One-stop multi-mode image auxiliary management system for intravenous thrombolysis

Abstract

The invention relates to the technical field of image auxiliary management, and provides a one-stop multi-modal image auxiliary management system for venous thrombolysis, which comprises a multi-modal image acquisition module, a multi-modal image fusion module, a blood vessel image analysis module and a thrombolysis process optimization module. The multi-mode image acquisition module acquires ischemia area blood vessel images of computed tomography and magnetic resonance imaging; the multi-modal image fusion module carries out registration and fusion on the blood vessel images of different modalities to generate a comprehensive blood vessel image; the blood vessel image analysis module comprises an image preprocessing unit, a blood vessel visualization evaluation unit, a blood vessel feature evaluation unit and a blood vessel continuity analysis unit, and is used for analyzing the comprehensive blood vessel image and evaluating a blood vessel visualization index and a feature parameter; and the thrombolysis process optimization module comprises a parameter analysis unit, a multi-modal parameter optimization unit and an equipment control unit, determines a multi-modal image parameter combination by utilizing a reinforcement learning algorithm based on a blood vessel image quality evaluation value and a statistical analysis technology, and executes equipment control.

Description

Technical Field

[0001] The present invention relates to the field of impact-assisted management technology, and more specifically, to a one-stop multimodal image-assisted management system for intravenous thrombolysis. Background Art

[0002] During intravenous thrombolysis for acute ischemic stroke, multimodal medical imaging technologies, such as computed tomography (CT) and magnetic resonance imaging (MRI), provide key imaging evidence for clinical decision-making. CT can quickly display changes in brain tissue structure, while MRI has unique advantages in soft tissue contrast. However, how to effectively integrate and analyze these multimodal imaging data and optimize imaging parameters in real time during treatment has become a key challenge to improving image quality and efficiency. Therefore, designing a system that can intelligently manage multimodal images, evaluate vascular conditions, and dynamically optimize imaging parameters is of great significance for improving the accuracy and effectiveness of imaging assistance during intravenous thrombolysis. Such a system can not only improve image quality, but also optimize the imaging process, ultimately providing more reliable imaging support for clinical decision-making.

[0003] Existing studies have preliminarily explored the application effects of multimodal imaging technology. For example, in document 1 (Feasibility study of multimodal imaging examination in acute ischemic stroke, Journal of Clinical Radiology, 2021, 40(2), 609-613), the effectiveness of multimodal imaging examinations combining CT and MRI in diagnosis and prognosis evaluation was verified in patients with acute ischemic stroke (AIS). Studies have shown that combining multimodal MRI imaging techniques such as DWI, 3D-TOF and ASL after CT scan can not only significantly improve image quality, but also optimize thrombolytic effects. In particular, DWI and 3D-TOF can obtain multidimensional information on vascular morphology and cerebral blood perfusion, which helps to identify responsible vessels and accurately evaluate ischemic lesions and low perfusion areas. This study provides strong support for further optimizing the application of multimodal imaging in the treatment and prognosis evaluation of AIS.

[0004] However, the existing technology still has many shortcomings in this regard. First, most existing systems can only process single-modality medical images, and it is difficult to fully utilize the complementary advantages of multimodality images such as CT and MRI, which limits the comprehensive evaluation of vascular structure; second, the current image analysis methods often lack a systematic evaluation of vascular structure, which may affect the evaluation of vascular image quality; third, there is a general lack of intelligent parameter optimization mechanism, and it is impossible to adjust imaging parameters in real time according to specific circumstances, which not only affects image quality, but also may reduce imaging efficiency; in addition, in reality, image acquisition, analysis, and equipment control are often separated, and there is a lack of a unified, closed-loop management system, which not only increases the complexity of operation, but also may affect the timeliness and effectiveness of image acquisition. These technical defects restrict the effectiveness of multimodality medical imaging in clinical applications.

[0005] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a one-stop multimodal image-assisted management system for intravenous thrombolysis, which realizes the management of multimodal images, the assessment of vascular conditions and the dynamic optimization of imaging parameters through the technical means of multimodal image fusion, vascular image analysis and thrombolysis process optimization, solves the problems of single modality processing, lack of systematic evaluation, unintelligent parameter optimization and fragmented management system in the prior art, and improves the accuracy and effectiveness of image assistance in the process of intravenous thrombolysis.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A one-stop multimodal image-assisted management system for intravenous thrombolysis, comprising a multimodal image acquisition module, a multimodal image fusion module, a vascular image analysis module and a thrombolysis process optimization module, wherein the multimodal image acquisition module is connected to the multimodal image fusion module, the multimodal image fusion module is connected to the vascular image analysis module, and the vascular image analysis module is connected to the thrombolysis process optimization module, wherein the multimodal image fusion module is used to integrate and analyze data from electronic computer tomography and nuclear magnetic resonance imaging modalities to generate a comprehensive vascular image, and the vascular image analysis module is used to process and analyze the comprehensive vascular image and evaluate the quality and characteristic parameters of the vascular image, and comprises an image preprocessing unit, a vascular visualization evaluation unit, and a vascular characteristic evaluation unit. The image preprocessing unit is connected to the vascular visualization evaluation unit, the vascular feature evaluation unit, the vascular continuity analysis unit and the comprehensive evaluation unit respectively. The vascular visualization evaluation unit, the vascular feature evaluation unit and the vascular continuity analysis unit are all connected to the comprehensive evaluation unit. The comprehensive evaluation unit builds a vascular image quality evaluation model according to the image data processed by the image preprocessing unit, the visualization index of the ischemic area blood vessels obtained by the vascular visualization evaluation unit, the characteristic parameters of the ischemic area blood vessels obtained by the vascular feature evaluation unit and the continuity parameters of the ischemic area blood vessels obtained by the vascular continuity analysis unit to obtain the vascular image quality evaluation value. The formula of the vascular image quality evaluation model is:

