Ultrasonic tumor examination system for obstetrics and gynecology department

Abstract

The invention discloses an ultrasonic tumor examination system for the obstetrics and gynecology department. The ultrasonic tumor examination system comprises a data acquisition module, a data processing module, a data analysis module and a result output module. The data acquisition module synchronously acquires pressure distribution, tumor morphological characteristics and blood flow data; the data processing module maps the pressure field to an ultrasonic image to generate a superposition graph, so that a three-dimensional space model is constructed, and deformation caused by breathing or vascular pulsation is corrected; the data analysis module divides regions according to tumor boundaries determined by pressure, calculates hemodynamic and morphological parameters of each region, and outputs malignancy probability classification through a convolutional neural network; and the result output module generates a space-function thermodynamic diagram and a diagnosis report, and also can form a four-dimensional dynamic report in combination with historical data, thereby improving the tumor diagnosis accuracy, realizing accurate matching of multi-dimensional information, and providing a more comprehensive basis for tumor property judgment.

Description

TECHNICAL FIELD

[0001] The application relates to an ultrasonic examination system, in particular to a gynecological tumor ultrasonic examination system. BACKGROUND

[0002] The gynecological tumor ultrasonic examination is an important clinical technology for assisting doctors in judging the position, size, shape and nature (benign or malignant) of a tumor by acquiring the image and related physiological information of the tumor of the female reproductive system (such as the uterus, ovary, cervix and the like) through an ultrasonic probe. With the advantages of non-invasiveness, real-time and repeatability, the gynecological tumor ultrasonic examination is widely used in tumor screening, diagnosis and efficacy monitoring and is a key link in the gynecological tumor diagnosis and treatment process.

[0003] In the prior art, the gynecological tumor ultrasonic examination mainly relies on B-mode ultrasonic to acquire the morphological features (such as the edge and echo) of a tumor, combines color Doppler ultrasonic to analyze the internal blood flow signals (such as the flow rate and direction) of the tumor, and some systems acquire the hardness information of the tissue through pressure palpation. The examination results are usually presented in the form of two-dimensional images or simple three-dimensional models and a static diagnosis report containing the size and shape description of the tumor is generated.

[0004] The existing gynecological tumor ultrasonic examination technology has a single type of acquired data, mainly relies on single-dimensional ultrasonic image information, is difficult to simultaneously integrate various data reflecting the physical characteristics of the tissue, and leads to the presentation of the final diagnosis result being based on limited information. This single data support mode makes the judgment of the nature of the tumor susceptible to one-sided information and has the possibility of inaccurate results, which is difficult to meet the demand of the clinical diagnosis accuracy. Meanwhile, the transition area between the tumor and the normal tissue is affected by the respiration and blood vessel pulsation, which further leads to the fuzzy division of the tumor boundary and the insufficient boundary positioning accuracy, thereby limiting the diagnosis accuracy and clinical practicability. SUMMARY

[0005] The application overcomes the shortcomings of the prior art and provides a gynecological tumor ultrasonic examination system.

[0006] To achieve the above object, the technical scheme adopted by the application is as follows: a gynecological tumor ultrasonic examination system, comprising a data acquisition module, a data processing module, a data analysis module and a result output module.

[0007] The data acquisition module uses an ultrasonic probe to acquire the pressure distribution in real time during palpation and generates a two-dimensional matrix of the pressure field; acquires the morphological features of the tumor through B-mode ultrasonic and acquires the blood flow velocity and direction data through color Doppler ultrasonic.

[0008] The data processing module maps the two-dimensional matrix of the pressure field to the two-dimensional plane of the ultrasound image to generate a pressure-image overlay; based on the pressure-image overlay, a three-dimensional spatial model of the tumor is generated by a volume rendering algorithm to clearly define the transition area between the tumor and normal tissue; and the three-dimensional spatial model deformation caused by respiration or blood vessel pulsation is corrected in combination with a tissue elastic modulus database.

