Medical image processing method and system

Abstract

A medical image processing method for determining a fractional flow reserve of a stenotic lesion through a coronary artery from medical image data is disclosed. The medical image data includes a set of images of a coronary artery region of a patient. The coronary artery region includes the stenotic lesion. The method comprises the following steps: reconstructing a three-dimensional coronary artery tree model of a patient according to medical image data; determining a stenosis lesion size according to the three-dimensional coronary artery tree model of the patient; simulating blood flow in the three-dimensional coronary artery tree model of the patient to determine a model flow; predicting a model pressure drop across the stenotic lesion using an analytical model depending on the stenotic lesion size according to the model flow; a fractional flow reserve through the stenotic lesion is determined based on the model pressure drop.

Description

[0001] This application is a divisional application of Chinese invention patent application No. 201680027187.1, filed on May 6, 2016, entitled "Medical Image Processing Method and System". Technical Field

[0002] This invention relates to a drug delivery device.

[0003] This invention relates to medical image processing. More specifically, the embodiments disclosed in this application relate to processing medical images of a patient's coronary artery region to assess coronary artery stenosis by determining the fractional flow reserve (FFR). Background Technology

[0004] Cardiovascular disease is a leading cause of death in both developed and developing countries, accounting for 30% of all deaths globally. The number of people with cardiovascular disease is expected to increase significantly due to the aging of the world's population. Coronary artery disease (CAD) is the most common cardiovascular disease, caused by the buildup of plaque in the coronary arteries. This plaque narrows the arteries and ultimately affects the heart's blood supply. Therefore, CAD is associated with the development of cardiovascular events and accounts for 45% of cardiovascular disease deaths.

[0005] Therefore, characterizing and quantifying the significance of coronary artery stenosis is crucial for managing patients and preventing death from coronary artery disease (CAD). In clinical cardiology, anatomical and hemodynamic parameters are commonly used to quantify the severity of CAD. Rapidly evolving non-invasive imaging techniques, such as computed tomography (CTA), are valuable due to their relatively low cost and medical complications. Most non-invasive imaging techniques focus on the anatomical significance and calculate the diameter of the stenotic lesion (DS) as a parameter. DS represents the diameter of the stenotic lesion area relative to the “normal” segment, but it rarely provides, or does not provide, hemodynamic information about the stenosis. The current gold standard for assessing the functional severity of coronary artery stenosis is the fractional flow reserve (FFR). FFR can be calculated as the ratio of pressure distal to the stenotic lesion in a congested state to the aortic pressure. When coronary artery stenosis results in an FFR ≤ 0.80, revascularization is generally recommended. However, FFR can only be measured via invasive coronary catheterization, which has high medical costs and several complications.

[0006] The limitations of invasive coronary artery catheterization have prompted us to find a new approach to assess the functional significance of coronary artery stenosis directly from medical images obtained through non-invasive imaging techniques.

[0007] In current clinical practice, cardiologists and radiologists obtain only anatomical information about coronary artery stenosis from CTA images. Information regarding the functional significance of coronary artery lesions is lacking. However, anatomical parameters of diameter stenosis (DS) have been found to lead to a much higher false-positive rate compared to free flow refractory (FFR), often overestimating the significance of stenosis for ischemia. Summary of the Invention

[0008] According to a first aspect of the present invention, a medical image processing method is provided, which determines the fractional flow reserve (FRR) through a stenotic lesion in a coronary artery based on medical image data. The medical image data includes a set of images of a patient's coronary artery region. The coronary artery region includes the stenotic lesion. The method includes: reconstructing a three-dimensional coronary artery tree model of the patient based on the medical image data; determining the size of the stenotic lesion based on the patient's three-dimensional coronary artery tree model; simulating blood flow in the patient's three-dimensional coronary artery tree model to determine the model flow rate; predicting, based on the model flow rate and using an analysis model dependent on the size of the stenotic lesion, the model pressure drop across the stenotic lesion; and determining the FRR through the stenotic lesion based on the model pressure drop.

[0009] Embodiments of this invention can utilize analytical models to allow real-time calculation of the hemodynamic response (FFR) based on the size of the stenotic lesion (proximal, distal, and minimum luminal area and length) and the flow rate through the congested coronary artery passing through the lesion. Since the size of the stenotic lesion can be obtained via non-invasive CTA, and computational fluid dynamics (CFD) can help derive the flow rate through the congested coronary artery passing through the lesion, it is feasible to establish a framework for analyzing the hemodynamic significance of coronary artery stenosis using analytical models.

