A system and method for determining fractional flow reserve

Abstract

Embodiments of the present specification provide a system and method for determining a blood flow reserve fraction. The method comprises: obtaining a medical image of a subject, the medical image comprising a coronary artery of the subject; determining, based on the medical image, an equivalent resistance of the coronary artery using a first machine learning model, wherein the equivalent resistance of the coronary artery comprises resistance values of a plurality of points on the coronary artery; determining, based on the medical image, a boundary condition resistance of the coronary artery using a second machine learning model, wherein the boundary condition resistance is used to describe a boundary condition of blood flowing out of a distal end of the coronary artery; and determining a blood flow reserve fraction of the subject based on the equivalent resistance and the boundary condition resistance of the coronary artery.

Description

Technical Field

[0001] This specification relates to the field of medical technology, and in particular to a system and method for determining fractional flow reserve. Background Technology

[0002] Coronary atherosclerotic heart disease (CAD) is a major chronic disease that seriously threatens the health and quality of life of Chinese residents. Fractional flow reserve (FFR) can effectively assess the degree of ischemia caused by plaque stenosis, providing a reference for subsequent treatment. Compared to invasive FFR, FFR obtained based on medical imaging techniques (e.g., coronary CT scans) (FTCT-based FFR is called FFRCT) is more patient-friendly due to its non-invasive examination method. However, this method typically estimates pressure and flow velocity at various points in the vessel using computational fluid dynamics (CFD). This calculation is extremely time-consuming (greater than 40 minutes per case) and highly sensitive to coronary artery segmentation results, making it prone to failure and difficult to be widely accepted in clinical practice.

[0003] Therefore, it is desirable to provide an efficient and accurate system and method for determining fractional flow reserve. Summary of the Invention

[0004] One embodiment of this specification provides a method for determining fractional flow reserve. The method includes: acquiring a medical image of a subject, the medical image including the coronary arteries of the subject; determining, based on the medical image, the equivalent resistance of the coronary arteries using a first machine learning model, wherein the equivalent resistance of the coronary arteries includes equivalent resistance values ​​at multiple points on the coronary arteries; determining, based on the medical image, the boundary condition resistance of the coronary arteries using a second machine learning model, wherein the boundary condition resistance describes the boundary conditions for blood outflow from the distal end of the coronary arteries; and determining the fractional flow reserve of the subject based on the equivalent resistance and the boundary condition resistance of the coronary arteries.

[0005] One embodiment of this specification provides a system for determining fractional flow reserve. The system includes an acquisition module configured to acquire a medical image of a subject, the medical image including the coronary arteries of the subject; a first determination module configured to determine the equivalent resistance of the coronary arteries based on the medical image using a first machine learning model, wherein the equivalent resistance of the coronary arteries includes resistance values ​​at multiple points on the coronary arteries; a second determination module configured to determine the boundary condition resistance of the coronary arteries based on the medical image using a second machine learning model, wherein the boundary condition resistance describes the boundary conditions for blood outflow from the ends of the coronary arteries; and a third determination module configured to determine the fractional flow reserve of the subject based on the equivalent resistance and the boundary condition resistance of the coronary arteries.

[0006] One embodiment of this specification provides a system for determining the fractional flow reserve. The system includes at least one storage device for storing computer instructions; and at least one processor for executing the computer instructions to implement the method for determining the fractional flow reserve described above.

[0007] Some of the additional features of this application will be described in the following description. These additional features will be apparent to those skilled in the art from the following description and accompanying drawings, or from an understanding of the production or operation of the embodiments. The features of this application can be implemented and obtained by practicing or using various aspects of the methods, means, and combinations set forth in the detailed examples below. Attached Figure Description

[0008] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0009] Figure 1 This is a schematic diagram illustrating an exemplary FFR determination system based on some embodiments shown in this specification;

[0010] Figure 2 This is a schematic diagram of an exemplary FFR determination system shown in some embodiments of this specification;

[0011] Figure 3 This is a schematic flowchart illustrating an exemplary process for determining FFR based on some embodiments of this specification;

[0012] Figure 4 This is an exemplary schematic diagram illustrating the determination of the equivalent resistance of a coronary artery using a first machine learning model according to some embodiments of this specification;

[0013] Figure 5This is a schematic diagram of the training process of exemplary first and second machine learning models according to some embodiments of this specification;

[0014] Figure 6 This is a schematic flowchart illustrating an exemplary process for determining FFR based on some embodiments of this specification;

[0015] Figure 7 This is a schematic diagram of the exemplary circuit topology corresponding to a coronary artery, as shown in some embodiments of this specification;

[0016] Figure 8 This is a schematic diagram of the exemplary circuit topology corresponding to a coronary artery, as shown in some embodiments of this specification. Detailed Implementation

[0017] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0018] It should be understood that the terms “system,” “device,” “unit,” and / or “module” used herein are one way to distinguish different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0019] As indicated in this specification and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of expressly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0020] Flowcharts are used in this specification to illustrate the operations performed by the system according to embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0021] Currently, to improve the computational efficiency of FFR (Flow Reserve), machine learning models are being used to determine FFR. Typically, the method for calculating FFR using machine learning models involves extracting vascular features as input to the model and directly predicting FFR. During the machine learning model training process, the fractional flow reserve is obtained through CFD simulation and used as the model training label (gold standard). However, upstream and downstream information of the coronary arteries, including vascular branching information, is crucial for FFR prediction. In current methods, the machine learning model only learns the influence of vascular features on FFR without incorporating the topology of the coronary artery tree, resulting in low accuracy of the obtained FFR. Therefore, this specification provides a method for calculating FFR that combines the topology of the coronary artery tree with a machine learning model, which can improve the accuracy and efficiency of FFR calculation.

[0022] Figure 1 These are schematic diagrams illustrating exemplary FFR determination systems based on some embodiments shown in this specification, depicting application scenarios. For example... Figure 1 As shown, the FFR determination system 100 may include a medical device 110, a processing device 120, a terminal device 130, a storage device 140, and a network 150. In some embodiments, the processing device 120 may be part of the medical device 110. The connections between the components in the FFR determination system 100 may be variable. Figure 1 As shown, medical device 110 can be connected to processing device 120 via network 150. Alternatively, medical device 110 can be directly connected to processing device 120. As another example, storage device 140 can be connected to processing device 120 directly or via network 150. As yet another example, terminal device 130 can be directly connected to processing device 120 (as shown by the dashed arrow connecting terminal device 130 and processing device 120) or connected to processing device 120 via network 150.