[0009] The data analysis module divides the tumor into multiple local regions based on the tumor boundary determined by the pressure in the three-dimensional spatial model; for each local region, the hemodynamic parameters and morphological parameters are calculated; and the blood flow parameters and morphological parameters of the local region are input into a convolutional neural network to output a malignant probability classification of each region.

[0010] The result output module maps the malignant probability classification of each local region to the three-dimensional spatial model to generate a spatial-function heat map and generate a search and diagnosis report.

[0011] In a preferred embodiment of the present application, the tumor morphological features include edge regularity and internal echo uniformity; the data acquisition module inserts synchronization markers in the pressure distribution data stream based on the ultrasound image frame header through timestamp alignment technology, and triggers pressure sampling through hardware interrupts to realize the synchronous acquisition of pressure distribution, tumor morphological features, and blood flow velocity and direction data, with a timestamp error of ≤10 ms.

[0012] In a preferred embodiment of the present application, the construction process of the pressure-image overlay includes the following steps:

[0013] S1, the space alignment of the pressure sensor coordinate system of the ultrasound probe and the ultrasound image coordinate system is performed by using an affine transformation matrix, and the transformation formula is:

[0014] ; in the formula, T is an affine transformation matrix determined by matching feature points of a calibration plate, and contains translation, rotation, and scaling parameters; is the pressure sensor coordinate system of the ultrasound probe; is the ultrasound image coordinate system;

[0015] S2, each pixel point of the ultrasound image is located to the pressure sensor coordinate system by reverse transformation, and the pressure value is calculated by using bilinear interpolation; wherein the calculation formula of the pressure value is:

[0016] ; in the formula, is the interpolation weight; are the pressure values of the adjacent four nodes in the pressure sensor array;

[0017] S3, mapping the interpolated pressure value to an Alpha channel of the ultrasound image to generate an RGB-Alpha four-channel overlay, and the mapping rule is: ; wherein, are a global minimum value and a global maximum value of the pressure field, respectively; is a transparency value.

[0018] In a preferred embodiment of the present application, the process of generating a tumor three-dimensional space model and correcting deformation based on the pressure-image overlay includes the following steps:

[0019] A1, interlayer space registration is performed on the pressure-image overlays at different depths, and a relative displacement vector of adjacent layers is obtained through feature point matching;

[0020] A2, a moving cube algorithm is used to voxelize the registered two-dimensional overlay sequence to generate an initial three-dimensional grid model, and the voxel size is adaptively adjusted according to the ultrasonic imaging resolution;

[0021] A3, a transition region between the tumor and the normal tissue is identified based on the pressure field gradient and the ultrasonic echo intensity change rate;

[0022] A4, a deformation timing model is established through the real-time acquisition of the respiratory phase signal and the blood vessel pulsation signal, the instantaneous displacement of each voxel is calculated, the tissue elastic modulus database is called, and the deformation is corrected based on Hooke's law.

[0023] In a preferred embodiment of the present application, the hemodynamic parameters calculated by the data analysis module include a resistance index and a pulsatility index, wherein the calculation formulae of the resistance index and the pulsatility index are:

[0024] ; ; wherein, is the resistance index; is the peak flow velocity in the systole; is the flow velocity at the end of diastole; is the pulsatility index; is the average flow velocity;

[0025] The morphological parameters include the sphericity of a local region of the tumor ; wherein, V is the volume of the local region of the tumor, and A is the surface area of the region.

[0026] In a preferred embodiment of the present application, in step S3, the mapping relationship between the pressure value and the transparency value adopts a segmented nonlinear function, and when the pressure value is in a pressure range corresponding to the transition region between the tumor and the normal tissue, the transparency change rate is increased.

