A method and device for predicting blood flow reserve fraction based on coronary magnetic resonance imaging

Abstract

The embodiment of the present invention relates to a method and device for predicting blood flow reserve fraction based on coronary magnetic resonance imaging, the method comprising: constructing a prediction model and training; after training, receiving the rest / congestion imaging sequence and angiography of the subject, and constructing a vascular tree; and identifying the inner wall edge and centerline of the rest / congestion imaging and the vascular tree, and sampling the vascular tree; and fitting the rest / congestion flow waveform of the sampling point; and estimating the CFR according to the rest / congestion flow waveform of each sampling point; and scaling the resting flow waveform with the CFR value of each sampling point; and inputting a vector sequence consisting of the sampling point vector composed of the coordinates, inner diameter and scaled flow waveform of each sampling point into the prediction model for prediction; and marking the vascular tree and angiography based on the prediction vector and the sampling point set; and displaying the marked vascular tree and angiography. The present invention can improve the comfort of the subject, and the present invention can be used as a continuous monitoring means.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for predicting blood flow reserve fraction based on coronary magnetic resonance imaging. Background Art

[0002] Fractional Flow Reserve (FFR) refers to the ratio of the maximum blood flow that can be obtained in the myocardial area supplied by the coronary artery to the maximum blood flow that can be obtained in the same area under normal conditions in theory when there is a stenosis in the coronary artery (hereinafter referred to as the coronary artery), that is, the ratio of the mean pressure in the stenotic distal coronary artery under the state of maximum myocardial hyperemia to the mean pressure in the aorta at the coronary artery orifice. The FFR score at a certain position of the coronary artery can be used to identify whether a stenosis has occurred at the current position. The current conventional method of determining the FFR score is an invasive and invasive method. This invasive method will cause certain vascular trauma to the subject every time an FFR test is performed, and a sufficiently long vascular recovery period must be separated between each two tests. In other words, this invasive method is not suitable as a sustainable monitoring method with a short time interval. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, electronic device and computer-readable storage medium for predicting blood flow reserve fraction based on coronary magnetic resonance imaging in response to the defects of the prior art. The present invention pre-constructs a first prediction model for predicting blood flow reserve fraction, and constructs a first data set to train the prediction model; and after the model training is completed, the model input data is prepared according to a group of rest imaging sequences, hyperemia imaging sequences and angiography of any subject, and the first prediction model predicts the blood flow reserve fraction according to the model input data, and marks the blood flow reserve fraction on the vascular tree and angiography based on the prediction results, and displays the marked vascular tree and angiography. The non-invasive blood flow reserve fraction prediction mechanism provided by the present invention can avoid damage to the subject's blood vessels, improve the subject's comfort, and can be used as a sustainable monitoring method with a short time interval.

[0004] To achieve the above object, a first aspect of an embodiment of the present invention provides a method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging, the method comprising:

[0005] Constructing a blood flow reserve fraction prediction model as a corresponding first prediction model; constructing a first data set for model training; and performing model training on the first prediction model based on the first data set;

[0006] After the model training is completed, the first rest imaging sequence, the first hyperemia imaging sequence and the first angiography of any subject are received; the first rest imaging sequence and the first hyperemia imaging sequence are respectively the coronary magnetic resonance phase contrast movie imaging sequences of the subject in the resting and hyperemia states, and the first angiography is the three-dimensional coronary magnetic resonance angiography of the subject; the first rest imaging sequence is composed of a plurality of the first rest imagings in chronological order; the first hyperemia imaging sequence is composed of a plurality of the first hyperemia imagings in chronological order; the time interval between every two adjacent first hyperemia imagings is consistent with the time interval between every two adjacent first rest imagings; the first rest imaging includes a plurality of first pixel points, each of which corresponds to a first pixel point vector; the first hyperemia imaging includes a plurality of second pixel points, each of which corresponds to a second pixel point vector; the first and second pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity directions;

[0007] A three-dimensional coronary vessel tree is constructed based on the first angiography image to obtain a corresponding first vessel tree; and the inner wall edge of the coronary vessel and the center line of the coronary vessel are identified for each of the first rest images, each of the first hyperemic images and the first vessel tree; and the first vessel tree is sampled based on a preset vessel center point sampling rule to obtain a corresponding first sampling point set; the first sampling point set includes a plurality of first sampling points; each of the first sampling points corresponds to a first sampling point vector; the first sampling point vector includes a first sampling point coordinate and a first sampling point inner diameter; the first sampling point coordinate is the relative displacement coordinate between the first sampling point and the center point of the coronary vessel entrance; the first sampling point inner diameter is the vessel inner diameter at the first sampling point;

[0008] According to the first sampling point set, the first vascular tree, the first rest imaging sequence and the first hyperemia imaging sequence, a rest / hyperemia flow waveform fitting process of the sampling points is performed to obtain a corresponding first rest flow waveform set and a first hyperemia flow waveform set; the first rest flow waveform set includes a plurality of first rest flow waveforms; the first hyperemia flow waveform set includes a plurality of first hyperemia flow waveforms; the first rest flow waveform and the first hyperemia flow waveform correspond to the first sampling point one by one;

[0009] performing coronary blood flow reserve fraction estimation processing according to the first hyperemic flow waveform and the first resting flow waveform corresponding to each of the first sampling points to obtain the corresponding first sampling point CFR value; and performing flow amplitude scaling processing on the corresponding first resting flow waveform according to the CFR value of each of the first sampling points to obtain the corresponding first sampling point scaled flow waveform;

[0010] The first sampling point coordinates, the first sampling point inner diameter and the first sampling point scaled flow waveform corresponding to each of the first sampling points form a corresponding first sampling point vector; and all the obtained first sampling point vectors form a corresponding first vector sequence;

[0011] The first prediction model predicts the blood flow reserve fraction according to the first vector sequence to obtain a corresponding first prediction vector; the first prediction vector includes a plurality of first sampling point FFR prediction values; the first sampling point FFR prediction values ​​correspond to the first sampling points one by one; the first sampling point FFR prediction value is a pressure ratio between the corresponding first sampling point and the center point of the coronary entrance, and the value is between 0 and 1;

[0012] Based on the first prediction vector and the first sampling point set, the first vascular tree and the first angiography are marked with blood flow reserve scores, and the marked points below a preset blood flow reserve score threshold are highlighted to obtain the corresponding first marked vascular tree and first marked angiography; and the first marked vascular tree and the first marked angiography are displayed.

[0013] Preferably, the first prediction model includes a data preprocessing module and an FFR prediction module; the FFR prediction module is implemented based on a class of deep learning models; the deep learning model includes at least an RNN model, a bi-RNN model, an LSTM model, a Bi-LSTM model and a Transformer model;

[0014] The data preprocessing module is used to normalize all sampling point coordinates and all sampling point inner diameters in the vector sequence input by the model to obtain corresponding multiple sampling point normalized coordinates and multiple sampling point normalized inner diameters; and based on a preset sampling frequency, perform flow sampling on each scaled flow waveform in the vector sequence input by the model to obtain a corresponding flow time series sequence Y, wherein the flow time series sequence Y is composed of multiple flow sampling values ​​y i The flow samples are sorted in order, 1≤time index i≤total number of flow samples; and each flow sample value y iand the corresponding normalized coordinates of the sampling points and the normalized inner diameter of the sampling points form a corresponding single-point moment feature; and all the single-point moment features corresponding to the sampling point vectors in the vector sequence input by the model are sorted in chronological order to form a corresponding sampling point time series feature; and when the FFR prediction module is implemented based on the RNN model, the bi-RNN model, the LSTM model or the Bi-LSTM model, all the obtained sampling point time series features are sorted in the order of the corresponding sampling point vectors to form a corresponding sampling point time series feature sequence and send it to the FFR prediction module; and when the FFR prediction module is implemented based on the Transformer model, all the obtained sampling point time series features are spliced ​​according to the feature channels to obtain a corresponding sampling point time series feature tensor and send it to the FFR prediction module;

[0015] The FFR prediction module is used to predict the blood flow reserve fraction according to the sampling point time series feature sequence when it is implemented based on the RNN model, bi-RNN model, LSTM model or Bi-LSTM model, obtain the corresponding prediction vector and output it; it is also used to predict the blood flow reserve fraction according to the sampling point time series feature tensor when it is implemented based on the Transformer model, obtain the corresponding prediction vector and output it; the prediction vector includes multiple sampling point FFR prediction values; the sampling point FFR prediction values ​​correspond one-to-one to the sampling point vectors in the vector sequence input by the model.

[0016] Preferably, the first data set includes a plurality of first data records; the first data record includes a first sample vector sequence and a first label prediction vector; the first sample vector sequence includes a plurality of first sample sampling point vectors; the first sample sampling point vector includes a first sample sampling point coordinate, a first sample sampling point inner diameter and a first sample sampling point scaled flow waveform; the first label prediction vector includes a plurality of first sampling point label FFR prediction values; the first sampling point label FFR prediction value corresponds one-to-one to the first sample sampling point vector; the first sampling point label FFR prediction value is a pressure ratio between the corresponding first sample sampling point and the center point of the coronary artery entrance, and the value is between 0 and 1;

[0017] The blood vessel center point sampling rule is as follows: starting from the center point of the coronary artery entrance, the current blood vessel center line position is recorded as the corresponding current sampling point at each preset first interval distance along the blood vessel center line of each trunk or branch blood vessel; the difference between the three-dimensional coordinates of the current sampling point and the center point of the coronary artery entrance is used as the corresponding current sampling point coordinate, and a tangent line of the blood vessel center line is made on the current sampling point, and a perpendicular line of the current tangent line is made through the current sampling point, and the straight-line distance between the two intersection points of the current perpendicular line and the edge of the blood vessel inner wall is used as the corresponding current sampling point inner diameter; and a corresponding current sampling point vector is formed by the obtained current sampling point coordinates and the current sampling point inner diameter; and after sampling of all the trunks or branches of the coronary artery is completed, a corresponding sampling point set is formed by all the obtained sampling points.