[0023] Medical device 110 may be a non-invasive scanning imaging device for disease diagnosis or research purposes. In some embodiments, medical device 110 may scan an object within a detection area or scanning area to obtain scan data of that object. In some embodiments, medical device 110 may include a single-modal scanner and / or a multimodal scanner. A single-modal scanner may include, for example, an ultrasound scanner, an X-ray scanner, a computed tomography (CT) scanner, a magnetic resonance imaging (MRI) scanner, an optical coherence tomography (OCT) scanner, an ultrasound (US) scanner, an intravascular ultrasound (IVUS) scanner, or any combination thereof. A multimodal scanner may include, for example, an X-ray imaging-magnetic resonance imaging (X-MRI) scanner, a positron emission tomography-X-ray imaging (PET-X-ray) scanner, a single-photon emission computed tomography-magnetic resonance imaging (SPECT-MRI) scanner, a positron emission tomography-computed tomography (PET-CT) scanner, a digital subtraction angiography-magnetic resonance imaging (DSA-MRI) scanner, etc. In some embodiments, medical device 110 is a CT scanner. In some embodiments, the processing device 120 may be integrated into the medical device 110, or the medical device 110 and the processing device 120 may function through the same entity. The medical devices described above are for illustrative purposes only and are not intended to limit the scope of this specification.

[0024] Processing device 120 can process data and / or information obtained from medical device 110, terminal device 130, storage device 140, or other components of FFR determination system 100. For example, processing device 120 can acquire a medical image of an object. The medical image may include the object's coronary arteries. Based on the medical image, processing device 120 can determine the equivalent resistance and boundary condition resistance of the coronary arteries using a first machine learning model and a second machine learning model, respectively. The equivalent resistance of the coronary arteries includes resistance values ​​at multiple points on the coronary arteries, and the boundary condition resistance describes the boundary conditions at which blood flows out of the ends of the coronary arteries. Further, processing device 120 can determine the object's fractional flow reserve based on the equivalent resistance and boundary condition resistance of the coronary arteries.

[0025] In some embodiments, the processing device 120 may be local or remote. For example, the processing device 120 may access information and / or data from the medical device 110, the terminal device 130, and / or the storage device 140 via the network 150.

[0026] Terminal device 130 may include mobile device 131, tablet computer 132, laptop computer 133, etc., or any combination thereof. In some embodiments, terminal device 130 may be part of processing device 120.

[0027] Storage device 140 may store data, instructions, and / or any other information. In some embodiments, storage device 140 may store data obtained from medical device 110, processing device 120, and / or terminal device 130, such as medical images generated by medical device 110.

[0028] Network 150 may include any suitable network capable of facilitating information and / or data exchange. In some embodiments, at least one component of the FFR determination system 100 (e.g., medical device 110, processing device 120, terminal device 130, storage device 140) may exchange information and / or data with at least one other component of the FFR determination system 100 via network 150. For example, processing device 120 may acquire medical images of an object from medical device 110 via network 150. Similarly, terminal device 130 may acquire the FFR of an object from processing device 120 via network 150.

[0029] It should be noted that the FFR determination system 100 is provided for illustrative purposes only and is not intended to limit the scope of this specification. Various modifications or variations can be made by those skilled in the art based on the description herein. For example, the FFR determination system 100 may also include input and / or output devices. As another example, the FFR determination system 100 may perform similar or different functions on other devices. However, these changes and modifications will not depart from the scope of this specification.

[0030] Figure 2 This is a schematic diagram of an exemplary FFR determination system according to some embodiments of this specification.

[0031] like Figure 2 As shown, in some embodiments, the FFR determination system 200 may include an acquisition module 210, a first determination module 220, a second determination module 230, and a third determination module 240. In some embodiments, the FFR determination system 200 may further include a model training module 250. In some embodiments, the functions corresponding to the FFR determination system 200 may be executed by the processing device 120; for example, the acquisition module 210, the first determination module 220, the second determination module 230, the third determination module 240, and the model training module 250 may be modules within the processing device 120.

[0032] The acquisition module 210 can be configured to acquire medical images of an object. The medical images may include the object's coronary arteries. Further description of acquiring medical images can be found elsewhere in this book (e.g., Figure 3 (310 in the middle).

[0033] The first determining module 220 can be configured to determine the equivalent resistance of the coronary arteries based on medical images and using a first machine learning model, wherein the equivalent resistance of the coronary arteries includes resistance values ​​at multiple points on the coronary arteries. Further description of determining the equivalent resistance of the coronary arteries using the first machine learning model can be found elsewhere in this book (e.g., Figure 3 The 320 in the text will not be elaborated upon here.

[0034] The second determination module 230 can be configured to determine the boundary condition resistance of the coronary arteries based on medical images and using a second machine learning model, wherein the boundary condition resistance is used to describe the boundary conditions at the ends of blood flowing out of the coronary arteries. Further description of using a second machine learning model to determine the boundary condition resistance of the coronary arteries can be found elsewhere in this book (e.g., Figure 3 The details of 330 in the text will not be elaborated upon here.

[0035] The third determination module 240 can be configured to determine the fractional flow reserve of an object based on the equivalent resistance and boundary condition resistance of the coronary arteries. Further description of determining the fractional flow reserve of an object can be found elsewhere in this book (e.g., Figure 3 The 340 in the text will not be elaborated upon here.

[0036] The model training module 250 can be configured to train machine learning models (e.g., a first machine learning model and a second machine learning model). Further descriptions of training the first and second machine learning models can be found elsewhere in this book (e.g., Figure 5 (The rest is omitted here.)

[0037] It should be understood that Figure 2 The system and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of both.

[0038] It should be noted that the above description of the system and its modules is for illustrative purposes only and should not be construed as limiting this specification to the scope of the illustrated embodiments. It is understood that those skilled in the art, after understanding the principles of this system, may arbitrarily combine the various modules or construct subsystems connected to other modules without departing from these principles. For example, in some embodiments, Figure 2The modules disclosed above can be different modules within a single system, or a single module can implement the functions of two or more of the modules described above. For example, the modules can share a single storage module, or each module can have its own separate storage module. In some embodiments, the model training module 250 and other modules can be implemented by different systems. For example, the model training module 250 can be implemented by the computing device of a machine learning model supplier, while other modules can be implemented by the computing device of a machine learning model user. Such variations are all within the scope of protection of this specification.

[0039] Figure 3 This is a schematic flowchart illustrating an exemplary determination of FFR according to some embodiments of this specification. In some embodiments, one or more steps of process 300 may be performed... Figure 1 The FFR determination system 100 shown is implemented or is implemented by... Figure 2 The FFR determination system 200 is shown to be executed. For example, process 300 can be executed by a module of processing device 120. Figure 3 As shown, process 300 may include the following steps.