[0027] In a preferred embodiment of the present application, in step A4, when establishing the deformation time sequence model, the Kalman filtering algorithm is used to fuse the respiratory phase signal and the blood vessel pulsation signal, including the following steps:

[0028] A401, model the respiratory motion as a low-frequency sinusoidal signal, with a frequency range of 0.1-0.5 Hz;

[0029] A402, model the blood vessel pulsation as a high-frequency pulse signal, with a frequency synchronized with the heart rate;

[0030] A403, predict the voxel displacement through a state transition matrix, and correct the predicted value using an observation matrix to improve the accuracy of the three-dimensional space model deformation correction.

[0031] In a preferred embodiment of the present application, in step S1, the calibration process of the affine transformation matrix uses a checkerboard calibration board, and the implementation steps include:

[0032] S101, place the calibration board in the detection area of the ultrasound probe, and collect pressure field data and ultrasound images;

[0033] S102, extract the coordinate pairs of the corner points of the calibration board in the two coordinate systems;

[0034] S103, solve the transformation matrix T containing rotation, translation and scaling parameters based on the least squares method;

[0035] S104, verify the matrix accuracy by re-projecting the error ≤0.5 pixels.

[0036] In a preferred embodiment of the present application, the result output module further includes calling the hospital information system to obtain the historical ultrasound examination data of the patient, spatially registering the current three-dimensional space model with the historical model, generating a tumor volume change rate, a malignant probability grading evolution trend chart, and superimposing the evolution trend chart on the space-function heat map to form a four-dimensional dynamic diagnosis report.

[0037] The present application solves the defects in the background art, and has the following beneficial effects:

[0038] (1) Through the time stamp alignment technology, the palpation pressure distribution, B-mode ultrasonic morphological features and color Doppler blood flow data are synchronously collected, and the space alignment of the pressure sensor coordinate system and the ultrasound image coordinate system is realized through affine transformation, solving the problem of asynchronous data collection and spatial misalignment of pressure, morphology and blood flow in the prior art, and realizing accurate matching of multi-dimensional information, providing a more comprehensive basis for tumor property judgment.

[0039] (2) Based on the pressure-image overlay, an initial three-dimensional mesh model is generated by moving cube algorithm, and Kalman filter is combined to fuse respiration and blood vessel pulsation, the model deformation is corrected by calling the tissue elastic modulus database, and the transition area of tumor and normal tissue is accurately identified by pressure field gradient and echo intensity change rate, which overcomes the defects that the three-dimensional model in the prior art is easily disturbed by respiration and blood vessel pulsation, and the tumor boundary is blurred. The corrected three-dimensional model can accurately reflect the spatial form and infiltration range of the tumor, especially for gynecological tumors which are deep in position and are greatly affected by physiological movement, the spatial positioning accuracy is significantly improved, and reliable basis is provided for operation planning.

[0040] (3) The tumor boundary determined by pressure is taken as a reference to divide local areas, the hemodynamic parameters and morphological parameters of each area are calculated, and the area-specific malignant probability grading is output through convolutional neural network, which breaks through the limitation of analyzing the tumor as a whole in the prior art, and local analysis can avoid that the high-risk area is covered by the overall characteristics, so that the accuracy of malignant probability grading is improved, and the misdiagnosis of borderline tumors is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings;

[0042] Figure 1 is a system flowchart of the preferred embodiment of the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0044] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below.

[0045] As shown in Figure 1 , a gynecological tumor ultrasound examination system comprises a data acquisition module, a data processing module, a data analysis module and a result output module.

[0046] The data acquisition module uses the ultrasonic probe to collect the pressure distribution in real time during palpation, and generates a two-dimensional matrix of the pressure field; the B-mode ultrasound is used to obtain the morphological characteristics of the tumor, and the color Doppler ultrasound is used to obtain the blood flow velocity and direction data. The pressure distribution can reflect the difference in tissue hardness, and combined with the hemodynamic information, the tumor properties can be more comprehensively evaluated. The timestamp alignment technology ensures the spatiotemporal consistency of the pressure, morphology and blood flow data, and avoids misdiagnosis caused by data misplacement. For example, the precise correlation between the rapidly changing blood flow signal and the pressure response can more accurately distinguish between benign tumors (regular blood flow and uniform pressure distribution) and malignant tumors (turbulent blood flow and abnormal pressure gradient).