[0009]

[0010] Where V is the vascular image quality assessment value, VI is the vascular visualization index, VFI is the vascular characteristic index, and L c is the total length of the continuous segment that can be tracked by the target vessel, L t is the theoretical total length of the target vessel, θ i is the angle between adjacent segments on the target vessel centerline, θ mean is the average value of the angle, n is the number of angle measurement points;

[0011] The thrombolysis process optimization module includes a parameter analysis unit, a multimodal parameter optimization unit and an equipment control unit, which is used to obtain the influence intensity of imaging parameters on the vascular image quality evaluation value based on the vascular image quality evaluation value output by the vascular image analysis module, the current electronic computed tomography parameters, the nuclear magnetic resonance imaging parameters and the statistical analysis technology, and use the reinforcement learning algorithm to determine the combination of electronic computed tomography and nuclear magnetic resonance imaging parameters to perform multimodal imaging equipment control, wherein the parameter analysis unit is connected to the multimodal parameter optimization unit, and the multimodal parameter optimization unit is connected to the equipment control unit.

[0012] As a further solution of the present invention, the multimodal image acquisition module is based on the medical imaging device interface standard, and synchronously acquires multimodal medical images of ischemic area blood vessels including computed tomography angiography and magnetic resonance angiography through imaging equipment, and uses communication technology to transmit the acquired image data to the multimodal image fusion module.

[0013] As a further solution of the present invention, a multimodal image fusion module uses image fusion technology to align and fuse vascular images from different imaging modalities to generate a comprehensive vascular image, which includes spatial resolution information of computed tomography angiography and tissue contrast information of magnetic resonance angiography.

[0014] As a further solution of the present invention, the image preprocessing unit preprocesses the comprehensive vascular image through adaptive image enhancement technology and noise suppression algorithm to improve image quality and vascular contrast, obtain optimized image data of the ischemic area blood vessels, including enhanced vascular structure information, noise-reduced image background information and optimized vascular edge information, and uses communication technology to transmit the optimized vascular image data to the vascular visualization evaluation unit, the vascular feature evaluation unit and the vascular continuity analysis unit.

[0015] As a further solution of the present invention, the vascular visualization evaluation unit performs vascular structure enhancement on the preprocessed comprehensive image through a vascular segmentation and enhancement algorithm to obtain the morphological characteristics of the blood vessels in the ischemic area, including extracting the vascular centerline using a skeletonization algorithm, calculating the vascular continuous length ratio and the vascular branch complexity; calculating the vascular visibility through image enhancement and segmentation technology; calculating the vascular branch complexity using a branch point detection algorithm; obtaining the background complexity using image processing technology, and presetting the vascular visibility reference value; the vascular visualization evaluation unit obtains the vascular visualization index by constructing a vascular morphology evaluation model, wherein the formula of the vascular morphology evaluation model is:

[0016]

[0017] In the formula, VI is the vascular visualization index, VD is the vascular visibility, and VD ref is the reference value of vascular visibility, BC is the complexity of vascular branches, CL is the ratio of vascular continuous length, B is the background complexity, and σ is the integral variable.

[0018] As a further solution of the present invention, the vascular feature evaluation unit analyzes the optimized vascular image data through a computer vision algorithm to obtain image feature parameters of the ischemic region blood vessels, including vascular contrast, contrast-to-noise ratio and spatial resolution. The vascular feature evaluation unit constructs a vascular feature evaluation model to calculate and obtain a vascular feature index, wherein the formula of the vascular feature evaluation model is:

[0019]

[0020] Where VFI is the vascular characteristic index, VC is the vascular contrast, CNR is the contrast-to-noise ratio, SR is the spatial resolution, and t is the integral variable.

[0021] As a further solution of the present invention, the vascular continuity analysis unit calculates the total length of the continuous segments that can be tracked in the target blood vessel in the ischemic area, the theoretical total length of the target blood vessel, the angles between adjacent segments on the centerline of the target blood vessel and their average values, and the number of angle measurement points through image processing and a blood vessel tracking algorithm, and transmits these parameters to the comprehensive evaluation unit.