[0018] Preferably, the constructing of the first data set for model training specifically includes:

[0019] Step 41, recruiting a plurality of volunteers without coronary artery stenosis symptoms and a plurality of volunteers with coronary artery stenosis symptoms of varying degrees to form a first volunteer set;

[0020] Wherein, the first volunteer set includes a plurality of first volunteers;

[0021] Step 42, taking the first of the first volunteers in the first volunteer set as the corresponding current volunteer;

[0022] Step 43, performing magnetic resonance phase contrast movie imaging on the current volunteer in a resting state to obtain a corresponding first sample resting imaging sequence; performing magnetic resonance phase contrast movie imaging on the current volunteer in a hyperemic state to obtain a corresponding first sample hyperemic imaging sequence; and performing three-dimensional coronary magnetic resonance angiography on the current volunteer to obtain a corresponding first sample angiography image;

[0023] Wherein, the first sample rest imaging sequence is formed by a plurality of the first sample rest imaging in chronological order; the first sample hyperemia imaging sequence is formed by a plurality of the first sample hyperemia imaging in chronological order; the time interval between each two adjacent first sample hyperemia imaging is consistent with the time interval between each two adjacent first sample rest imaging; the first sample rest imaging includes a plurality of first sample pixel points, each of which corresponds to a first sample pixel point vector; the first sample hyperemia imaging includes a plurality of second sample pixel points, each of which corresponds to a second sample pixel point vector; the first and second sample pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction;

[0024] Step 44, constructing a three-dimensional coronary vessel tree based on the first sample angiogram to obtain a corresponding first sample vessel tree; and identifying the inner wall edge of the coronary vessel and the centerline of the coronary vessel for each of the first sample rest images, each of the first sample hyperemia images, and the first sample vessel tree; and sampling the first sample vessel tree based on the vessel center point sampling rule to obtain a corresponding first sample sampling point set;

[0025] Wherein, the first sample sampling point set includes a plurality of first sample sampling points; each of the first sample sampling points corresponds to a first sample sampling point vector; the first sample sampling point vector includes the first sample sampling point coordinates and the first sample sampling point inner diameter; the first sample sampling point coordinates are the relative displacement coordinates between the first sample sampling point and the center point of the coronary vessel entrance; the first sample sampling point inner diameter is the inner diameter of the blood vessel at the first sample sampling point;

[0026] Step 45, performing sampling point rest / congestion flow waveform fitting processing according to the first sample sampling point set, the first sample vascular tree, the first sample rest imaging sequence, and the first sample congestion imaging sequence to obtain a corresponding first sample rest flow waveform set and a first sample congestion flow waveform set;

[0027] Wherein, the first sample resting flow waveform set includes a plurality of first sample resting flow waveforms; the first sample congestion flow waveform set includes a plurality of first sample congestion flow waveforms; the first sample resting flow waveform, the first sample congestion flow waveform and the first sample sampling point correspond one to one;

[0028] Step 46, performing coronary blood flow reserve fraction estimation processing according to the first sample hyperemia flow waveform and the first sample resting flow waveform corresponding to each of the first sample sampling points to obtain the corresponding first sample sampling point CFR value; and performing flow amplitude scaling processing on the corresponding first sample resting flow waveform according to each of the first sample sampling point CFR values ​​to obtain the corresponding first sample sampling point scaled flow waveform;

[0029] Step 47, forming a corresponding first sample sampling point vector from the first sample sampling point coordinates, the first sample sampling point inner diameter and the first sample sampling point scaled flow waveform corresponding to each first sample sampling point; and forming a corresponding first sample vector sequence from all the obtained first sample sampling point vectors;

[0030] Step 48, using computational fluid dynamics simulation software to model the fluid dynamics simulation equations for solving the pressure distribution of each point on the three-dimensional coronary tree structure to obtain the corresponding first simulation equations; and setting the spatial boundary conditions of the first simulation equations according to the first sample vascular tree and the first sample sampling point set, and setting the temporal boundary conditions of the first simulation equations according to the first sample hyperemia flow waveforms corresponding to all the first sample sampling points; and after the setting of the spatial and temporal boundary conditions, performing simulation operation on the first simulation equations based on the computational fluid dynamics simulation software to obtain the pressure time series sequence of each of the first sample sampling points on the first sample vascular tree; and taking the pressure average of the pressure time series sequence of each of the first sample sampling points as the corresponding first sample sampling point pressure value; and taking the pressure ratio of the first sample sampling point pressure value of each of the first sample sampling points to the first sample sampling point pressure value corresponding to the center point of the coronary entrance as a corresponding first sampling point label FFR prediction value; and forming a corresponding first label prediction vector from all the obtained first sampling point label FFR prediction values;

[0031] Step 49, the first sample vector sequence and the first label prediction vector corresponding to the current volunteer form a corresponding first data record; and identify whether the current volunteer is the last first volunteer of the first volunteer set. If not, the next first volunteer of the first volunteer set is taken as the new current volunteer and the process returns to step 43. If so, all the first data records obtained form the corresponding first data set.

[0032] Preferably, the performing model training on the first prediction model based on the first data set specifically includes:

[0033] Step 51, setting an integer greater than zero as the corresponding first consecutive verification pass number; and setting a first counter initialized to 0;

[0034] Step 52: randomly select a plurality of the first data records from the first data set according to a preset test set ratio to form a corresponding current test set, and form a corresponding current verification set from all the remaining first data records;

[0035] Wherein, the current test set and the current validation set are both composed of a plurality of the first data records, and a ratio of the total number of the first data records of the current test set to the total number of the first data records of the first data set satisfies the test set ratio;

[0036] Step 53, extracting the first of the first data records in the current test set as the corresponding current data record;

[0037] Step 54, inputting the first sample vector sequence of the current data record into the first prediction model, and the first prediction model performs blood flow reserve fraction prediction according to the first sample vector sequence input this time to obtain a corresponding first training prediction vector;

[0038] Wherein, the first training prediction vector includes a plurality of first sample sampling point FFR prediction values; the first sample sampling point FFR prediction values ​​correspond one-to-one to the first sample sampling point vectors of the first sample vector sequence;

[0039] Step 55, bringing the first training prediction vector and the first label prediction vector of the current data record into a preset first model loss function to calculate and obtain a corresponding first loss value;

[0040] Wherein, the first model loss function is implemented based on the L1 loss function or the L2 loss function;

[0041] Step 56, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current data record is the last first data record of the current test set, if so, turning to step 57, if not, extracting the next first data record of the current test set as the new current data record and returning to step 54 to continue training; if the first loss value does not satisfy the first loss value range, performing a round of model parameter optimization on the first prediction model in the direction of minimizing the first model loss function based on a preset first model parameter optimizer, and returning to step 54 to continue training at the end of this round of parameter optimization;

[0042] Step 57, traverse all the first data records of the current verification set; and during the traversal, take the first data record currently traversed as the corresponding current verification record; and record the first sample vector sequence and the first label prediction vector of the current verification record as the corresponding current sample vector sequence and the current label prediction vector; and use the first prediction model to predict the blood flow reserve fraction according to the current sample vector sequence to obtain the corresponding current training prediction vector; and form a corresponding verification-label vector pair with the current training prediction vector and the current label prediction vector obtained this time; and at the end of the traversal, bring all the obtained verification-label vector pairs into the preset first error function to calculate and obtain the corresponding first error value;

[0043] Wherein, the first error function is implemented based on the MSE error function;

[0044] Step 58, identifying whether the first error value does not exceed a preset first error threshold; if the first error value does not exceed the first error threshold, adding 1 to the first counter, and identifying whether the first counter after adding 1 is greater than the first continuous verification pass number, if so, go to step 59, otherwise return to step 52 to continue training; if the first error value exceeds the first error threshold, clear the first counter and return to step 52 to continue training;

[0045] Step 59, solidify the model parameters of the first prediction model and confirm that the model training is completed.

[0046] Preferably, the sampling point resting / congestion flow waveform fitting process specifically includes:

[0047] Step 61, taking the sampling point set, vascular tree, rest imaging sequence and congestion imaging sequence input by the rest / congestion flow waveform fitting process of the current sampling point as the corresponding second sampling point set, second vascular tree, second rest imaging sequence and second congestion imaging sequence;

[0048] Wherein, the second sampling point set includes a plurality of second sampling points; each of the second sampling points corresponds to a second sampling point vector; the second sampling point vector includes a second sampling point coordinate and a second sampling point inner diameter; the second sampling point coordinate is the relative displacement coordinate between the second sampling point and the center point of the coronary vessel entrance; the second sampling point inner diameter is the vessel inner diameter at the second sampling point; the second rest imaging sequence is composed of a plurality of second rest imaging in chronological order; the second hyperemia imaging sequence is composed of a plurality of second hyperemia imaging in chronological order; the time interval between each two adjacent second hyperemia imaging is consistent with the time interval between each two adjacent second rest imaging; the second rest imaging includes a plurality of third pixel points, each of which corresponds to a third pixel point vector; the second hyperemia imaging includes a plurality of fourth pixel points, each of which corresponds to a fourth pixel point vector; the third and fourth pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction;

[0049] Step 62, taking the first second sampling point of the second sampling point set as the corresponding current sampling point;

[0050] Step 63, calculating the straight-line distance from the current sampling point to the center point of the coronary artery entrance of the second vascular tree to obtain a corresponding first distance, and calculating the blood vessel centerline length from the current sampling point to the center point of the coronary artery entrance of the second vascular tree to obtain a corresponding first length; and estimating the cross-sectional area of ​​the blood vessel inner wall at the current sampling point based on the inner diameter of the second sampling point of the current sampling point to obtain a corresponding first cross-sectional area;

[0051] Step 64, taking each of the third pixel points on the second resting image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as the corresponding resting matching pixel point; and sorting the third pixel point vectors of all the resting matching pixel points in chronological order to form a corresponding first sequence;

[0052] Step 65, taking each of the fourth pixels on the second hyperemia image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as the corresponding hyperemia matching pixel; and sorting the fourth pixel vectors of all the obtained hyperemia matching pixel points in chronological order to form a corresponding second sequence;

[0053] Step 66, performing curve fitting on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity coding and the positive / negative phase velocity direction of each of the third pixel point vectors in the first sequence to obtain a corresponding first resting velocity waveform; and performing curve fitting on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity coding and the positive / negative phase velocity direction of each of the fourth pixel point vectors in the second sequence to obtain a corresponding first hyperemia velocity waveform;

[0054] Step 67, according to the calculation method that the blood flow rate is approximately equal to the product of the blood vessel cross-sectional area and the blood flow velocity, the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first resting velocity waveform is used as the second resting flow waveform corresponding to the current sampling point, and the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first hyperemia velocity waveform is used as the second hyperemia flow waveform corresponding to the current sampling point;

[0055] Step 68, identifying whether the current sampling point is the last second sampling point of the second sampling point set; if so, going to step 69; if not, taking the next second sampling point of the second sampling point set as the new current sampling point and returning to step 63;

[0056] Step 69, all the second resting flow waveforms obtained form a corresponding resting flow waveform set for the current time; and all the second congestion flow waveforms obtained form a corresponding congestion flow waveform set for the current time; and the resting flow waveform set for the current time and the congestion flow waveform set for the current time obtained are output as the processing results of the resting / congestion flow waveform fitting processing of the current sampling point.

[0057] Preferably, the coronary blood flow reserve fraction estimation process specifically includes:

[0058] The hyperemic flow waveform and resting flow waveform inputted in this coronary blood flow reserve fraction estimation process are used as the corresponding current hyperemic flow waveform and current resting flow waveform; and the waveform amplitudes of the current hyperemic flow waveform and the current resting flow waveform are respectively averaged to obtain the corresponding hyperemic average flow and resting average flow; and the ratio of the hyperemic average flow to the resting average flow is used as the corresponding current sampling point CFR value; and the current sampling point CFR value obtained this time is outputted as the processing result of this coronary blood flow reserve fraction estimation process.

[0059] Preferably, the flow amplitude scaling process specifically includes:

[0060] The sampling point CFR value and resting flow waveform input for this flow amplitude scaling process are used as the corresponding current sampling point CFR value and current resting flow waveform; the waveform obtained by multiplying the current sampling point CFR value with the waveform amplitude of the current resting flow waveform is used as the corresponding current sampling point scaled flow waveform; and the current sampling point scaled flow waveform obtained this time is output as the processing result of this flow amplitude scaling process.