[0040] Step 310: Acquire a medical image of the object, the medical image including the object's coronary arteries. In some embodiments, step 310 may be performed by processing device 120 or acquisition module 210.

[0041] The object can include the human body, an animal, or a part thereof. The following explanation uses the human body as an example.

[0042] Medical images can be images obtained by scanning the coronary arteries in the heart region of a human body, such as CT angiography images, CT plain scan images, etc. In some embodiments, the coronary arteries of an object can be scanned by medical device 110 to obtain medical images. In some embodiments, medical images can be generated in advance and stored in a storage device (e.g., storage device 140 or external storage device), and processing device 120 can retrieve medical images from the storage device.

[0043] Step 320: Based on the medical image, using a first machine learning model, determine the equivalent resistance of the coronary artery, wherein the equivalent resistance of the coronary artery includes resistance values ​​at multiple points on the coronary artery. In some embodiments, step 320 may be performed by processing device 120 or first determining module 220.

[0044] In some embodiments, the processing device 120 can determine the coronary artery segmentation image, lesion detection result, and lumen segmentation image of an object based on medical images. Specifically, the processing device 120 can determine the coronary artery segmentation image of the object based on medical images. Further, the processing device 120 can determine the lesion detection result and lumen segmentation image based on the coronary artery segmentation image. For each of the plurality of segmented image blocks, the processing device 120 can determine the feature information of the segmented image block based on the coronary artery segmentation image, the lesion detection result, and the lumen segmentation image. Based on the feature information of each segmented image block, the processing device 120 can use the first machine learning model to determine the equivalent resistance of the coronary artery.

[0045] A coronary artery segmentation image refers to an image generated by segmenting the coronary arteries in a medical image, which can indicate the coronary arteries of an object. In some embodiments, different coronary artery branches can be displayed in different ways in a coronary artery segmentation image. A coronary artery branch can include a segment between two adjacent bifurcation points, a segment between the origin of the coronary artery and its adjacent bifurcation point, and a segment between the terminal end of each coronary artery and its adjacent bifurcation point. For example, different colors can be used to display different coronary artery branches in a coronary artery segmentation image. Alternatively, different branch labels (e.g., labels "1", "2", etc.) can be used to display different coronary artery branches in a coronary artery segmentation image. In some embodiments, a coronary artery segmentation image can be obtained manually by a user (e.g., a radiologist) from a medical image by segmenting the coronary arteries. In some embodiments, a coronary artery segmentation image can be obtained automatically by a processing device 120 from a medical image by segmenting the coronary arteries. For example, the processing device 120 can use an image segmentation algorithm or a machine learning model to segment the coronary arteries from a medical image to obtain a coronary artery segmentation image.

[0046] Lesion detection results can indicate lesions within the coronary arteries. Lesion detection results can be achieved using any lesion detection algorithm (e.g., a lesion detection model). A lumen segmentation image refers to an image generated by segmenting the lumens of the coronary arteries in a medical image; it can indicate one or more lumens of the coronary arteries. Lumen segmentation can be achieved using any lumen segmentation algorithm (e.g., a lumen segmentation model). In some embodiments, lesion detection results and lumen segmentation images can be generated based on coronary artery segmentation images and medical images. For example, the corresponding region of the coronary artery can be determined in a medical image based on the coronary artery segmentation image, and lesion detection and lumen segmentation can be performed on that region to obtain lesion detection results and lumen segmentation images.

[0047] In some embodiments, a lumen of a coronary artery refers to the segment of the coronary artery through which blood flows from the inlet to the outlet. This is merely an example, such as... Figure 7As shown, coronary artery segments L1+L2+L3 form one lumen, coronary artery segments L1+L2+L4 form one lumen, and coronary artery segments L1+L5 form one lumen.

[0048] Further, the processing device 120 can determine multiple segmented image blocks in the coronary artery segmentation image. For example, the processing device 120 can extract the centerline of the coronary artery and label the centerline corresponding to different coronary artery branches according to the aforementioned coronary artery branch labels. The processing device 120 can sample and extract multiple segmented image blocks along the centerline of the coronary artery. For example, the processing device 120 can sequentially divide the coronary artery into multiple coronary artery segments along the vessel centerline at a preset size (e.g., 1 mm). For each coronary artery segment, the processing device 120 can determine an image region containing that coronary artery segment as a segmented image block corresponding to that coronary artery segment. In some embodiments, two adjacent segmented image blocks can partially overlap. For each segmented image block, the processing device 120 can determine the feature information of the segmented image block based on the coronary artery segmentation image, lesion detection results, and lumen segmentation image. In some embodiments, the feature information of the segmented image block may include the normalized world coordinates of the segmented image block, branch weight features, lesion features, cross-sectional features (including cross-sectional area, minor axis, and major axis) along the tangent direction perpendicular to the vessel centerline, and other features. The normalized world coordinates of a segmented image patch can represent its location within the entire coronary artery. The branch weight feature of the image patch indicates the ratio of the volume of the branch coronary artery containing the segmented image patch to the total volume of the coronary artery. Exemplary lesion features may include whether the image patch contains a lesion (i.e., whether it has stenosis), the number of upstream (downstream) lesions, the stenosis rate of the nearest upstream (downstream) lesion, the extent of the nearest upstream (downstream) lesion, the distance to the nearest upstream (downstream) lesion, the stenosis rate of the most severe upstream lesion, the integral of the stenosis rate of the most severe upstream lesion, the extent of the most severe upstream lesion, and the distance to the most severe upstream lesion. Exemplary cross-sectional features may include the average cross-sectional area upstream (downstream), the minimum cross-sectional area upstream (downstream), the minimum minor axis of the cross-section upstream (downstream), the average minor axis of the cross-section upstream (downstream), the maximum minor axis of the cross-section upstream (downstream), the distance to the minimum minor axis of the cross-section upstream (downstream), the minor axis and area of ​​the segmented image patch, the median cross-sectional area upstream (downstream), the median minor axis of the cross-section upstream (downstream), the ratio of the cross-sectional area of ​​the current segmented image patch to the cross-sectional area of ​​the upstream (downstream), and the ratio of the minor axis of the cross-section of the current segmented image patch to the minor axis of the cross-section upstream (downstream). Other features may include the distance from the segmented image patch to the nearest upstream bifurcation point, the length of the coronary artery segment corresponding to the segmented image patch along the blood flow direction, the label of the centerline of the coronary artery segment corresponding to the segmented image patch, the number of upstream and downstream branches of the segmented image patch, and the label of the combined centerline of the lumen in which the segmented image patch is located. A combined centerline of a lumen refers to the line connecting the centerlines of the coronary artery segments contained in the lumen.