[0022] As a further solution of the present invention, the parameter analysis unit uses statistical analysis technology to construct a multivariate regression model to obtain the influence intensity of the electronic computer tomography parameters and the nuclear magnetic resonance imaging parameters on the vascular image quality evaluation value, wherein the steps of constructing the multivariate regression model are:

[0023] Step A1, the parameter analysis unit sets multimodal imaging parameters as independent variables, including X-ray tube voltage, X-ray tube current, scanning time, slice thickness of electronic computed tomography and repetition time, echo time and flip angle of magnetic resonance imaging; sets vascular image quality assessment value as dependent variable, collects parameter settings used in historical intravenous thrombolysis and corresponding vascular image quality assessment values ​​as training set;

[0024] Step A2, the parameter analysis unit estimates the model parameters using the least squares method, ensuring that all variables are defined. After the model input configuration is completed, the estimation process is run to obtain the regression coefficient of each multimodal imaging parameter;

[0025] Step A3: After the parameter estimation is completed, the parameter analysis unit checks the generated determination coefficient and statistics to confirm the goodness of fit of the model. If the index does not meet the standard, the interaction terms and nonlinear terms between the multimodal imaging parameters are introduced and the estimation process is run again for verification.

[0026] Step A4, the parameter analysis unit analyzes and records the regression coefficients of the final model, and uses the regression coefficients to indicate the influence of the X-ray tube voltage, X-ray tube current, scanning time, slice thickness of the electronic computed tomography and the repetition time, echo time, and flip angle of the magnetic resonance imaging on the vascular image quality assessment value.

[0027] As a further solution of the present invention, the multimodal parameter optimization unit uses a reinforcement learning algorithm based on the result obtained by the parameter analysis unit to find a multimodal imaging parameter combination, wherein finding a multimodal imaging parameter combination includes the following steps:

[0028] Step B1, define the state space of reinforcement learning as the current vascular image quality assessment value and multimodal image parameters, including X-ray tube voltage, X-ray tube current, scanning time, layer thickness of electronic computed tomography, repetition time, echo time and flip angle of magnetic resonance imaging; the action space is the adjustable range of these parameters, and the reward function is based on the improvement of the vascular image quality assessment value relative to the previous state, as well as the contrast of the blood vessels in the ischemic area of ​​the comprehensive vascular image and the edge clarity of the vascular wall structure, wherein the formula of the reward function is:

[0029]

[0030] Among them, R is the reward value, V new is the new vascular image quality assessment value, V old is the old vascular image quality assessment value, C vessel is the vascular contrast, C wall is the gradient amplitude of the vascular wall area, w1 and w2 are weight factors;

[0031] Step B2, initialize the Q table, select actions according to the current state, i.e., adjust the image parameters, use the ε-greedy strategy to balance exploration and exploitation, execute the selected actions and observe the new state and reward;

[0032] Step B3, update the Q value according to the Bellman equation and optimize the decision strategy;

[0033] Step B4, repeating steps B2 and B3 until a preset number of iterations is reached and the change in the Q value is less than a predetermined threshold;

[0034] Step B5: Select a multimodal imaging parameter combination from the final learned strategy.

[0035] As a further solution of the present invention, the device control unit receives the multimodal imaging parameter combination from the multimodal parameter optimization unit, automatically adjusts the parameter settings of each imaging device through the standardized device control interface, and feeds back the adjusted parameters to the multimodal imaging acquisition module.

[0036] Compared with the prior art, the one-stop multimodal image-assisted management system for intravenous thrombolysis of the present invention has the following beneficial effects:

[0037] 1. The present invention introduces a vascular image quality assessment model, which comprehensively considers the vascular visualization index, vascular characteristic index and vascular continuity parameter, and performs quantitative assessment through a formula, thereby improving the accuracy, reliability and repeatability of vascular image quality assessment, providing clinicians with a more objective and accurate imaging basis, and helping to improve the accuracy of intravenous thrombolytic therapy.

[0038] 2. The present invention introduces a parameter optimization mechanism based on reinforcement learning. By defining a state space including vascular image quality assessment values ​​and multimodal imaging parameters, as well as a reward function based on the improvement of vascular image quality and vascular structure clarity, a reinforcement learning model is constructed, which enables the system to dynamically adjust the imaging parameters of CT and MRI according to real-time feedback. Through iterative optimization, the system continuously learns and improves the parameter selection strategy, and finally realizes the intelligent and personalized adjustment of imaging parameters. This design can improve image quality and imaging efficiency, and is particularly suitable for clinical scenarios such as intravenous thrombolytic therapy, which have extremely high requirements for time and image quality, and provides patients with more timely and accurate imaging support. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is the DWI image of a patient with acute ischemic stroke in Reference 1.

[0040] Figure 2 This is the 3D-TOF image of a patient with acute ischemic stroke in Reference 1.

[0041] Figure 3 This is the ADC graph of patients with acute ischemic stroke in Reference 1.

[0042] Figure 4 This is the CBF map of patients with acute ischemic stroke in literature 1.

[0043] Figure 5 The schematic diagram of the structure of a one-stop multimodal image-assisted management system for intravenous thrombolysis according to the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Example 1

[0046] A one-stop multimodal image-assisted management system for intravenous thrombolysis includes a multimodal image acquisition module, a multimodal image fusion module, a vascular image analysis module and a thrombolysis process optimization module.