[0010] In one embodiment, simulating blood flow in a three-dimensional coronary tree model of a patient to determine the flow rate determined through modeling includes simulating steady-state blood flow. Simulating steady-state blood flow may include iteratively determining steady-state fluid resistance and pressure values.

[0011] The embodiments of the present invention use simplified computational fluid dynamics (CFD), which may involve steady-state flow simulations and / or novel boundary conditions, reducing the complexity of pulsating flow simulations using reduced-order models as boundary conditions.

[0012] In one embodiment, determining the steady-state fluid resistance and pressure values ​​iteratively includes updating the steady-state fluid resistance and pressure values ​​iteratively using an under-relaxation scheme.

[0013] In one embodiment, the patient's three-dimensional coronary tree model includes at least one inlet and multiple outlets, and the method further includes estimating the total inflow into the patient's three-dimensional coronary tree model based on the medical image data, and wherein determining steady-state fluid resistance and pressure values ​​iteratively includes iteratively updating the steady-state fluid resistance and pressure values ​​until the total outflow from the multiple outlets matches the total inflow.

[0014] The simulated blood flow can be congested blood flow. Congested flow refers to a physiological state in which the resistance of downstream vessels is minimized and the blood flow through the coronary vessels is maximized. It involves flow conditions modeled as steady-state flow or pulsating flow.

[0015] In one embodiment, based on the flow rate determined by modeling, the analysis model predicts the pressure drop across the stenotic lesion determined by modeling, depending on the stenotic lesion size, including determining the radius of the non-viscous fluid core depending on the stenotic lesion length and flow rate, and comparing the radius of the non-viscous fluid core with a threshold to determine whether the inlet effect is included in the analysis model.

[0016] In one embodiment, the analytical model is based on an improved version of the Bernoulli equation.

[0017] According to another aspect of the present invention, a method for assessing the functional significance of coronary artery stenosis based on medical image data is provided. The method includes, according to the method described above, determining the fractional flow reserve (FVR) of the stenosis passing through the coronary artery based on medical image data; and assessing the functional significance of the coronary artery stenosis by comparing the FVR of the stenosis with a threshold.

[0018] In one embodiment, the functional significance of coronary artery stenosis is assessed by comparing the fractional flow reserve of the stenosis with a threshold, including classifying the stenosis as ischemic if the fractional flow reserve of the stenosis through the coronary artery is less than the threshold.

[0019] In one embodiment, the fractional flow reserve (FLR) of a stenotic lesion through a coronary artery is compared to a threshold to assess the functional significance of the lesion, including classifying the stenotic lesion as non-ischemic if the FLR is greater than the threshold. In one embodiment, the threshold is 0.8.

[0020] According to another aspect of the present invention, a medical image processing system is provided that evaluates the fractional flow reserve (FRR) through a coronary artery stenosis based on medical image data. The medical image data includes a set of images of a patient's coronary artery region. The coronary artery region includes the stenosis. The system includes a computer processor and a data storage device having a three-dimensional model reconstruction module, a stenosis size determination module, a blood flow simulation module, an analysis modeling module, and a FRR determination module, including non-transient instructions operable by the processor to: reconstruct a three-dimensional coronary artery tree model of the patient based on the medical image data; determine the stenosis size based on the patient's three-dimensional coronary artery tree model; simulate blood flow in the patient's three-dimensional coronary artery tree model to determine a flow rate determined by modeling; predict, based on the flow rate determined by modeling, a pressure drop across the stenosis determined by modeling, depending on the stenosis size, using an analysis model; and determine the FRR through the stenosis based on the pressure drop determined by modeling.

[0021] According to another aspect, a non-transitory computer-readable medium is provided. The computer-readable medium stores program instructions for causing at least one processor to perform operations of the methods disclosed above. Attached Figure Description

[0022] In the following description, embodiments of the invention will be illustrated by way of non-limiting examples with reference to the accompanying drawings, wherein: Figure 1 This is a block diagram illustrating a medical image processing system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for processing medical image data according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for assessing the functional significance of coronary artery disease according to an embodiment of the present invention; Figure 4a , 4b An example is shown of applying the method according to embodiments of the present invention to medical image data for two patients with ischemic coronary artery disease; Figure 5a , 5b An example is shown of applying the method according to embodiments of the present invention to medical image data for two patients with non-ischemic coronary artery disease; Figure 6a This shows the measured FFR. B A graph showing the correlation between the FFR calculated using a method according to an embodiment of the present invention; Figure 6b This shows the FFR measured on a per-vessel basis.B A graph of the Bland-Altman plot of the FFR calculated using a method according to an embodiment of the present invention; Figure 7 This is a graph showing the area under each vascular performance curve according to one embodiment of the present invention. Detailed Implementation

[0023] Figure 1 A medical image processing system according to an embodiment of the present invention is illustrated. The image processing system 100 includes a processor 110, which may be referred to as a central processing unit (CPU), and is connected to a storage device including a storage unit 120 and a memory unit 150. The processor 110 may be implemented as one or more CPU chips. The memory unit 150 is implemented as random access memory (RAM). The medical image processing system also includes an input / output (I / O) device 160 and a network interface 170.

