CN111297388A

CN111297388A - Fractional flow reserve measurement method and device

Info

Publication number: CN111297388A
Application number: CN202010260601.1A
Authority: CN
Inventors: 张贺晔; 郭赛迪; 张冬
Original assignee: National Sun Yat Sen University
Current assignee: National Sun Yat Sen University
Priority date: 2020-04-03
Filing date: 2020-04-03
Publication date: 2020-06-19
Anticipated expiration: 2040-04-03
Also published as: CN111297388B

Abstract

The invention discloses a method and a device for measuring fractional flow reserve, which comprises the following steps: establishing a corresponding relation between a blood vessel characteristic vector of a coronary artery and a blood flow reserve fraction through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve; acquiring a current blood vessel feature vector of a patient; and determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation. The method can accurately predict the fractional flow reserve and has wide application range.

Description

Fractional flow reserve measurement method and device

Technical Field

The invention relates to the field of medical detection, in particular to a fractional flow reserve measuring method and a fractional flow reserve measuring device.

Background

Coronary angiography and intravascular ultrasound are considered as the 'gold standard' for diagnosing coronary heart disease, but the coronary angiography and intravascular ultrasound can only carry out imaging evaluation on the stenosis degree of lesion, and the influence of the stenosis on the far-end blood flow is unknown; fractional Flow Reserve (FFR) has now become a recognized indicator of functional assessment of coronary stenosis, the most important function of which is the accurate assessment of the functional consequences of an unknown-affected coronary stenosis.

Fractional Flow Reserve (FFR) is the ratio of the maximum blood flow that can be obtained in the myocardial region supplied by the target measurement vessel in the presence of a stenotic lesion in the coronary artery to the maximum blood flow that can be obtained theoretically normally in the same region. FFR is mainly obtained by calculating the ratio of the distal pressure of the coronary stenosis to the aortic root pressure. The stenotic distal pressure can be measured by a pressure guidewire at maximum perfused blood flow (by intracoronary or intravenous injection of papaverine or adenosine or ATP).

FFR＝P_d/P_a(P_dDistal coronary stenosis pressure, P, measured for catheter guidance_aAortic pressure measured for pressure guidewire) in general, FFR refers to the concept that in the maximal hyperemic state, there is no "resting FFR".

The normal epicardial coronary artery has little resistance to blood flow, with a normal value of FFR of 1.0; a FFR value of less than 1.0 indicates the presence of a stenosis in the current epicardial coronary artery.

In the case of FFR < 0.75, myocardial ischemia is caused by the typical stenosis, and in the case of FFR > 0.75, myocardial ischemia is very unlikely to be caused by the typical stenosis.

The coronary artery CTA can accurately evaluate the coronary stenosis degree and distinguish the plaque property of the vessel wall, is a non-invasive and simple-operation inspection method for diagnosing coronary artery lesion, and can be used as a preferred method for screening high risk groups. Therefore, if intervention is performed on the blood vessels of patients with coronary heart disease, the patient's coronary artery should be evaluated for CTA at the early stage. Coronary Chronic Total Occlusion (CTO) if evaluated with CTA, the results of the evaluation certainly have some valuable information.

The non-invasively obtained FFR (CTFFR) is calculated by CCTA (computed tomography angiography) of coronary artery CT (computed tomography), not only is no need of additional image examination or medicine, but also has good correlation with the FFR measured during radiography, and the integrated technology can fundamentally avoid unnecessary coronary angiography and blood circulation reconstruction treatment. The results of the Defacto trial also clearly show that in coronary CT, analysis of the CTFFR results provides physiological information about those lesions that actually restrict blood flow and increase the risk to the patient. CTFFR combines the advantages of coronary CTA and FFR, can evaluate coronary stenosis from both structural and functional aspects, and becomes a brand-new noninvasive detection system providing anatomical and functional information of coronary lesions.

However, the existing detection systems for measuring the fraction of stored blood flow generally have the following disadvantages: the disadvantages of invasive FFR techniques: there is a risk of surgery, the use of vasodilators is toxic and may cause allergy to the patient, and it is expensive.

The non-invasive FFR technology based on the hemodynamics simulation has the following defects: the calculation amount is large, and a high-performance computer is needed; the time consumption is long, and the real-time requirement cannot be met.

Disclosure of Invention

In view of the above, the present application is proposed to provide fractional flow reserve measurement methods and devices that overcome or at least partially address the above problems, comprising:

a fractional flow reserve measurement method comprising the steps of:

establishing a corresponding relation between a blood vessel characteristic vector of a coronary artery and a blood flow reserve fraction through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve;

acquiring a current blood vessel characteristic vector of a patient, specifically acquiring the current blood vessel characteristic vector according to a coronary artery computed tomography image of the patient;

and determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation.

Further, the vessel feature vector includes:

the local geometric characteristics of the blood vessel are as follows: the radius of each cross section of the vessel;

the upstream and downstream geometric characteristics of the blood vessel are as follows: stenosis proximal radius, stenosis entrance length, minimum radius zone length, stenosis exit length, radius reduction ratio:

in the formula, r_rRepresents a radius reduction ratio; r is_sA minimum radius representing stenosis; r is_pRepresents the normal radius of the segment near the stenosis; r is_dIndicating the normal radius distal to the stenosis.

Further, the first generation network and the first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and the second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the step of the sample data being the vessel feature vector and the fractional flow reserve comprises:

independently performing iterative training on the first generation network and the first discriminant network by maximizing the discriminative power of the discriminant network and minimizing the distribution loss function of the generation network until the discriminant output probability value of the fractional flow reserve generated by the first generation network in the first discriminant network is close to 0.5;

according to the artificial neuron parameters of the first three-layer network layer in the first judgment network, even the artificial neuron parameters of the first three-layer network layer in the second generation network;

and independently performing iterative training on the second generation network and the second judgment network by maximizing the difference capability of the judgment network and minimizing the distribution loss function of the generation network until the judgment output probability value of the fractional flow reserve generated by the second generation network in the second judgment network is close to 0.5.