[0047] The morphological characteristics of the tumor include edge regularity and internal echo uniformity; the data acquisition module inserts synchronization markers in the pressure distribution data stream based on the timestamp alignment technology and the frame header of the ultrasound image, and triggers pressure sampling through hardware interruption, thereby realizing the synchronous acquisition of the pressure distribution, the morphological characteristics of the tumor, and the blood flow velocity and direction data, with a timestamp error of ≤10 ms.

[0048] The data processing module maps the two-dimensional matrix of the pressure field to the two-dimensional plane of the ultrasound image, and generates a pressure-image overlay; based on the pressure-image overlay, a three-dimensional spatial model of the tumor is generated through a volume rendering algorithm, and the transition area between the tumor and the normal tissue is clearly defined; combined with the tissue elastic modulus database, the deformation of the three-dimensional spatial model caused by respiration or blood vessel pulsation is corrected.

[0049] The construction process of the pressure-image overlay includes the following steps:

[0050] S1, the affine transformation matrix is used to align the pressure sensor coordinate system of the ultrasonic probe with the ultrasound image coordinate system in space, and the transformation formula is:

[0051] In the formula, T is the affine transformation matrix, which is determined by matching the feature points of the calibration board and contains translation, rotation and scaling parameters; is the pressure sensor coordinate system of the ultrasonic probe; is the ultrasound image coordinate system;

[0052] Further, in step S1, the calibration process of the affine transformation matrix uses a checkerboard calibration board, and the implementation steps include:

[0053] S101, place the calibration board in the detection area of the ultrasonic probe, and collect the pressure field data and the ultrasound image;

[0054] S102, extract the coordinate pairs of the corner points of the calibration board in the two coordinate systems;

[0055] S103, solve the transformation matrix T containing rotation, translation and scaling parameters based on the least squares method;

[0056] S104, verify the matrix accuracy by re-projection error ≤ 0.5 pixels.

[0057] The checkerboard calibration board can efficiently extract feature point pairs due to its regular corner point distribution. The least squares method can optimize the affine transformation matrix containing rotation, translation and scaling, ensuring pixel-level matching of pressure data and ultrasound images. The accuracy of re-projection error ≤ 0.5 pixels ensures the accuracy of the overlay image and avoids misjudgment of pressure values due to coordinate offset.

[0058] S2, locate each pixel point of the ultrasound image to the pressure sensor coordinate system through inverse transformation, and calculate the pressure value by bilinear interpolation; wherein the calculation formula of the pressure value is:

[0059] ; in the formula, is the interpolation weight; are the pressure values of the adjacent four nodes in the pressure sensor array;

[0060] S3, map the interpolated pressure value to the Alpha channel of the ultrasound image to generate an RGB-Alpha four-channel overlay image, and the mapping rule is: ; in the formula, are the global minimum and maximum values of the pressure field, respectively; is the transparency value.

[0061] Bilinear interpolation smoothes the transition of pressure field data through weighted average of the adjacent four pressure nodes, avoiding the "mosaic" effect caused by insufficient sensor resolution. The segmented nonlinear transparency mapping highlights the pressure changes in the transition area between the tumor and normal tissue, which is often the malignant infiltration edge in clinical practice. Enhancing the transparency change rate can help doctors quickly locate the suspicious boundary and improve the accuracy of biopsy sampling.

[0062] In step S3, the mapping relationship between the pressure value and the transparency value adopts a segmented nonlinear function. When the pressure value is in the pressure range corresponding to the transition area between the tumor and normal tissue, the transparency change rate is increased.

[0063] The data analysis module divides the tumor into multiple local areas based on the tumor boundary determined by the pressure in the three-dimensional space model. For each local area, the hemodynamic parameters and morphological parameters are calculated. The blood flow parameters and morphological parameters of the local area are input into the convolutional neural network, and the malignant probability classification of each area is output.

