Coronary artery radiography ray dose regulation and control method based on image acquisition quality

Abstract

The invention relates to the technical field of image processing, in particular to a coronary artery radiography ray dose regulation and control method and system based on image acquisition quality, and the method comprises the steps: obtaining a radiography image corresponding to a preset default radiography parameter; performing image enhancement operation on the contrast image to obtain a contrast enhanced image, and performing prediction operation based on diagnostic value on the contrast enhanced image to obtain a diagnostic value score; the method comprises the following steps of: performing dose increasing operation on default radiography parameters to obtain updated radiography parameters, acquiring updated radiography images corresponding to the updated radiography parameters, replacing the radiography images by using the updated radiography images, and replacing the default radiography parameters by using the updated radiography parameters. Returning to the step of generating the confrontation model by using the pre-trained low-dose imaging-enhanced diagnosis; until the diagnostic value score is higher than the high value threshold. According to the invention, the radiation dose can be greatly reduced by improving the diagnostic value in the contrast image to reduce the image quality.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for regulating the X-ray dose of coronary angiography based on the quality of image acquisition. Background Art

[0002] Coronary Angiography (CAG) is an important evidence for diagnosing Coronary Artery Disease (CAD). By injecting a contrast agent into the coronary artery and using X-ray imaging, doctors can accurately identify the locations of vascular stenosis, calcification, and plaques, providing a key basis for stent implantation or bypass surgery.

[0003] This technology relies on ionizing radiation, and the risk of long-term exposure of patients and medical staff to radiation cannot be ignored. Currently, to clearly view coronary angiography images accurately, it is necessary to increase the X-ray dose, but increasing the X-ray dose will harm the physical health of patients. Therefore, balancing the quality of angiography images and the amount of X-ray dose has become an urgent problem to be solved.

[0004] In the current industry, the general pursuit of angiography images is to use the minimum radiation dose as much as possible on the basis of clear angiography images. However, even on the basis of only being able to identify angiography images, the radiation dose is still too high for human body harm. When the angiography image is complex or difficult to confirm, there will be a situation of sacrificing the health of patients to improve the diagnostic accuracy. Summary of the Invention

[0005] The present invention provides a method for regulating the X-ray dose of coronary angiography based on the quality of image acquisition. Its main purpose is to greatly reduce the radiation dose by increasing the diagnostic value in the angiography image while reducing the image quality.

[0006] To achieve the above object, a method for regulating the X-ray dose of coronary angiography based on the quality of image acquisition provided by the present invention includes:

[0007] During the process of performing single-exposure angiography on a pre-identified target heart region using a pre-constructed coronary angiography device, obtaining an angiography image corresponding to preset default angiography parameters;

[0008] Using a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to perform image enhancement operation on the angiography image to obtain an enhanced angiography image, and performing a prediction operation based on the diagnostic value on the enhanced angiography image to obtain a diagnostic value score;

[0009] Judging whether the diagnostic value score is higher than a preset high-value threshold;

[0010] When the diagnostic value score is lower than or equal to the high-value threshold, perform a dose increase operation on the default contrast parameters according to a preset dose adjustment rule to obtain updated contrast parameters, acquire the updated contrast image corresponding to the updated contrast parameters, replace the contrast image with the updated contrast image, and replace the default contrast parameters with the updated contrast parameters, and return to the above step of using the pre-trained generative adversarial model for low-dose imaging enhancement-diagnosis;

[0011] When the diagnostic value score is higher than the high-value threshold, end the process of single-exposure contrast imaging of the target heart region according to the default contrast parameters to obtain the contrast image and the contrast-enhanced image.

[0012] Optionally, before using the pre-trained generative adversarial model for low-dose imaging enhancement-diagnosis, the method further includes:

[0013] Obtain a set of high- and low-dose image samples;

[0014] Obtain a generative adversarial model for low-dose imaging enhancement-diagnosis, where the generative adversarial model includes a generator based on a multi-scale feature fusion network and an image enhancement network, and a discriminator based on a differentiable diagnostic network;

[0015] Obtain a generator loss function, where the generator loss function is expressed as:

[0016]

[0017] In the formula, represents the generator loss function, represents the adversarial loss, represents the structural similarity loss, Dose_Penalty represents that the penalty dose exceeds the preset dose, and λ adv 、λ struct and λ dose all represent weight coefficients;

[0018] Obtain a discriminator loss function, where the discriminator loss function is expressed as:

[0019]

[0020] In the formula, represents the discriminator loss function, I real represents the real high-dose image, I low represents the real low-dose image, I real and I low all belong to the set of high- and low-dose image samples, G(·) represents the output result of the generator, and D(·) represents the output result of the discriminator, represents the expectation operator;

[0021] Using the high - low dose image sample set, according to the generator loss function and the discriminator loss function, train the generative adversarial model to obtain a trained generative adversarial model.

[0022] Optionally, the obtaining of the high - low dose image sample set includes:

[0023] Obtain a human equivalent phantom;

[0024] Use the coronary angiography device to perform angiography operations on the human equivalent phantom with a preset standard radiation dose and a preset minimum radiation dose respectively to obtain a standard angiogram and a minimum angiogram;

[0025] Perform a multi - level low - dose data generation operation on the standard radiation dose and the standard angiogram based on the Poisson noise model to obtain a multi - level low - dose image set;

[0026] Summarize the standard angiogram, the multi - level low - dose image set and the minimum angiogram to obtain a high - low dose image sample set.

[0027] Optionally, the using of the pre - trained generative adversarial model based on low - dose imaging enhancement - diagnosis to perform image enhancement on the angiogram to obtain an enhanced angiogram includes:

[0028] Perform a normalization operation on the angiogram according to a preset pixel scaling range to obtain a normalized angiogram;

[0029] Perform a noise convolution operation on the normalized angiogram based on the Poisson noise model to obtain a predicted noise distribution, and obtain a denoised image signal according to the difference between the angiogram and the predicted noise distribution;

[0030] Use the multi - scale feature fusion network to perform a vascular feature recognition operation on the denoised image signal in a multi - scale vascular branch field of view to obtain a hierarchical vascular feature set;

[0031] Use a pre - constructed residual block to perform feature fusion on the hierarchical vascular feature set to obtain a fused vascular feature set, and perform an up - sampling operation on the fused vascular feature set to obtain an enhanced angiogram.

[0032] Optionally, the performing of a prediction operation based on diagnostic value on the enhanced angiogram to obtain a diagnostic value score includes:

[0033] Perform an authenticity detection on the enhanced angiogram to obtain an authenticity probability;

[0034] Determine whether the authenticity probability is greater than a preset real image threshold;

[0035] When the authenticity probability is greater than the real image threshold, use the differentiable diagnosis network to perform a diagnostic feature extraction operation on the contrast-enhanced image to obtain a set of diagnostic features;

[0036] Perform a lesion presence prediction operation on the set of diagnostic features to obtain a lesion presence probability, and perform a vascular continuity scoring operation on the set of diagnostic features to obtain a vascular continuity score;

[0037] Obtain the contrast-to-noise ratio of the contrast-enhanced image, and perform a weighted calculation on the contrast-to-noise ratio, the lesion presence probability, and the vascular continuity score to obtain a diagnostic value score.

