidal impedance change of all grid cells within each anatomical ROI output by the EIT image anatomical region segmentation module 4.

[0067] The report generation unit is used to automatically generate structured clinical reports, which associate quantitative parameters with the corresponding anatomical region names.

[0068] In this embodiment, the regional parameter calculation unit provides accurate average impedance and tidal impedance variation, making lung function assessment objective and reliable, and providing key quantitative indicators for clinical practice. The report generation unit automatically generates structured clinical reports, precisely linking quantitative parameters with anatomical regions, enabling physicians to quickly obtain clinical information with clear anatomical localization, improving diagnostic efficiency and accuracy.

[0069] In one specific embodiment, the present invention requires the configuration of hardware devices before implementation, including a data acquisition terminal, a processing unit, an EIT system, and an ultrasound system.

[0070] The data acquisition unit uses a customized flexible patch, which integrates 16 or 32 EIT electrode arrays and a linear ultrasound transducer array (frequency range of 2-5MHz). The flexible patch is made of medical-grade silicone, which has good biocompatibility and fit, and can be attached to the patient's chest area.

[0071] The processing unit uses an embedded GPU or a high-performance industrial computer to run real-time image processing, registration and segmentation algorithms, and the processing speed meets the EIT imaging frame rate requirements.

[0072] The EIT system supports multi-frequency measurements with an imaging frame rate of no less than 10 frames per second, used to capture dynamic impedance changes and provide temporal resolution for lung function monitoring.

[0073] The ultrasound system supports B-mode imaging, which can provide clear two-dimensional ultrasound images of the ribs, pleural line, and diaphragm line, with a frequency range of 2-5MHz, for acquiring high-resolution anatomical information.

[0074] Then, anatomical landmarks are identified, including rib identification, pleural line identification, and diaphragmatic line identification.

[0075] Rib recognition involves preprocessing ultrasound images (including filtering and contrast enhancement) and then using Hough transform or a pre-trained U-Net convolutional neural network model to automatically detect and locate the arc-shaped upper edge of the rib shadow. The aforementioned U-Net network model achieves higher accuracy rib recognition through pixel-level segmentation.

[0076] Pleural line identification involves: enhancing the pleural line features in the intercostal region between adjacent ribs using a phase-symmetric filter, and then extracting them using Radon transform or a linear structure deep learning model to determine the upper boundary of the lung.

[0077] Diaphragm line identification includes: in the basal region of the lungs, by identifying the strong echo interface formed by the lung tissue and the liver / spleen tissue, using a dynamic programming algorithm to find the globally optimal smooth curve, and identifying this curve as the diaphragm line, so as to determine the lower boundary of the lungs and the separation line between the left and right lungs.

[0078] Secondly, high-precision flexible registration and mapping are performed, including initialization, anchor point extraction, flexible transformation, and constraint mapping.

[0079] Initialization includes: during the design and manufacturing stage of the flexible patch, obtaining the precise physical coordinates of each EIT electrode and ultrasonic element in the patch coordinate system using high-precision measuring equipment, and forming a basic coordinate mapping table.

[0080] Anchor point extraction includes: during system operation, automatically extracting at least two stable anatomical anchor points from ultrasound images as the first anchor point set, and simultaneously inferring corresponding functional points from EIT impedance images based on impedance distribution characteristics as the second anchor point set. The aforementioned anatomical anchor points include the center point of the second intercostal space and the apex of the diaphragm.

[0081] The elastic transformation includes: using the aforementioned basic coordinate mapping as initial values, and utilizing the extracted anchor point pairs, calculating the Thin Plate Spline (TPS) transformation model. This TPS model can precisely correct nonlinear spatial distortions caused by the curvature of the human thoracic cavity, soft tissue elasticity, and inconsistent patch fit.

[0082] The constraint mapping includes: using the solved TPS transform, accurately mapping all anatomical landmarks (rib point set, pleural line point set, diaphragm line point set) identified in the ultrasound image to the EIT finite element mesh space. Subsequently, a high-precision elastic registration algorithm is executed. The registration algorithm of this invention includes two stages and aims to solve the problem of nonlinear spatial distortion caused by the curvature of the human thoracic wall, soft tissue elasticity, and inconsistent patch fit.