[0064] The process of generating a tumor three-dimensional space model based on the pressure-image overlay image and correcting deformation includes the following steps:

[0065] The process of generating a three-dimensional spatial model of the tumor and correcting deformation based on the pressure-image overlay is the core step of achieving precise spatial positioning of the tumor, and the specific steps are as follows:

[0066] A1. The pressure-image overlays at different depths (i.e., two-dimensional slices of ultrasound at different scanning depths) have spatial misalignment caused by probe movement and tissue sliding, which needs to be corrected through inter-layer registration. The specific operation includes:

[0067] Preferably, select characteristic points with high stability such as high echo points at the edge of the tumor, pressure gradient mutation points (such as sudden change positions of pressure at the junction between the tumor and normal tissue), and vascular bifurcation points. The form and position of these points can be traced in adjacent layers, reducing registration errors.

[0068] Use the SIFT (Scale-Invariant Feature Transform) algorithm to extract the local descriptors (including position, scale, and rotation angle) of the feature points in each layer, and use the K-Nearest Neighbor matching method to find the corresponding relationship of the feature points in adjacent layers. Then, use the RANSAC (Random Sample Consensus) algorithm to remove mismatched points (such as false correspondences caused by noise or artifacts) and retain matching pairs with a confidence level of ≥95%.

[0069] Based on the matched feature point pairs, calculate the relative displacement vectors (unit: mm) of adjacent layers in x, y, and z dimensions to ensure that each layer of the overlay is arranged in the actual anatomical position in three-dimensional space, laying a foundation for spatial consistency for subsequent three-dimensional reconstruction.

[0070] A2. The registered two-dimensional overlay sequence needs to be converted into three-dimensional body data, and the specific process is as follows:

[0071] Adjust the voxel size adaptively according to the ultrasound imaging resolution (such as a conventional gynecological ultrasound resolution of 512x512 pixels with a pixel spacing of 0.1 mm). For example, when the ultrasound lateral resolution is 0.1 mm, the voxel size in x and y directions is set to 0.1 mm, and the z direction (depth direction) size is set according to the inter-layer spacing (usually 0.5-1 mm), ensuring that the three-dimensional size of the voxel is consistent with the actual anatomical size.

[0072] This algorithm traverses the voxel grid to determine whether the vertices of each cubic unit (composed of adjacent 8 voxels) are located on the tumor boundary (based on a joint threshold of pressure value and echo intensity: e.g., vertices with pressure > 2 times the mean value of normal tissue and echo intensity > threshold T are determined as "tumor points"). Then, calculate the boundary isosurface through linear interpolation to generate a triangular mesh of the tumor surface.

[0073] All the triangular facets are spliced to form an initial three-dimensional mesh model containing tumor morphology, pressure distribution, and echo characteristics. Each vertex in the model is associated with the original data of the corresponding position, such as pressure value and echo intensity, providing a basis for subsequent region identification.

[0074] A3, the transition area is the key boundary of tumor infiltration into normal tissue, which needs to be determined by combining pressure field and echo characteristics:

[0075] For each voxel in the three-dimensional model, calculate the pressure difference value of the voxel and its adjacent six voxels (front and back, left and right, and up and down) to obtain the pressure field gradient (unit: kPa / mm). The region with a gradient value >0.5 kPa / mm is marked as a "pressure sudden change area", which indicates the boundary of tissue hardness change.

[0076] For the echo intensity (gray value) of each voxel, calculate the gray difference value of the adjacent voxels as a proportion of its own gray value (i.e. change rate). The region with a change rate >30% is marked as an "echo sudden change area", reflecting the significant difference in tissue density.

[0077] Take the intersection of "pressure sudden change area" and "echo sudden change area", and set the transition area width threshold (usually 1-3 voxels) based on clinical data (such as tumor infiltration boundary confirmed by pathology). The final transition area is marked with a translucent yellow color in the three-dimensional model, clearly distinguishing the tumor core area, transition area, and normal tissue.