Further, the first generation network comprises: a first input layer, a first deconvolution layer G₁A second deconvolution layer G₁And a third deconvolution layer G₁And a fourth deconvolution layer G₁And a first output layer, wherein the number of artificial neurons in the first input layer is the same as the number of vessel feature vectors in the sample data, and a first deconvolution layer G₁The number of artificial neurons in (A) is 512, and the second deconvolution layer G₁The number of artificial neurons of 256, the third deconvolution layer G₁The number of artificial neurons of 128, the fourth deconvolution layer G₁The number of artificial neurons in (1) is 64, and the number of artificial neurons in the first output layer is 1.

Further, the first discrimination network includes: second input layer, first winding layer D₁A second convolution layer D₁And a third convolution layer D₁And a fourth convolution layer D₁And a second output layer, wherein the number of artificial neurons of the second input layer is the same as the total number of fractional flow reserve generated by the first generation layer through sample data, and the first convolution layer D₁The number of artificial neurons of (D) is 32, and the second convolutional layer D₁The number of artificial neurons of (D) is 64, and the third convolutional layer D₁The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₁The number of artificial neurons in the second output layer is 512, and the number of artificial neurons in the second output layer is 1.

Further, the second generation network includes: third input layer, first deconvolution layer G₂A second deconvolution layer G₂And a third deconvolution layer G₂And a fourth deconvolution layer G₂The fifth deconvolution layer G₂And a third output layer, wherein the number of artificial neurons in the third input layer is the same as the number of vessel feature vectors in sample data, and the first deconvolution layer G₂The number of artificial neurons of (2) is 32, and the second deconvolution layer G₂The number of artificial neurons of (2) is 64, and the third deconvolution layer G₂The number of artificial neurons of (2) is 256, and the fourth deconvolution layer G₂The number of artificial neurons of (2) is 64, and the fifth deconvolution layer G₂Has an artificial neuron number of 32, and has an artificial neuron number of the third output layerIs 1.

Further, the second determination network includes: fourth input layer, first winding layer D₂A second convolution layer D₂And a third convolution layer D₂And a fourth convolution layer D₂And a fourth output layer, wherein the number of artificial neurons in the fourth input layer is the same as the total number of the fractional flow reserve generated by the second generation network through sample data and the actual fractional flow reserve corresponding to the sample data, and the first convolution layer D is a layer including a plurality of layers including a plurality of₂The number of artificial neurons of (D) is 32, and the second convolutional layer D₂The number of artificial neurons of (D) is 64, and the third convolutional layer D₂The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₂The number of artificial neurons in the fourth output layer is 512, and the number of artificial neurons in the fourth output layer is 1.

Further, the iterative training step of the first discriminant network or the second discriminant network includes:

setting a blood flow reserve fraction calculation value calculated by the blood vessel feature vector as a false sample set, and setting all class labels of the false sample set as 0;

setting invasive measurement values of fractional flow reserve obtained through invasive detection as a true sample set, and setting all class labels of the true sample set as 1;

inputting the blood vessel characteristic vector and the invasive measurement value of the blood flow reserve fraction, and adjusting the weight by comparing the direct difference value between the value output by the discrimination network and 1 to make the value output by the discrimination network approach to 1.

inputting the blood vessel characteristic vector and the blood flow reserve fraction calculation value, comparing the value output by the discrimination network with the direct difference value of 0, and adjusting the weight to enable the value output by the discrimination network to be close to 1.

Further, the step of iteratively training the first generation network and the second generation network comprises:

fixing the calculation parameters of the discrimination network;

inputting a blood vessel characteristic vector, carrying out initial calculation to obtain an initial fractional flow reserve calculation value, and setting a label of the initial fractional flow reserve calculation value as 1;

and adjusting the weight by comparing the difference between the calculated value of the fractional flow reserve output by the generating network and the invasive measurement value of the fractional flow reserve, so that the calculated value of the fractional flow reserve output by the generating network is close to the invasive measurement value of the fractional flow reserve.

fixing the calculation parameters of the discrimination network;

and inputting the blood vessel characteristic vector and the fractional flow reserve calculation value output by the generation network into a discrimination network, and adjusting the weight of the generation network according to the direct difference value between the value output by the discrimination network and 1 to enable the fractional flow reserve calculation value output by the generation network to be close to 1 in the discrimination result of the discrimination network.

A fractional flow reserve measurement apparatus comprising:

the establishing module is used for establishing the corresponding relation between the blood vessel characteristic vector of the coronary artery and the blood flow reserve fraction through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve;

an obtaining module, configured to obtain a current blood vessel feature vector of a patient, and specifically, obtain the current blood vessel feature vector according to a coronary computed tomography image of the patient;

and the measuring module is used for determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation.

An apparatus comprising a processor, a memory and a computer program stored on the memory and capable of running on the processor, the computer program when executed by the processor implementing the steps of a fractional flow reserve measurement method as described above.

A computer-readable storage medium, having stored thereon a computer program which, when being executed by a processor, carries out the steps of a fractional flow reserve measurement method as set out above.

The application has the following advantages:

in the embodiment of the application, the corresponding relation between the blood vessel characteristic vector of the coronary artery and the blood flow reserve fraction is established through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve; acquiring a current blood vessel characteristic vector of a patient, specifically acquiring the current blood vessel characteristic vector according to a coronary artery computed tomography image of the patient; and determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation. The method can accurately predict the blood flow reserve fraction, has wide application range and higher degree of freedom, generates a training method of a generation network (G network) and a discrimination network (D network) of an antagonistic network through the blood vessel characteristics extracted from a coronary CT image, avoids the surgical risk, does not need a vasodilator, is safer for patients, and has low calculation complexity and high calculation speed compared with a noninvasive FFR (flow induced regression) technology based on hemodynamics simulation, and can meet the real-time requirement.

Drawings

In order to more clearly illustrate the technical solutions of the present application, the drawings needed to be used in the description of the present application will be briefly introduced below, and it is apparent that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive labor.