[0047] The multimodal image acquisition module in the embodiment of the present invention is connected to the multimodal image fusion module, the multimodal image fusion module is connected to the vascular image analysis module, and the vascular image analysis module is connected to the thrombolysis process optimization module.

[0048] The multimodal image acquisition module is based on the medical imaging device interface standard. It synchronously acquires multimodal medical images of ischemic area blood vessels including computed tomography angiography and magnetic resonance angiography through imaging devices, and uses communication technology to transmit the acquired image data to the multimodal image fusion module.

[0049] The multimodal image fusion module uses image fusion technology to register and fuse vascular images from different imaging modalities to generate a comprehensive vascular image, which contains spatial resolution information of computed tomography angiography and tissue contrast information of magnetic resonance angiography.

[0050] The multimodal image fusion module is used to integrate and analyze data from electronic computed tomography and magnetic resonance imaging modalities to generate comprehensive vascular images. The vascular image analysis module is used to process and analyze comprehensive vascular images and evaluate the quality and characteristic parameters of vascular images. It includes an image preprocessing unit, a vascular visualization evaluation unit, a vascular characteristic evaluation unit, a vascular continuity analysis unit and a comprehensive evaluation unit.

[0051] The image preprocessing unit in the embodiment of the present invention is respectively connected to the blood vessel visualization evaluation unit, the blood vessel characteristic evaluation unit, the blood vessel continuity analysis unit and the comprehensive evaluation unit, and the blood vessel visualization evaluation unit, the blood vessel characteristic evaluation unit and the blood vessel continuity analysis unit are all connected to the comprehensive evaluation unit.

[0052] The image preprocessing unit in the embodiment of the present invention preprocesses the comprehensive vascular image through adaptive image enhancement technology and noise suppression algorithm to improve image quality and vascular contrast, obtain optimized image data of the ischemic area blood vessels, including enhanced vascular structure information, denoised image background information and optimized vascular edge information, and uses communication technology to transmit the optimized vascular image data to the vascular visualization evaluation unit, the vascular feature evaluation unit and the vascular continuity analysis unit.

[0053] The vascular visualization evaluation unit enhances the vascular structure of the preprocessed comprehensive image through the vascular segmentation and enhancement algorithm to obtain the morphological characteristics of the blood vessels in the ischemic area, including extracting the vascular centerline using the skeletonization algorithm, calculating the vascular continuous length ratio and vascular branch complexity; calculating the vascular visibility through image enhancement and segmentation technology; calculating the vascular branch complexity using the branch point detection algorithm; obtaining the background complexity using image processing technology, and presetting the vascular visibility reference value; the vascular visualization evaluation unit obtains the vascular visualization index by constructing a vascular morphology evaluation model, wherein the formula of the vascular morphology evaluation model is:

[0054]

[0055] In the formula, VI is the vascular visualization index, VD is the vascular visibility, and VD ref is the reference value of vascular visibility, BC is the complexity of vascular branches, CL is the ratio of vascular continuous length, B is the background complexity, and σ is the integral variable.

[0056] The ratio of the vascular visibility to the vascular visibility reference value in the embodiment of the present invention is weighted by adjusting its weight through a cube root operation to ensure that a reasonable score can be obtained under low visibility conditions; the vascular branch complexity adopts a natural logarithmic function to maintain sensitivity to changes in branch complexity while avoiding excessive influence of an excessively high vascular branch complexity value on the total score; an S-type function is used to evaluate the vascular continuous length ratio, and when the vascular continuous length ratio is large, the function approaches 1, indicating good vascular continuity; when the vascular continuous length ratio is small, the function approaches 0, indicating poor vascular continuity. This design can effectively distinguish different degrees of vascular continuity; the integral term takes into account the influence of background complexity on vascular visualization, and the integral variable makes the influence of background complexity smoothly considered in the entire image range. The integral result decreases with increasing B value, reflecting the negative effect of background complexity on vascular visualization.

[0057] Calculating blood vessel visibility by using image enhancement and segmentation technology in an embodiment of the present invention includes the following steps:

[0058] Step C1: Preprocess the vascular image by first performing Gaussian filtering using a 3x3 or 5x5 Gaussian kernel to remove noise, and then applying histogram equalization technology to enhance the image contrast;

[0059] Step C2: Apply Frangi filter to enhance the vascular structure of the preprocessed image, set the scale range to 1 to 10 pixels, the step length to 1 pixel, calculate the Hessian matrix for each scale and extract the eigenvalue, calculate the vascular metric according to the formula, so as to highlight the vascular structure. The formula for calculating the vascular metric is:

[0060]

[0061] In the formula, η1, η2, η3 are the eigenvalues ​​of the Hessian matrix, α, β, c are the control parameters;

[0062] Step C3: Adaptively binarize the Frangi filtered image using the Otsu threshold method to segment the image into a blood vessel area and a background area;

[0063] Step C4: In the original grayscale image, the blood vessel area and background area pixels are extracted according to the mask of the binary image, the average grayscale value of the two areas is calculated, and the difference between the two is taken as the contrast difference to quantify the distinction between the blood vessel and the background;

[0064] Step C5: Apply 3x3 Sobel operator to the binary image for edge detection, perform convolution in the x and y directions respectively, calculate the gradient amplitude, and finally calculate the average value of the gradient amplitude of all edge pixels as the edge clarity index;

[0065] Step C6: Normalize the contrast difference and edge clarity to the range of [0, 1], set the weight coefficient, and calculate the blood vessel visibility by weighted summation.