[0038] Optionally, the operation of increasing the dose of the default contrast parameters according to a preset dose adjustment rule to obtain updated contrast parameters includes:

[0039] Obtain an adjustment formula in the preset dose adjustment rule, where the adjustment formula is expressed as:

[0040]

[0041] In the formula, the Δη represents an adjustment amount, ΔS represents a difference in diagnostic value scores, S threshold represents the high value threshold, CNR represents the contrast-to-noise ratio, BMI norm represents the standard body mass index, HR represents the heart rate, and α, β, γ, and δ all represent weight parameters;

[0042] According to the dose adjustment rule, determine the numerical range of the difference in diagnostic value scores;

[0043] If the difference in diagnostic value scores belongs to a preset first numerical range, according to the adjustment formula, perform a mild adjustment on the default contrast parameters to obtain updated contrast parameters, where the mild adjustment is expressed as:

[0044] η new = η current + 0.05·Δη

[0045] In the formula, η new represents the updated contrast parameters, and η current represents the default contrast parameters;

[0046] If the difference in diagnostic value scores belongs to a preset second numerical range, according to the adjustment formula, perform a moderate adjustment on the default contrast parameters to obtain updated contrast parameters, where the moderate adjustment is expressed as:

[0047] η new = η current + 0.1· Δη + 0.02·BMI norm

[0048] When the difference in the diagnostic value scores belongs to a preset third numerical interval, according to the adjustment formula, a severe adjustment is performed on the default contrast parameters to obtain updated contrast parameters, where the severe adjustment is expressed as:

[0049] η new = min(η current + 0.3, η safe )

[0050] In the formula, min(·) represents the minimum value function, and η safe represents the preset safe dose.

[0051] Optionally, the process of ending a single-exposure angiography of the target heart region according to the default contrast parameters includes:

[0052] Monitor the changes of the default contrast parameters to obtain a parameter update sequence, and calculate the numerical change rate of the parameter update sequence;

[0053] Obtain the time point when the numerical change rate is zero. When the time point has passed a preset static exposure duration, end the process of single-exposure angiography of the target heart region;

[0054] During the process of single-exposure angiography of the target heart region, continuously obtain the cumulative dose of the coronary angiography device to the target heart region;

[0055] Judge whether the cumulative dose is greater than a preset safety threshold;

[0056] When the cumulative dose is greater than the safety threshold, end the process of single-exposure angiography of the target heart region.

[0057] Optionally, after obtaining the angiography image and the contrast-enhanced angiography image, the method further includes:

[0058] Obtain the CT image of the target heart region, and construct a key-value pair with the CT image, the standard body mass index, and the updated contrast parameters to obtain a CT image-contrast parameter sample;

[0059] Store the CT image-contrast parameter sample in a pre-constructed empirical database;

[0060] According to a preset time frequency, using each CT image-contrast parameter sample stored in the experience database, train a pre-constructed initialization contrast parameter prediction model to obtain a trained initialization contrast parameter prediction model;

[0061] Obtain the physical information indicators of a new patient, and obtain the CT image of the new patient. Using the trained initialization contrast parameter prediction model, according to the CT image of the new patient and the physical information indicators of the new patient, predict the default contrast parameters of the new patient.

[0062] To achieve the above object, the present invention also provides a coronary angiography ray dose regulation system based on image acquisition quality, including:

[0063] An initialization contrast module, configured to obtain a contrast image corresponding to preset default contrast parameters during the process of performing single-exposure angiography on a pre-confirmed target heart region using a pre-constructed coronary angiography device;

[0064] An image enhancement and value recognition module, configured to use a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to perform an image enhancement operation on the contrast image to obtain a contrast-enhanced image, and perform a prediction operation based on diagnostic value on the contrast-enhanced image to obtain a diagnostic value score;

[0065] A contrast parameter update module, configured to determine whether the diagnostic value score is higher than a preset high-value threshold, and when the diagnostic value score is lower than or equal to the high-value threshold, perform a dose increase operation on the default contrast parameters according to a preset dose adjustment rule to obtain updated contrast parameters, obtain an updated contrast image corresponding to the updated contrast parameters, and use the updated contrast image to replace the contrast image, and use the updated contrast parameters to replace the default contrast parameters, and return to the above step of using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis;

[0066] A contrast image output module, configured to end the process of performing single-exposure angiography on the target heart region according to the default contrast parameters when the diagnostic value score is higher than the high-value threshold, to obtain the contrast image and the contrast-enhanced image.

[0067] Optionally, the generative adversarial model based on low-dose imaging enhancement-diagnosis includes:

[0068] Obtain a set of high- and low-dose image samples;

[0069] Obtain a generative adversarial model based on low-dose imaging enhancement-diagnosis, wherein the generative adversarial model includes a generator based on a multi-scale feature fusion network and an image enhancement network, and a discriminator based on a differentiable diagnosis network;

[0070] Obtain the generator loss function, where the generator loss function is expressed as:

[0071]

[0072] In the formula, represents the generator loss function, represents the adversarial loss, represents the structural similarity loss, Dose_Penalty represents that the penalty dose exceeds the preset dose, and λ adv , λ struct and λ dose represent weight coefficients;

[0073] Obtain the discriminator loss function, where the discriminator loss function is expressed as:

[0074]

[0075] In the formula, represents the discriminator loss function, I real represents the real high-dose image, I low represents the real low-dose image, I real and I low belong to the high-low dose image sample set, G(·) represents the output result of the generator, D(·) represents the output result of the discriminator, represents the expectation operator;

[0076] Utilize the high-low dose image sample set, and train the generative adversarial model according to the generator loss function and the discriminator loss function to obtain a trained generative adversarial model.

[0077] To solve the above problems, the present invention also provides an electronic device, which includes:

[0078] A memory that stores at least one instruction;

[0079] A processor that executes the instructions stored in the memory to implement the above-mentioned coronary angiography ray dose regulation method based on image acquisition quality.

[0080] To solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned coronary angiography ray dose regulation method based on image acquisition quality.

[0081] To solve the problems described in the background art, the present invention first performs single-exposure angiography on the target heart area using common coronary angiography equipment, and then updates the angiography parameters to the optimal state during the single-exposure angiography. Herein, the optimal state means that the angiography image cannot be observed by medical staff, but relatively useful diagnostic features can be recognized by a computer. The present invention uses the generator of a generative adversarial network to perform image enhancement as realistically as possible, and uses the discriminator to judge as realistically as possible whether the enhanced image is real and rich in diagnostic information. Since the requirement for the clarity of the angiography image is basically abandoned, the radiation dose is greatly reduced, thereby ensuring the health and safety of patients and operators. Therefore, the present invention greatly reduces the radiation dose by improving the diagnostic value in the angiography image while reducing the image quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 FIG. is a schematic flowchart of a method for regulating the radiation dose of coronary angiography based on image acquisition quality provided by an embodiment of the present invention;

[0083] Figure 2 FIG. is a functional module diagram of a system for regulating the radiation dose of coronary angiography based on image acquisition quality provided by an embodiment of the present invention;

[0084] Figure 3 FIG. is a schematic structural diagram of an electronic device for implementing the method for regulating the radiation dose of coronary angiography based on image acquisition quality provided by an embodiment of the present invention.

[0085] DESCRIPTION OF REFERENCE NUMERALS:

[0086] 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0087] The implementation, functional features, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0088] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0089] An embodiment of the present application provides a method for regulating the radiation dose of coronary angiography based on image acquisition quality. The execution subject of the method for regulating the radiation dose of coronary angiography based on image acquisition quality includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for regulating the radiation dose of coronary angiography based on image acquisition quality can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to: a single server, a server cluster, a cloud server, or a cloud server cluster, etc.