Fig. 1 is a flowchart illustrating steps of a fractional flow reserve measurement method according to an embodiment of the present application;

fig. 2 is a schematic network structure diagram of a first generation network according to an embodiment of the present application;

fig. 3 is a schematic network structure diagram of a first discriminant network according to an embodiment of the present disclosure;

fig. 4 is a schematic network structure diagram of a second generation network provided in an embodiment of the present application;

fig. 5 is a schematic network structure diagram of a second decision network according to an embodiment of the present application;

FIG. 6 is a schematic flowchart of iterative training of a discriminant network according to an embodiment of the present disclosure;

FIG. 7 is a schematic flow chart of iterative training of a generation network according to an embodiment of the present application;

fig. 8 is a schematic block diagram of a fractional flow reserve measurement apparatus according to an embodiment of the present application;

fig. 9 is a schematic structural diagram of a computer device according to an embodiment of the present invention.

Detailed Description

In order to make the aforementioned objects, features and advantages of the present application more comprehensible, the present application is described in further detail with reference to the accompanying drawings and the detailed description. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that in any embodiment of the present invention, a cascade is an effective structure that can be used to automatically associate tasks and improve the performance of the multitask model. For the multi-task problem, the cascading operation designs the tasks into a cascading mode, the former tasks transmit effective information to the latter tasks, and the latter tasks are modeled according to the transmitted effective information.

FFR (fractional flow reserve) of a coronary artery is defined as the maximum blood flow QS after a myocardial vasogenic hyperemia in the area innervated by a stenotic coronary artery_maxMaximum blood flow QN of the region than if no stenosis were present_maxI.e. by

In the formula P_dPressure distal to coronary stenosis, P_aIs aortic pressure, R_sAnd R_NIs microcirculation resistance, P_VCentral venous pressure. In the normal case, P_VAlmost close to zero and constant. When vasodilators such as intravenous or intra-arterial adenosine are used, the coronary artery is maximally engorged, i.e. the resistance to microcirculation is minimized, and R is considered to be_s＝R_N，P_VRelative to P_aAnd P_dThe suspicion is ignored, i.e. the above formula can be simplified as:

wherein, P_dAnd P_aAll can be measured by a pressure guidewire.

Referring to fig. 1, the present invention provides a fractional flow reserve measurement method, including the following steps:

s110, establishing a corresponding relation between a blood vessel characteristic vector of a coronary artery and a blood flow reserve fraction through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve;

s120, obtaining a current blood vessel characteristic vector of the patient, specifically obtaining the current blood vessel characteristic vector according to a coronary artery computed tomography image of the patient;

and S130, determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation.

Next, the fractional flow reserve measurement method in the present exemplary embodiment will be further described.

As described in the above step S110, establishing a corresponding relationship between the blood vessel feature vector of the coronary artery and the fractional flow reserve through the artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve.

For example: and analyzing the state rule of the blood vessel characteristic vectors corresponding to the blood flow reserve fraction by using an artificial neural network algorithm, and finding out the mapping rule among the blood vessel characteristic vectors corresponding to the blood flow reserve fraction of the patient through the self-learning and self-adaptive characteristics of the artificial neural network.

For example: the method can utilize an artificial neural network algorithm, collect a large number of vascular feature vectors of different volunteers (including but not limited to one or more of age, weight, sex, disease conditions and the like), select the vascular feature vectors corresponding to the blood flow reserve scores of a plurality of volunteers as sample data, learn and train the neural network, fit the relationship between the vascular feature vectors corresponding to the blood flow reserve scores through adjusting the network structure and the weights among network nodes, and finally enable the neural network to accurately fit the corresponding relationship between the vascular feature vectors corresponding to the blood flow reserve scores of different patients.

It should be noted that the blood vessel feature vector is generally obtained by integrating a plurality of feature values in a target designated region, where the target designated region may be a coronary artery, but is not limited to a coronary artery, and the target designated region may be changed correspondingly according to a change of a test purpose; the fractional flow reserve is a fractional flow reserve measurement value measured invasively corresponding to the vessel feature vector.

In one embodiment, the vessel feature vector comprises:

Specifically, first, upstream and downstream stenosis regions (including all stenosis with a radius reduction greater than 10%) are identified by an automated detection algorithm. The stenosis is then ranked based on the degree of reduction in radius. Finally, the four most significant stenosis upstream and four most significant downstream stenosis downstream along the main branch path are selected, for each of these stenosis, the features extracted are: 1) a stenosis proximal radius, i.e., a stenosis minimum radius and a stenosis distal radius; 2) the stenosis entrance length, i.e., the length along the centerline between the beginning of the stenosis and the beginning of the segment with the smallest radius; 3) minimum radius zone length, i.e., the length along the centerline between the beginning and end of the segment having the smallest radius; 4) a narrow exit length, i.e., a length along a centerline between the end of the segment having the smallest radius and the end of the narrow; 5) radius reduction ratio:

all the above characteristic valuesAnd combined to form a feature vector.

In an embodiment, the correspondence includes: and (4) functional relation.

Preferably, the vessel feature vector of the coronary artery is an input parameter of the functional relationship, and the fractional flow reserve is an output parameter of the functional relationship;

determining a current fractional flow reserve corresponding to the current vessel feature vector, further comprising:

and when the corresponding relation comprises a functional relation, inputting the current blood vessel feature vector into the functional relation, and determining the output parameter of the functional relation as the current blood flow reserve fraction.

Therefore, the flexibility and convenience of determining the current fractional flow reserve can be improved through the corresponding relations in various forms.

Referring to fig. 2 to 5, as an example, the network structure is a CGAN network structure, and includes a first generation network, a second generation network, a first discrimination network, and a second discrimination network, where the first generation network, the second generation network, the first discrimination network, and the second discrimination network are respectively formed by deep convolutional neural networks.

Referring to fig. 2, in one embodiment, the first production network comprises: a first input layer, a first deconvolution layer G₁A second deconvolution layer G₁And a third deconvolution layer G₁And a fourth deconvolution layer G₁And a first output layer, wherein the number of artificial neurons in the first input layer is the same as the number of vessel feature vectors in the sample data, and a first deconvolution layer G₁The number of artificial neurons in (A) is 512, and the second deconvolution layer G₁The number of artificial neurons of 256, the third deconvolution layer G₁The number of artificial neurons of 128, the fourth deconvolution layer G₁The number of artificial neurons in the first output layer is 64, the number of artificial neurons in the first output layer is 1, and the output value is a fractional flow reserve value obtained by initial calculation and is used as the input of the first discrimination network.