[0049] Furthermore, the processing device 120 can determine the equivalent resistance of the coronary artery using a first machine learning model based on the feature information of each segmented image block. Specifically, the processing device 120 can input the feature information and the segmented image block of each segmented image block into the first machine learning model, and the first machine learning model can output the equivalent resistance corresponding to each segmented image block. In some embodiments, the processing device 120 can determine the original image block corresponding to each segmented image block in the medical image based on the correspondence between the coronary artery segmented image and the medical image. The original image block corresponding to each segmented image block can also be used as one of the inputs to the first machine learning model, and is input into the first machine learning model along with other inputs. In some embodiments, the processing device 120 can also acquire the physiological parameter features of the object. Exemplary physiological parameter features may include diastolic blood pressure, systolic blood pressure, cardiac output, age, gender, etc. The physiological parameter features of the object can also be used as one of the inputs to the first machine learning model, and is input into the first machine learning model along with other inputs.

[0050] In some embodiments, the first machine learning model may refer to a model used to determine the equivalent resistance of a point on a coronary artery. In this specification, a segmented image patch of the aforementioned blood vessel can be considered to correspond to a point on the blood vessel. In some embodiments, the first machine learning model may include Convolutional Neural Networks (CNNs) and Transformer models. CNNs can be configured to extract appearance features from image patches (e.g., original image patches, segmented image patches). Transformer models can be configured to fuse the appearance features of image patches and the aforementioned feature information, and obtain the equivalent resistance of each image patch based on the fused image patch features. This is merely an example. Figure 4 This is an exemplary schematic diagram illustrating the determination of the equivalent resistance of a coronary artery using a first machine learning model, according to some embodiments of this specification. Figure 4As shown, the processing device 120 can segment the medical image 410 to obtain a coronary artery segmentation image, and extract the centerline from the coronary artery segmentation image to obtain a coronary artery centerline image 420. Further, the processing device 120 can divide the coronary artery into multiple segmented image blocks along the coronary artery centerline to obtain a segmented image block sequence 430. Then, the processing device 120 can extract feature information for each segmented image block. The processing device 120 can input the segmented image block sequence 430 and the feature information of the segmented image blocks into a first machine learning model. The CNN of the first machine learning model can extract the apparent features of each segmented image block. The apparent features of each segmented image block and the feature information of that segmented image block can be input into a Transformer model, which can output the equivalent resistance R1, R2, ..., Rn corresponding to each image block.

[0051] In some embodiments, processing device 120 may acquire a first machine learning model from one or more components of FFR determination system 100 (e.g., storage device 140, terminal device 130) or an external source via a network (e.g., network 150) or a network (e.g., network 150). For example, the first machine learning model may be trained in advance by a computing device (e.g., processing device 120) and stored in a storage device (e.g., storage device 140). Processing device 120 may access the storage device to acquire the first machine learning model. As an example only, processing device 120 may acquire multiple first training samples. Each first training sample may include sample feature information of sample segmented image blocks in a coronary artery segmentation image of a sample object and sample equivalent resistance of each sample segmented image block. Further, processing device 120 may generate a first machine learning model based on multiple first training samples by training an initial first model. For information on training the first machine learning model, please refer to other parts of this specification (e.g., Figure 5 (This will not be elaborated upon here.)

[0052] Step 330: Based on the medical image, a second machine learning model is used to determine the boundary condition resistance of the coronary artery, wherein the boundary condition resistance is used to describe the boundary conditions of blood outflow from the distal end of the coronary artery. In some embodiments, step 330 may be performed by the processing device 120 or the second determining module 230.

[0053] In some embodiments, the processing device 120 can identify one or more lumens in the lumen segmentation image. For each lumen, the processing device 120 can determine the feature information of each point on the lumen based on the coronary artery segmentation image, the lesion detection result, and the lumen segmentation image.

[0054] In some embodiments, a point on the lumen may be similar to or identical to a segmented image block described in step 320. In some embodiments, the feature information of a point on the lumen may include the lesion features of that point, cross-sectional features along a line perpendicular to the vessel centerline (including cross-sectional area, minor axis, and major axis), and other features. In some embodiments, the lesion features and cross-sectional features of a point on the lumen may be the same as or similar to the lesion features and cross-sectional features of a segmented image block described in step 320. Other features may include the distance from the point to the nearest upstream bifurcation point, the length of the coronary artery segment corresponding to the point along the blood flow direction, the label of the coronary artery centerline corresponding to the point, the number of upstream and downstream branches at the point, and the combined centerline label of the lumen in which the point is located, etc.

[0055] Furthermore, the processing device 120 can determine the boundary condition resistance of the coronary artery using a second machine learning model based on the feature information of each point on each lumen. Specifically, the processing device 120 can input the feature information of each point into the second machine learning model, and the second machine learning model can output the boundary condition resistance of the coronary artery. In some embodiments, the processing device 120 can also acquire the physiological parameter features of the object described in step 320. The physiological parameter features of the object can be input into the second machine learning model together with the feature information of each point.

[0056] In some embodiments, the second machine learning model may refer to a model used to determine the boundary condition resistance of the coronary artery. In some embodiments, the second machine learning model may include a deep learning model, a traditional machine learning model, etc. Exemplary traditional machine learning models may include linear regression models, random forest models, gradient ascending tree models, support vector machine models, etc., or any combination thereof.

[0057] In some embodiments, the processing device 120 may acquire a second machine learning model in a manner similar to that used to acquire the first machine learning model. As an example only, the processing device 120 may acquire multiple second training samples. Each second training sample includes sample feature information of sample points on the sample lumen of the sample object and the sample boundary condition resistance of each sample lumen. The sample boundary condition resistance of each sample lumen can be determined by a boundary condition resistance equation. The boundary condition resistance equation relates to the branch vessel weights corresponding to the end of the sample lumen, the blood pressure of the sample object, and the cardiac output. Further, the processing device 120 may generate a second machine learning model based on multiple second training samples by training an initial second model. For information on training the second machine learning model, please refer to other parts of this specification (e.g., Figure 5 (This will not be elaborated upon here.)

[0058] Step 340: Based on the equivalent resistance and boundary condition resistance of the coronary artery, determine the fractional flow reserve of the object. In some embodiments, step 340 may be performed by processing device 120 or third determining module 240.

[0059] In some embodiments, the fractional flow reserve of an object may include the fractional flow reserve at multiple points on the coronary artery. In some embodiments, the processing device 120 may generate a circuit topology corresponding to the coronary artery based on a medical image. Further, the processing device 120 may determine the total downstream resistance at each bifurcation point of the coronary artery based on the circuit topology, the equivalent resistance of the coronary artery, and the boundary condition resistance. Then, for each point of the coronary artery, the processing device 120 may determine the fractional flow reserve at that point based on the equivalent resistance of the coronary artery and the total downstream resistance at each bifurcation point. A description of determining the fractional flow reserve can be found elsewhere in this specification (e.g., Figure 6 (This will not be elaborated upon here.)