[0061] Preferably, the step of marking the first vascular tree and the first angiogram with a blood flow reserve fraction based on the first prediction vector and the first sampling point set and highlighting the marked points below a preset blood flow reserve fraction threshold to obtain the corresponding first marked vascular tree and first marked angiogram specifically includes:

[0062] Drawing a blood vessel section on the first blood vessel tree based on the first sampling point coordinates and the first sampling point inner diameter of each of the first sampling points of the first sampling point set to obtain a corresponding first sampling point blood vessel section; and taking each of the first sampling point blood vessel sections as a corresponding current section; and taking the first sampling point FFR prediction value corresponding to the current section in the first prediction vector as a corresponding current section FFR value; and identifying whether the current section FFR value is lower than the blood flow reserve fraction threshold, and if so, at the current section, based on the current section FFR value corresponding to a preset first color mark, and if not, at the current section, based on the current section FFR value corresponding to a preset second color mark; the brightness of the first color is greater than the second color;

[0063] and taking the angiogram voxel points on the first angiogram corresponding to each of the first sampling points of the first sampling point set as the corresponding first sampling voxel points; and taking each of the first sampling voxel points as the corresponding current voxel point; and taking the first sampling point FFR prediction value corresponding to the current voxel point in the first prediction vector as the corresponding current voxel point FFR value; and identifying whether the current voxel point FFR value is lower than the blood flow reserve fraction threshold, and if so, at the current voxel point based on the current voxel point FFR value corresponding to the first color mark, and if not, at the current voxel point based on the current voxel point FFR value corresponding to the second color mark;

[0064] The first vascular tree and the first angiography image that have been marked are used as the corresponding first marked vascular tree and the first marked angiography image.

[0065] A second aspect of an embodiment of the present invention provides a device for implementing the method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging as described in the first aspect, the device comprising: a model preparation module, a data receiving module, a first data preprocessing module, a second data preprocessing module, a third data preprocessing module, a fourth data preprocessing module, a model prediction module and a prediction result processing module;

[0066] The model preparation module is used to construct a blood flow reserve fraction prediction model recorded as a corresponding first prediction model; and to construct a first data set for model training; and to perform model training on the first prediction model based on the first data set;

[0067] The data receiving module is used to receive the first rest imaging sequence, the first hyperemia imaging sequence and the first angiography of any subject after the model training is completed; the first rest imaging sequence and the first hyperemia imaging sequence are respectively the coronary magnetic resonance phase contrast movie imaging sequences of the subject in the rest and hyperemia states, and the first angiography is the three-dimensional coronary magnetic resonance angiography of the subject; the first rest imaging sequence is composed of a plurality of the first rest imagings in chronological order; the first hyperemia imaging sequence is composed of a plurality of the first hyperemia imagings in chronological order; the time interval between every two adjacent first hyperemia imagings is consistent with the time interval between every two adjacent first rest imagings; the first rest imaging includes a plurality of first pixel points, each of which corresponds to a first pixel point vector; the first hyperemia imaging includes a plurality of second pixel points, each of which corresponds to a second pixel point vector; the first and second pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity directions;

[0068] The first data preprocessing module is used to construct a three-dimensional coronary vessel tree based on the first angiogram to obtain a corresponding first vessel tree; and to identify the inner wall edge of the coronary vessel and the center line of the coronary vessel for each of the first rest images, each of the first hyperemic images and the first vessel tree; and to sample the first vessel tree based on a preset vessel center point sampling rule to obtain a corresponding first sampling point set; the first sampling point set includes a plurality of first sampling points; each of the first sampling points corresponds to a first sampling point vector; the first sampling point vector includes a first sampling point coordinate and a first sampling point inner diameter; the first sampling point coordinate is the relative displacement coordinate between the first sampling point and the center point of the coronary vessel entrance; the first sampling point inner diameter is the vessel inner diameter at the first sampling point;

[0069] The second data preprocessing module is used for performing sampling point rest / congestion flow waveform fitting processing according to the first sampling point set, the first vascular tree, the first rest imaging sequence and the first congestion imaging sequence to obtain the corresponding first rest flow waveform set and first congestion flow waveform set; the first rest flow waveform set includes a plurality of first rest flow waveforms; the first congestion flow waveform set includes a plurality of first congestion flow waveforms; the first rest flow waveform and the first congestion flow waveform correspond to the first sampling point one by one;

[0070] The third data preprocessing module is used to perform coronary blood flow reserve fraction estimation processing according to the first hyperemic flow waveform and the first resting flow waveform corresponding to each of the first sampling points to obtain the corresponding first sampling point CFR value; and perform flow amplitude scaling processing on the corresponding first resting flow waveform according to each of the first sampling point CFR values ​​to obtain the corresponding first sampling point scaled flow waveform;

[0071] The fourth data preprocessing module is used to form a corresponding first sampling point vector from the first sampling point coordinates, the first sampling point inner diameter and the first sampling point scaled flow waveform corresponding to each first sampling point; and form a corresponding first vector sequence from all the obtained first sampling point vectors;

[0072] The model prediction module is used to predict the blood flow reserve fraction according to the first vector sequence by the first prediction model to obtain a corresponding first prediction vector; the first prediction vector includes a plurality of first sampling point FFR prediction values; the first sampling point FFR prediction values ​​correspond to the first sampling points one by one; the first sampling point FFR prediction value is the pressure ratio between the corresponding first sampling point and the center point of the coronary entrance, and the value is between 0 and 1;

[0073] The prediction result processing module is used to mark the blood flow reserve score of the first vascular tree and the first angiography image based on the first prediction vector and the first sampling point set, and highlight the marked points that are lower than a preset blood flow reserve score threshold to obtain the corresponding first marked vascular tree and first marked angiography image; and display the first marked vascular tree and the first marked angiography image.

[0074] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0075] The processor is used to be coupled to the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;

[0076] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0077] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions. When the computer instructions are executed by a computer, the computer executes the instructions of the method described in the first aspect above.

[0078] The embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for predicting blood flow reserve fraction based on coronary magnetic resonance imaging. As can be seen from the above content, the embodiment of the present invention pre-constructs a first prediction model for predicting blood flow reserve fraction, and constructs a first data set to train the prediction model; and after the model training is completed, the model input data is prepared according to a group of rest imaging sequences, hyperemia imaging sequences and angiography images of any subject, and the first prediction model predicts the blood flow reserve fraction according to the model input data, and the blood flow reserve fraction is marked on the vascular tree and angiography image based on the prediction results, and the marked vascular tree and angiography image are displayed. The non-invasive blood flow reserve fraction prediction mechanism provided by the embodiment of the present invention avoids damage to the subject's blood vessels, improves the subject's comfort, and establishes a sustainable monitoring method with a short time interval. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 A schematic diagram of a method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging provided in Example 1 of the present invention;

[0080] Figure 2 A module structure diagram of a device for predicting blood flow reserve fraction based on coronary magnetic resonance imaging provided in Embodiment 2 of the present invention;

[0081] Figure 3 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0083] Embodiment 1 of the present invention provides a method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging, such as Figure 1 As shown in the schematic diagram of a method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging provided in the first embodiment of the present invention, the method mainly comprises the following steps:

[0084] Step 1, constructing a blood flow reserve fraction prediction model and recording it as a corresponding first prediction model; and constructing a first data set for model training; and performing model training on the first prediction model based on the first data set;

[0085] Specifically comprising: step 11, constructing a blood flow reserve fraction prediction model recorded as a corresponding first prediction model;

[0086] Here, the first prediction model of the embodiment of the present invention includes a data preprocessing module and an FFR prediction module; wherein the FFR prediction module is implemented based on a class of deep learning models; the deep learning model at least includes an RNN model, a bi-RNN model, an LSTM model, a Bi-LSTM model and a Transformer model;

[0087] The data preprocessing module of the first prediction model of the embodiment of the present invention is used to normalize the coordinates of all sampling points and the inner diameters of all sampling points in the vector sequence input by the model to obtain corresponding normalized coordinates of multiple sampling points and normalized inner diameters of multiple sampling points; and based on a preset sampling frequency, flow sampling is performed on each scaled flow waveform in the vector sequence input by the model to obtain a corresponding flow time series sequence Y, and the flow time series sequence Y is composed of multiple flow sampling values ​​y i The flow samples are sorted in order, 1≤time index i≤total number of flow samples; and each flow sample value y iand the corresponding normalized coordinates of the sampling points and the normalized inner diameter of the sampling points form a corresponding single-point moment feature; and all the single-point moment features corresponding to each sampling point vector in the vector sequence input by the model are sorted in chronological order to form a corresponding sampling point time series feature; and when the FFR prediction module is implemented based on the RNN model, the bi-RNN model, the LSTM model or the Bi-LSTM model, all the obtained sampling point time series features are sorted in the order of the corresponding sampling point vectors to form a corresponding sampling point time series feature sequence and sent to the FFR prediction module; and when the FFR prediction module is implemented based on the Transformer model, all the obtained sampling point time series features are spliced ​​according to the feature channels to obtain a corresponding sampling point time series feature tensor and send it to the FFR prediction module;

[0088] Here, the preset sampling frequency is a preset sampling frequency parameter;

[0089] The FFR prediction module of the first prediction model of the embodiment of the present invention is used to predict the blood flow reserve fraction according to the sampling point time series feature sequence when it is implemented based on the RNN model, bi-RNN model, LSTM model or Bi-LSTM model, obtain the corresponding prediction vector and output it; it is also used to predict the blood flow reserve fraction according to the sampling point time series feature tensor when it is implemented based on the Transformer model, obtain the corresponding prediction vector and output it; wherein the prediction vector includes a plurality of sampling point FFR prediction values; the sampling point FFR prediction values ​​correspond one-to-one to the sampling point vectors in the vector sequence input by the model;

[0090] Step 12, and construct a first data set for model training;

[0091] Wherein, the first data set includes a plurality of first data records; the first data record includes a first sample vector sequence and a first label prediction vector; the first sample vector sequence includes a plurality of first sample sampling point vectors; the first sample sampling point vector includes a first sample sampling point coordinate, a first sample sampling point inner diameter and a first sample sampling point scaled flow waveform; the first label prediction vector includes a plurality of first sampling point label FFR prediction values; the first sampling point label FFR prediction value corresponds one-to-one to the first sample sampling point vector; the first sampling point label FFR prediction value is a pressure ratio between the corresponding first sample sampling point and the center point of the coronary artery entrance, and the value is between 0 and 1;

[0092] Specifically, it includes: step 121, recruiting a plurality of volunteers without coronary artery stenosis symptoms and a plurality of volunteers with coronary artery stenosis symptoms of varying degrees to form a first volunteer group;

[0093] Wherein, the first volunteer set includes multiple first volunteers;

[0094] Step 122, taking the first first volunteer in the first volunteer set as the corresponding current volunteer;

[0095] Step 123, performing a Phase Contrast Cine Magnetic Resonance Imaging (PC-MRI) on the current volunteer in a resting state to obtain a corresponding first sample resting imaging sequence; performing a Phase Contrast Cine Magnetic Resonance Imaging on the current volunteer in a hyperemic state to obtain a corresponding first sample hyperemic imaging sequence; and performing a three-dimensional coronary magnetic resonance angiography (MRA) on the current volunteer to obtain a corresponding first sample angiography image;

[0096] Wherein, the first sample rest imaging sequence is formed by a plurality of first sample rest imaging in chronological order; the first sample hyperemia imaging sequence is formed by a plurality of first sample hyperemia imaging in chronological order; the time interval between each two adjacent first sample hyperemia imaging is consistent with the time interval between each two adjacent first sample rest imaging; the first sample rest imaging includes a plurality of first sample pixel points, each of which corresponds to a first sample pixel point vector; the first sample hyperemia imaging includes a plurality of second sample pixel points, each of which corresponds to a second sample pixel point vector; the first and second sample pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction;