[0090] Refer to Figure 1 As shown, it is a schematic flowchart of a method for regulating the X-ray dose of coronary angiography based on the image acquisition quality provided by an embodiment of the present invention. In this embodiment, the method for regulating the X-ray dose of coronary angiography based on the image acquisition quality includes:

[0091] S1. During the process of performing single-exposure angiography on a pre-confirmed target heart region using a pre-built coronary angiography device, obtain a contrast image corresponding to preset default angiography parameters.

[0092] Among them, the coronary angiography device is a medical device used to display the coronary artery structure and blood flow conditions through X-ray imaging and contrast agents.

[0093] Among them, the target heart region refers to the heart region of a certain patient, and the observation angle of the target heart region is determined by the most effective radiation angle considered by medical staff.

[0094] Among them, the process of single-exposure angiography is not the process within a single exposure frame period, but a continuous exposure process, and during the continuous exposure process, the previously set default angiography parameters can be adjusted, so that the contrast image changes accordingly.

[0095] Among them, the default angiography parameters include ray energy, ray dose, and single-exposure duration. The contrast image is the output result of the coronary angiography device.

[0096] Specifically, in the embodiment of the present invention, only the process of single-exposure angiography is considered, and efforts are made to complete the adjustment of angiography parameters within a single exposure period to obtain a contrast image with diagnostic information.

[0097] S2. Use a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to perform image enhancement operations on the contrast image to obtain a contrast-enhanced image, and perform prediction operations based on diagnostic value on the contrast-enhanced image to obtain a diagnostic value score.

[0098] Among them, the generative adversarial model based on low-dose imaging enhancement-diagnosis is a neural network model, including a generator and a discriminator. The generator is used to achieve "low-dose imaging enhancement" and perform image enhancement operations on the low-dose imaging results. The discriminator is used to achieve "diagnosis", determine whether the enhanced image generated by the generator is real, and guide the generator to pay more attention to enhancing some contrast information that can be used for diagnosis during image enhancement.

[0099] Among them, the image enhancement operation refers to the process of reducing noise and improving some key information in the image, such as the accuracy of blood vessel distribution. The contrast-enhanced image is the enhanced result of the contrast image, for example, making the tiny blood vessels invisible in the contrast image visible, or clearly showing the blood vessel structure of the coronary artery.

[0100] Among them, the diagnostic value refers to the degree of support of medical images for accurately identifying lesions (such as stenosis and calcification).

[0101] Among them, the prediction operation refers to a neural network prediction function in the discriminator, which can be implemented through a fully connected network or a random forest network. The diagnostic value score is the output result of the prediction operation, which helps to guide the generator to generate contrast-enhanced images with a higher diagnostic value score.

[0102] Specifically, in the embodiment of the present invention, before using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis, the method further includes:

[0103] Obtain a set of high-dose and low-dose image samples;

[0104] Obtain a generative adversarial model based on low-dose imaging enhancement-diagnosis, where the generative adversarial model includes a generator based on a multi-scale feature fusion network and an image enhancement network, and a discriminator based on a differentiable diagnosis network;

[0105] Obtain the generator loss function, where the generator loss function is expressed as:

[0106]

[0107] In the formula, represents the generator loss function, represents the adversarial loss, represents the structural similarity loss, Dose_Penalty represents that the penalty dose exceeds the preset dose, and λ adv 、λ struct and λ dose all represent weight coefficients;

[0108] Obtain the discriminator loss function, where the discriminator loss function is expressed as:

[0109]

[0110] In the formula, represents the discriminator loss function, I real represents the real high-dose image, I low represents the real low-dose image, I real and I lowBoth belong to the high- and low-dose image sample set, G(·) represents the output result of the generator, and D(·) represents the output result of the discriminator. represents the expectation operator;

[0111] Using the high- and low-dose image sample set, according to the generator loss function and the discriminator loss function, train the generative adversarial model to obtain a trained generative adversarial model.

[0112] Among them, the high- and low-dose image sample set refers to a set of contrast images corresponding to radiating agents of different dose levels.

[0113] Among them, the multi-scale feature fusion network is a neural network structure that can capture the global vascular tree and local plaque details, and is used to solve the problem that the original contrast image may lose small blood vessels, resulting in the generator being unable to restore the real structure.

[0114] Among them, the image enhancement network is a neural network structure that can reduce the noise in the image and improve the clarity of the target object part.

[0115] Among them, the differentiable diagnosis network is a technology that embeds clinical diagnosis logic into the deep learning framework and directly correlates image generation with the diagnosis result through gradient optimization, and is used to enable the model to reverse and optimize the image generation process from the diagnosis result through "differentiability" in mathematics, and is especially suitable for the GAN (generative adversarial network) low-dose imaging enhancement of this solution.

[0116] Among them, the generator loss function is a loss function that measures the gap between the structure (contrast-enhanced image) generated by the generator and the real result (contrast image). The adversarial loss refers to forcing the generator to generate images consistent with the real image distribution through the feedback of the discriminator to deceive the discriminator. The structural similarity loss refers to quantifying the similarity in brightness, contrast, and structure between the contrast-enhanced image and the real contrast image, and retaining details such as vascular topology. The penalty for the dose exceeding the preset dose is a constraint condition. Since the contrast-enhanced image generated by the generator has a certain guiding role for the adjustment of the next contrast parameter, the preset dose is used to ensure that the updated contrast parameter after adjustment does not exceed the preset safety limit, giving priority to ensuring the safety of patients.

[0117] Among them, the discriminator loss function is a loss function that measures the gap between the contrast-enhanced image generated by the generator and the real contrast image. The real high-dose image is the real contrast image corresponding to the standard dose in the high-low dose image sample set. Similarly, the real low-dose image is the real contrast image corresponding to 10% of the standard dose in the high-low dose image sample set. The expectation operator represents the weighted average of all possible values of a random variable according to the probability distribution. In the GAN of the present invention, it specifically refers to the statistical average of the real data distribution (such as the real high-dose image) or the generated distribution (such as the contrast-enhanced image).

[0118] Among them, the training refers to an iterative optimization process of adjusting model parameters (such as the weight coefficients) to minimize each loss function, enabling the model to learn from the input data and improve the performance of specific tasks.

[0119] Specifically, since the present invention pursues a reduction in radiation dose. Therefore, the dose levels of "high and low doses" in the high-low dose image sample set are less than or equal to 100% of the standard dose. In addition, studies have shown that: when the radiation dose is less than 10% of the standard dose, the associated features between the contrast image and the coronary artery cannot be clearly identified. Therefore, the high-low dose image sample set is in the range of 10% to 100%.

[0120] Specifically, in the embodiment of the present invention, the multi-scale feature fusion network first performs global average pooling on the contrast image of the coronary artery to enhance the contrast of the main blood vessels. Then it performs medium-scale pooling to enhance the sharpness of the edges of secondary branches. Finally, it performs local feature extraction to adaptively enhance small blood vessels.

[0121] Specifically, in the embodiment of the present invention, using the high-low dose image sample set with data augmentation, according to the preset generator loss function and discriminator loss function, the network framework of the pre-constructed generative adversarial model is trained to obtain a trained generative adversarial model. Among them, the specific training process is a common process of neural networks and will not be elaborated here.