[0060] It should be noted that the above description of process 300 is for illustrative purposes only and does not limit the scope of this specification. Those skilled in the art can make various modifications and changes to process 300 under the guidance of this specification. However, these modifications and changes remain within the scope of this specification.

[0061] Figure 5 This is a schematic flowchart illustrating the training of exemplary first and second machine learning models according to some embodiments of this specification. In some embodiments, one or more steps of process 500 may be performed... Figure 1 The FFR determination system 100 shown is implemented or is implemented by... Figure 2 The FFR determination system 200 shown is executed. For example, process 500 can be executed by processing device 120 or model training module 250. Figure 5 As shown, process 500 may include the following steps.

[0062] Step 510: Determine the first hyperparameters for training the first machine learning model and the second hyperparameters for training the second machine learning model.

[0063] Hyperparameters refer to parameters that are determined before model training and will not be updated during model training. In some embodiments, the first hyperparameter can be one or more hyperparameters in a CFD equation (e.g., the Navier-Stokes equation). In some embodiments, the first hyperparameter can be one or more hyperparameters in equations such as the naive resistance equation, the curvature resistance equation, the Bernoulli resistance equation, and the elliptic coefficient resistance equation. For example, the naive resistance equation, the curvature resistance equation, the Bernoulli resistance equation, and the elliptic coefficient resistance equation can be expressed as the following equations (1)-(4): Where R1-R4 are the equivalent resistances of the coronary artery segments, l is the length of the coronary artery segment along the blood flow direction, s is the cross-sectional area of ​​the coronary artery segment, r is the radius of the coronary artery segment, c is the curvature of the coronary artery segment, and when the coronary artery segment contains a lesion, s in The cross-sectional area at the upstream origin of the lesion; when the coronary artery segment does not contain the lesion, s in ρ is the average cross-sectional area of ​​the vessel within a certain distance (e.g., 10-20 mm) upstream of the coronary artery segment, d is the diameter of the coronary artery segment, and ρ and k are hyperparameters. When the coronary artery segment does not contain lesions, the value of ρ can be 0 or close to 0.

[0064] In some embodiments, the second hyperparameter may be one or more hyperparameters in the boundary condition resistance equation. In some embodiments, the boundary condition resistance equation may be related to the branch vessel weights corresponding to the lumen termination of the coronary artery, the blood pressure of the object, and the cardiac output. For example, the boundary condition resistance equation may be expressed as the following equation (5): Among them, R bc P is the boundary condition resistance of the current cavity. mean Let Q be blood pressure, Q be cardiac output, i be the lumen number of the coronary artery, n be the total number of coronary artery lumens, and l be the number of blood vessels. i Let l be the weight of the branch vessel corresponding to the end of the i-th lumen. e represents the weight of the branch vessel corresponding to the current lumen terminal, and β and γ are the second hyperparameters.

[0065] In some embodiments, the processing device 120 acquires reference blood flow reserve scores for multiple reference objects. The reference blood flow reserve scores are invasive blood flow reserve scores. In this specification, invasive blood flow reserve scores refer to blood flow reserve scores obtained invasively. For example, the processing device 120 can acquire the reference blood flow reserve scores of multiple reference objects from the storage device 140 of the FFR system 100 or other storage devices. The processing device 120 can determine the predicted blood flow reserve score for each reference object based on the initial values ​​of a first hyperparameter and a second hyperparameter. In some embodiments, the initial values ​​of the first and second hyperparameters can be manually determined by the user or automatically determined by the processing device 120. Further, the processing device 120 can update the initial values ​​of the first and second hyperparameters to ensure that the similarity between the predicted blood flow reserve score and the reference blood flow reserve score for each reference object meets a preset condition.

[0066] Specifically, the reference flow reserve fraction for each reference object may include the reference flow reserve fractions at multiple points on the coronary artery of that reference object. The processing device 120 can determine the measurement site for each reference object based on the lesion detection results and coronary artery segmentation images. For example, the processing device 120 can determine a point located a certain distance (e.g., 2 cm) downstream of the lesion as the measurement site. For each measurement site, the processing device 120 can determine the predicted flow reserve fraction corresponding to that measurement site based on the initial values ​​of the first hyperparameter and the second hyperparameter. Further, the processing device 120 can use genetic algorithms, gradient algorithms, simulated annealing algorithms, etc., to update the initial values ​​of the first and second hyperparameters so that the similarity between the predicted flow reserve fraction and the reference flow reserve fraction at each reference object's measurement site meets a preset condition. For example, the preset condition may be that the difference between the predicted flow reserve fraction and the reference flow reserve fraction is less than a certain threshold.

[0067] For example, the simulated annealing algorithm may include the following three steps: (1) Set the search parameter set and p(n) and search space [0.01p(n), 100p(n)], the initial temperature T is 100, the cooling factor is 0.99, the step size change is 0.1, the error function is the mean-square error (MSE) between the reference blood flow reserve and the predicted blood flow reserve, and the error tolerance is 0.05; (2) Calculate the current temperature (T*cooling factor), update each parameter within the parameter range within the step size range, simulate the newly generated parameter model to obtain the simulation results, and update the error function. Compare the errors before and after, if the error decreases, accept the new parameters, if the error increases, then use the new parameters. (2) Accept the current set of parameters with a certain probability; (3) Repeat step (2) until the error is less than the error tolerance.

[0068] In some embodiments, the processing device 120 can determine the initial equivalent resistance of multiple points on the coronary artery of the reference object according to equations (1)-(4), and determine the initial boundary condition resistance of the reference object according to equation (5). Then, the processing device 120 can determine the predicted fractional flow reserve corresponding to the measurement site based on the initial equivalent resistance and the initial boundary condition resistance of the multiple points on the coronary artery. Further, the processing device 120 can update the initial values ​​of the first hyperparameter in equations (1)-(4) and the second hyperparameter in equation (5) so that the similarity between the predicted fractional flow reserve and the reference fractional flow reserve of each measurement site of the reference object meets a preset condition. When the preset condition is met, the processing device 120 can specify the updated first hyperparameter and second hyperparameter as the first hyperparameter for training the first machine learning model and the second hyperparameter for training the second machine learning model.