Referring to fig. 3, in an embodiment, the first discriminant network includes: first, theTwo input layers, a first winding layer D₁A second convolution layer D₁And a third convolution layer D₁And a fourth convolution layer D₁And a second output layer, wherein the number of artificial neurons of the second input layer is the same as the total number of fractional flow reserve generated by the first generation layer through sample data, and the first convolution layer D₁The number of artificial neurons of (D) is 32, and the second convolutional layer D₁The number of artificial neurons of (D) is 64, and the third convolutional layer D₁The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₁The number of artificial neurons in the second output layer is 512, the number of artificial neurons in the second output layer is 1, wherein the output layer contains a single node of a linear activation function, the activation function is a Sigmoid function, the obtained value is a judgment probability value for input, binary classification is solved by using the Sigmoid function, and a value from 0 to 1 is output.

Referring to fig. 4, in an embodiment, the second generation network includes: third input layer, first deconvolution layer G₂A second deconvolution layer G₂And a third deconvolution layer G₂And a fourth deconvolution layer G₂The fifth deconvolution layer G₂And a third output layer, wherein the number of artificial neurons in the third input layer is the same as the number of vessel feature vectors in sample data, and the first deconvolution layer G₂The number of artificial neurons of (2) is 32, and the second deconvolution layer G₂The number of artificial neurons of (2) is 64, and the third deconvolution layer G₂The number of artificial neurons of (2) is 256, and the fourth deconvolution layer G₂The number of artificial neurons of (2) is 64, and the fifth deconvolution layer G₂The number of artificial neurons in the third output layer is 32, and the number of artificial neurons in the third output layer is 1.

Referring to fig. 5, in an embodiment, the second determination network includes: fourth input layer, first winding layer D₂A second convolution layer D₂And a third convolution layer D₂And a fourth convolution layer D₂And a fourth output layer, wherein the number of artificial neurons of the fourth input layer and the fractional flow reserve generated by the second generation network through sample data are calculated to obtain a ratio of the number of artificial neurons to the fractional flow reserveThe total number of actual fractional flow reserve corresponding to the sample data is the same, and the first convolution layer D₂The number of artificial neurons of (D) is 32, and the second convolutional layer D₂The number of artificial neurons of (D) is 64, and the third convolutional layer D₂The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₂The number of artificial neurons in the fourth output layer is 512, the number of artificial neurons in the fourth output layer is 1, wherein the output layer contains a single node of a linear activation function, the activation function is a Sigmoid function, the obtained value is a judgment probability value for input, binary classification is solved by using the Sigmoid function, and a value from 0 to 1 is output.

In an embodiment, the first generated network and the first discriminant network are trained through sample data to obtain network parameters of a second generated network, and the second generated network and the second network are trained through the sample data to obtain the corresponding relationship; wherein the step of the sample data being the vessel feature vector and the fractional flow reserve comprises:

it should be noted that, because the discrimination probability value of the first discrimination network to the real sample is high at the initial operation, the discrimination probability value of the output of the sample generated by the first discrimination network is low. Through mutual loop iterative learning, the output discrimination probability is fed back to the first generation network, and the first generation network continuously optimizes network parameters to ensure that the first discrimination network cannot judge whether the first discrimination network is true or false; the first judging network also continuously optimizes network parameters, so that the judging capability is improved, and the judging probability values of true and false samples have differences. Finally, training and iterating for multiple times until the discrimination output probability value from the sample generated by the first generation network to the first discrimination network is close to 0.5, namely that the true and false samples are difficult to distinguish, and finishing the training.

it should be noted that, the first three-layer neuron parameters of the second generation network are obtained by the first discrimination network after training, and the output obtained value is the fractional flow reserve value obtained by initial calculation and is used as the input of the second discrimination network.

It should be noted that, since the first three layers of neuron parameters of the second generation network are obtained by being transmitted by the first discrimination network, the output discrimination probability value of the second discrimination network for the sample generated by the second generation network is low at the initial operation. Through mutual loop iterative learning, the output discrimination probability is fed back to a second generation network, and the second generation network continuously optimizes network parameters so that the second discrimination network cannot judge whether the second discrimination network is true or false; and the second judgment network also continuously optimizes network parameters, so that the judgment capability is improved, and the judgment probability values of true and false samples have differences. And finally, training and iterating for multiple times until the judgment output probability value from the sample generated by the second generation network to the second judgment network is close to 0.5, namely that the true and false samples are difficult to distinguish, and finishing the training.

In an embodiment, a specific process of "establishing a correspondence between the blood vessel feature vector of the coronary artery and the fractional flow reserve" in step S110 may be further described with reference to the following description.

The following steps are described: acquiring sample data for establishing a corresponding relation between the blood vessel feature vector and the blood flow reserve fraction;

in an advanced embodiment, a specific process of acquiring sample data for establishing a correspondence between a blood vessel feature vector of the coronary artery and a fractional flow reserve may be further described in conjunction with the following description.

The following steps are described: collecting the vessel feature vectors and the fractional flow reserve of patients of different cardiac conditions;

for example: data collection: collecting blood vessel characteristic vectors and corresponding blood flow reserve fractions of patients with different health conditions; collecting blood vessel characteristic vectors and corresponding blood flow reserve fractions of patients of different ages; and collecting the blood vessel characteristic vectors of patients of different sexes and corresponding blood flow reserve fractions.

Therefore, the operation data are collected through multiple ways, the quantity of the operation data is increased, the learning capacity of the artificial neural network is improved, and the accuracy and the reliability of the determined corresponding relation are improved.

The following steps are described: analyzing the blood vessel characteristic vector, and selecting data related to the fractional flow reserve as the blood vessel characteristic vector by combining with prestored expert experience information (for example, selecting the blood vessel characteristic vector influencing the fractional flow reserve as an input parameter, and using a specified parameter as an output parameter);

for example: the blood vessel characteristic vector in the relevant data of the diagnosed volunteer is used as an input parameter, and the fractional flow reserve in the relevant data is used as an output parameter.