[0097] Step 124, constructing a three-dimensional coronary vessel tree based on the first sample angiography to obtain a corresponding first sample vessel tree; and identifying the inner wall edge of the coronary vessel and the center line of the coronary vessel for each first sample rest image, each first sample hyperemia image, and the first sample vessel tree; and sampling the first sample vessel tree based on a preset vessel center point sampling rule to obtain a corresponding first sample sampling point set;

[0098] Here, in the embodiment of the present invention, the construction of a three-dimensional blood vessel tree based on a three-dimensional angiography image can be processed in a variety of ways, one of which is to use three-dimensional modeling software to achieve it;

[0099] The embodiment of the present invention can identify the inner wall edge of the coronary artery and the centerline of the coronary artery for rest imaging, hyperemia imaging and vascular tree respectively in a variety of ways. One of the ways is to use a coronary rest imaging / coronary hyperemia imaging / coronary tree vascular inner wall edge recognition model and a coronary rest imaging / coronary hyperemia imaging / coronary tree vascular centerline recognition model that are implemented based on a visual image model and have completed model training to perform processing;

[0100] The blood vessel center point sampling rule of the embodiment of the present invention is to start from the center point of the coronary artery entrance, and record the current blood vessel center line position as the corresponding current sampling point at each preset first interval distance along the blood vessel center line of each trunk or branch blood vessel downward; and use the difference between the three-dimensional coordinates of the current sampling point and the center point of the coronary artery entrance as the corresponding current sampling point coordinates, and make a tangent to the blood vessel center line at the current sampling point, and make a perpendicular line to the current tangent through the current sampling point, and use the straight-line distance between the two intersection points of the current perpendicular line and the edge of the inner wall of the blood vessel as the corresponding current sampling point inner diameter; and form a corresponding current sampling point vector with the obtained current sampling point coordinates and the current sampling point inner diameter; and form a corresponding sampling point set with all the obtained sampling points after sampling of all the trunks or branches of the coronary artery is completed; here, the first interval distance is a preset distance parameter, and the distance parameter is not a straight-line distance parameter but a blood-flow distance parameter, that is, the length of a straight or non-straight blood-flow distance;

[0101] The first sample sampling point set obtained based on the above-mentioned blood vessel center point sampling rule includes multiple first sample sampling points; each first sample sampling point corresponds to a first sample sampling point vector; the first sample sampling point vector includes the first sample sampling point coordinates and the first sample sampling point inner diameter; the first sample sampling point coordinates are the relative displacement coordinates of the first sample sampling point and the center point of the coronary blood vessel entrance; the first sample sampling point inner diameter is the blood vessel inner diameter at the first sample sampling point;

[0102] Step 125, performing sampling point rest / congestion flow waveform fitting processing according to the first sample sampling point set, the first sample vascular tree, the first sample rest imaging sequence and the first sample congestion imaging sequence to obtain the corresponding first sample rest flow waveform set and the first sample congestion flow waveform set;

[0103] The first sample resting flow waveform set includes a plurality of first sample resting flow waveforms; the first sample congestion flow waveform set includes a plurality of first sample congestion flow waveforms; the first sample resting flow waveforms, the first sample congestion flow waveforms correspond to the first sample sampling points one by one;

[0104] Specifically, it includes: step 1251, taking the sampling point set, vascular tree, rest imaging sequence and congestion imaging sequence input by the rest / congestion flow waveform fitting processing of the current sampling point as the corresponding second sampling point set, second vascular tree, second rest imaging sequence and second congestion imaging sequence;

[0105] Wherein, the second sampling point set includes a plurality of second sampling points; each second sampling point corresponds to a second sampling point vector; the second sampling point vector includes the second sampling point coordinates and the second sampling point inner diameter; the second sampling point coordinates are the relative displacement coordinates between the second sampling point and the center point of the coronary vessel entrance; the second sampling point inner diameter is the vessel inner diameter at the second sampling point; the second rest imaging sequence is composed of a plurality of second rest imaging in chronological order; the second congestion imaging sequence is composed of a plurality of second congestion imaging in chronological order; the time interval between each two adjacent second congestion imaging is consistent with the time interval between each two adjacent second rest imaging; the second rest imaging includes a plurality of third pixel points, each of which corresponds to a third pixel point vector; the second congestion imaging includes a plurality of fourth pixel points, each of which corresponds to a fourth pixel point vector; the third and fourth pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction;

[0106] Step 1252, taking the first second sampling point of the second sampling point set as the corresponding current sampling point;

[0107] Step 1253, the straight-line distance from the current sampling point to the center point of the coronary artery entrance of the second vascular tree is calculated to obtain a corresponding first distance, and the blood vessel centerline length from the current sampling point to the center point of the coronary artery entrance of the second vascular tree is calculated to obtain a corresponding first length; and the cross-sectional area of ​​the inner wall of the blood vessel at the current sampling point is estimated based on the second sampling point inner diameter of the current sampling point to obtain a corresponding first cross-sectional area;

[0108] Step 1254, taking the third pixel points on each second resting image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as corresponding resting matching pixel points; and sorting the third pixel point vectors of all the resting matching pixel points in chronological order to form a corresponding first sequence;

[0109] Step 1255, taking the fourth pixel points on each second hyperemia image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as corresponding hyperemia matching pixel points; and sorting the fourth pixel point vectors of all the obtained hyperemia matching pixel points in chronological order to form a corresponding second sequence;

[0110] Step 1256, curve fitting is performed on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity code and the positive / negative phase velocity direction of each third pixel point vector in the first sequence to obtain the corresponding first resting velocity waveform; and curve fitting is performed on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity code and the positive / negative phase velocity direction of each fourth pixel point vector in the second sequence to obtain the corresponding first hyperemia velocity waveform;

[0111] Step 1257, according to the calculation method that the blood flow rate is approximately equal to the product of the cross-sectional area of ​​the blood vessel and the blood flow velocity, the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first resting velocity waveform is used as the second resting flow waveform corresponding to the current sampling point, and the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first hyperemia velocity waveform is used as the second hyperemia flow waveform corresponding to the current sampling point;

[0112] Step 1258, identifying whether the current sampling point is the last second sampling point of the second sampling point set; if so, go to step 1259; if not, take the next second sampling point of the second sampling point set as the new current sampling point and return to step 1253;

[0113] Step 1259, all the obtained second resting flow waveforms form a corresponding resting flow waveform set for the current time; and all the obtained second hyperemia flow waveforms form a corresponding hyperemia flow waveform set for the current time; and the resting flow waveform set for the current time and the hyperemia flow waveform set for the current time obtained this time are output as the processing result of the resting / hyperemia flow waveform fitting processing of the current sampling point;

[0114] Here, the processing result of the resting / congestion flow waveform fitting processing of this sampling point is the first sample resting flow waveform set and the first sample congestion flow waveform set;

[0115] Step 126, performing coronary flow reserve (CFR) estimation processing according to the first sample hyperemic flow waveform and the first sample resting flow waveform corresponding to each first sample sampling point to obtain the corresponding CFR value of the first sample sampling point; and performing flow amplitude scaling processing on the corresponding first sample resting flow waveform according to the CFR value of each first sample sampling point to obtain the corresponding scaled flow waveform of the first sample sampling point;

[0116] Specifically, it includes: step 1261, performing coronary blood flow reserve fraction estimation processing according to the first sample hyperemia flow waveform and the first sample resting flow waveform corresponding to each first sample sampling point to obtain the corresponding CFR value of the first sample sampling point;

[0117] Specifically, it includes: using the hyperemic flow waveform and the resting flow waveform inputted in the current coronary blood flow reserve fraction estimation process as the corresponding current hyperemic flow waveform and the current resting flow waveform; and respectively calculating the mean of the waveform amplitudes of the current hyperemic flow waveform and the current resting flow waveform to obtain the corresponding hyperemic average flow and the resting average flow; and using the ratio of the hyperemic average flow to the resting average flow as the corresponding current sampling point CFR value; and outputting the current sampling point CFR value obtained this time as the processing result of the current coronary blood flow reserve fraction estimation process;

[0118] Here, the processing result of this coronary blood flow reserve fraction estimation process is the CFR value of the first sample sampling point;

[0119] Step 1262, and performing flow amplitude scaling processing on the corresponding first sample resting flow waveform with the CFR value of each first sample sampling point to obtain the corresponding first sample sampling point scaled flow waveform;

[0120] Specifically, it includes: using the sampling point CFR value and resting flow waveform input by this flow amplitude scaling processing as the corresponding current sampling point CFR value and current resting flow waveform; multiplying the current sampling point CFR value with the waveform amplitude of the current resting flow waveform to obtain a waveform as the corresponding current sampling point scaled flow waveform; and outputting the current sampling point scaled flow waveform obtained this time as the processing result of this flow amplitude scaling processing.

[0121] Here, the processing result of this flow amplitude scaling process is the scaled flow waveform of the first sample sampling point;

[0122] Step 127, a first sample sampling point vector is formed by the first sample sampling point coordinates, the first sample sampling point inner diameter and the first sample sampling point scaled flow waveform corresponding to each first sample sampling point; and a corresponding first sample vector sequence is formed by all the obtained first sample sampling point vectors;

[0123] Step 128, using computational fluid dynamics (CFD) simulation software to model the fluid dynamics simulation equations for solving the pressure distribution of each point on the three-dimensional coronary tree structure to obtain the corresponding first simulation equations; and setting the spatial boundary conditions of the first simulation equations according to the first sample vascular tree and the first sample sampling point set, and setting the temporal boundary conditions of the first simulation equations according to the first sample hyperemia flow waveforms corresponding to all the first sample sampling points; and after the setting of the spatial and temporal boundary conditions, performing simulation operations on the first simulation equations based on the computational fluid dynamics simulation software to obtain the pressure time series sequence of each first sample sampling point on the first sample vascular tree; and taking the pressure average of the pressure time series sequence of each first sample sampling point as the corresponding first sample sampling point pressure value; and taking the pressure ratio of the first sample sampling point pressure value of each first sample sampling point to the first sample sampling point pressure value corresponding to the center point of the coronary entrance as a corresponding first sampling point label FFR prediction value; and forming a corresponding first label prediction vector from all the obtained first sampling point label FFR prediction values;

[0124] Here, after the three-dimensional structure of the coronary artery and the flow time series at each sampling point are known, calculating the fluid dynamics pressure parameters at each sampling point of the three-dimensional coronary artery through CFD simulation is a public technical means, and correspondingly, there are also many types of software for CFD simulation available for use; this public technical means is used to solve the fluid dynamics simulation equation group of the pressure distribution at each point on the three-dimensional structure. The equation group is commonly composed of the Wiener-Stokes equation + the incompressible fluid continuity equation based on the premise that blood is an incompressible Newtonian fluid; the space-time boundary configuration process of this public technical means can be understood by referring to the configuration instructions of the public CFD simulation software, and will not be further elaborated here;

[0125] Step 129, the first sample vector sequence and the first label prediction vector corresponding to the current volunteer form a corresponding first data record; and identify whether the current volunteer is the last first volunteer of the first volunteer set, if not, the next first volunteer of the first volunteer set is used as the new current volunteer and the process returns to step 123, if so, all the first data records obtained form a corresponding first data set;

[0126] Step 13, performing model training on the first prediction model based on the first data set;

[0127] Specifically, it includes: step 131, setting an integer greater than zero as the corresponding first continuous verification pass number; and setting a first counter initialized to 0;

[0128] Step 132, randomly selecting a plurality of first data records from the first data set according to a preset test set ratio to form a corresponding current test set, and forming a corresponding current verification set from all remaining first data records;