[0069] In some embodiments, for each reference object, the processing device 120 determines the aortic inlet boundary conditions (e.g., cardiac cycle of 1 second, cardiac output of 83 ml / s) and aortic outlet boundary condition resistance of that reference object. The processing device 120 can also determine the boundary condition resistance of the coronary artery of the reference object according to the boundary condition resistance equation (e.g., equation (5)). Then, the processing device 120 can use the CFD equation to determine the initial equivalent voltage at each point on the coronary artery of the reference object based on the aortic inlet boundary condition, aortic outlet boundary condition, and coronary artery boundary condition resistance of the reference object. The processing device 120 can determine the predicted fractional flow reserve at each point on the coronary artery by normalizing the initial equivalent voltage at each point on the coronary artery based on the coronary inlet equivalent voltage, and obtain the predicted fractional flow reserve corresponding to the measurement site. Further, the processing device 120 can update the initial value of the first hyperparameter in the CFD equation and the initial value of the second hyperparameter in equation (5) so that the similarity between the predicted fractional flow reserve and the reference fractional flow reserve at each measurement site of the reference object meets a preset condition. When the preset condition is met, the processing device 120 can specify the updated first hyperparameter and second hyperparameter as the first hyperparameter for training the first machine learning model and the second hyperparameter for training the second machine learning model.

[0070] Step 520: Based on the first hyperparameter, determine multiple first training samples.

[0071] In some embodiments, each first training sample may include sample feature information of a segmented image block in the coronary artery segmentation image of the sample object and the sample equivalent resistance of each segmented image block. In some embodiments, the processing device 120 may acquire a sample medical image of the sample object from one or more components of the FFR determination system 100 (e.g., medical device 110, storage device 140, etc.). The processing device 120 may obtain the sample feature information of the coronary artery segmentation image block of the sample object based on the sample medical image in a manner similar to that described in step 320 for determining the feature information of the segmented image block.

[0072] The equivalent resistance of each first training sample can be used as a training label, which can be determined based on a first hyperparameter. In some embodiments, the processing device 120 can directly specify the first hyperparameter as a target first hyperparameter. In some embodiments, the processing device 120 can further adjust the first hyperparameter based on the feature information of the sample object (e.g., physiological feature parameters of the sample object) to obtain the target first hyperparameter. The processing device 120 can construct CFD equations, naive resistance equations, curvature resistance equations, Bernoulli resistance equations, and elliptic coefficient resistance equations, etc., based on the target first hyperparameter, and determine the equivalent resistance of multiple first training samples based on the constructed equations.

[0073] In some embodiments, for each sample segmented image block, the processing device 120 can determine the sample equivalent resistance of the segmented image block using at least one of the naive resistance equation, curvature resistance equation, Bernoulli resistance equation, and elliptic coefficient resistance equation. For example, the processing device 120 can determine the sample equivalent resistance of the sample segmented image block according to the following formula (6): R t =α1R1+α2R2+α3R3+α4R4 (6) Among them, R t Let R1-R4 be the equivalent resistance of the sample segmented image block determined by formulas (1)-(4), and α1-α4 be the weighting coefficients of R1-R4. In some embodiments, α1-α4 can be set by the user based on experience. In some embodiments, when the sample segmented image block has no lesions, the value of α1 can be larger. When the sample segmented image block has lesions, the values ​​of α3 and α4 can be larger.

[0074] In some embodiments, the processing device 120 can determine the sample equivalent resistance of each sample segmented image block based on CFD equations. For example, the processing device 120 can acquire the aortic inlet boundary conditions and outlet boundary conditions of the sample object. The processing device 120 can also use boundary condition resistance equations (e.g., equation (5) described in step 510) to determine the boundary condition resistance of the coronary arteries of the sample object. Further, the processing device 120 can use CFD equations to determine the equivalent voltage at each point on the coronary arteries of the sample object based on the aortic inlet and outlet boundary conditions and the boundary condition resistance of the coronary arteries. Then, the processing device 120 can acquire the circuit topology corresponding to the coronary arteries of the sample object and determine the sample equivalent resistance of each sample segmented image block based on the circuit topology and the equivalent voltage at each point on the coronary arteries.

[0075] Step 530: Based on multiple first training samples, generate a first machine learning model by training a first initial model.

[0076] During training, the sample feature information of the segmented image patch can be used as the model input, the sample equivalent resistance of each segmented image patch can be used as the training label, and the model parameters of the first initial model can be iteratively updated to satisfy the iteration termination condition. In some embodiments, the processing device 120 can use any suitable loss function (e.g., MSE) and a suitable optimizer (e.g., the Adam optimizer) to train the first initial model to obtain the first machine learning model.

[0077] Step 540: Based on the second hyperparameter, determine multiple second training samples.

[0078] In some embodiments, each second training sample includes sample feature information of sample points on the lumen of the sample object and sample boundary condition resistance of each sample lumen. In some embodiments, the processing device 120 may acquire sample medical images of the sample object from one or more components of the FFR determination system 100 (e.g., medical device 110, storage device 140, etc.). The processing device 120 may obtain sample feature information of sample points on the lumen based on the sample medical images in a manner similar to determining the feature information of points on the lumen as described in step 330. In some embodiments, the sample medical images used to acquire the second training samples and the sample medical images used to acquire the first training samples may be the same; that is, data from the same sample object can be used to acquire the first and second training samples. In this way, the first and second training samples can be acquired simultaneously using less data, thereby improving the efficiency of acquiring the first and second training samples.

[0079] The sample boundary condition resistance of each second training sample can be used as a training label, which can be determined based on the second hyperparameter. In some embodiments, the processing device 120 can directly specify the second hyperparameter as the target second hyperparameter. In some embodiments, the processing device 120 can further adjust the second hyperparameter based on the feature information of the sample object (e.g., physiological feature parameters of the sample object, etc.) to obtain the target second hyperparameter. The processing device 120 can construct a boundary condition resistance equation based on the target second hyperparameter, and determine multiple second training samples based on the constructed boundary condition resistance equation.

[0080] In some embodiments, the sample boundary condition resistance of each sample lumen can be determined by the boundary condition resistance equation. In some embodiments, the boundary condition resistance equation is related to the branch vessel weights corresponding to the end of the sample lumen, the blood pressure of the sample object, and the cardiac output. For example, the processing device 120 can directly determine the sample boundary condition resistance of each sample lumen according to formula (5) described in step 510.

[0081] Step 550: Based on multiple second training samples, a second machine learning model is generated by training a second initial model.

[0082] During training, the sample feature information of the segmented image patch can be used as model input, and the sample equivalent resistance of each segmented image patch can be used as training label. The model parameters of the first initial model can be iteratively updated to satisfy the iteration termination condition. In some embodiments, the processing device 120 can use any suitable loss function (e.g., MSE) to train the second initial model to obtain the second machine learning model.