The following steps are described: and taking a data pair formed by the blood flow reserve fraction and the selected blood vessel characteristic vector as sample data.

For example: and using part of the obtained input and output parameter pairs as training sample data and using part of the obtained input and output parameter pairs as test sample data.

Therefore, the collected blood vessel feature vectors are analyzed and processed to obtain sample data, the operation process is simple, and the reliability of the operation result is high.

The following steps are described: analyzing the characteristics and the rules of the blood vessel characteristic vectors, and determining the network structure and the network parameters of the artificial neural network according to the characteristics and the rules;

for example: the basic structure of the network, the input and output node number of the network, the number of hidden layers of the network, the number of hidden nodes, the initial weight of the network and the like can be preliminarily determined by analyzing the multi-view coronary angiography data characteristics and the morphological parameter characteristics of coronary stenosis.

Optionally, a specific process of training the network structure and the network parameters in the step of training and testing the network structure and the network parameters and determining the corresponding relationship between the blood vessel feature vector and the fractional flow reserve may be further described in conjunction with the following description.

Selecting a part of data in the sample data as a training sample, inputting the blood vessel feature vector in the training sample into the network structure, and training through a loss function of the network structure, an activation function and the network parameters to obtain an actual training result;

specifically, a loss function is minimized through a gradient descent algorithm, network parameters are updated, a current multi-view neural network is trained, and an actual training result is obtained;

determining whether an actual training error between the actual training result and a corresponding fractional flow reserve in the training sample satisfies a preset training error; determining that the training of the network structure and the network parameters is completed when the actual training error meets the preset training error;

specifically, when the actual training error satisfies the preset training error, and the currently trained model converges, it is determined that the training of the network structure and the network parameters is completed.

More optionally, training the network structure and the network parameters further includes:

when the actual training error does not meet the set training error, updating the network parameters through an error loss function of the network structure; activating a function and the updated network parameters to retrain through the loss function of the network structure until the retrained actual training error meets the set training error;

for example: and if the test error meets the requirement, finishing the network training test.

Therefore, the reliability of the network structure and the network parameters is further verified by using the test sample for testing the network structure and the network parameters obtained by training.

Optionally, a specific process of testing the network structure and the network parameters in the step of training and testing the network structure and the network parameters and determining the corresponding relationship between the blood vessel feature vector and the fractional flow reserve may be further described in conjunction with the following description.

Selecting another part of data in the sample data as a test sample, inputting the blood vessel feature vector in the test sample into the trained network structure, and testing by using the loss function, the activation function and the trained network parameters to obtain an actual test result; determining whether an actual test error between the actual test result and a corresponding fractional flow reserve in the test sample satisfies a set test error; and when the actual test error meets the set test error, determining that the test on the network structure and the network parameters is finished.

It should be noted that, in the CGAN network framework, the first generation network, the second generation network, the first discrimination network, and the second discrimination network use a small batch gradient descent method for joint training, and both generation network frameworks have their own antagonistic loss. Therefore, the calculation formula for using the feature matching loss function for the first discriminant network is as follows:

in the formula, D_1-featureRepresenting the output characteristics of the first discrimination network, L_1-featureIs the feature matching loss function, x and of the first discriminant network

Respectively true data and false data, P, generated by the first generation network_riAnd P_fiCharacteristic distributions of real data and dummy data, respectively.

The penalty function for the first generation network is calculated as follows:

in the formula (I), the compound is shown in the specification,

is the countermeasure loss for the first generation network,

is the countermeasure loss of the first discriminant network and λ is the weight of the penalty term. λ is set to 10.

Feature matching function L of second discrimination network_2-featureThe calculation formula of (a) is as follows:

in the formula (I), the compound is shown in the specification,

and

respectively true and false data of the generation of the second generation network, P_lsAnd P_usFeature distributions, D, representing real and false data, respectively_2-featureA characteristic output representing the second decision network.

Similar to the countermeasure loss of the first generation network and the first discrimination network, L is_2-featureCountermeasures to impairments by introducing a second generating network and a second discriminating networkThe calculation formula is as follows:

the above-mentioned counter-loss can be regarded as an unsupervised loss since it uses only the generation data of the second generation network, not the real data.

Finally, real data s is utilized_lDefining a supervised loss function L based on Dice loss_G2supThe calculation formula is as follows:

simultaneous update of supervision losses while the algorithm is running

And fight against loss

It should be noted that the training flows of the first discriminant network and the second discriminant network are similar, so only the training flow of the first discriminant network is briefly described here.

Referring to fig. 6, in an embodiment, the step of iteratively training the first discriminant network or the second discriminant network includes:

it should be noted that, before calculating the fractional flow reserve calculation value, parameters of the generation network need to be fixed, so as to avoid introducing unnecessary uncertain unknown conditions due to changes of network parameters when the fractional flow reserve calculation value is generated by the generation network, thereby causing errors and further causing the result of iterative training to generate an inclination;

the test conditions (blood vessel characteristic values) of the invasive measurement values of fractional flow reserve correspond to the calculation conditions (blood vessel characteristic values) of the calculated values of fractional flow reserve in the previous step;

It should be noted that, because the label of the input sample has only 0 or 1, a value between 0 and 1 can be obtained according to the discrimination results of the true sample set and the false sample set, and because the data truth of the true and false sample sets in the training parameters is known, the discrimination network can record the correction result by artificially shifting the weight according to the discrimination result, so as to perfect the discrimination network, it should be noted that, in this step, the weight is adjusted by calculating the direct difference between the discrimination result output by the discrimination network and 1, and when the direct difference between the discrimination result and 1 is close to 0, the network training is judged to be completed.

It should be noted that, when the direct difference between the above-mentioned determination result and 0 is close to 1, that is, it is determined that the network training is completed, this step may be used alternatively with the corresponding step in the previous embodiment in the iterative training process or may be trained in a manner that any one step is used as the verification step.