[0066] In the embodiment of the present invention, the skeletonization algorithm is used to extract the blood vessel centerline, and the blood vessel continuous length ratio and blood vessel branch complexity are calculated, including the following steps:

[0067] Step D1: Applying the Zhang-Suen thinning algorithm to the binary vascular image obtained in the previous step, iteratively applying the thinning rule until the image no longer changes, thereby extracting the vascular skeleton and obtaining a single-pixel width vascular centerline representation;

[0068] Step D2: Use the 8-connectivity analysis algorithm to scan the skeleton image and mark all connected non-zero pixels as the same vascular segment to achieve the identification and marking of continuous vascular segments;

[0069] Step D3: traverse all marked blood vessel segments, count the number of pixels in each segment, find the blood vessel segment with the largest number of pixels, and record its length as the longest continuous blood vessel segment length;

[0070] Step D4: Count the number of all non-zero pixels in the skeleton image to obtain the total vessel length, which represents the overall length of the entire vascular network;

[0071] Step D5: Calculate the ratio of the longest continuous blood vessel segment length to the total blood vessel length to obtain the blood vessel continuous length ratio.

[0072] Calculating the complexity of blood vessel branches using a branch point detection algorithm in an embodiment of the present invention includes the following steps:

[0073] Step E1: Based on the vascular skeleton image obtained in the previous step, a 3x3 sliding window is used to scan the entire image. For each pixel, the number of non-zero pixels in its 8-neighborhood is checked. When the number is greater than 2, the pixel is marked as a branch point.

[0074] Step E2: Counting the total number of pixels marked as branch points in the entire skeleton image, wherein the total number of pixels reflects the branching degree of the vascular network;

[0075] Step E3: Calculate the branch point density, i.e., the total number of branch points divided by the total number of pixels in the image, to standardize the branch complexity of images of different sizes;

[0076] Step E4: Set a maximum branch point density threshold, then divide the calculated branch point density by the threshold and take the minimum value 1.

[0077] The use of image processing technology to obtain background complexity in the embodiment of the present invention includes the following steps:

[0078] Step F1: Apply the Canny edge detection algorithm to the original blood vessel image, set the low threshold to 50 and the high threshold to 150, and obtain the edge information of the overall image;

[0079] Step F2: using the previously obtained vascular binary image as a mask, calculate the density of edge pixels in the non-vascular area, that is, the number of background edge pixels divided by the total number of background pixels. The density value reflects the complexity of the background.

[0080] Step F3: Apply the local binary pattern operator to the original image, calculate the 8-bit binary code using a 3x3 neighborhood, then calculate the LBP histogram, divide the LBP code into 59 uniform pattern bins, and normalize the histogram;

[0081] Step F4: Calculate the entropy of the normalized LBP histogram as a texture complexity index, and perform weighted average of the texture complexity index and the edge density to obtain background complexity.

[0082] The vascular feature evaluation unit in the embodiment of the present invention analyzes the optimized vascular image data through a computer vision algorithm to obtain the image feature parameters of the ischemic region blood vessels, including vascular contrast, contrast-to-noise ratio and spatial resolution. The vascular feature evaluation unit constructs a vascular feature evaluation model to calculate and obtain a vascular feature index, wherein the formula of the vascular feature evaluation model is:

[0083]

[0084] Where VFI is the vascular characteristic index, VC is the vascular contrast, CNR is the contrast-to-noise ratio, SR is the spatial resolution, and t is the integral variable.

[0085] The blood vessel continuity analysis unit in the embodiment of the present invention calculates the total length of the continuous segments that can be tracked in the target blood vessel in the ischemic area, the theoretical total length of the target blood vessel, the angles between adjacent segments on the center line of the target blood vessel and their average values, and the number of angle measurement points through image processing and blood vessel tracking algorithms, and transmits these parameters to the comprehensive evaluation unit.

[0086] The comprehensive evaluation unit in the embodiment of the present invention constructs a vascular image quality evaluation model based on the image data processed by the image preprocessing unit, the visualization index of the ischemic region blood vessels obtained by the vascular visualization evaluation unit, the characteristic parameters of the ischemic region blood vessels obtained by the vascular characteristic evaluation unit, and the continuity parameters of the ischemic region blood vessels obtained by the vascular continuity analysis unit to obtain a vascular image quality evaluation value, wherein the formula of the vascular image quality evaluation model is:

[0087]

[0088] Where V is the vascular image quality assessment value, VI is the vascular visualization index, VFI is the vascular characteristic index, and L c is the total length of the continuous segment that can be tracked by the target vessel, L t is the theoretical total length of the target vessel, θ i is the angle between adjacent segments on the target vessel centerline, θ mean is the average value of the angle, and n is the number of angle measurement points.