[0078] A4, respiratory motion (low frequency) and blood vessel pulsation (high frequency) can cause dynamic deformation of the three-dimensional tumor model, which needs to be corrected through multi-modal signal fusion and mechanical model:

[0079] Respiratory phase signals are collected by chest and abdominal respiration sensors (sampling rate 10 Hz), and heart rate signals are obtained by electrocardiogram monitoring (sampling rate 100 Hz), and are aligned with the ultrasound scanning data through time stamp (error ≤10 ms), ensuring the consistency of signal timing.

[0080] Dual signals are fused using Kalman filter algorithm: respiratory motion is modeled as a low-frequency sinusoidal signal (amplitude varies with respiratory depth, usually 2-5 mm) of 0.1-0.5 Hz, and blood vessel pulsation is modeled as a high-frequency pulse signal (frequency 60-100 times / minute, amplitude 0.5-1 mm) synchronized with heart rate; the theoretical displacement of each voxel at each time is predicted through a state transition matrix, and the predicted value is corrected by an observation matrix combined with the actually collected signal, reducing noise interference.

[0081] The elastic modulus database of the organization (including the elastic modulus values of different tissue types, such as tumor tissue about 50~100kPa, normal uterine muscle layer about 10~30kPa) is called, and the actual deformation amount of the voxel due to breathing / pulsation is calculated according to Hooke's law (stress=elastic modulus*strain): for tumor tissue with high elastic modulus, the deformation amount is reduced in proportion to the stiffness; for normal tissue with low elastic modulus, the deformation amount is adjusted in proportion to the flexibility. Finally, by inversely compensating the instantaneous displacement of each voxel, a stable three-dimensional model that eliminates the interference of physiological motion is obtained.

[0082] Through the above steps, the generated three-dimensional spatial model of the tumor can not only accurately restore the shape, size and spatial relationship with the surrounding tissue of the tumor, but also eliminate the interference of dynamic physiological factors, providing reliable spatial anatomical basis for subsequent tumor property analysis and clinical decision-making.

[0083] Layer registration and voxelization convert two-dimensional superimposed images into a three-dimensional model, and the voxel size is self-adaptive to the ultrasonic resolution, taking into account both accuracy and computational efficiency. Transition area identification combines pressure gradient and echo intensity (tissue density difference) to accurately delineate the tumor boundary. Kalman filtering dynamically corrects the voxel displacement by fusing low-frequency breathing (0.1~0.5Hz) and high-frequency pulsation signals, avoiding model distortion caused by patient breathing or blood vessel pulsation, and ensuring the consistency of the three-dimensional model with the real anatomical structure.

[0084] Further, in step A4, when establishing the deformation time sequence model, Kalman filtering algorithm is used to fuse the respiratory phase signal and the blood vessel pulsation signal, including the following steps:

[0085] A401, model the respiratory motion as a low-frequency sinusoidal signal, with a frequency range of 0.1~0.5Hz;

[0086] A402, model the blood vessel pulsation as a high-frequency pulse signal, with a frequency synchronized with the heart rate;

[0087] A403, predict the voxel displacement through the state transition matrix, and correct the predicted value using the observation matrix to improve the accuracy of the deformation correction of the three-dimensional spatial model.

[0088] The hemodynamic parameters calculated by the data analysis module include resistance index and pulsatility index, wherein the calculation formula of the resistance index and the pulsatility index is:

[0089] ; ; wherein, is the resistance index; is the peak flow velocity in systole; is the end-diastolic flow velocity; is the pulsatility index; is the average flow velocity;

[0090] The morphological parameter includes sphericity of a local region of the tumor ; in the formula, V is a volume of a local region of the tumor; and A is a region surface area.