It should be noted that the training procedure of the first generation network and the training procedure of the second generation network are similar, so only the training procedure of the first generation network is briefly described here.

Referring to fig. 7, in an embodiment, the step of iteratively training the first generation network and the second generation network includes:

fixing the calculation parameters of the discrimination network;

determining the discriminant model as a quantitative condition by fixing calculation parameters of the discriminant network, and adjusting generation weights of the generation network by discriminating a fractional flow reserve calculation value generated by the generation network, wherein the calculation parameters generally include but are not limited to the discrimination weights of the discrimination network;

setting the label of the fractional flow reserve calculation value to 1, namely, indicating that the fractional flow reserve calculation value is regarded as a invasive measurement value of the fractional flow reserve under the condition of the current feature vector when discrimination is performed, performing authenticity discrimination through the discrimination network, and recording the discrimination result of the discrimination network, wherein the discrimination result is a value between 0 and 1, namely, a discrimination weight;

It should be noted that, an invasive measurement value of the blood flow reserve fraction of the same blood vessel feature vector as the calculated value of the blood flow reserve fraction is obtained, a difference between the calculated value of the blood flow reserve fraction and the invasive measurement value of the blood flow reserve fraction is calculated, and a generation weight in the generation network is adjusted according to the difference to perfect the generation network.

fixing the calculation parameters of the discrimination network;

Note that the examination weight of the generation network is adjusted by calculating a direct difference between the discrimination result of the discrimination network and 1 to obtain a deviation ratio of the fractional flow reserve calculated by the generation network.

As described in the above step S120, the current blood vessel feature vector of the patient is obtained, specifically, the current blood vessel feature vector is obtained according to the computed tomography image of the coronary artery of the patient, and the pressure P at the distal end of the coronary stenosis is measured by the pressure guide wire in an invasive manner_dAortic pressure P_aAnd by the formula

And (6) calculating.

As described in step S130 above, the current fractional flow reserve corresponding to the current vessel feature vector is determined according to the correspondence relationship.

For example: vessel feature vectors of coronary arteries of a patient are identified in real time.

Therefore, the current blood flow reserve fraction of the coronary artery is effectively identified according to the current blood vessel feature vector based on the corresponding relation, so that accurate judgment basis is provided for diagnosis of doctors, and the judgment result is good in accuracy.

In an alternative example, the determining the current fractional flow reserve corresponding to the vessel feature vector in step S130 may include: and determining the blood flow reserve fraction corresponding to the blood vessel feature vector which is the same as the current blood vessel feature vector in the corresponding relation as the current blood flow reserve fraction.

In an optional example, the determining the current fractional flow reserve corresponding to the vessel feature vector in step S130 may further include: when the corresponding relation can comprise a functional relation, inputting the current blood vessel feature vector into the functional relation, and determining the output parameter of the functional relation as the current fractional flow reserve.

Therefore, the current blood flow reserve fraction is determined according to the current blood vessel feature vector based on the corresponding relation or the functional relation, the determination mode is simple and convenient, and the reliability of the determination result is high.

For example, the artificial neural network obtained by training is used to detect the fractional flow reserve of each sample in the test set.

In an alternative embodiment, the method may further include: a process of verifying whether the current fractional flow reserve matches an actual fractional flow reserve.

Optionally, when a verification result that the current fractional flow reserve is not consistent with the actual fractional flow reserve is received and/or it is determined that there is no blood vessel feature vector in the correspondence that is the same as the current blood vessel feature vector, at least one maintenance operation of updating, correcting, and relearning the correspondence may be performed.

For example: the actual fractional flow reserve cannot be known by the device, and the feedback operation of the doctor is needed, namely, if the device intelligently judges the fractional flow reserve, the doctor feeds back that the fractional flow reserve is not in accordance with the actual state through the operation, and the device can know the fractional flow reserve.

And verifying whether the current fractional flow reserve and the actual fractional flow reserve match (for example, displaying the actual fractional flow reserve through an AR display module to verify whether the determined current fractional flow reserve and the actual fractional flow reserve match).

And when the current blood flow reserve fraction does not accord with the actual blood flow reserve fraction and/or the corresponding relation does not have a blood vessel feature vector which is the same as the current blood vessel feature vector, performing at least one maintenance operation of updating, correcting and relearning on the corresponding relation.

For example: the current fractional flow reserve may be determined according to the current vessel feature vector according to the maintained correspondence. For example: and determining the blood flow reserve fraction corresponding to the blood vessel feature vector which is the same as the current blood vessel feature vector in the maintained corresponding relation as the current blood flow reserve fraction.

Therefore, the accuracy and the reliability of the determination of the blood flow reserve fraction are improved by maintaining the corresponding relation between the determined blood vessel feature vector and the blood flow reserve fraction.

Referring to fig. 8, the present invention provides a fractional flow reserve measurement apparatus, including:

an establishing module 810, configured to establish a correspondence between a blood vessel feature vector of a coronary artery and a fractional flow reserve through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve;

an obtaining module 820, configured to obtain a current blood vessel feature vector of a patient, specifically, obtain the current blood vessel feature vector according to a coronary computed tomography image of the patient;

and a measuring module 830, configured to determine, according to the correspondence, a current fractional flow reserve corresponding to the current blood vessel feature vector.

In one embodiment, the vessel feature vector comprises:

in the formula, r_rRepresents a radius reduction ratio; r is_sA minimum radius representing stenosis; r is_pIndicating normality of the segment approaching stenosisA radius; r is_dIndicating the normal radius distal to the stenosis.