[0089] The thrombolysis process optimization module includes a parameter analysis unit, a multimodal parameter optimization unit and an equipment control unit, which is used to obtain the influence of imaging parameters on the vascular image quality assessment value based on the vascular image quality assessment value output by the vascular image analysis module, the current electronic computed tomography parameters, the magnetic resonance imaging parameters and the statistical analysis technology, and use the reinforcement learning algorithm to determine the combination of electronic computed tomography and magnetic resonance imaging parameters to perform multimodal imaging equipment control.

[0090] The parameter analysis unit in the embodiment of the present invention is connected to the multimodal parameter optimization unit, and the multimodal parameter optimization unit is connected to the device control unit.

[0091] The parameter analysis unit in the embodiment of the present invention uses statistical analysis technology to construct a multivariate regression model to obtain the influence intensity of the electronic computer tomography parameters and the nuclear magnetic resonance imaging parameters on the vascular image quality evaluation value, wherein the steps of constructing the multivariate regression model are:

[0092] Step A1, the parameter analysis unit sets multimodal imaging parameters as independent variables, including X-ray tube voltage, X-ray tube current, scanning time, slice thickness of electronic computed tomography and repetition time, echo time and flip angle of magnetic resonance imaging; sets vascular image quality assessment value as dependent variable, collects parameter settings used in historical intravenous thrombolysis and corresponding vascular image quality assessment values ​​as training set;

[0093] Step A2, the parameter analysis unit estimates the model parameters using the least squares method, ensuring that all variables are defined. After the model input configuration is completed, the estimation process is run to obtain the regression coefficient of each multimodal imaging parameter;

[0094] Step A3: After the parameter estimation is completed, the parameter analysis unit checks the generated determination coefficient and statistics to confirm the goodness of fit of the model. If the index does not meet the standard, the interaction terms and nonlinear terms between the multimodal imaging parameters are introduced and the estimation process is run again for verification.

[0095] Step A4, the parameter analysis unit analyzes and records the regression coefficients of the final model, and uses the regression coefficients to indicate the influence of the X-ray tube voltage, X-ray tube current, scanning time, slice thickness of the electronic computed tomography and the repetition time, echo time, and flip angle of the magnetic resonance imaging on the vascular image quality assessment value.

[0096] The following is a Python code example that builds a multivariate regression model to obtain the influence of CT parameters and MRI parameters on vascular image quality assessment values. Please note that this example is only a starting point and may need to be adjusted according to actual conditions and device interfaces in actual applications.

[0097] import numpy as np

[0098] import pandas as pd

[0099] from sklearn.linear_model import LinearRegression

[0100] from sklearn.preprocessing import PolynomialFeatures

[0101] from sklearn.metrics import r2_score

[0102] #Assuming the data set, X is the multimodal image parameter, Y is the image quality assessment value

[0103] #X column order: [CT tube voltage, CT tube current, scan time, slice thickness, MRI repetition time, MRI echo time, MRI flip angle]

[0104] #Y is the image quality assessment value

[0105] data=pd.read_csv('imaging_data.csv')

[0106] X=data[['CT_voltage','CT_current','scan_time','slice_thickness','MRI_TR','MRI_TE','MRI_flip_angle']]

[0107] Y = data['image_quality']

[0108] # Least squares method to build a multivariate linear regression model

[0109] linear_regressor=LinearRegression()

[0110] linear_regressor.fit(X,Y)

[0111] #Output the regression coefficient of each image parameter

[0112] print("Regression coefficient:",linear_regressor.coef_)

[0113] #Model goodness of fit check

[0114] Y_pred=linear_regressor.predict(X)

[0115] r2=r2_score(Y,Y_pred)

[0116] print("The coefficient of determination of the model (R 2 ):",r2)

[0117] #If the coefficient of determination is not ideal, introduce polynomial features (interaction terms and nonlinear terms)

[0118] if r2<0.8:#Assume 0.8 is the goodness of fit standard

[0119] poly=PolynomialFeatures(degree=2,interaction_only=True)

[0120] X_poly=poly.fit_transform(X)

[0121] poly_regressor=LinearRegression()

[0122] poly_regressor.fit(X_poly,Y)

[0123] #Output updated model goodness of fit

[0124] Y_poly_pred=poly_regressor.predict(X_poly)

[0125] r2_poly=r2_score(Y,Y_poly_pred)

[0126] print("The updated model's coefficient of determination (R 2 with polynomial features):",r2_poly)

[0127] #Output updated regression coefficients

[0128] print("Polynomial regression coefficient:",poly_regressor.coef_)

[0129] This code is only used as an example. In actual applications, appropriate modifications and adjustments need to be made according to specific circumstances.