[0091] The local region division (such as division according to a pressure gradient) can analyze tumor heterogeneity in a targeted manner, and avoid that a whole model masks local malignant characteristics. Blood flow parameters (R i , P i ) reflect tumor angiogenesis activity, and the morphological parameter (sphericity) is related to a tumor growth mode (a benign tumor is mostly round, and the sphericity is close to 1). A convolutional neural network (CNN) automatically extracts a multi-parameter correlation feature, and compared with a traditional threshold method, a malignant probability classification is more in line with a clinical complex scene (such as discrimination of a borderline tumor), and assists a doctor in formulating an individualized diagnosis and treatment scheme.

[0092] The result output module maps the malignant probability classification of each local region to a three-dimensional space model, generates a space-function heat map, and generates a search diagnosis report.

[0093] The result output module further includes calling a hospital information system to obtain historical ultrasound examination data of a patient, performing spatial registration on a current three-dimensional space model and a historical model, generating a tumor volume change rate and a malignant probability classification evolution trend chart, and then superimposing the evolution trend chart on the space-function heat map to form a four-dimensional dynamic diagnosis report.

[0094] The space-function heat map maps the malignant probability classification to the three-dimensional model, directly displays a tumor “space distribution-function state” correlation, and helps a doctor to locate a high-risk area. The four-dimensional dynamic report (a time dimension) quantifies tumor progression (such as a volume growth rate and a malignant classification evolution) by comparing historical data, and provides an objective basis for efficacy evaluation (such as tumor shrinkage after chemotherapy) or recurrence monitoring. Compared with a traditional static ultrasound report, the four-dimensional dynamic mode is more in line with the continuity characteristics of tumor biological behavior, and improves the scientific nature of clinical decision-making.

[0095] According to the ideal embodiments of the present application, through the above description, relevant personnel can make various changes and modifications without deviating from the technical idea of the present application. The technical scope of the present application is not limited to the content in the specification, and must be determined according to the scope of claims.

Claims

1. A gynecological tumor ultrasound examination system, characterized in that, The application relates to a tumor diagnosis system based on real-time palpation pressure and ultrasound imaging. The system comprises a data acquisition module, a data processing module, a data analysis module and a result output module. The data acquisition module uses an ultrasonic probe to collect pressure distribution in real time during palpation, and generates a pressure field two-dimensional matrix; tumor morphological features are obtained through B-mode ultrasound, and blood flow velocity and direction data are obtained through color Doppler ultrasound; The data processing module maps the pressure field two-dimensional matrix to the two-dimensional plane of the ultrasonic image, and generates a pressure-image superposition graph; based on the pressure-image superposition graph, a three-dimensional space model of the tumor is generated through a volume rendering algorithm, and the transition area between the tumor and normal tissue is determined; in combination with a tissue elastic modulus database, deformation of the three-dimensional space model caused by respiration or blood vessel pulsation is corrected; The data analysis module divides the tumor into multiple local areas based on the tumor boundary determined by pressure in the three-dimensional space model; for each local area, hemodynamic parameters and morphological parameters are calculated; the blood flow parameters and morphological parameters of the local area are input into a convolutional neural network, and a malignant probability classification of each area is output; The result output module maps the malignant probability classification of each local area to the three-dimensional space model, generates a space-function heat map, and generates a retrieval diagnosis report.

2. The gynecological tumor ultrasound examination system according to claim 1, characterized in that: The tumor morphological features include edge regularity and internal echo uniformity; the data acquisition module inserts synchronization marks in the pressure distribution data stream through a timestamp alignment technology, and triggers pressure sampling through a hardware interrupt, so that the pressure distribution, tumor morphological features, blood flow velocity and direction data are synchronously collected, and the timestamp error is less than or equal to 10 ms.