[0122] In detail, in the embodiment of the present invention, the obtaining of the high-low dose image sample set includes:

[0123] Obtain a human equivalent phantom;

[0124] Use the coronary angiography device to perform contrast operations on the human equivalent phantom with a preset standard radiation dose and a preset minimum radiation dose respectively to obtain a standard contrast image and a minimum contrast image;

[0125] Perform multi-level low-dose data generation operations on the standard radiation dose and the standard contrast image based on the Poisson noise model to obtain a multi-level low-dose image set;

[0126] Summarize the standard contrast images, the multi-level low-dose image set, and the lowest contrast image to obtain a high-low dose image sample set.

[0127] Among them, the human equivalent phantom is provided by the National Institute for Radiation Protection and Nuclear Safety, Chinese Center for Disease Control and Prevention. The phantom is shaped by human tissue equivalent materials and has the same density as the corresponding tissues of a normal human body. It is divided into three sections: head and neck, chest and abdomen, and abdomen and buttocks, and includes preformed heart, liver, lungs, and bones, etc.

[0128] Among them, the contrast operation is the same as the contrast step in S1 and is implemented by a coronary angiography device.

[0129] Among them, the standard radiation dose and the lowest radiation dose are 100% and 10% of the standard dose respectively, and the standard dose is the dose evaluated by a professional model or medical staff according to various human body indicators. The standard contrast image and the lowest contrast image are the contrast images of the standard radiation dose and the lowest radiation dose respectively.

[0130] Among them, the Poisson noise model is a statistical model that describes the randomness of X-ray photon counting. The Poisson noise model usually assumes that the number of photons received by the detector follows a Poisson distribution, so the noise in the contrast image at low doses is more significant. The present invention is used to simulate the noise characteristics of low-dose medical images.

[0131] Among them, the multi-level low-dose data generation operation refers to the process of predicting the contrast images at different levels of doses such as 90%, 60%, and 30% based on the 100% standard radiation dose and the standard contrast image through the Poisson noise model. The multi-level low-dose image set is the set of contrast images at different levels of doses such as 90%, 60%, and 30%.

[0132] Specifically, in the embodiment of the present invention, a contrast simulation experiment at 100% and 10% doses is carried out through a human equivalent phantom to avoid harm to real people, and then through the multi-level low-dose data generation operation based on the Poisson noise model, some contrast images between 100% and 10% doses are predicted, so as to further expand the database. Finally, the standard contrast images, the multi-level low-dose image set, and the lowest contrast image are summarized to form a high-low dose image sample set.

[0133] In detail, in the embodiment of the present invention, the image enhancement operation is performed on the contrast image by using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to obtain a contrast-enhanced image, including:

[0134] Perform a normalization operation on the contrast image according to a preset pixel scaling range to obtain a normalized contrast image;

[0135] Perform a noise convolution operation on the normalized contrast image based on the Poisson noise model to obtain a predicted noise distribution, and obtain a denoised image signal according to the difference between the contrast image and the predicted noise distribution;

[0136] Use the multi-scale feature fusion network to perform a vascular feature recognition operation on the denoised image signal in a multi-scale vascular branch field of view to obtain a hierarchical vascular feature set;

[0137] Use a pre-constructed residual block to perform feature fusion on the hierarchical vascular feature set to obtain a fused vascular feature set, and perform an upsampling operation on the fused vascular feature set to obtain a contrast-enhanced image.

[0138] Wherein, the pixel scaling range refers to linearly or non-linearly mapping the pixel values in the contrast image to a specified interval (such as [0,1] or [-1,1]) to adapt to the input requirements of the generative adversarial network model.

[0139] Wherein, the normalization refers to an operation that eliminates the dimensional difference of data and improves the stability of model training through standardization (such as Z-score: mean 0, variance 1) or maximum-minimum scaling. The normalized contrast image is the normalization result of the contrast image.

[0140] Wherein, the noise convolution operation is a feature extraction operation for noise objects in the Poisson noise model. The predicted noise distribution is the feature extraction result of the noise convolution operation.

[0141] Wherein, the difference refers to the inverse fitting between signals, which is used to delete the components of the predicted noise distribution in the contrast image to achieve the noise reduction operation. The denoised image signal is the result of the noise reduction operation.

[0142] Wherein, the vascular feature recognition is the operation of "first performing global average pooling on the coronary angiogram image to enhance the contrast of the main trunk vessels. Then performing mesoscale pooling to enhance the sharpness of the secondary branch edges. Finally performing local feature extraction to adaptively enhance the fine vessels" as described above. The hierarchical vascular feature set is the result set of the vascular feature recognition operation.

[0143] Wherein, the residual block is a neural network module that realizes residual learning through skip connections (adding the input and output) to solve the problem of gradient disappearance in deep networks and improve feature reuse and training stability.

[0144] Wherein, the feature fusion refers to the operation of integrating features from different levels, branches or modalities through concatenation, weighting or attention mechanisms to comprehensively utilize multi-source information. The fused vascular feature set is the feature fusion result of the hierarchical vascular feature set.

[0145] Among them, the upsampling operation refers to an operation that increases the spatial resolution of an image or feature map through methods such as interpolation or transposed convolution, and is used to restore details from low-resolution data. The contrast-enhanced image is the result of the upsampling process.

[0146] Specifically, in the embodiments of the present invention, first, the contrast image is transformed into a normalized contrast image suitable for the input format of the model through the pixel scaling range. Then, the Poisson noise model is used to separate the noise and the signal to obtain the denoised image signal. Then, the multi-scale feature fusion network is used to extract the features of blood vessels or plaques in the contrast image from different fields of view to obtain a hierarchical blood vessel feature set. The present invention fuses the hierarchical blood vessel feature set into a fused blood vessel feature set through residual blocks to realize the comprehensive utilization of multi-source information. Finally, by performing an upsampling operation on the fused blood vessel feature set, the data details are restored to obtain a contrast-enhanced image.

[0147] Specifically, in the embodiments of the present invention, the operation of predicting the diagnostic value for the contrast-enhanced image to obtain a diagnostic value score includes:

[0148] Performing authenticity detection on the contrast-enhanced image to obtain an authenticity probability;

[0149] Judging whether the authenticity probability is greater than a preset real image threshold;

[0150] When the authenticity probability is greater than the real image threshold, using the differentiable diagnosis network to perform a diagnostic feature extraction operation on the contrast-enhanced image to obtain a diagnostic feature set;

[0151] Performing a lesion presence prediction operation on the diagnostic feature set to obtain a lesion presence probability, and performing a blood vessel continuity scoring operation on the diagnostic feature set to obtain a blood vessel continuity score;

[0152] Obtaining the contrast-to-noise ratio of the contrast-enhanced image, and performing a weighted calculation on the contrast-to-noise ratio, the lesion presence probability, and the blood vessel continuity score to obtain a diagnostic value score.

[0153] Among them, the authenticity detection refers to the process of judging whether the contrast-enhanced image is a real contrast image rather than a generated or tampered contrast-enhanced image through an algorithm, so as to ensure the fidelity and credibility of the contrast-enhanced image. The authenticity probability is the output result of the authenticity detection.

[0154] Among them, the real image threshold is configured to be 90%.

[0155] Among them, the diagnostic feature extraction operation is the feature extraction operation for the "diagnosis" object in the differentiable diagnosis network. The diagnostic feature set is the feature extraction result of the diagnostic feature extraction operation.

[0156] Among them, the lesion presence prediction operation is another neural network function in the discriminator, which is used to determine whether there is a lesion area in the contrast-enhanced image. If there is a lesion area, it indicates that the contrast-enhanced image has diagnostic value. The lesion presence probability is the prediction result of the lesion presence prediction operation.