[0129] The test set ratio is a preset ratio value, such as 80%; the current test set and the current validation set are both composed of a plurality of first data records, and the ratio of the total number of first data records of the current test set to the total number of first data records of the first data set satisfies the test set ratio;

[0130] Step 133, extracting the first first data record of the current test set as the corresponding current data record;

[0131] Step 134, inputting the first sample vector sequence of the current data record into the first prediction model, and the first prediction model performs blood flow reserve fraction prediction based on the first sample vector sequence input this time to obtain a corresponding first training prediction vector;

[0132] Wherein, the first training prediction vector includes a plurality of first sample sampling point FFR prediction values; the first sample sampling point FFR prediction values ​​correspond one-to-one to the first sample sampling point vectors of the first sample vector sequence;

[0133] Step 135, bringing the first training prediction vector and the first label prediction vector of the current data record into a preset first model loss function to calculate and obtain a corresponding first loss value;

[0134] Wherein, the first model loss function is implemented based on the L1 loss function or the L2 loss function;

[0135] Step 136, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current data record is the last first data record of the current test set, if so, turning to step 137, if not, extracting the next first data record of the current test set as a new current data record and returning to step 134 to continue training; if the first loss value does not satisfy the first loss value range, performing a round of model parameter optimization on the first prediction model in the direction of minimizing the first model loss function based on a preset first model parameter optimizer, and returning to step 134 to continue training at the end of this round of parameter optimization;

[0136] Here, the first loss value range is a preset loss value range parameter; the first model parameter optimizer includes at least an SDG optimizer;

[0137] Step 137, traverse all first data records of the current verification set; and during the traversal, use the currently traversed first data record as the corresponding current verification record; and record the first sample vector sequence and the first label prediction vector of the current verification record as the corresponding current sample vector sequence and the current label prediction vector; and use the first prediction model to predict the blood flow reserve fraction according to the current sample vector sequence to obtain the corresponding current training prediction vector; and use the current training prediction vector and the current label prediction vector obtained this time to form a corresponding verification-label vector pair; and at the end of the traversal, bring all the obtained verification-label vector pairs into the preset first error function to calculate and obtain the corresponding first error value;

[0138] Wherein, the first error function is implemented based on a mean-square error (MSE) function;

[0139] Step 138, identifying whether the first error value does not exceed the preset first error threshold; if the first error value does not exceed the first error threshold, then adding 1 to the first counter, and identifying whether the first counter after adding 1 is greater than the first continuous verification pass number, if so, go to step 139, if not, return to step 132 to continue training; if the first error value exceeds the first error threshold, then clear the first counter and return to step 132 to continue training;

[0140] Here, the first error threshold is a preset error value parameter;

[0141] Step 139, solidify the model parameters of the first prediction model and confirm that the model training is completed.

[0142] Step 2, after the model training is completed, receiving the first rest imaging sequence, the first hyperemia imaging sequence and the first angiogram of any subject;

[0143] Among them, the first resting imaging sequence and the first congestion imaging sequence are coronary magnetic resonance phase contrast movie imaging sequences of the subject in the resting and congestion states, respectively, and the first angiogram is a three-dimensional coronary magnetic resonance angiography imaging of the subject; the first resting imaging sequence is composed of multiple first resting images in chronological order; the first congestion imaging sequence is composed of multiple first congestion images in chronological order; the time interval between every two adjacent first congestion images is consistent with the time interval between every two adjacent first resting images; the first resting imaging includes multiple first pixel points, each first pixel point corresponds to a first pixel point vector; the first congestion imaging includes multiple second pixel points, each second pixel point corresponds to a second pixel point vector; the first and second pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity directions.

[0144] Step 3, constructing a three-dimensional coronary vascular tree based on the first angiography image to obtain a corresponding first vascular tree; and identifying the inner wall edge of the coronary artery and the center line of the coronary artery for each first resting image, each first congestion image and the first vascular tree respectively; and sampling the first vascular tree based on a preset vascular center point sampling rule to obtain a corresponding first sampling point set.

[0145] Here, the first sampling point set obtained based on the blood vessel center point sampling rule of the embodiment of the present invention includes multiple first sampling points; each first sampling point corresponds to a first sampling point vector; the first sampling point vector includes the first sampling point coordinates and the first sampling point inner diameter; the first sampling point coordinates are the relative displacement coordinates of the first sampling point and the center point of the coronary vessel entrance; the first sampling point inner diameter is the blood vessel inner diameter at the first sampling point.

[0146] Step 4, performing sampling point rest / congestion flow waveform fitting processing according to the first sampling point set, the first vascular tree, the first rest imaging sequence and the first congestion imaging sequence to obtain the corresponding first rest flow waveform set and first congestion flow waveform set;

[0147] The first resting flow waveform set includes a plurality of first resting flow waveforms; the first congestion flow waveform set includes a plurality of first congestion flow waveforms; the first resting flow waveforms and the first congestion flow waveforms correspond to the first sampling points one by one;

[0148] It should be noted that the processing of the sampling point resting / congestion flow waveform fitting processing in the current step 4 is similar to the processing steps of the previous step 125, specifically including:

[0149] Step 41, taking the sampling point set, vascular tree, rest imaging sequence and congestion imaging sequence input by the rest / congestion flow waveform fitting process of the current sampling point as the corresponding second sampling point set, second vascular tree, second rest imaging sequence and second congestion imaging sequence;

[0150] Wherein, the second sampling point set includes a plurality of second sampling points; each second sampling point corresponds to a second sampling point vector; the second sampling point vector includes the second sampling point coordinates and the second sampling point inner diameter; the second sampling point coordinates are the relative displacement coordinates between the second sampling point and the center point of the coronary vessel entrance; the second sampling point inner diameter is the vessel inner diameter at the second sampling point; the second rest imaging sequence is composed of a plurality of second rest imaging in chronological order; the second congestion imaging sequence is composed of a plurality of second congestion imaging in chronological order; the time interval between each two adjacent second congestion imaging is consistent with the time interval between each two adjacent second rest imaging; the second rest imaging includes a plurality of third pixel points, each of which corresponds to a third pixel point vector; the second congestion imaging includes a plurality of fourth pixel points, each of which corresponds to a fourth pixel point vector; the third and fourth pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction;

[0151] Step 42, taking the first second sampling point of the second sampling point set as the corresponding current sampling point;

[0152] Step 43, calculating the straight-line distance from the current sampling point to the center point of the coronary artery entrance of the second vascular tree to obtain the corresponding first distance, and calculating the blood vessel centerline length from the current sampling point to the center point of the coronary artery entrance of the second vascular tree to obtain the corresponding first length; and estimating the cross-sectional area of ​​the inner wall of the blood vessel at the current sampling point based on the second sampling point inner diameter of the current sampling point to obtain the corresponding first cross-sectional area;

[0153] Step 44, taking the third pixel points on each second resting image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as corresponding resting matching pixel points; and sorting the third pixel point vectors of all the resting matching pixel points in chronological order to form a corresponding first sequence;

[0154] Step 45, taking the fourth pixel points on each second hyperemia image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as corresponding hyperemia matching pixel points; and sorting the fourth pixel point vectors of all the obtained hyperemia matching pixel points in chronological order to form a corresponding second sequence;

[0155] Step 46, curve fitting is performed on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity coding and the positive / negative phase velocity direction of each third pixel point vector in the first sequence to obtain the corresponding first resting velocity waveform; and curve fitting is performed on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity coding and the positive / negative phase velocity direction of each fourth pixel point vector in the second sequence to obtain the corresponding first hyperemia velocity waveform;

[0156] Step 47, according to the calculation method that the blood flow rate is approximately equal to the product of the cross-sectional area of ​​the blood vessel and the blood flow velocity, the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first resting velocity waveform is used as the second resting flow waveform corresponding to the current sampling point, and the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first hyperemia velocity waveform is used as the second hyperemia flow waveform corresponding to the current sampling point;

[0157] Step 48, identifying whether the current sampling point is the last second sampling point of the second sampling point set; if so, go to step 49; if not, take the next second sampling point of the second sampling point set as the new current sampling point and return to step 43;

[0158] Step 49, all the obtained second resting flow waveforms form a corresponding resting flow waveform set for the current time; and all the obtained second congestion flow waveforms form a corresponding congestion flow waveform set for the current time; and the resting flow waveform set and the congestion flow waveform set obtained this time are output as the processing results of the resting / congestion flow waveform fitting processing of the current sampling point.

[0159] Here, the processing result of the resting / congestion flow waveform fitting processing of this sampling point is the first resting flow waveform set and the first congestion flow waveform set.

[0160] Step 5, performing coronary blood flow reserve fraction estimation processing according to the first hyperemic flow waveform and the first resting flow waveform corresponding to each first sampling point to obtain the corresponding first sampling point CFR value; and performing flow amplitude scaling processing on the corresponding first resting flow waveform according to the CFR value of each first sampling point to obtain the corresponding first sampling point scaled flow waveform;

[0161] Specifically, it includes: step 51, performing coronary blood flow reserve fraction estimation processing according to the first hyperemic flow waveform and the first resting flow waveform corresponding to each first sampling point to obtain the corresponding first sampling point CFR value;

[0162] It should be noted that the processing procedure of the coronary blood flow reserve fraction estimation processing of the current step 51 is similar to the processing steps of the previous step 1261, specifically including: taking the hyperemic flow waveform and the resting flow waveform inputted in the current coronary blood flow reserve fraction estimation processing as the corresponding current hyperemic flow waveform and the current resting flow waveform; and respectively calculating the mean of the waveform amplitudes of the current hyperemic flow waveform and the current resting flow waveform to obtain the corresponding hyperemic average flow and the resting average flow; and taking the ratio of the hyperemic average flow to the resting average flow as the corresponding current sampling point CFR value; and outputting the current sampling point CFR value obtained this time as the processing result of the current coronary blood flow reserve fraction estimation processing; here, the processing result of the current coronary blood flow reserve fraction estimation processing is the CFR value of the first sampling point;

[0163] Step 52, performing flow amplitude scaling processing on the corresponding first resting flow waveform using the CFR value of each first sampling point to obtain the corresponding first sampling point scaled flow waveform;

[0164] It should be noted that the processing procedure of the flow amplitude scaling processing of the current step 52 is similar to the processing steps of the previous step 1262, specifically including: taking the sampling point CFR value and the resting flow waveform input for this flow amplitude scaling processing as the corresponding current sampling point CFR value and the current resting flow waveform; multiplying the current sampling point CFR value with the waveform amplitude of the current resting flow waveform to obtain a waveform as the corresponding current sampling point scaled flow waveform; and outputting the current sampling point scaled flow waveform obtained this time as the processing result of this flow amplitude scaling processing; here, the processing result of this flow amplitude scaling processing is the first sampling point scaled flow waveform.

[0165] Step 6: The first sampling point coordinates, the first sampling point inner diameter and the first sampling point scaled flow waveform corresponding to each first sampling point form a corresponding first sampling point vector; and all the obtained first sampling point vectors form a corresponding first vector sequence.

[0166] Step 7: The first prediction model predicts the blood flow reserve fraction according to the first vector sequence to obtain a corresponding first prediction vector.

[0167] Here, the first prediction vector includes multiple first sampling point FFR prediction values; the first sampling point FFR prediction values ​​correspond to the first sampling points one by one; the first sampling point FFR prediction value is the pressure ratio between the corresponding first sampling point and the center point of the coronary entrance, and the value is between 0 and 1.