3. The gynecological tumor ultrasound examination system according to claim 1, characterized in that: The construction process of the pressure-image superposition graph comprises the following steps: S1, an affine transformation matrix is used to perform space alignment on the pressure sensor coordinate system of the ultrasonic probe and the ultrasonic image coordinate system, and the transformation formula is: ; where T is an affine transformation matrix, determined by matching the feature points of the calibration board, containing translation, rotation, scaling parameters; is the coordinate system of the pressure sensor of the ultrasound probe; is the coordinate system of the ultrasound image; S2, each pixel point of the ultrasonic image is located to the pressure sensor coordinate system through reverse transformation, and a pressure value is calculated through bilinear interpolation; wherein the calculation formula of the pressure value is: wherein are interpolation weights; are pressure values of the four adjacent nodes in the pressure sensor array; S3, mapping the interpolated pressure values to the Alpha channel of the ultrasound image to generate an RGB-Alpha four-channel overlay, the mapping rule being: ; wherein, are the global minimum and maximum values of the pressure field, respectively; is the transparency value.

4. The gynecological tumor ultrasound examination system according to claim 1, characterized in that: The process of generating the three-dimensional space model of the tumor based on the pressure-image superposition graph and correcting deformation comprises the following steps: A1, space registration is performed on the pressure-image superposition graphs at different depths, and a relative displacement vector of adjacent layers is obtained through feature point matching; A2, a moving cube algorithm is used to perform voxelization processing on the registered two-dimensional superposition graph sequence, and an initial three-dimensional grid model is generated, and the voxel size is adaptively adjusted according to the ultrasonic imaging resolution; A3, the transition area between the tumor and normal tissue is identified based on the pressure field gradient and the ultrasonic echo intensity change rate; A4, a deformation timing model is established through real-time acquisition of the respiration phase signal and the blood vessel pulsation signal, the instantaneous displacement of each voxel is calculated, the tissue elastic modulus database is called, and the deformation is corrected based on Hooke's law.

5. The gynecological tumor ultrasound examination system according to claim 1, characterized in that: The hemodynamic parameters calculated by the data analysis module include resistance index and pulsatility index, wherein the calculation formulas of the resistance index and the pulsatility index are: ; ; wherein, is the resistance index; is the peak systolic flow velocity; is the end diastolic flow velocity; is the pulsatility index; is the mean flow velocity; Morphological parameters include sphericity of a local region of the tumor ; where V is the volume of a local region of the tumor and A is the surface area of the region.

6. The gynecological tumor ultrasound examination system according to claim 3, characterized in that: In step S3, the mapping relationship between the pressure value and the transparency value adopts a segmented nonlinear function, and when the pressure value is in the pressure range corresponding to the transition area between the tumor and normal tissue, the transparency change rate is increased.

7. The gynecological tumor ultrasound examination system according to claim 4, characterized in that: In step A4, when establishing the deformation time sequence model, Kalman filtering algorithm is used to fuse the respiratory phase signal and the blood vessel pulsation signal, including the following steps: A401, model the respiratory motion as a low-frequency sinusoidal signal, the frequency range is 0.1~0.5Hz; A402, model the blood vessel pulsation as a high-frequency pulse signal, the frequency is synchronized with the heart rate; A403, predict the voxel displacement through the state transition matrix, and correct the predicted value by using the observation matrix, so as to improve the accuracy of the three-dimensional space model deformation correction.

8. The gynecological tumor ultrasound examination system according to claim 3, characterized in that: In step S1, the calibration process of the affine transformation matrix adopts a checkerboard calibration board, and the implementation steps include: S101, place the calibration board in the detection area of the ultrasonic probe, and collect the pressure field data and the ultrasonic image; S102, extract the coordinate pairs of the corner points of the calibration board in the two coordinate systems; S103, solve the transformation matrix T containing rotation, translation and scaling parameters based on the least square method; S104, verify the matrix accuracy by the re-projection error≤0.5 pixels.

9. The gynecological tumor ultrasound examination system according to claim 1, characterized in that: The result output module further includes calling a hospital information system to obtain historical ultrasonic examination data of the patient, performing spatial registration on the current three-dimensional space model and the historical model, generating a tumor volume change rate and a malignant probability classification evolution trend diagram, and then superimposing the evolution trend diagram on the space-function heat map to form a four-dimensional dynamic diagnosis report.

Images