[0157] Among them, the contrast-to-noise ratio is the ratio of the mean difference between the target area (such as blood vessels) and the background in the medical image to the noise levels of both.

[0158] Among them, the diagnostic value score refers to a comprehensive score generated by quantitatively evaluating the degree of support of the contrast-enhanced image for clinical diagnosis, combining image quality indicators (such as CNR, SSIM) and the probability of lesion detection, and is used to determine whether the enhanced image meets the diagnostic requirements.

[0159] Specifically, in the embodiment of the present invention, the primary task of the discriminator is to detect the authenticity of the contrast-enhanced image generated by the generator to prevent the generator from generating false enhancement information. When the authenticity probability is greater than 90%, it indicates that the contrast-enhanced image is true and credible, and then the feature extraction operation based on diagnostic features can be performed through the differentiable diagnosis network to obtain the diagnostic feature set. In addition, the present invention can also predict the presence of lesions and the vascular continuity in the diagnostic feature set through a neural network, and obtain the lesion presence probability and the vascular continuity score respectively.

[0160] Specifically, in the embodiment of the present invention, the contrast-to-noise ratio is obtained by calculating the ratio of the pixels of the "blood vessel" part to the pixels of the "background" part in the contrast-enhanced image. Since the contrast-to-noise ratio, the lesion presence probability, and the vascular continuity score are all evaluation indicators for determining whether the contrast-enhanced image has diagnostic value. Therefore, the diagnostic value score can be obtained by performing a weighted calculation on the contrast-to-noise ratio, the lesion presence probability, and the vascular continuity score. The specific weight coefficients are obtained by experimental statistics by technicians.

[0161] S3. Determine whether the diagnostic value score is higher than a preset high-value threshold.

[0162] In the embodiment of the present invention, the high-value threshold is configured to be 90%.

[0163] Specifically, in the embodiment of the present invention, determining whether the diagnostic value score is higher than a preset high-value threshold helps to determine whether the contrast-enhanced image has a therapeutic guiding effect.

[0164] When the diagnostic value score is lower than or equal to the high-value threshold, S4. According to the preset dose adjustment rule, perform a dose increase operation on the default contrast parameters to obtain updated contrast parameters, acquire the updated contrast image corresponding to the updated contrast parameters, and use the updated contrast image to replace the contrast image, and use the updated contrast parameters to replace the default contrast parameters, and return to the above steps of using the pre-trained generative adversarial model for low-dose imaging-enhanced diagnosis.

[0165] Among them, when the diagnostic value score is lower than or equal to the high-value threshold, it indicates that the current contrast-enhanced image has no reference significance for disease judgment, and the dose needs to be increased to improve the effective information in the contrast image.

[0166] Among them, the dose adjustment rule means that the modification degree of the default contrast parameters is different for different scenarios. The dose increase operation is the execution process of the dose adjustment rule. The updated contrast parameters are the correction results of the default contrast parameters. The updated contrast image is the image automatically generated by the coronary angiography device according to the default contrast parameters.

[0167] Specifically, in the embodiment of the present invention, the step of performing a dose increase operation on the default contrast parameters according to the preset dose adjustment rule to obtain updated contrast parameters includes:

[0168] Obtain the adjustment formula in the preset dose adjustment rule, where the adjustment formula is expressed as:

[0169]

[0170] In the formula, the Δη represents the adjustment amount, ΔS represents the difference in diagnostic value scores, S threshold represents the high-value threshold, CNR represents the contrast-to-noise ratio, BMI norm represents the standard body mass index, HR represents the heart rate, and α, β, γ, and δ all represent weight parameters;

[0171] According to the dose adjustment rule, judge the numerical range of the difference in diagnostic value scores;

[0172] If the difference in diagnostic value scores belongs to the preset first numerical range, according to the adjustment formula, perform a mild adjustment on the default contrast parameters to obtain updated contrast parameters, where the mild adjustment is expressed as:

[0173] η new =η current +0.05·Δη

[0174] In the formula, η new represents the updated contrast parameters, η currentrepresenting the default contrast parameters;

[0175] When the difference in diagnostic value scores belongs to a preset second numerical range, according to the adjustment formula, moderately adjust the default contrast parameters to obtain updated contrast parameters, where the moderate adjustment is expressed as:

[0176] η new = η current + 0.1· Δη + 0.02·BMI norm

[0177] When the difference in diagnostic value scores belongs to a preset third numerical range, according to the adjustment formula, severely adjust the default contrast parameters to obtain updated contrast parameters, where the severe adjustment is expressed as:

[0178] η new = min(η current + 0.3, η safe )

[0179] In the formula, min(·) represents the minimum value function, and η safe represents the preset safety dose.

[0180] Among them, the adjustment formula refers to the formula for calculating the amount by which the default contrast parameters need to change. Among them, the adjustment amount is the result of substituting values into the adjustment formula. The difference in diagnostic value scores is the difference between the high-value threshold and the default contrast parameters. The standard body mass index refers to the internationally common weight status classification index, and the calculation formula is weight (kg) divided by the square of height (m). The weight parameters such as α, β, γ, and δ are used to measure the importance of each parameter item.

[0181] Among them, the safety dose is configured as 0.5.

[0182] Among them, the first numerical range is the numerical range less than 0.1. The mild adjustment is an adjustment method calculated according to the mild adjustment formula. The mild adjustment formula is a formula configured according to the experience of medical staff. Similarly for the subsequent moderate adjustment formula and severe adjustment formula.

[0183] Among them, the second numerical range is the numerical range of 0.1 - 0.3. The moderate adjustment is an adjustment method calculated according to the moderate adjustment formula.

[0184] Among them, the third numerical range is the numerical range greater than or equal to 0.3. The severe adjustment is an adjustment method calculated according to the severe adjustment formula.

[0185] Specifically, in the embodiments of the present invention, according to a preset dose adjustment rule, fine adjustment is performed to achieve fine-tuning of the dose. Priority is given to optimizing the kVp (such as +5 kV) rather than the mA to reduce the noise amplification effect. Through moderate adjustment, the kVp (+10 kV) and mA (+10%) are synchronously increased to balance the penetration and photon flux. Through severe adjustment, a direct switch to a preset safe dose is achieved to avoid over-frequent iteration.

[0186] Further, in the embodiments of the present invention, if a dose increase operation is performed on the default contrast parameters according to the dose adjustment rule to obtain updated contrast parameters, the updated contrast parameters can directly replace the default contrast parameters during the current exposure contrast process, and then the coronary angiography device continues to work to obtain updated contrast images.

[0187] In the embodiments of the present invention, to ensure that the updated contrast images also conform to the previous contrast image judgment process, the updated contrast images are used to replace the previous contrast images.

[0188] When the diagnostic value score is higher than the high value threshold, S5. According to the updated contrast parameters, end the process of single-exposure contrast of the target heart region to obtain the contrast image and the contrast-enhanced image.

[0189] Among them, when the diagnostic value score is higher than the high value threshold, it indicates that the authenticity probability and diagnostic value of the contrast-enhanced image of the contrast image are both higher than 90%, and the increase of the radiation dose can be stopped, and the current exposure contrast process can be ended.