[0130] The multimodal parameter optimization unit in the embodiment of the present invention uses a reinforcement learning algorithm to find a multimodal imaging parameter combination based on the result obtained by the parameter analysis unit, wherein finding a multimodal imaging parameter combination includes the following steps:

[0131] Step B1, define the state space of reinforcement learning as the current vascular image quality assessment value and multimodal image parameters, including X-ray tube voltage, X-ray tube current, scanning time, layer thickness of electronic computed tomography, repetition time, echo time and flip angle of magnetic resonance imaging; the action space is the adjustable range of these parameters, and the reward function is based on the improvement of the vascular image quality assessment value relative to the previous state, as well as the contrast of the blood vessels in the ischemic area of ​​the comprehensive vascular image and the edge clarity of the vascular wall structure, wherein the formula of the reward function is:

[0132]

[0133] Among them, R is the reward value, V new is the new vascular image quality assessment value, V old is the old vascular image quality assessment value, C vessel is the vascular contrast, C wall is the gradient amplitude of the vascular wall area, w1 and w2 are weight factors;

[0134] Step B2, initialize the Q table, select actions according to the current state, i.e., adjust the image parameters, use the ε-greedy strategy to balance exploration and exploitation, execute the selected actions and observe the new state and reward;

[0135] Step B3, update the Q value according to the Bellman equation and optimize the decision strategy;

[0136] Step B4, repeating steps B2 and B3 until a preset number of iterations is reached and the change in the Q value is less than a predetermined threshold;

[0137] Step B5: Select a multimodal imaging parameter combination from the final learned strategy.

[0138] The device control unit in the embodiment of the present invention receives the multimodal imaging parameter combination from the multimodal parameter optimization unit, automatically adjusts the parameter settings of each imaging device through the standardized device control interface, and feeds back the adjusted parameters to the multimodal imaging acquisition module.

[0139] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0140] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A one-stop multimodal image-assisted management system for intravenous thrombolysis, comprising a multimodal image acquisition module, a multimodal image fusion module, a vascular image analysis module and a thrombolysis process optimization module, characterized in that: The multimodal image acquisition module is connected to the multimodal image fusion module, the multimodal image fusion module is connected to the vascular image analysis module, and the vascular image analysis module is connected to the thrombolysis process optimization module. The multimodal image fusion module is used to integrate and analyze data from electronic computer tomography and magnetic resonance imaging modalities to generate comprehensive vascular images. The vascular image analysis module is used to process and analyze the comprehensive vascular images and evaluate the quality and characteristic parameters of vascular images. It includes an image preprocessing unit, a vascular visualization evaluation unit, a vascular characteristic evaluation unit, a vascular continuity analysis unit and a comprehensive evaluation unit. The image preprocessing unit is respectively connected to the vascular visualization evaluation unit, the vascular characteristic evaluation unit, the vascular continuity analysis unit and the comprehensive evaluation unit. The visualization evaluation unit, the vascular characteristic evaluation unit, the vascular continuity analysis unit and the comprehensive evaluation unit are connected. The vascular visualization evaluation unit, the vascular characteristic evaluation unit and the vascular continuity analysis unit are all connected to the comprehensive evaluation unit. The comprehensive evaluation unit builds a vascular image quality evaluation model according to the image data processed by the image preprocessing unit, the visualization index of the ischemic region blood vessels obtained by the vascular visualization evaluation unit, the characteristic parameters of the ischemic region blood vessels obtained by the vascular characteristic evaluation unit and the continuity parameters of the ischemic region blood vessels obtained by the vascular continuity analysis unit, and obtains the vascular image quality evaluation value. The formula of the vascular image quality evaluation model is: Where V is the vascular image quality assessment value, VI is the vascular visualization index, VFI is the vascular characteristic index, and L c is the total length of the continuous segment that can be tracked by the target vessel, L t is the theoretical total length of the target vessel, θ i is the angle between adjacent segments on the target vessel centerline, θ mean is the average value of the angle, n is the number of angle measurement points; The thrombolysis process optimization module includes a parameter analysis unit, a multimodal parameter optimization unit and an equipment control unit, which is used to obtain the influence intensity of imaging parameters on the vascular image quality evaluation value based on the vascular image quality evaluation value output by the vascular image analysis module, the current electronic computed tomography parameters, the nuclear magnetic resonance imaging parameters and the statistical analysis technology, and use the reinforcement learning algorithm to determine the combination of electronic computed tomography and nuclear magnetic resonance imaging parameters to perform multimodal imaging equipment control, wherein the parameter analysis unit is connected to the multimodal parameter optimization unit, and the multimodal parameter optimization unit is connected to the equipment control unit.

2. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The multimodal image acquisition module is based on the medical imaging device interface standard. It synchronously acquires multimodal medical images of ischemic area blood vessels including computed tomography angiography and magnetic resonance angiography through imaging devices, and uses communication technology to transmit the acquired image data to the multimodal image fusion module.

3. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The multimodal image fusion module uses image fusion technology to register and fuse vascular images from different imaging modalities to generate a comprehensive vascular image, which contains spatial resolution information of computed tomography angiography and tissue contrast information of magnetic resonance angiography.

4. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The image preprocessing unit preprocesses the comprehensive vascular image through adaptive image enhancement technology and noise suppression algorithm to improve image quality and vascular contrast, and obtains optimized image data of the ischemic area blood vessels, including enhanced vascular structure information, noise-reduced image background information and optimized vascular edge information, and uses communication technology to transmit the optimized vascular image data to the vascular visualization evaluation unit, vascular feature evaluation unit and vascular continuity analysis unit.

5. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The vascular visualization evaluation unit enhances the vascular structure of the preprocessed comprehensive image through the vascular segmentation and enhancement algorithm to obtain the morphological characteristics of the blood vessels in the ischemic area, including extracting the vascular centerline using the skeletonization algorithm, calculating the vascular continuous length ratio and vascular branch complexity; calculating the vascular visibility through image enhancement and segmentation technology; calculating the vascular branch complexity using the branch point detection algorithm; obtaining the background complexity using image processing technology, and presetting the vascular visibility reference value; the vascular visualization evaluation unit obtains the vascular visualization index by constructing a vascular morphology evaluation model, wherein the formula of the vascular morphology evaluation model is: In the formula, VI is the vascular visualization index, VD is the vascular visibility, and VD ref is the reference value of vascular visibility, BC is the complexity of vascular branches, CL is the ratio of vascular continuous length, B is the background complexity, and σ is the integral variable.

6. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The vascular feature evaluation unit analyzes the optimized vascular image data through a computer vision algorithm to obtain the image feature parameters of the ischemic area blood vessels, including vascular contrast, contrast-to-noise ratio, and spatial resolution. The vascular feature evaluation unit constructs a vascular feature evaluation model to calculate and obtain a vascular feature index, wherein the formula of the vascular feature evaluation model is: Where VFI is the vascular characteristic index, VC is the vascular contrast, CNR is the contrast-to-noise ratio, SR is the spatial resolution, and t is the integral variable.

7. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The vascular continuity analysis unit calculates the total length of the traceable continuous segments of the target vessel in the ischemic area, the theoretical total length of the target vessel, the angles between adjacent segments on the centerline of the target vessel and their average values, and the number of angle measurement points through image processing and a vascular tracking algorithm, and transmits these parameters to the comprehensive evaluation unit.

8. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The parameter analysis unit uses statistical analysis technology to construct a multivariate regression model to obtain the influence intensity of the electronic computer tomography parameters and the nuclear magnetic resonance imaging parameters on the vascular image quality evaluation value, wherein the steps of constructing the multivariate regression model are: Step A1, the parameter analysis unit sets multimodal imaging parameters as independent variables, including X-ray tube voltage, X-ray tube current, scanning time, slice thickness of electronic computed tomography and repetition time, echo time and flip angle of magnetic resonance imaging; sets vascular image quality assessment value as dependent variable, collects parameter settings used in historical intravenous thrombolysis and corresponding vascular image quality assessment values ​​as training set; Step A2, the parameter analysis unit estimates the model parameters using the least squares method, ensuring that all variables are defined. After the model input configuration is completed, the estimation process is run to obtain the regression coefficient of each multimodal imaging parameter; Step A3: After the parameter estimation is completed, the parameter analysis unit checks the generated determination coefficient and statistics to confirm the goodness of fit of the model. If the index does not meet the standard, the interaction terms and nonlinear terms between the multimodal imaging parameters are introduced and the estimation process is run again for verification. Step A4, the parameter analysis unit analyzes and records the regression coefficients of the final model, and uses the regression coefficients to indicate the influence of the X-ray tube voltage, X-ray tube current, scanning time, slice thickness of the electronic computed tomography and the repetition time, echo time, and flip angle of the magnetic resonance imaging on the vascular image quality assessment value.

9. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The multimodal parameter optimization unit uses a reinforcement learning algorithm based on the result obtained by the parameter analysis unit to find a multimodal imaging parameter combination, wherein finding a multimodal imaging parameter combination includes the following steps: Step B1, define the state space of reinforcement learning as the current vascular image quality assessment value and multimodal image parameters, including X-ray tube voltage, X-ray tube current, scanning time, layer thickness of electronic computed tomography, repetition time, echo time and flip angle of magnetic resonance imaging; the action space is the adjustable range of these parameters, and the reward function is based on the improvement of the vascular image quality assessment value relative to the previous state, as well as the contrast of the blood vessels in the ischemic area of ​​the comprehensive vascular image and the edge clarity of the vascular wall structure, wherein the formula of the reward function is: Among them, R is the reward value, V new is the new vascular image quality assessment value, V old is the old vascular image quality assessment value, C vessel is the vascular contrast, C wall is the gradient amplitude of the vascular wall area, w1 and w2 are weight factors; Step B2, initialize the Q table, select actions according to the current state, i.e., adjust the image parameters, use the ε-greedy strategy to balance exploration and exploitation, execute the selected actions and observe the new state and reward; Step B3, update the Q value according to the Bellman equation and optimize the decision strategy; Step B4, repeating steps B2 and B3 until a preset number of iterations is reached and the change in the Q value is less than a predetermined threshold; Step B5: Select a multimodal imaging parameter combination from the final learned strategy.

10. The one-stop multimodal image-assisted management system for intravenous thrombolysis according to claim 1, characterized in that: The device control unit receives the multimodal imaging parameter combination from the multimodal parameter optimization unit, automatically adjusts the parameter settings of each imaging device through the standardized device control interface, and feeds back the adjusted parameters to the multimodal imaging acquisition module.

Images