[0083] Phase 1 is the initial affine registration based on physical coordinates, which includes: establishing the basic mapping model, calculating the initial transformation matrix, and initial registration.

[0084] Establishing the basic mapping model includes: obtaining the precise physical coordinates of each EIT electrode and ultrasonic element in the patch coordinate system using high-precision measurement equipment.

[0085] The calculation of the initial transformation matrix includes: using the coordinate correspondence of all EIT electrodes and ultrasound array elements, calculating the optimal affine transformation matrix Taffine, providing a global, linear mapping basis from the ultrasound image coordinate system to the EIT finite element mesh coordinate system.

[0086] Initial registration includes mapping points PU in the ultrasound image to their initial estimated locations PE_initial=Taffine·PU in the EIT grid.

[0087] Phase two is elastic deformation correction based on common anatomical anchor points, including: selection and extraction of common anatomical anchor points, selection of elastic transformation model, and final registration and mapping.

[0088] The selection and extraction of common anatomical anchor points include: Anchor point A (center of the second intercostal space): The center point AU of the second intercostal space along the parasternal line is extracted by ultrasound, and the impedance mutation area at the upper edge of the heart is inferred by EIT as the corresponding point AE.

[0089] Anchor point B (diaphragm apex): The intersection point BU of the diaphragm and chest wall is extracted by ultrasound, and EIT infers the extreme point of the impedance gradient at the lung-abdomen boundary as the corresponding point BE. The elastic transformation model includes: Model selection: TPS is adopted as the elastic transformation model, and its mapping function F(PU) maps the point PU in the ultrasonic coordinate system to the EIT coordinate system.

[0090] Mathematical expression: F(PU)=A·PU+b+∑wi·φ(‖PU-CUi‖), where φ(r)=r²logr is the TPS kernel function.

[0091] Model solution: Substitute the anchor point pairs (AU,AE), (BU,BE), and (CU,CE) into the model and solve the linear equation system to determine the weights wi of the affine components A and b and the non-rigid components.

[0092] The final registration and mapping process involves applying the solved TPS transformation function F to all anatomical landmarks identified in the ultrasound images to achieve a high-precision, nonlinear mapping of anatomical structures to the EIT grid.

[0093] Finally, EIT image segmentation is performed. The segmentation module of this invention achieves accurate and anatomically consistent region segmentation through a hierarchical constrained graph cut model. This includes: graph construction, energy function and hierarchical constraints, and algorithm execution flow.

[0094] The construction of the graph includes: Node: Each finite element mesh element of the EIT image is defined as a node of the graph.

[0095] Edges: Connections between adjacent grid cells are defined as n-links. Connections between each node and the source and sink nodes are defined as t-links.

[0096] Energy functions and hierarchical constraints include: Region Term: Based on EIT image data, the penalty node is penalized when the label assigned to it does not match its observed impedance value.

[0097] Boundary terms: Hierarchical constraint design for boundary terms: a) Hard constraint: Set the capacity of the n-link edges corresponding to the mapped rib and heart regions to infinity to ensure that the segmentation boundary will never cross these regions.

[0098] b) Strong soft constraint: Set the capacity of the n-link edges corresponding to the mapped pleural line and diaphragm line to a fixed high value C_high (100 times higher than the cost of ordinary boundary) to guide the segmentation boundary to closely fit the anatomical line.

[0099] c) Ordinary image gradient constraints: For mesh edges that do not involve anatomical constraints, their capacity is calculated based on the EIT image gradient.

[0100] The algorithm execution process includes: Input: EIT finite element mesh, anatomical constraint information mapped onto the mesh, and EIT impedance image data.

[0101] Graph construction: Set the weights of nodes, n-links, and t-links.

[0102] Optimization solution: Use a maximum flow / minimum cut algorithm (such as the Boykov-Kolmogorov algorithm) to solve for the minimum cut.

[0103] Output: Anatomical region labels for each grid cell, enabling ROI segmentation that conforms to individualized anatomical structures.