[0083] In some embodiments, steps 510, 520, and 540 may be omitted. The processing device 120 can directly obtain the pre-generated first and second training samples. For example, in step 510, the first and / or second hyperparameters may be preset (e.g., determined by the user). The processing device 120 can directly obtain the first and / or second hyperparameters to determine the first and / or second training samples.

[0084] Figure 6 This is a schematic flowchart illustrating an exemplary determination of FFR according to some embodiments of this specification. In some embodiments, one or more steps of process 600 may be performed... Figure 1 The FFR determination system 100 shown is implemented or is implemented by... Figure 2 The FFR determination system 200 shown is executed. For example, process 600 can be executed by processing device 120 or third determination module 240. Figure 6 As shown, process 600 may include the following steps.

[0085] Step 610: Based on the medical image, generate the circuit topology corresponding to the coronary artery.

[0086] In some embodiments, the processing device 120 may generate the circuit topology corresponding to the coronary artery using a DC-based lumped parameter model. Specifically, the processing device 120 may equate the heart of the object to a power supply voltage (DC), equate a segment of the blood vessel between every two bifurcation points of the coronary artery to a resistor R, and connect a boundary condition resistor R before grounding each end of the coronary artery. bc This is just an example. Figure 7 This is a schematic diagram of the exemplary circuit topology corresponding to a coronary artery, based on some embodiments of this specification. For example... Figure 7 As shown, the coronary artery comprises five branches L1-L5, which correspond to resistors R1-R5 in the circuit topology. The three ends of the coronary artery are connected to boundary condition resistors R1 and R5, respectively. bc1 -R bc3 .

[0087] In some embodiments, the processing device 120 may generate the circuit topology corresponding to the coronary artery using an AC-based lumped parameter model. For example, the processing device 120 may replace the resistor in the topology generated by the DC-based lumped parameter model with an RLC circuit. For example, the RLC circuit may be a circuit obtained by connecting an inductor and a resistor in series and then connecting them in parallel with a capacitor.

[0088] Step 620: Based on the circuit topology, the equivalent resistance of the coronary artery, and the boundary condition resistance, determine the total downstream resistance at each bifurcation point of the coronary artery.

[0089] In some embodiments, the processing device 120 can trace back upstream from each end of the coronary artery to sequentially determine the downstream total resistance of each bifurcation point. The downstream total resistance of a bifurcation point refers to the total resistance corresponding to the downstream point resistance and the downstream boundary condition resistance. In some embodiments, the processing device 120 can determine the downstream total resistance of each bifurcation point based on the circuit structure downstream of each bifurcation point. For example, as... Figure 7 As shown, the total resistance downstream of the bifurcation point C2 is The total resistance downstream of the bifurcation point C1 is

[0090] Step 630: For each point in the coronary artery, determine the fractional flow reserve at that point based on the equivalent resistance of the coronary artery and the total downstream resistance at each bifurcation point.

[0091] In some embodiments, for each point in the coronary artery, the processing device 120 can determine a first resistance and a second resistance at that point based on the equivalent resistance of the coronary artery and the total downstream resistance of each bifurcation point. The first resistance is the total downstream resistance at that point. The second resistance is the sum of the total resistance of the coronary branch to which the point is located and the total downstream resistance of the adjacent downstream bifurcation points at that point. In this specification, the adjacent downstream bifurcation point of a point in a coronary artery branch refers to the bifurcation point where the downstream end of the coronary artery branch connects. The total resistance of a coronary artery branch is the sum of the equivalent resistances of all points on that coronary artery branch. The processing device 120 can obtain the equivalent voltage of the adjacent upstream bifurcation point at that point. Further, the processing device 120 can determine the fractional flow reserve at that point based on the first resistance, the second resistance, and the equivalent voltage of the adjacent upstream bifurcation point. In this specification, the adjacent upstream bifurcation point of a point in a coronary artery branch refers to the bifurcation point where the upstream initiation of the coronary artery branch connects or the heart. That is, when a coronary artery branch is the starting segment of a coronary artery, the upstream nearest bifurcation point of a point on that coronary artery branch refers to the heart to which the starting end of the coronary artery connects; when a coronary artery branch is not the starting segment of a coronary artery, the upstream nearest bifurcation point of a point on that coronary artery branch refers to the bifurcation point to which the upstream starting end of the coronary artery branch connects. Specifically, for each point, the processing device 120 can determine the equivalent voltage of that point based on the first resistance, the second resistance, and the equivalent voltage of the upstream nearest bifurcation point. Then, the processing device 120 can determine the ratio of the equivalent voltage of that point to the power supply voltage as the fractional flow reserve of that point.

[0092] Just as an example, Figure 8 This is a schematic diagram of the exemplary circuit topology corresponding to a coronary artery, based on some embodiments of this specification. For example... Figure 8 As shown, the power supply voltage is U0, the total resistances of coronary branches L1-L5 are R1-R5 respectively, and the downstream total resistances of bifurcation points C1 and C2 are R1-R5 respectively. c1 and R c2 Coronary branch L1 includes points L11-L14. Processing device 120 can calculate the fractional flow reserve for each point sequentially, starting from point L11. For example, processing device 120 can determine the equivalent voltage U at point L12 according to formula (7). L12 : Among them, R L13 R L14 The equivalent resistances of points L13 and L14 are respectively, and the first resistance of point L12 is R. L13 +R L14 +R c1 The second resistor is R1+R c1 .

[0093] For example, coronary branch L1 includes points L31-L35. Processing device 120 can determine the equivalent voltage U at point L32 according to formula (8). L32 : Among them, R L33 R L34 R L35 The equivalent resistances of points L33, L34, and L35 are respectively, and the first resistance of point L32 is R. L33 +R L34 +R L35 +R c2 The second resistor is R3+R c1 U c1 The equivalent voltage of C1 at the bifurcation point. In some embodiments, the equivalent voltage of the bifurcation point can be the equivalent voltage of a point upstream of the bifurcation point, for example, the equivalent voltage of the bifurcation point C1 can be the equivalent voltage of point L14.

[0094] Processing device 120 can determine U L12 / U0 represents the fractional blood flow reserve at point L12, U L32 / U0 represents the fractional flow reserve at point L32.

[0095] In some embodiments of this specification, the fractional flow reserve (FFR) of an object can be determined based on the equivalent resistance and boundary condition resistance of the coronary artery. The beneficial effects of these embodiments include, but are not limited to: (1) This application trains first and second machine learning models based on a large amount of data. By using these models, the accuracy and efficiency of determining the equivalent resistance and boundary condition resistance of the coronary artery can be greatly improved; (2) As described elsewhere in this application, in current methods, the machine learning model only learns the influence of vascular features on FFR without combining the topological structure of the coronary artery tree, resulting in low accuracy of the obtained FFR. In contrast, this application, after obtaining the equivalent resistance and boundary condition resistance of the coronary artery, further combines the circuit topology of the coronary artery to determine the FFR at each point, thereby significantly improving the accuracy of the FFR and having high clinical applicability; (3) The training labels for the first and second machine learning models in this application can be determined without using CFD equations, thus avoiding the solution of complex fluid dynamics differential equations, greatly improving the speed of training label determination and robustness to coronary artery segmentation.