[0168] Step 8, based on the first prediction vector and the first sampling point set, the first vascular tree and the first angiogram are marked with a blood flow reserve fraction, and the marked points that are lower than a preset blood flow reserve fraction threshold are highlighted to obtain the corresponding first marked vascular tree and the first marked angiogram; and the first marked vascular tree and the first marked angiogram are displayed;

[0169] Specifically, it includes: step 81, based on the first prediction vector and the first sampling point set, the first vascular tree and the first angiogram are marked with a blood flow reserve score, and the marked points that are lower than a preset blood flow reserve score threshold are highlighted to obtain a corresponding first marked vascular tree and a first marked angiogram;

[0170] Specifically, it includes: step 811, drawing a blood vessel section on the first blood vessel tree based on the first sampling point coordinates and the first sampling point inner diameter of each first sampling point of the first sampling point set to obtain a corresponding first sampling point blood vessel section; and taking each first sampling point blood vessel section as the corresponding current section; and taking the first sampling point FFR prediction value corresponding to the current section in the first prediction vector as the corresponding current section FFR value; and identifying whether the current section FFR value is lower than the blood flow reserve fraction threshold, and if so, the current section FFR value corresponding to the preset first color mark at the current section is determined, and if not, the current section FFR value corresponding to the preset second color mark at the current section is determined;

[0171] Here, the blood flow reserve score threshold is a preset blood flow reserve score value; the first and second colors are two preset colors, wherein the brightness of the first color is greater than that of the second color;

[0172] Step 812, and use the angiogram voxel points corresponding to each first sampling point of the first sampling point set on the first angiogram as the corresponding first sampling voxel points; and use each first sampling voxel point as the corresponding current voxel point; and use the first sampling point FFR prediction value corresponding to the current voxel point in the first prediction vector as the corresponding current voxel point FFR value; and identify whether the current voxel point FFR value is lower than the blood flow reserve fraction threshold, if so, then the current voxel point FFR value corresponding to the first color mark is based on the current voxel point at the current voxel point, if not, then the current voxel point FFR value corresponding to the second color mark is based on the current voxel point at the current voxel point;

[0173] Step 813, and taking the marked first vascular tree and the first angiography image as the corresponding first marked vascular tree and the first marked angiography image;

[0174] Step 82: Display the first marked blood vessel tree and the first marked angiogram.

[0175] Figure 2The module structure diagram of a device for predicting blood flow reserve fraction based on coronary magnetic resonance imaging provided in the second embodiment of the present invention is a terminal device or server for implementing the aforementioned method embodiment, and can also be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiment. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 2 As shown, the device includes: a model preparation module 201, a data receiving module 202, a first data preprocessing module 203, a second data preprocessing module 204, a third data preprocessing module 205, a fourth data preprocessing module 206, a model prediction module 207 and a prediction result processing module 208.

[0176] The model preparation module 201 is used to construct a blood flow reserve fraction prediction model recorded as a corresponding first prediction model; and to construct a first data set for model training; and to perform model training on the first prediction model based on the first data set.

[0177] The data receiving module 202 is used to receive the first resting imaging sequence, the first congestion imaging sequence and the first angiography of any subject after the model training is completed; the first resting imaging sequence and the first congestion imaging sequence are coronary magnetic resonance phase contrast movie imaging sequences of the subject in the resting and congestion states, respectively, and the first angiography is the three-dimensional coronary magnetic resonance angiography of the subject; the first resting imaging sequence is composed of multiple first resting images in chronological order; the first congestion imaging sequence is composed of multiple first congestion images in chronological order; the time interval between every two adjacent first congestion images is consistent with the time interval between every two adjacent first resting images; the first resting imaging includes multiple first pixel points, each of which corresponds to a first pixel point vector; the first congestion imaging includes multiple second pixel points, each of which corresponds to a second pixel point vector; the first and second pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction.

[0178] The first data preprocessing module 203 is used to construct a three-dimensional coronary vascular tree based on the first angiography image to obtain a corresponding first vascular tree; and to identify the inner wall edge of the coronary vessel and the center line of the coronary vessel for each first resting image, each first congestion image and the first vascular tree respectively; and to sample the first vascular tree based on a preset vascular center point sampling rule to obtain a corresponding first sampling point set; the first sampling point set includes multiple first sampling points; each first sampling point corresponds to a first sampling point vector; the first sampling point vector includes a first sampling point coordinate and a first sampling point inner diameter; the first sampling point coordinate is the relative displacement coordinate of the first sampling point and the center point of the coronary vessel entrance; the first sampling point inner diameter is the vascular inner diameter at the first sampling point.

[0179] The second data preprocessing module 204 is used to perform resting / congestive flow waveform fitting processing on the sampling points according to the first sampling point set, the first vascular tree, the first resting imaging sequence and the first congestive imaging sequence to obtain the corresponding first resting flow waveform set and first congestive flow waveform set; the first resting flow waveform set includes multiple first resting flow waveforms; the first congestive flow waveform set includes multiple first congestive flow waveforms; the first resting flow waveform and the first congestive flow waveform correspond one-to-one to the first sampling point.

[0180] The third data preprocessing module 205 is used to estimate the coronary blood flow reserve fraction according to the first congestion flow waveform and the first resting flow waveform corresponding to each first sampling point to obtain the corresponding first sampling point CFR value; and to perform flow amplitude scaling processing on the corresponding first resting flow waveform according to the CFR value of each first sampling point to obtain the corresponding first sampling point scaled flow waveform.

[0181] The fourth data preprocessing module 206 is used to form a corresponding first sampling point vector from the first sampling point coordinates, the first sampling point inner diameter and the first sampling point scaled flow waveform corresponding to each first sampling point; and form a corresponding first vector sequence from all the obtained first sampling point vectors.

[0182] The model prediction module 207 is used to predict the blood flow reserve fraction according to the first vector sequence by the first prediction model to obtain the corresponding first prediction vector; the first prediction vector includes multiple first sampling point FFR prediction values; the first sampling point FFR prediction value corresponds to the first sampling point one by one; the first sampling point FFR prediction value is the pressure ratio between the corresponding first sampling point and the center point of the coronary entrance, and the value is between 0 and 1.

[0183] The prediction result processing module 208 is used to mark the blood flow reserve score of the first vascular tree and the first angiography image based on the first prediction vector and the first sampling point set, and highlight the marked points whose blood flow reserve score is lower than the preset blood flow reserve score threshold to obtain the corresponding first marked vascular tree and first marked angiography image; and display the first marked vascular tree and the first marked angiography image.

[0184] An embodiment of the present invention provides a device for predicting blood flow reserve fraction based on coronary magnetic resonance imaging, which can execute the method steps in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here.

[0185] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the model preparation module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above-mentioned module is determined. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0186] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0187] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above method embodiments are generated. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) methods. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, a hard disk, a tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0188] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. The electronic device may be a terminal device or a server for implementing the method of the aforementioned embodiment, or may be a terminal device or a server for implementing the method of the aforementioned embodiment connected to the aforementioned terminal device or server. Figure 3 As shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver 303. Various instructions may be stored in the memory 302 to complete various processing functions and implement the processing steps described in the aforementioned embodiment method. Preferably, the electronic device involved in the embodiment of the present invention also includes: a power supply 304, a system bus 305 and a communication port 306. The system bus 305 is used to realize the communication connection between components. The above-mentioned communication port 306 is used for connecting and communicating between the electronic device and other peripherals.

[0189] exist Figure 3The system bus 305 mentioned in the figure can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include a random access memory (RAM), and may also include a non-volatile memory (Non-Volatil e Memory), such as at least one disk storage.

[0190] The above-mentioned processor can be a general-purpose processor, including a central processing unit CPU, a network processor (Network Processor, NP), a graphics processing unit (Graphics Processing Unit, GPU), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0191] It should be noted that an embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the method and processing process provided in the above embodiments.

[0192] An embodiment of the present invention further provides a chip for executing instructions, wherein the chip is used to execute the processing steps described in the aforementioned method embodiment.

[0193] The embodiment of the present invention provides a method, device, electronic device and computer-readable storage medium for predicting blood flow reserve fraction based on coronary magnetic resonance imaging. As can be seen from the above content, the embodiment of the present invention pre-constructs a first prediction model for predicting blood flow reserve fraction, and constructs a first data set to train the prediction model; and after the model training is completed, the model input data is prepared according to a group of rest imaging sequences, hyperemia imaging sequences and angiography images of any subject, and the first prediction model predicts the blood flow reserve fraction according to the model input data, and the blood flow reserve fraction is marked on the vascular tree and angiography image based on the prediction results, and the marked vascular tree and angiography image are displayed. The non-invasive blood flow reserve fraction prediction mechanism provided by the embodiment of the present invention avoids damage to the subject's blood vessels, improves the subject's comfort, and establishes a sustainable monitoring method with a short time interval.

[0194] The professionals should further realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0195] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0196] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging, characterized in that: The method comprises: Constructing a blood flow reserve fraction prediction model as a corresponding first prediction model; constructing a first data set for model training; and performing model training on the first prediction model based on the first data set; After the model training is completed, the first rest imaging sequence, the first hyperemia imaging sequence and the first angiography of any subject are received; the first rest imaging sequence and the first hyperemia imaging sequence are respectively the coronary magnetic resonance phase contrast movie imaging sequences of the subject in the resting and hyperemia states, and the first angiography is the three-dimensional coronary magnetic resonance angiography of the subject; the first rest imaging sequence is composed of a plurality of the first rest imagings in chronological order; the first hyperemia imaging sequence is composed of a plurality of the first hyperemia imagings in chronological order; the time interval between every two adjacent first hyperemia imagings is consistent with the time interval between every two adjacent first rest imagings; the first rest imaging includes a plurality of first pixel points, each of which corresponds to a first pixel point vector; the first hyperemia imaging includes a plurality of second pixel points, each of which corresponds to a second pixel point vector; the first and second pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity directions; A three-dimensional coronary vessel tree is constructed based on the first angiography image to obtain a corresponding first vessel tree; and the inner wall edge of the coronary vessel and the center line of the coronary vessel are identified for each of the first rest images, each of the first hyperemic images and the first vessel tree; and the first vessel tree is sampled based on a preset vessel center point sampling rule to obtain a corresponding first sampling point set; the first sampling point set includes a plurality of first sampling points; each of the first sampling points corresponds to a first sampling point vector; the first sampling point vector includes a first sampling point coordinate and a first sampling point inner diameter; the first sampling point coordinate is the relative displacement coordinate between the first sampling point and the center point of the coronary vessel entrance; the first sampling point inner diameter is the vessel inner diameter at the first sampling point; According to the first sampling point set, the first vascular tree, the first rest imaging sequence and the first hyperemia imaging sequence, a rest / hyperemia flow waveform fitting process of the sampling points is performed to obtain a corresponding first rest flow waveform set and a first hyperemia flow waveform set; the first rest flow waveform set includes a plurality of first rest flow waveforms; the first hyperemia flow waveform set includes a plurality of first hyperemia flow waveforms; the first rest flow waveform and the first hyperemia flow waveform correspond to the first sampling point one by one; performing coronary blood flow reserve fraction estimation processing according to the first hyperemic flow waveform and the first resting flow waveform corresponding to each of the first sampling points to obtain the corresponding first sampling point CFR value; and performing flow amplitude scaling processing on the corresponding first resting flow waveform according to the CFR value of each of the first sampling points to obtain the corresponding first sampling point scaled flow waveform; The first sampling point coordinates, the first sampling point inner diameter and the first sampling point scaled flow waveform corresponding to each of the first sampling points form a corresponding first sampling point vector; and all the obtained first sampling point vectors form a corresponding first vector sequence; The first prediction model predicts the blood flow reserve fraction according to the first vector sequence to obtain a corresponding first prediction vector; the first prediction vector includes a plurality of first sampling point FFR prediction values; the first sampling point FFR prediction values ​​correspond to the first sampling points one by one; the first sampling point FFR prediction value is a pressure ratio between the corresponding first sampling point and the center point of the coronary entrance, and the value is between 0 and 1; Based on the first prediction vector and the first sampling point set, the first vascular tree and the first angiography are marked with blood flow reserve scores, and the marked points below a preset blood flow reserve score threshold are highlighted to obtain the corresponding first marked vascular tree and first marked angiography; and the first marked vascular tree and the first marked angiography are displayed.

2. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to claim 1, characterized in that: The first prediction model includes a data preprocessing module and an FFR prediction module; the FFR prediction module is implemented based on a class of deep learning models; the deep learning model includes at least an RNN model, a bi-RNN model, an LSTM model, a Bi-LSTM model and a Transformer model; The data preprocessing module is used to normalize all sampling point coordinates and all sampling point inner diameters in the vector sequence input by the model to obtain corresponding multiple sampling point normalized coordinates and multiple sampling point normalized inner diameters; Based on a preset sampling frequency, each scaled flow waveform in the vector sequence input by the model is sampled to obtain a corresponding flow time series sequence Y, wherein the flow time series sequence Y is composed of a plurality of flow sampling values ​​y i Sort by order, 1≤time index i≤total number of traffic sampling values; And each flow sampling value y i The corresponding normalized coordinates of the sampling point and the normalized inner diameter of the sampling point form a corresponding single-point moment feature; All the single-point time features corresponding to each sampling point vector in the vector sequence input by the model are sorted in chronological order to form a corresponding sampling point time series feature; When the FFR prediction module is implemented based on the RNN model, the bi-RNN model, the LSTM model or the Bi-LSTM model, all the obtained sampling point time series features are sorted in the order of the corresponding sampling point vectors to form a corresponding sampling point time series feature sequence, which is sent to the FFR prediction module; and when the FFR prediction module is implemented based on the Transformer model, all the obtained sampling point time series features are spliced ​​according to the feature channels to obtain a corresponding sampling point time series feature tensor, which is sent to the FFR prediction module; The FFR prediction module is used to predict the blood flow reserve fraction according to the sampling point time series feature sequence when it is implemented based on the RNN model, bi-RNN model, LSTM model or Bi-LSTM model, obtain the corresponding prediction vector and output it; it is also used to predict the blood flow reserve fraction according to the sampling point time series feature tensor when it is implemented based on the Transformer model, obtain the corresponding prediction vector and output it; the prediction vector includes multiple sampling point FFR prediction values; the sampling point FFR prediction values ​​correspond one-to-one to the sampling point vectors in the vector sequence input by the model.

3. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to claim 1, characterized in that: The first data set includes a plurality of first data records; the first data record includes a first sample vector sequence and a first label prediction vector; the first sample vector sequence includes a plurality of first sample sampling point vectors; the first sample sampling point vector includes a first sample sampling point coordinate, a first sample sampling point inner diameter and a first sample sampling point scaled flow waveform; the first label prediction vector includes a plurality of first sampling point label FFR prediction values; the first sampling point label FFR prediction value corresponds one-to-one to the first sample sampling point vector; the first sampling point label FFR prediction value is a pressure ratio between the corresponding first sample sampling point and the center point of the coronary artery entrance, and the value is between 0 and 1; The blood vessel center point sampling rule is as follows: starting from the center point of the coronary artery entrance, the current blood vessel center line position is recorded as the corresponding current sampling point at each preset first interval distance along the blood vessel center line of each trunk or branch blood vessel; the difference between the three-dimensional coordinates of the current sampling point and the center point of the coronary artery entrance is used as the corresponding current sampling point coordinate, and a tangent line of the blood vessel center line is made on the current sampling point, and a perpendicular line of the current tangent line is made through the current sampling point, and the straight-line distance between the two intersection points of the current perpendicular line and the edge of the blood vessel inner wall is used as the corresponding current sampling point inner diameter; and a corresponding current sampling point vector is formed by the obtained current sampling point coordinates and the current sampling point inner diameter; and after sampling of all the trunks or branches of the coronary artery is completed, a corresponding sampling point set is formed by all the obtained sampling points.

4. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to claim 3, characterized in that: The constructing of a first data set for model training specifically includes: Step 41, recruiting a plurality of volunteers without coronary artery stenosis symptoms and a plurality of volunteers with coronary artery stenosis symptoms of varying degrees to form a first volunteer set; Wherein, the first volunteer set includes a plurality of first volunteers; Step 42, taking the first of the first volunteers in the first volunteer set as the corresponding current volunteer; Step 43, performing magnetic resonance phase contrast movie imaging on the current volunteer in a resting state to obtain a corresponding first sample resting imaging sequence; performing magnetic resonance phase contrast movie imaging on the current volunteer in a hyperemic state to obtain a corresponding first sample hyperemic imaging sequence; and performing three-dimensional coronary magnetic resonance angiography on the current volunteer to obtain a corresponding first sample angiography image; Wherein, the first sample rest imaging sequence is formed by a plurality of the first sample rest imaging in chronological order; the first sample hyperemia imaging sequence is formed by a plurality of the first sample hyperemia imaging in chronological order; the time interval between each two adjacent first sample hyperemia imaging is consistent with the time interval between each two adjacent first sample rest imaging; the first sample rest imaging includes a plurality of first sample pixel points, each of which corresponds to a first sample pixel point vector; the first sample hyperemia imaging includes a plurality of second sample pixel points, each of which corresponds to a second sample pixel point vector; the first and second sample pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction; Step 44, constructing a three-dimensional coronary vessel tree based on the first sample angiogram to obtain a corresponding first sample vessel tree; and identifying the inner wall edge of the coronary vessel and the centerline of the coronary vessel for each of the first sample rest images, each of the first sample hyperemia images, and the first sample vessel tree; and sampling the first sample vessel tree based on the vessel center point sampling rule to obtain a corresponding first sample sampling point set; Wherein, the first sample sampling point set includes a plurality of first sample sampling points; each of the first sample sampling points corresponds to a first sample sampling point vector; the first sample sampling point vector includes the first sample sampling point coordinates and the first sample sampling point inner diameter; the first sample sampling point coordinates are the relative displacement coordinates between the first sample sampling point and the center point of the coronary vessel entrance; the first sample sampling point inner diameter is the inner diameter of the blood vessel at the first sample sampling point; Step 45, performing sampling point rest / congestion flow waveform fitting processing according to the first sample sampling point set, the first sample vascular tree, the first sample rest imaging sequence, and the first sample congestion imaging sequence to obtain a corresponding first sample rest flow waveform set and a first sample congestion flow waveform set; Wherein, the first sample resting flow waveform set includes a plurality of first sample resting flow waveforms; the first sample congestion flow waveform set includes a plurality of first sample congestion flow waveforms; the first sample resting flow waveform, the first sample congestion flow waveform and the first sample sampling point correspond one to one; Step 46, performing coronary blood flow reserve fraction estimation processing according to the first sample hyperemia flow waveform and the first sample resting flow waveform corresponding to each of the first sample sampling points to obtain the corresponding first sample sampling point CFR value; and performing flow amplitude scaling processing on the corresponding first sample resting flow waveform according to each of the first sample sampling point CFR values ​​to obtain the corresponding first sample sampling point scaled flow waveform; Step 47, forming a corresponding first sample sampling point vector from the first sample sampling point coordinates, the first sample sampling point inner diameter and the first sample sampling point scaled flow waveform corresponding to each first sample sampling point; and forming a corresponding first sample vector sequence from all the obtained first sample sampling point vectors; Step 48, using computational fluid dynamics simulation software to model the fluid dynamics simulation equations for solving the pressure distribution of each point on the three-dimensional coronary tree structure to obtain the corresponding first simulation equations; and setting the spatial boundary conditions of the first simulation equations according to the first sample vascular tree and the first sample sampling point set, and setting the temporal boundary conditions of the first simulation equations according to the first sample hyperemia flow waveforms corresponding to all the first sample sampling points; and after the setting of the spatial and temporal boundary conditions, performing simulation operation on the first simulation equations based on the computational fluid dynamics simulation software to obtain the pressure time series sequence of each of the first sample sampling points on the first sample vascular tree; and taking the pressure average of the pressure time series sequence of each of the first sample sampling points as the corresponding first sample sampling point pressure value; and taking the pressure ratio of the first sample sampling point pressure value of each of the first sample sampling points to the first sample sampling point pressure value corresponding to the center point of the coronary entrance as a corresponding first sampling point label FFR prediction value; and forming a corresponding first label prediction vector from all the obtained first sampling point label FFR prediction values; Step 49, the first sample vector sequence and the first label prediction vector corresponding to the current volunteer form a corresponding first data record; and identify whether the current volunteer is the last first volunteer of the first volunteer set. If not, the next first volunteer of the first volunteer set is taken as the new current volunteer and the process returns to step 43. If so, all the first data records obtained form the corresponding first data set.

5. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to claim 3, characterized in that: The performing model training on the first prediction model based on the first data set specifically includes: Step 51, setting an integer greater than zero as the corresponding first consecutive verification pass number; and setting a first counter initialized to 0; Step 52: randomly select a plurality of the first data records from the first data set according to a preset test set ratio to form a corresponding current test set, and form a corresponding current verification set from all the remaining first data records; Wherein, the current test set and the current validation set are both composed of a plurality of the first data records, and a ratio of the total number of the first data records of the current test set to the total number of the first data records of the first data set satisfies the test set ratio; Step 53, extracting the first of the first data records in the current test set as the corresponding current data record; Step 54, inputting the first sample vector sequence of the current data record into the first prediction model, and the first prediction model performs blood flow reserve fraction prediction according to the first sample vector sequence input this time to obtain a corresponding first training prediction vector; Wherein, the first training prediction vector includes a plurality of first sample sampling point FFR prediction values; the first sample sampling point FFR prediction values ​​correspond one-to-one to the first sample sampling point vectors of the first sample vector sequence; Step 55, bringing the first training prediction vector and the first label prediction vector of the current data record into a preset first model loss function to calculate and obtain a corresponding first loss value; Wherein, the first model loss function is implemented based on the L1 loss function or the L2 loss function; Step 56, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current data record is the last first data record of the current test set, if so, turning to step 57, if not, extracting the next first data record of the current test set as the new current data record and returning to step 54 to continue training; if the first loss value does not satisfy the first loss value range, performing a round of model parameter optimization on the first prediction model in the direction of minimizing the first model loss function based on a preset first model parameter optimizer, and returning to step 54 to continue training at the end of this round of parameter optimization; Step 57, traverse all the first data records of the current verification set; and during the traversal, take the first data record currently traversed as the corresponding current verification record; and record the first sample vector sequence and the first label prediction vector of the current verification record as the corresponding current sample vector sequence and the current label prediction vector; and use the first prediction model to predict the blood flow reserve fraction according to the current sample vector sequence to obtain the corresponding current training prediction vector; and form a corresponding verification-label vector pair with the current training prediction vector and the current label prediction vector obtained this time; and at the end of the traversal, bring all the obtained verification-label vector pairs into the preset first error function to calculate and obtain the corresponding first error value; Wherein, the first error function is implemented based on the MSE error function; Step 58, identifying whether the first error value does not exceed a preset first error threshold; if the first error value does not exceed the first error threshold, adding 1 to the first counter, and identifying whether the first counter after adding 1 is greater than the first continuous verification pass number, if so, go to step 59, otherwise return to step 52 to continue training; if the first error value exceeds the first error threshold, clear the first counter and return to step 52 to continue training; Step 59, solidify the model parameters of the first prediction model and confirm that the model training is completed.

6. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to any one of claims 1 and 4, characterized in that: The sampling point rest / congestion flow waveform fitting process specifically includes: Step 61, taking the sampling point set, vascular tree, rest imaging sequence and congestion imaging sequence input by the rest / congestion flow waveform fitting process of the current sampling point as the corresponding second sampling point set, second vascular tree, second rest imaging sequence and second congestion imaging sequence; Wherein, the second sampling point set includes a plurality of second sampling points; each of the second sampling points corresponds to a second sampling point vector; the second sampling point vector includes a second sampling point coordinate and a second sampling point inner diameter; the second sampling point coordinate is the relative displacement coordinate between the second sampling point and the center point of the coronary vessel entrance; the second sampling point inner diameter is the vessel inner diameter at the second sampling point; the second rest imaging sequence is composed of a plurality of second rest imaging in chronological order; the second hyperemia imaging sequence is composed of a plurality of second hyperemia imaging in chronological order; the time interval between each two adjacent second hyperemia imaging is consistent with the time interval between each two adjacent second rest imaging; the second rest imaging includes a plurality of third pixel points, each of which corresponds to a third pixel point vector; the second hyperemia imaging includes a plurality of fourth pixel points, each of which corresponds to a fourth pixel point vector; the third and fourth pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity direction; Step 62, taking the first second sampling point of the second sampling point set as the corresponding current sampling point; Step 63, calculating the straight-line distance from the current sampling point to the center point of the coronary artery entrance of the second vascular tree to obtain a corresponding first distance, and calculating the blood vessel centerline length from the current sampling point to the center point of the coronary artery entrance of the second vascular tree to obtain a corresponding first length; and estimating the cross-sectional area of ​​the blood vessel inner wall at the current sampling point based on the inner diameter of the second sampling point of the current sampling point to obtain a corresponding first cross-sectional area; Step 64, taking each of the third pixel points on the second resting image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as the corresponding resting matching pixel point; and sorting the third pixel point vectors of all the resting matching pixel points in chronological order to form a corresponding first sequence; Step 65, taking each of the fourth pixels on the second hyperemia image whose straight-line distance to the center point of the coronary artery entrance matches the first distance and whose blood vessel centerline length to the center point of the coronary artery entrance matches the first length as the corresponding hyperemia matching pixel; and sorting the fourth pixel vectors of all the obtained hyperemia matching pixel points in chronological order to form a corresponding second sequence; Step 66, performing curve fitting on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity coding and the positive / negative phase velocity direction of each of the third pixel point vectors in the first sequence to obtain a corresponding first resting velocity waveform; and performing curve fitting on the blood flow velocity-time waveform at the current sampling point according to the positive / negative phase velocity coding and the positive / negative phase velocity direction of each of the fourth pixel point vectors in the second sequence to obtain a corresponding first hyperemia velocity waveform; Step 67, according to the calculation method that the blood flow rate is approximately equal to the product of the blood vessel cross-sectional area and the blood flow velocity, the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first resting velocity waveform is used as the second resting flow waveform corresponding to the current sampling point, and the waveform obtained by multiplying the first cross-sectional area by the waveform amplitude of the first hyperemia velocity waveform is used as the second hyperemia flow waveform corresponding to the current sampling point; Step 68, identifying whether the current sampling point is the last second sampling point of the second sampling point set; if so, going to step 69; if not, taking the next second sampling point of the second sampling point set as the new current sampling point and returning to step 63; Step 69, all the second resting flow waveforms obtained form a corresponding resting flow waveform set for the current time; and all the second congestion flow waveforms obtained form a corresponding congestion flow waveform set for the current time; and the resting flow waveform set for the current time and the congestion flow waveform set for the current time obtained are output as the processing results of the resting / congestion flow waveform fitting processing of the current sampling point.

7. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to any one of claims 1 and 4, characterized in that: The coronary blood flow reserve fraction estimation process specifically includes: The hyperemic flow waveform and resting flow waveform inputted in this coronary blood flow reserve fraction estimation process are used as the corresponding current hyperemic flow waveform and current resting flow waveform; and the waveform amplitudes of the current hyperemic flow waveform and the current resting flow waveform are respectively averaged to obtain the corresponding hyperemic average flow and resting average flow; and the ratio of the hyperemic average flow to the resting average flow is used as the corresponding current sampling point CFR value; and the current sampling point CFR value obtained this time is outputted as the processing result of this coronary blood flow reserve fraction estimation process.

8. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to any one of claims 1 and 4, characterized in that: The flow amplitude scaling process specifically includes: The sampling point CFR value and resting flow waveform input for this flow amplitude scaling process are used as the corresponding current sampling point CFR value and current resting flow waveform; the waveform obtained by multiplying the current sampling point CFR value with the waveform amplitude of the current resting flow waveform is used as the corresponding current sampling point scaled flow waveform; and the current sampling point scaled flow waveform obtained this time is output as the processing result of this flow amplitude scaling process.

9. The method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to claim 1, characterized in that: The step of marking the first vascular tree and the first angiogram by blood flow reserve fraction based on the first prediction vector and the first sampling point set and highlighting the marked points below a preset blood flow reserve fraction threshold to obtain the corresponding first marked vascular tree and first marked angiogram specifically includes: Drawing a blood vessel section on the first blood vessel tree based on the first sampling point coordinates and the first sampling point inner diameter of each of the first sampling points of the first sampling point set to obtain a corresponding first sampling point blood vessel section; and taking each of the first sampling point blood vessel sections as a corresponding current section; and taking the first sampling point FFR prediction value corresponding to the current section in the first prediction vector as a corresponding current section FFR value; and identifying whether the current section FFR value is lower than the blood flow reserve fraction threshold, and if so, at the current section, based on the current section FFR value corresponding to a preset first color mark, and if not, at the current section, based on the current section FFR value corresponding to a preset second color mark; the brightness of the first color is greater than the second color; and taking the angiogram voxel points on the first angiogram corresponding to each of the first sampling points of the first sampling point set as the corresponding first sampling voxel points; and taking each of the first sampling voxel points as the corresponding current voxel point; and taking the first sampling point FFR prediction value corresponding to the current voxel point in the first prediction vector as the corresponding current voxel point FFR value; and identifying whether the current voxel point FFR value is lower than the blood flow reserve fraction threshold, and if so, at the current voxel point based on the current voxel point FFR value corresponding to the first color mark, and if not, at the current voxel point based on the current voxel point FFR value corresponding to the second color mark; The first vascular tree and the first angiography image that have been marked are used as the corresponding first marked vascular tree and the first marked angiography image.

10. A device for executing the method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to any one of claims 1 to 9, characterized in that: The device comprises: a model preparation module, a data receiving module, a first data preprocessing module, a second data preprocessing module, a third data preprocessing module, a fourth data preprocessing module, a model prediction module and a prediction result processing module; The model preparation module is used to construct a blood flow reserve fraction prediction model recorded as a corresponding first prediction model; and to construct a first data set for model training; and to perform model training on the first prediction model based on the first data set; The data receiving module is used to receive the first rest imaging sequence, the first hyperemia imaging sequence and the first angiography of any subject after the model training is completed; the first rest imaging sequence and the first hyperemia imaging sequence are respectively the coronary magnetic resonance phase contrast movie imaging sequences of the subject in the rest and hyperemia states, and the first angiography is the three-dimensional coronary magnetic resonance angiography of the subject; the first rest imaging sequence is composed of a plurality of the first rest imagings in chronological order; the first hyperemia imaging sequence is composed of a plurality of the first hyperemia imagings in chronological order; the time interval between every two adjacent first hyperemia imagings is consistent with the time interval between every two adjacent first rest imagings; the first rest imaging includes a plurality of first pixel points, each of which corresponds to a first pixel point vector; the first hyperemia imaging includes a plurality of second pixel points, each of which corresponds to a second pixel point vector; the first and second pixel point vectors are both composed of positive / negative phase velocity encoding and positive / negative phase velocity directions; The first data preprocessing module is used to construct a three-dimensional coronary vessel tree based on the first angiogram to obtain a corresponding first vessel tree; and to identify the inner wall edge of the coronary vessel and the center line of the coronary vessel for each of the first rest images, each of the first hyperemic images and the first vessel tree; and to sample the first vessel tree based on a preset vessel center point sampling rule to obtain a corresponding first sampling point set; the first sampling point set includes a plurality of first sampling points; each of the first sampling points corresponds to a first sampling point vector; the first sampling point vector includes a first sampling point coordinate and a first sampling point inner diameter; the first sampling point coordinate is the relative displacement coordinate between the first sampling point and the center point of the coronary vessel entrance; the first sampling point inner diameter is the vessel inner diameter at the first sampling point; The second data preprocessing module is used for performing sampling point rest / congestion flow waveform fitting processing according to the first sampling point set, the first vascular tree, the first rest imaging sequence and the first congestion imaging sequence to obtain the corresponding first rest flow waveform set and first congestion flow waveform set; the first rest flow waveform set includes a plurality of first rest flow waveforms; the first congestion flow waveform set includes a plurality of first congestion flow waveforms; the first rest flow waveform and the first congestion flow waveform correspond to the first sampling point one by one; The third data preprocessing module is used to perform coronary blood flow reserve fraction estimation processing according to the first hyperemic flow waveform and the first resting flow waveform corresponding to each of the first sampling points to obtain the corresponding first sampling point CFR value; and perform flow amplitude scaling processing on the corresponding first resting flow waveform according to each of the first sampling point CFR values ​​to obtain the corresponding first sampling point scaled flow waveform; The fourth data preprocessing module is used to form a corresponding first sampling point vector from the first sampling point coordinates, the first sampling point inner diameter and the first sampling point scaled flow waveform corresponding to each first sampling point; and form a corresponding first vector sequence from all the obtained first sampling point vectors; The model prediction module is used to predict the blood flow reserve fraction according to the first vector sequence by the first prediction model to obtain a corresponding first prediction vector; the first prediction vector includes a plurality of first sampling point FFR prediction values; the first sampling point FFR prediction values ​​correspond to the first sampling points one by one; the first sampling point FFR prediction value is the pressure ratio between the corresponding first sampling point and the center point of the coronary entrance, and the value is between 0 and 1; The prediction result processing module is used to mark the blood flow reserve score of the first vascular tree and the first angiography image based on the first prediction vector and the first sampling point set, and highlight the marked points that are lower than a preset blood flow reserve score threshold to obtain the corresponding first marked vascular tree and first marked angiography image; and display the first marked vascular tree and the first marked angiography image.

11. An electronic device, characterized in that: include: memory, processors, and transceivers; The processor is used to couple with the memory, read and execute instructions in the memory, so as to implement the method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging according to any one of claims 1 to 9; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer executes the method for predicting blood flow reserve fraction based on coronary magnetic resonance imaging as described in any one of claims 1 to 9.

Images