[0104] The system of this invention was evaluated in a preclinical trial involving 20 patients with pleural effusion. The system successfully identified the diaphragmatic line using ultrasound images and accurately segmented the pleural effusion region on EIT images through elastic registration and layered constraint segmentation. Quantitative analysis of this ROI was performed, calculating the mean impedance value and tidal impedance change of the effusion region. Correlation analysis between the effusion index calculated by EIT and the effusion volume measured by the gold standard CT showed a high correlation (Pearson r = -0.91). In contrast, analysis of the same data using a conventional geometric mean-based method showed a correlation of only -0.63 with CT. These experimental data fully demonstrate the superior effectiveness of this invention in improving the accuracy and clinical reliability of quantitative analysis of EIT.

[0105] The functions of each module in the EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration correspond to the steps in the embodiment of the EIT region segmentation and quantitative analysis method based on ultrasound anatomical calibration. Their functions and implementation processes will not be described in detail here.

[0106] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0107] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0108] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0109] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0110] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0111] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0112] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration, characterized in that, The method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration includes: Simultaneous acquisition of electrical impedance tomography data and lung ultrasound images; The ultrasound images were processed to identify anatomical landmarks; Initial affine registration is performed based on physical coordinates, and elastic registration is implemented through shared anatomical anchor points; The anatomical landmarks are mapped onto an electrical impedance tomography (EIT) grid; hierarchical constrained region segmentation is performed on the grid; and quantitative parameters are calculated based on the segmentation results.

2. The method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration as described in claim 1, characterized in that, Rib lines are identified using Hough transform or U-Net, pleural lines are identified using phase symmetry filters or linear structure deep learning models, and diaphragm lines are identified using dynamic programming.

3. The method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration as described in claim 1, characterized in that, The common anatomical anchor points include the center point of the second intercostal space and the apex of the diaphragm; the center point of the second intercostal space is extracted based on the parasternal line, and the apex of the diaphragm is extracted based on the intersection of the diaphragm and the chest wall.

4. The method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration as described in claim 1, characterized in that, The layered constraints include hard constraints and strong soft constraints; the hard constraints are set at the grid boundaries corresponding to the rib and heart regions; the strong soft constraints are set at the grid boundaries corresponding to the pleural line and diaphragm line.

5. The method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration as described in claim 1, characterized in that, The quantitative parameters include the average impedance and the change in moisture impedance.

6. The method for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration as described in claim 4, characterized in that, The cost of the strong soft constraint is 100 times the cost of the ordinary boundary.

7. A system for EIT region segmentation and quantitative analysis based on ultrasound anatomical calibration, characterized in that, The system includes: The multimodal data synchronous acquisition module is used to simultaneously acquire electrical impedance tomography data and lung ultrasound images; An ultrasound image processing and anatomical landmark recognition module is used to process the ultrasound images to identify the rib line, pleural line and diaphragmatic line; The multimodal image registration and mapping module is used to perform initial affine registration based on physical coordinates, implement elastic registration through shared anatomical anchor points, and map anatomical landmarks to electrical impedance tomography grids. The EIT image anatomical region segmentation module is used to perform hierarchical constrained region segmentation on the grid. The precise quantitative analysis module is used to calculate the average impedance and the change in humid impedance based on the segmentation results.

8. The EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration as described in claim 7, characterized in that, The ultrasound image processing and anatomical landmark recognition module includes: a submodule for recognizing rib lines using Hough transform or U-Net; a submodule for recognizing pleural lines using phase symmetry filters or linear structure deep learning models; and a submodule for recognizing diaphragmatic lines using dynamic programming.

9. The EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration as described in claim 7, characterized in that, The common anatomical anchor points include the center point of the second intercostal space and the center point of the fourth intercostal space; the center point of the second intercostal space is extracted based on the parasternal line, and the apex of the diaphragm is extracted based on the intersection of the diaphragm and the chest wall.

10. The EIT region segmentation and quantitative analysis system based on ultrasound anatomical calibration as described in claim 7, characterized in that, The layered constraints include hard constraints and strong soft constraints; the hard constraints are set at the grid boundaries corresponding to the rib and heart regions; the strong soft constraints are set at the grid boundaries corresponding to the pleural line and diaphragm line.

Images