[0096] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this specification. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this specification. Such modifications, improvements, and corrections are suggested in this specification and therefore remain within the spirit and scope of the exemplary embodiments described herein.

[0097] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Moreover, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0098] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the system components described above can be implemented using hardware devices, they can also be implemented solely using software solutions, such as installing the described system on existing servers or mobile devices.

[0099] Similarly, it should be noted that, in order to simplify the description disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments in this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0100] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0101] For each patent, patent application, patent application publication, and other material, such as articles, books, specifications, publications, and documents, referenced in this specification, the entire contents of which are incorporated herein by reference. This excludes historical application documents that are inconsistent with or conflict with the content of this specification, as well as documents that limit the broadest scope of the claims in this specification (currently or subsequently appended to this specification). It should be noted that in the event of any inconsistency or conflict between the descriptions, definitions, and / or terminology used in the supplementary materials to this specification and the content of this specification, the descriptions, definitions, and / or terminology used in this specification shall prevail.

[0102] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for determining fractional blood flow reserve, characterized in that, The method includes: Acquire a medical image of the object, the medical image including the coronary arteries of the object; Based on the medical image, a first machine learning model is used to determine the equivalent resistance of the coronary artery, wherein the equivalent resistance of the coronary artery includes the equivalent resistance values ​​of multiple points on the coronary artery; Based on the medical image, a second machine learning model is used to determine the boundary condition resistance of the coronary artery, wherein the boundary condition resistance describes the boundary conditions for blood outflow from the distal end of the coronary artery; and The fractional flow reserve of the object is determined based on the equivalent resistance of the coronary artery and the boundary condition resistance.

2. The method as described in claim 1, characterized in that, The determination of the equivalent resistance of the coronary artery using a first machine learning model based on the medical image includes: Based on the medical images, the coronary artery segmentation image, lesion detection results, and lumen segmentation image of the object are determined; Identify multiple segmented image blocks in the coronary artery segmentation image; For each of the plurality of segmented image blocks, based on the coronary artery segmentation image, the lesion detection result, and the lumen segmentation image, the feature information of the segmented image block is determined; and Based on the feature information of each segmented image block, the equivalent resistance of the coronary artery is determined using the first machine learning model.

3. The method as described in claim 1, characterized in that, Determining the fractional flow reserve of the object based on the equivalent resistance and the boundary condition resistance of the coronary artery includes: Based on the medical image, generate the circuit topology corresponding to the coronary artery; Based on the circuit topology, the equivalent resistance of the coronary artery, and the boundary condition resistance, the downstream total resistance at each bifurcation point of the coronary artery is determined. For each point of the coronary artery, the fractional flow reserve at that point is determined based on the equivalent resistance of the coronary artery and the total downstream resistance at each bifurcation point.

4. The method as described in claim 3, characterized in that, The determination of the fractional flow reserve at each bifurcation point based on the equivalent resistance of the coronary artery and the downstream total resistance at each bifurcation point includes: Based on the equivalent resistance of the coronary artery and the total downstream resistance of each bifurcation point, a first resistance and a second resistance of the point are determined, wherein the first resistance is the total downstream resistance of the point, and the second resistance is the sum of the total resistance of the coronary branch where the point is located and the total downstream resistance of the adjacent bifurcation point downstream of the point; Obtain the equivalent voltage of the upstream neighboring bifurcation point of the point; The fractional blood flow reserve of a point is determined based on the first resistance, the second resistance, and the equivalent voltage of the upstream adjacent bifurcation point of the point.

5. The method as described in claim 1, characterized in that, The first machine learning model was generated through the following training process: Multiple first training samples are obtained. Each first training sample includes sample feature information of sample segmentation image blocks in the coronary artery segmentation image of the sample object and sample equivalent resistance of each sample segmentation image block. Based on the plurality of first training samples, the first machine learning model is generated by training the first initial model.

6. The method as described in claim 5, characterized in that, The sample equivalent resistance of each sample segmented image block is determined by at least one of the naive resistance equation, curvature resistance equation, Bernoulli resistance equation, and elliptic coefficient resistance equation, or The sample equivalent resistance of each sample segmented image block is determined using computational fluid dynamics (CFD) equations.

7. The method as described in claim 1, characterized in that, The second machine learning model is generated through the following training process: Multiple second training samples are obtained. Each second training sample includes sample feature information of sample points on the sample lumen of the sample object and sample boundary condition resistance of each sample lumen. The sample boundary condition resistance of each sample lumen is determined by the boundary condition resistance equation, which is related to the branch vessel weights corresponding to the end of the sample lumen, the blood pressure and cardiac output of the sample object. Based on the multiple second training samples, the second machine learning model is generated by training the second initial model.

8. The method as described in claim 1, characterized in that, The training labels of the first machine learning model are determined based on a first hyperparameter, and the training labels of the second machine learning model are determined based on a second hyperparameter. The values ​​of the first hyperparameter and the second hyperparameter are determined through the following process: Obtain reference blood flow reserve fractions from multiple reference objects, wherein the reference blood flow reserve fractions are invasive blood flow reserve fractions; Based on the initial values ​​of the first hyperparameter and the second hyperparameter, the predicted blood flow reserve fraction for each of the reference objects is determined; as well as Update the initial values ​​of the first hyperparameter and the second hyperparameter so that the similarity between the predicted blood flow reserve fraction and the reference blood flow reserve fraction of each reference object meets a preset condition.

9. A system for determining fractional flow reserve, comprising: The acquisition module is configured to acquire a medical image of an object, the medical image including the coronary arteries of the object; The first determining module is configured to determine the equivalent resistance of the coronary artery based on the medical image and using a first machine learning model, wherein the equivalent resistance of the coronary artery includes the equivalent resistance values ​​of multiple points on the coronary artery; The second determining module is configured to determine the boundary condition resistance of the coronary artery based on the medical image and using a second machine learning model, wherein the boundary condition resistance is used to describe the boundary conditions at the end of blood flow out of the coronary artery; and The third determining module is configured to determine the fractional blood flow reserve of the object based on the equivalent resistance of the coronary artery and the boundary condition resistance.

10. A system for determining fractional flow reserve, comprising: At least one storage device for storing computer instructions; At least one processor is configured to execute the computer instructions to implement the method described in any one of claims 1-8.

Images