[0190] Specifically, in the embodiments of the present invention, the process of ending the single-exposure contrast of the target heart region according to the default contrast parameters includes:

[0191] Monitor the changes of the default contrast parameters to obtain a parameter update sequence, and calculate the numerical change rate of the parameter update sequence;

[0192] Obtain the time point when the numerical change rate is zero. When the time point passes through a preset static exposure duration, end the process of single-exposure contrast of the target heart region;

[0193] During the process of single-exposure contrast of the target heart region, continuously obtain the cumulative dose of the coronary angiography device to the target heart region;

[0194] Judge whether the cumulative dose is greater than a preset safety threshold;

[0195] When the cumulative dose is greater than the safety threshold, end the process of single-exposure contrast of the target heart region.

[0196] Among them, the change monitoring refers to the operation of recording each update process of the default contrast parameters. The parameter update sequence is the result of the change monitoring. Among them, the numerical change rate refers to the change amplitude of the updated contrast parameters per unit time.

[0197] Among them, the static exposure duration refers to the time length of the coronary angiography device for exposing one frame. It can indicate that the updated contrast parameters do not change again during the process of completing the angiography (indicating stability).

[0198] Among them, the cumulative dose refers to: the total radiation dose received by a patient in a single or multiple medical imaging examinations, which is jointly determined by kVp, mA, exposure time, and irradiation area.

[0199] Among them, the safety threshold is determined according to the cumulative dose of the patient within half a year.

[0200] Specifically, in the embodiment of the present invention, the time length of the single-exposure angiography is determined by the angiography eligibility (whether the contrast parameters can generate diagnostically valuable images) and the cumulative dose. If the angiography eligibility is low and the contrast parameters are continuously iterated during a single exposure, there may be a malfunction. For the safety of the patient, the current exposure angiography process is directly ended through the cumulative dose.

[0201] In detail, in the embodiment of the present invention, after obtaining the angiography image and the contrast-enhanced image, the method further includes:

[0202] Obtain the CT image of the target heart region, and construct a key-value pair with the CT image, the standard body mass index, and the updated contrast parameters to obtain a CT image-contrast parameter sample;

[0203] Store the CT image-contrast parameter sample in a pre-constructed experience database;

[0204] According to a preset time frequency, use each CT image-contrast parameter sample stored in the experience database to train a pre-constructed initialization contrast parameter prediction model to obtain a trained initialization contrast parameter prediction model;

[0205] Obtain the body information indicators of a new patient and obtain the CT image of the new patient. Use the trained initialization contrast parameter prediction model to predict the default contrast parameters of the new patient according to the CT image of the new patient and the body information indicators of the new patient.

[0206] Among them, the CT image refers to a tomographic anatomical image generated by computer reconstruction after the X-ray beam rotates around the body, which reflects tissue density differences in gray scale and is used to display internal structures and lesions such as organs and blood vessels.

[0207] Among them, the operation of constructing key-value pairs refers to the process of taking the updated contrast parameters as keys and the CT images and standard body mass index as values.

[0208] Among them, in the CT image-contrast parameter samples, the keys serve as the labels of the samples, and the values serve as the data content of the samples.

[0209] Among them, the empirical database refers to a database constructed through data alliance technology, which can store CT image-contrast parameter samples corresponding to various medical processes in each medical institution.

[0210] Among them, the time frequency is configured to be once a month.

[0211] Among them, the initialization of the contrast parameter prediction model is a regression network model used to learn the mapping relationship between the keys and values.

[0212] Among them, the new patient refers to a patient after the patient in the target heart region. The detection process of the patient in the target heart region can be regarded as the empirical data for the detection of new patients.

[0213] Among them, the body information index is equivalent to the standard body mass index, but the people are different.

[0214] Specifically, in the embodiments of the present invention, the default contrast parameters preset in the coronary angiography device should be adjusted before radiating the patient to reduce the number of updates of the contrast parameters. Most of the previous default contrast parameters were configured based on expert experience. After the method of the present invention, the default contrast parameters should be configured with extremely low radiation doses, and the previous expert experience fails. It is necessary to configure an initialization contrast parameter prediction model that identifies the default contrast parameters based on CT images and body information indicators. Thus, the accuracy of the initialization configuration of the default contrast parameters is improved.

[0215] To solve the problems described in the background art, the present invention first performs single-exposure angiography on the target heart region through a common coronary angiography device, and then updates the contrast parameters to the optimal state during the single-exposure angiography. Among them, the optimal state refers to a state where the angiography image cannot be observed by medical staff but can be recognized by a computer to have relatively useful diagnostic features; the present invention uses the generator of the generative adversarial network to perform image enhancement as realistically as possible, and uses the discriminator to judge as realistically as possible whether the enhanced image is real and rich in diagnostic information; since the requirement for the clarity of the angiography image is basically abandoned, the radiation dose is greatly reduced, thereby ensuring the health and safety of patients and operators. Therefore, the present invention greatly reduces the radiation dose by improving the diagnostic value in the angiography image while reducing the image quality.

[0216] Such asFigure 2 As shown, it is a functional block diagram of a coronary angiography ray dose regulation system based on image acquisition quality provided by an embodiment of the present invention.

[0217] The coronary angiography ray dose regulation system 100 based on image acquisition quality of the present invention can be installed in an electronic device. According to the functions achieved, the coronary angiography ray dose regulation system 100 based on image acquisition quality can include an initialization angiography module 101, an image enhancement and value recognition module 102, an angiography parameter update module 103, and an angiography image output module 104. The modules of the present invention can also be referred to as units, which refer to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device.

[0218] The initialization angiography module 101 is used to obtain an angiography image corresponding to preset default angiography parameters during the process of performing single-exposure angiography on a pre-confirmed target heart region using a pre-built coronary angiography device;

[0219] The image enhancement and value recognition module 102 is used to perform an image enhancement operation on the angiography image using a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to obtain an enhanced angiography image, and perform a prediction operation based on diagnostic value on the enhanced angiography image to obtain a diagnostic value score;

[0220] The angiography parameter update module 103 is used to determine whether the diagnostic value score is higher than a preset high-value threshold, and when the diagnostic value score is lower than or equal to the high-value threshold, perform a dose increase operation on the default angiography parameters according to a preset dose adjustment rule to obtain updated angiography parameters, obtain an updated angiography image corresponding to the updated angiography parameters, replace the angiography image with the updated angiography image, replace the default angiography parameters with the updated angiography parameters, and return to the above step of using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis;

[0221] The angiography image output module 104 is used to end the process of performing single-exposure angiography on the target heart region according to the default angiography parameters when the diagnostic value score is higher than the high-value threshold, and obtain the angiography image and the enhanced angiography image.

[0222] Among them, the construction process of the device of the generative adversarial model based on low-dose imaging enhancement-diagnosis:

[0223] Obtain a set of high- and low-dose image samples;

[0224] Obtain a generative adversarial model for low-dose imaging enhancement-diagnosis, wherein the generative adversarial model includes a generator based on a multi-scale feature fusion network and an image enhancement network, and a discriminator based on a differentiable diagnosis network;

[0225] Obtain a generator loss function, wherein the generator loss function is expressed as:

[0226]

[0227] In the formula, represents the generator loss function, represents the adversarial loss, represents the structural similarity loss, Dose_Penalty represents that the penalty dose exceeds the preset dose, and λ adv 、λ struct and λ dose represent weight coefficients;

[0228] Obtain a discriminator loss function, wherein the discriminator loss function is expressed as:

[0229]

[0230] In the formula, represents the discriminator loss function, I real represents the real high-dose image, I low represents the real low-dose image, I real and I low belong to the high-low dose image sample set, G(·) represents the output result of the generator, D(·) represents the output result of the discriminator, represents the expectation operator;

[0231] Utilize the high-low dose image sample set, and train the generative adversarial model according to the generator loss function and the discriminator loss function to obtain a trained generative adversarial model.