In one embodiment, the establishing module 810 includes:

the first iterative training submodule is used for independently performing iterative training on the first generation network and the first discriminant network through maximizing the difference capability of the discriminant network and minimizing the distribution loss function of the generation network until the discriminant output probability value of the fractional flow reserve generated by the first generation network in the first discriminant network is close to 0.5;

the neuron parameter generation submodule is used for generating artificial neuron parameters of a first three-layer network layer of the second generation network according to the artificial neuron parameters of the first three-layer network layer of the first discrimination network;

a second iterative training submodule, configured to perform iterative training on the second generation network and the second determination network independently by maximizing a difference capability of the determination network and minimizing a distribution loss function of the generation network until a determination output probability value of a fractional flow reserve generated by the second generation network in the second determination network is close to 0.5

In one embodiment, the first production network comprises: a first input layer, a first deconvolution layer G₁A second deconvolution layer G₁And a third deconvolution layer G₁And a fourth deconvolution layer G₁And a first output layer, wherein the number of artificial neurons in the first input layer is the same as the number of vessel feature vectors in the sample data, and a first deconvolution layer G₁The number of artificial neurons in (A) is 512, and the second deconvolution layer G₁The number of artificial neurons of 256, the third deconvolution layer G₁The number of artificial neurons of 128, the fourth deconvolution layer G₁The number of artificial neurons in (1) is 64, and the number of artificial neurons in the first output layer is 1.

In one embodiment, the first discriminant network comprises: second input layer, first winding layer D₁A second convolution layer D₁And a third convolution layer D₁And a fourth convolution layer D₁And a second output layer, wherein the second input layerThe number of artificial neurons is the same as the total number of fractional flow reserve generated by the first generation network through sample data, and the first convolution layer D₁The number of artificial neurons of (D) is 32, and the second convolutional layer D₁The number of artificial neurons of (D) is 64, and the third convolutional layer D₁The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₁The number of artificial neurons in the second output layer is 512, and the number of artificial neurons in the second output layer is 1.

In one embodiment, the second generation network comprises: third input layer, first deconvolution layer G₂A second deconvolution layer G₂And a third deconvolution layer G₂And a fourth deconvolution layer G₂The fifth deconvolution layer G₂And a third output layer, wherein the number of artificial neurons in the third input layer is the same as the number of vessel feature vectors in sample data, and the first deconvolution layer G₂The number of artificial neurons of (2) is 32, and the second deconvolution layer G₂The number of artificial neurons of (2) is 64, and the third deconvolution layer G₂The number of artificial neurons of (2) is 256, and the fourth deconvolution layer G₂The number of artificial neurons of (2) is 64, and the fifth deconvolution layer G₂The number of artificial neurons in the third output layer is 32, and the number of artificial neurons in the third output layer is 1.

In one embodiment, the second decision network includes: fourth input layer, first winding layer D₂A second convolution layer D₂And a third convolution layer D₂And a fourth convolution layer D₂And a fourth output layer, wherein the number of artificial neurons in the fourth input layer is the same as the total number of the fractional flow reserve generated by the second generation network through sample data and the actual fractional flow reserve corresponding to the sample data, and the first convolution layer D is a layer including a plurality of layers including a plurality of₂The number of artificial neurons of (D) is 32, and the second convolutional layer D₂The number of artificial neurons of (D) is 64, and the third convolutional layer D₂The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₂The number of artificial neurons in the fourth output layer is 512, and the number of artificial neurons in the fourth output layer is 1.

In an embodiment, the first iterative training sub-module or the second iterative training sub-module includes:

a first false sample set generation submodule, configured to set a fractional flow reserve calculated by the vessel feature vector as a false sample set, and set all class labels of the false sample set as 0;

a first true sample set generation submodule for setting a blood flow reserve fraction invasive measurement value obtained by invasive detection as a true sample set, and setting all class labels of the true sample set as 1;

and the first weight adjusting submodule is used for inputting the blood vessel characteristic vector and the invasive measurement value of the blood flow reserve fraction, and adjusting the weight by comparing the value output by the discrimination network with the direct difference value of 1 so as to enable the value output by the discrimination network to be close to 1.

a second false sample set generation submodule, configured to set a fractional flow reserve calculated by the vessel feature vector as a false sample set, and set all class labels of the false sample set as 0;

a second true sample set generation submodule for setting the invasive measurement value of fractional flow reserve obtained by invasive detection as a true sample set and setting all class labels of the true sample set as 1;

and the second weight adjusting submodule is used for inputting the blood vessel characteristic vector and the blood flow reserve fraction calculation value, comparing the value output by the discrimination network with a direct difference value of 0, and adjusting the weight to enable the value output by the discrimination network to be close to 1.

the first fixed calculation parameter submodule is used for fixing the calculation parameters of the discrimination network;

the first input submodule is used for inputting the blood vessel characteristic vector and carrying out initial calculation to obtain an initial fractional flow reserve calculation value, and the label of the initial fractional flow reserve calculation value is set to be 1;

and the third weight adjusting submodule is used for adjusting the weight by comparing the difference value between the calculated value of the fractional flow reserve output by the generating network and the invasive measurement value of the fractional flow reserve, so that the calculated value of the fractional flow reserve output by the generating network is close to the invasive measurement value of the fractional flow reserve.

the second fixed calculation parameter submodule is used for fixing the calculation parameters of the discrimination network;

the second input submodule is used for inputting the blood vessel characteristic vector and carrying out initial calculation to obtain an initial fractional flow reserve calculation value, and the label of the initial fractional flow reserve calculation value is set to be 1;

and the fourth weight adjusting submodule is used for inputting the blood vessel characteristic vector and the blood flow reserve fraction calculation value output by the generating network into the judging network, and adjusting the weight of the generating network according to the direct difference value between the value output by the judging network and 1 so as to enable the blood flow reserve fraction calculation value output by the generating network to be close to 1 in the judging result of the judging network.

Referring to fig. 9, a computer device of a fractional flow reserve measurement method according to the present invention is shown, which may specifically include the following:

the computer device 12 described above is embodied in the form of a general purpose computing device, and the components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.

Bus 18 represents one or more of any of several types of bus 18 structures, including a memory bus 18 or memory controller, a peripheral bus 18, an accelerated graphics port, and a processor or local bus 18 using any of a variety of bus 18 architectures. By way of example, such architectures include, but are not limited to, Industry Standard Architecture (ISA) bus 18, micro-channel architecture (MAC) bus 18, enhanced ISA bus 18, audio Video Electronics Standards Association (VESA) local bus 18, and Peripheral Component Interconnect (PCI) bus 18.