[0232] Specifically, each module in the coronary angiography X-ray dose regulation system 100 based on image acquisition quality in the embodiments of the present invention adopts the same technical means as those in the above-mentioned Figure 1 and can produce the same technical effects, which will not be elaborated here.

[0233] As Figure 3 shown, it is a schematic structural diagram of an electronic device for implementing the coronary angiography X-ray dose regulation method based on image acquisition quality provided by an embodiment of the present invention.

[0234] The electronic device 1 may include a processor 10, a memory 11, and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a program for regulating the X-ray dose of coronary angiography based on image acquisition quality.

[0235] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the electronic device 1, such as the mobile hard disk of the electronic device 1. In some other embodiments, the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 1. Further, the memory 11 further includes the internal storage unit of the electronic device 1 and also includes an external storage device. The memory 11 can not only be used to store application software installed in the electronic device 1 and various types of data, such as the code of the program for regulating the X-ray dose of coronary angiography based on image acquisition quality, but can also be used to temporarily store data that has been output or will be output.

[0236] In some embodiments, the processor 10 may be composed of integrated circuits. For example, it may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same or different functions, including a combination of one or more Central Processing Units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting various components of the entire electronic device through various interfaces and circuits, and by running or executing programs or modules stored in the memory 11 (such as the program for regulating the X-ray dose of coronary angiography based on image acquisition quality, etc.), and calling data stored in the memory 11, to execute various functions of the electronic device 1 and process data.

[0237] The bus 12 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to implement connection communication between the memory 11 and at least one processor 10, etc.

[0238] Figure 3 Only an electronic device with components is shown. Those skilled in the art can understand that Figure 3 the shown structure does not constitute a limitation on the electronic device 1, and it may include fewer or more components than shown, or combine certain components, or have a different component arrangement.

[0239] For example, although not shown, the electronic device 1 may further include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so as to implement functions such as charge management, discharge management, and power consumption management through the power management device. The power source may also include any components such as one or more DC or AC power sources, a recharge device, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device 1 may also include various sensors, a Bluetooth module, a Wi-Fi module, etc., which will not be elaborated here.

[0240] Further, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0241] Optionally, the electronic device 1 may further include a user interface. The user interface may be a display, an input unit (such as a keyboard), and optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visual user interface.

[0242] The program of the method for regulating the X-ray dose of coronary angiography based on image acquisition quality stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve the following:

[0243] During the process of performing single-exposure angiography on a pre-confirmed target heart region using a pre-built coronary angiography device, obtain an angiography image corresponding to preset default angiography parameters;

[0244] Use a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to perform image enhancement operations on the angiography image to obtain an enhanced angiography image, and perform prediction operations based on diagnostic value on the enhanced angiography image to obtain a diagnostic value score;

[0245] Determine whether the diagnostic value score is higher than a preset high-value threshold;

[0246] When the diagnostic value score is lower than or equal to the high-value threshold, perform a dose increase operation on the default angiography parameters according to a preset dose adjustment rule to obtain updated angiography parameters, obtain an updated angiography image corresponding to the updated angiography parameters, use the updated angiography image to replace the angiography image, use the updated angiography parameters to replace the default angiography parameters, and return to the above step of using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis;

[0247] When the diagnostic value score is higher than the high-value threshold, end the process of performing single-exposure angiography on the target heart region according to the default angiography parameters to obtain the angiography image and the enhanced angiography image.

[0248] Specifically, for the specific implementation method of the above instructions by the processor 10, reference can be made to Figures 1 to 3 the description of the relevant steps in the corresponding embodiments, which will not be elaborated here.

[0249] Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM, Read-Only Memory).

[0250] The present invention also provides a computer-readable storage medium. The readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, it can achieve the following:

[0251] During the process of performing single-exposure angiography on a pre-identified target heart region using a pre-built coronary angiography device, an angiography image corresponding to preset default angiography parameters is obtained;

[0252] Using a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis, perform an image enhancement operation on the angiography image to obtain an enhanced angiography image, and perform a prediction operation based on diagnostic value on the enhanced angiography image to obtain a diagnostic value score;

[0253] Determine whether the diagnostic value score is higher than a preset high-value threshold;

[0254] When the diagnostic value score is lower than or equal to the high-value threshold, perform a dose increase operation on the default angiography parameters according to a preset dose adjustment rule to obtain updated angiography parameters, obtain an updated angiography image corresponding to the updated angiography parameters, and use the updated angiography image to replace the angiography image, and use the updated angiography parameters to replace the default angiography parameters, and return to the above step of using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis;

[0255] When the diagnostic value score is higher than the high-value threshold, end the process of performing single-exposure angiography on the target heart region according to the default angiography parameters to obtain the angiography image and the enhanced angiography image.

[0256] In several embodiments provided by the present invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and there can be other division methods in actual implementation.

[0257] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0258] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware, or in the form of a combination of hardware and software functional modules.

[0259] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0260] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for controlling the radiation dose of coronary angiography based on image acquisition quality, characterized in that: The method comprises: In the process of performing single-exposure angiography of a pre-confirmed target heart region using a pre-constructed coronary angiography device, obtaining an angiography image corresponding to a preset default angiography parameter; Using a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis, performing an image enhancement operation on the contrast image to obtain a contrast-enhanced image, and performing a prediction operation based on the diagnostic value on the contrast-enhanced image to obtain a diagnostic value score; Determining whether the diagnostic value score is higher than a preset high value threshold; When the diagnostic value score is lower than or equal to the high value threshold, performing a dose increase operation on the default contrast parameter according to a preset dose adjustment rule to obtain an updated contrast parameter, obtaining an updated contrast image corresponding to the updated contrast parameter, replacing the contrast image with the updated contrast image, and replacing the default contrast parameter with the updated contrast parameter, and returning to the above step of using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis; When the diagnostic value score is higher than the high value threshold, the process of performing single-exposure angiography on the target heart area is terminated according to the default angiography parameters to obtain the angiography image and the contrast-enhanced image.

2. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 1, characterized in that: Prior to using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis, the method further includes: Obtain a collection of high and low dose image samples; Acquire a generative adversarial model based on low-dose imaging enhancement-diagnosis, wherein the generative adversarial model includes a generator based on a multi-scale feature fusion network and an image enhancement network, and a discriminator based on a differentiable diagnosis network; Get the generator loss function, where the generator loss function is expressed as: In the formula, represents the generator loss function, Represents resistance to loss, represents the structural similarity loss, Dose_Penalty represents the penalty dose exceeding the preset dose, and λ adv , struct and dose All represent weight coefficients; Obtain the discriminator loss function, wherein the discriminator loss function is expressed as: In the formula, represents the discriminator loss function, I real represents the real high-dose image, I low represents the real low-dose image, I real and I low All belong to the high-dose and low-dose image sample sets, G(·) represents the output result of the generator, and D(·) represents the output result of the discriminator. Indicates the expected operator; The generative adversarial model is trained using the high-dose and low-dose image sample sets according to the generator loss function and the discriminator loss function to obtain a trained generative adversarial model.

3. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 2, characterized in that: The step of acquiring a high-dose and low-dose image sample set comprises: Obtain a human equivalent phantom; Using the coronary angiography equipment to perform angiography operations of a preset standard radiation dose and a preset minimum radiation dose on the human equivalent phantom, respectively, to obtain a standard angiography image and a minimum angiography image; Performing a multi-level low-dose data generation operation based on a Poisson noise model on the standard radiation dose and the standard contrast image to obtain a multi-level low-dose image set; The standard angiography image, the multi-level low-dose image set and the minimum angiography image are summarized to obtain a high- and low-dose image sample set.

4. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 3, characterized in that: The method of using the pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to perform an image enhancement operation on the contrast image to obtain a contrast-enhanced image includes: According to a preset pixel scaling range, the angiography image is normalized to obtain a normalized angiography image; Performing a noise convolution operation based on the Poisson noise model on the normalized contrast image to obtain a predicted noise distribution, and acquiring a denoised image signal according to a difference between the contrast image and the predicted noise distribution; Using the multi-scale feature fusion network, a vascular feature recognition operation of a multi-scale vascular branch field of view is performed on the denoised image signal to obtain a hierarchical vascular feature set; The pre-constructed residual block is used to perform feature fusion on the hierarchical vascular feature set to obtain a fused vascular feature set, and an upsampling operation is performed on the fused vascular feature set to obtain a contrast-enhanced image.

5. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 4, characterized in that: The performing a prediction operation based on the diagnostic value on the contrast-enhanced image to obtain a diagnostic value score includes: Performing authenticity detection on the contrast-enhanced image to obtain an authenticity probability; Determining whether the authenticity probability is greater than a preset real image threshold; When the authenticity probability is greater than the real image threshold, using the differentiable diagnostic network to perform a diagnostic feature extraction operation on the contrast-enhanced image to obtain a diagnostic feature set; performing a lesion presence prediction operation on the diagnostic feature set to obtain a lesion presence probability, and performing a vascular continuity scoring operation on the diagnostic feature set to obtain a vascular continuity score; The contrast-to-noise ratio of the contrast-enhanced image is obtained, and the contrast-to-noise ratio, the probability of lesion existence and the vascular continuity score are weightedly calculated to obtain a diagnostic value score.

6. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 5, characterized in that: The step of performing a dose increase operation on the default angiography parameters according to a preset dose adjustment rule to obtain updated angiography parameters includes: Obtain an adjustment formula in a preset dosage adjustment rule, wherein the adjustment formula is expressed as: In the formula, Δη represents the adjustment amount, ΔS represents the difference in diagnostic value score, and S threshold represents the high value threshold, CNR represents the contrast-to-noise ratio, BMI norm represents the standard body mass index, HR represents the heart rate, α, β, γ and δ represent weight parameters; According to the dosage adjustment rule, determining the numerical interval of the diagnostic value score difference; If the diagnostic value score difference belongs to the preset first value interval, the default angiography parameters are slightly adjusted according to the adjustment formula to obtain updated angiography parameters, wherein the slight adjustment is expressed as: or new =the current +0.05·D Where η new represents the updated imaging parameters, η current represents the default angiography parameters; If the diagnostic value score difference belongs to the preset second value interval, the default angiography parameters are moderately adjusted according to the adjustment formula to obtain updated angiography parameters, wherein the moderate adjustment is expressed as: or new =the current +0.1·Dη+0.02·BMI norm If the diagnostic value score difference belongs to the preset third value interval, the default angiography parameters are heavily adjusted according to the adjustment formula to obtain updated angiography parameters, wherein the heavy adjustment is expressed as: or new =min(η current +0.3,h safe ) In the formula, min(·) represents the minimum function, η safe Indicates the preset safe dose.

7. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 6, characterized in that: The process of ending the single-exposure angiography of the target heart area according to the default angiography parameters includes: Monitor the changes of the default angiography parameters to obtain a parameter update sequence, and calculate the value change rate of the parameter update sequence; Obtaining a time point at which the value change rate is zero, and when the time point exceeds a preset static exposure time, ending the process of performing single exposure angiography on the target heart area; During the single-exposure angiography of the target heart region, the cumulative dose of the coronary angiography device to the target heart region is acquired in real time; Determining whether the cumulative dose is greater than a preset safety threshold; When the accumulated dose is greater than the safety threshold, the process of performing single-exposure angiography on the target heart area is terminated.

8. The method for controlling the radiation dose of coronary angiography based on image acquisition quality according to claim 7, characterized in that: After obtaining the contrast image and the contrast-enhanced image, the method further includes: Acquire a CT image of the target heart area, and construct a key-value pair with the CT image, a standard body mass index, and the updated angiography parameter to obtain a CT image-angiography parameter sample; storing the CT image-contrast parameter samples in a pre-built empirical database; According to a preset time frequency, the pre-constructed initialization contrast parameter prediction model is trained using each CT image-contrast parameter sample stored in the experience database to obtain a trained initialization contrast parameter prediction model; Obtain the physical information indicators of the new patient and the CT image of the new patient, and use the trained initialized angiography parameter prediction model to predict the default angiography parameters of the new patient according to the CT image of the new patient and the physical information indicators of the new patient.

9. A coronary angiography radiation dose control system based on image acquisition quality, characterized in that: The system comprises: An initialization angiography module is used to obtain an angiography image corresponding to a preset default angiography parameter during a single exposure angiography of a pre-confirmed target heart area using a pre-constructed coronary angiography device; An image enhancement and value recognition module, used to perform an image enhancement operation on the contrast image using a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis to obtain a contrast-enhanced image, and perform a prediction operation based on the diagnostic value on the contrast-enhanced image to obtain a diagnostic value score; a contrast parameter updating module, configured to determine whether the diagnostic value score is higher than a preset high value threshold, and when the diagnostic value score is lower than or equal to the high value threshold, perform a dose increase operation on the default contrast parameter according to a preset dose adjustment rule to obtain an updated contrast parameter, obtain an updated contrast image corresponding to the updated contrast parameter, replace the contrast image with the updated contrast image, replace the default contrast parameter with the updated contrast parameter, and return to the above-mentioned step of using a pre-trained generative adversarial model based on low-dose imaging enhancement-diagnosis; The contrast image output module is used to end the single-exposure contrast imaging process of the target heart area according to the default contrast parameters when the diagnostic value score is higher than the high-value threshold, and obtain the contrast image and the contrast-enhanced image.

10. The coronary angiography radiation dose control system based on image acquisition quality according to claim 9, characterized in that: The generative adversarial model based on low-dose imaging enhancement-diagnosis includes: Obtain a collection of high and low dose image samples; Acquire a generative adversarial model based on low-dose imaging enhancement-diagnosis, wherein the generative adversarial model includes a generator based on a multi-scale feature fusion network and an image enhancement network, and a discriminator based on a differentiable diagnosis network; Get the generator loss function, where the generator loss function is expressed as: In the formula, represents the generator loss function, Represents resistance to loss, represents the structural similarity loss, Dose_Penalty represents the penalty dose exceeding the preset dose, and λ adv , struct and dose All represent weight coefficients; Obtain the discriminator loss function, wherein the discriminator loss function is expressed as: In the formula, represents the discriminator loss function, I real represents the real high-dose image, I low represents the real low-dose image, I real and I low All belong to the high-dose and low-dose image sample sets, G(·) represents the output result of the generator, and D(·) represents the output result of the discriminator. Indicates the expected operator; The generative adversarial model is trained using the high-dose and low-dose image sample sets according to the generator loss function and the discriminator loss function to obtain a trained generative adversarial model.

Images