Computer device 12 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by computer device 12 and includes both volatile and nonvolatile media, removable and non-removable media.

The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)30 and/or cache memory 32. Computer device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (commonly referred to as "hard drives"). Although not shown in FIG. 9, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42, with the program modules 42 configured to carry out the functions of embodiments of the invention.

A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules 42, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally carry out the functions and/or methodologies of the described embodiments of the invention.

Computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, camera, etc.), with one or more devices that enable a user to interact with computer device 12, and/or with any devices (e.g., network card, modem, etc.) that enable computer device 12 to communicate with one or more other computing devices. Such communication may be through an input/output (I/O) interface 22. Also, computer device 12 may communicate with one or more networks (e.g., a Local Area Network (LAN)), a Wide Area Network (WAN), and/or a public network (e.g., the Internet) via network adapter 20. As shown, the network adapter 20 communicates with the other modules of the computer device 12 via the bus 18. It should be appreciated that although not shown in FIG. 9, other hardware and/or software modules may be used in conjunction with computer device 12, including but not limited to: microcode, device drivers, redundant processing units 16, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 34, etc.

The processing unit 16 executes programs stored in the system memory 28 to perform various functional applications and data processing, such as implementing the fractional flow reserve measurement method provided by the embodiments of the present invention.

That is, the processing unit 16 implements, when executing the program,: establishing a corresponding relation between a blood vessel characteristic vector of a coronary artery and a blood flow reserve fraction through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve; acquiring a current blood vessel characteristic vector of a patient, specifically acquiring the current blood vessel characteristic vector according to a coronary artery computed tomography image of the patient; and determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation.

In an embodiment of the present invention, the present invention further provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the fractional flow reserve measurement method as provided in all embodiments of the present application:

that is, the program when executed by the processor implements: establishing a corresponding relation between a blood vessel characteristic vector of a coronary artery and a blood flow reserve fraction through an artificial neural network; the artificial neural network comprises a first generation network, a second generation network, a first judgment network and a second judgment network; specifically, a first generation network and a first discrimination network are trained through sample data to obtain network parameters of a second generation network, and the second generation network and a second discrimination network are trained through the sample data to obtain the corresponding relation; wherein the sample data is the vessel feature vector and the fractional flow reserve; acquiring a current blood vessel characteristic vector of a patient, specifically acquiring the current blood vessel characteristic vector according to a coronary artery computed tomography image of the patient; and determining the current blood flow reserve fraction corresponding to the current blood vessel feature vector through the corresponding relation.

Any combination of one or more computer-readable media may be employed. The computer readable medium may be a computer-readable storage medium or a computer-readable signal medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPOM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider). The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

While preferred embodiments of the present application have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the true scope of the embodiments of the application.

Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.

The method and the device for measuring fractional flow reserve provided by the present application are introduced in detail, and specific examples are applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiments is only used to help understand the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A fractional flow reserve measurement method, comprising the steps of:

2. The method of claim 1, wherein the vessel feature vector comprises:

3. The method according to claim 1, wherein the training of the first generated network and the first discriminant network through sample data obtains network parameters of a second generated network, and the training of the second generated network and the second discriminant network through the sample data obtains the correspondence; wherein the step of the sample data being the vessel feature vector and the fractional flow reserve comprises:

4. The method of claim 3,

the first production network comprises: a first input layer, a first deconvolution layer G₁A second deconvolution layer G₁And a third deconvolution layer G₁And a fourth deconvolution layer G₁And a first output layer, wherein the number of artificial neurons in the first input layer is the same as the number of vessel feature vectors in the sample data, and a first deconvolution layer G₁The number of artificial neurons in (A) is 512, and the second deconvolution layer G₁The number of artificial neurons of 256, the third deconvolution layer G₁The number of artificial neurons of 128, the fourth deconvolution layer G₁The number of artificial neurons in (1) is 64, and the number of artificial neurons in the first output layer is 1.

5. The method of claim 3,

the first discrimination network includes: second input layer, first winding layer D₁A second convolution layer D₁And a third convolution layer D₁And a fourth convolution layer D₁And a second output layer, wherein the number of artificial neurons of the second input layer is the same as the total number of fractional flow reserve generated by the first generation layer through sample data, and the first convolution layer D₁The number of artificial neurons of (D) is 32, and the second convolutional layer D₁The number of artificial neurons of (D) is 64, and the third convolutional layer D₁The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₁The number of artificial neurons in the second output layer is 512, and the number of artificial neurons in the second output layer is 1.

6. The method of claim 3,

the second generation network includes: third input layer, first deconvolution layer G₂A second deconvolution layer G₂And a third deconvolution layer G₂And a fourth deconvolution layer G₂The fifth deconvolution layer G₂And a third output layer, wherein the number of artificial neurons in the third input layer is the same as the number of vessel feature vectors in sample data, and the first deconvolution layer G₂The number of artificial neurons of (2) is 32, and the second deconvolution layer G₂The number of artificial neurons of (2) is 64, and the third deconvolution layer G₂The number of artificial neurons of (2) is 256, and the fourth deconvolution layer G₂The number of artificial neurons of (2) is 64, and the fifth deconvolution layer G₂The number of artificial neurons in the third output layer is 32, and the number of artificial neurons in the third output layer is 1.

7. The method of claim 3,

the second determination network includes: fourth input layer, first winding layer D₂A second convolution layer D₂And a third convolution layer D₂And a fourth convolution layer D₂And a fourth output layer, wherein the number of artificial neurons in the fourth input layer is the same as the total number of the fractional flow reserve generated by the second generation network through sample data and the actual fractional flow reserve corresponding to the sample data, and the first convolution layer D is a layer including a plurality of layers including a plurality of₂The number of artificial neurons of (D) is 32, and the second convolutional layer D₂The number of artificial neurons of (D) is 64, and the third convolutional layer D₂The number of artificial neurons of (2) is 256, and the fourth convolutional layer D₂The number of artificial neurons in the fourth output layer is 512, and the number of artificial neurons in the fourth output layer is 1.

8. A fractional flow reserve measurement device, comprising:

9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method according to any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the method of any one of claims 1 to 7.