[0024] Storage 120 generally includes one or more disk drives for non-volatile storage of data. Storage 120 stores several program modules 130, which are loaded into memory 150 when selected for execution. The program modules include a 3D model reconstruction module 131, a stenosis lesion size determination module 132, a blood flow simulation module 133, an analysis modeling module 134, a fractional flow reserve (FFR) determination module 135, and a threshold comparison module 136. Each program module 130 stored in storage 120 includes non-transitory instructions that are operated by processor 120 to perform various operations in the methods described in more detail later. Storage 120 and memory 150 may, in some cases, be referred to as computer-readable storage media and / or non-transitory computer-readable media.

[0025] Storage 120 also stores medical image data 140. The medical image data includes a set of medical images of a patient's coronary artery region. The patient's coronary artery region may include stenotic lesions analyzed using methods described later. The medical image data 140 may be in DICOM format. The medical image data may be obtained using medical imaging techniques, such as computed tomography (CT) scans, computed tomography angiography (CTA) scans of the patient, or magnetic resonance imaging (MRI) scans of the patient.

[0026] I / O device 160 may include: a printer; a video monitor; a liquid crystal display (LCD); a plasma display; a touch screen display keyboard; a keypad; a switch; a dial; a mouse; a trackball; a voice recognition device; a card reader; or other known input devices.

[0027] Network interface 170 may take the form of a modem, modem unit, Ethernet card, Universal Serial Bus (USB) interface card, serial interface, token ring card, Fiber Distributed Data Interface (FDDI) card, Wireless Local Area Network (WLAN) card, wireless transceiver card and / or other air interface protocol radio transceiver card, and other known network devices. The wireless transceiver card uses protocols such as Code Division Multiple Access (CDMA), Global System for Mobile Communications (GSM), Long Term Evolution (LTE), WiMAX, Near Field Communication (NFC), and Radio Frequency Identification (RFID) to facilitate radio communication. These network interfaces 170 enable processor 110 to communicate with the Internet or more intranets. Through such a network connection, it is conceivable that processor 110 may receive information from the network or output information to the network during the execution of the operations described later. The information is typically represented as a sequence of instructions to be executed by processor 110, which can be received from and output to the network, for example, as computer data signals embodied in a carrier wave.

[0028] Network interface 170 allows medical image processing system 110 to receive medical image data 140 from medical imaging equipment for processing in accordance with the methods described later.

[0029] Processor 110 executes instructions, code, computer programs, and scripts accessed from hard disks, floppy disks, optical disks (all of these disk-based systems can be considered as storage 120), flash drives, memory 150, or network interface 170. Although only one processor 110 is shown, multiple processors may exist. Therefore, while instructions can be discussed as being executed by a processor, instructions can be executed simultaneously, serially, or otherwise by more than one processor.

[0030] It is understood that by programming and / or loading executable instructions onto the medical image processing system 100, at least one of the CPU 110, memory 150, and storage 120 is modified, partially transforming the image processing system into a purpose-specific machine or system with the novel functions taught in this application. A fundamental point for the fields of electrical engineering and software engineering is that functions achievable by loading executable software into a computer can be translated into hardware implementations using well-known design rules.

[0031] Although system 100 is described with reference to a computer, it should be understood that the system can be formed by two or more computers communicating with each other and cooperating to perform tasks. For example, but not as a limitation, an application can be partitioned to allow simultaneous and / or parallel processing of the application's instructions. Alternatively, data processed by the application can be partitioned to allow two or more computers to simultaneously and / or in parallel process different portions of the data set. In one embodiment, the functionality disclosed above can be provided by executing the application and / or multiple applications in a cloud computing environment. Cloud computing can include providing computing services using dynamically scalable computing resources via network connections. The cloud computing environment can be established by an enterprise and / or rented from a third-party provider as needed.

[0032] Figure 2 This is a flowchart illustrating a method for processing medical image data according to an embodiment of the present invention. Method 200 in the figure can be... Figure 1 This is performed on the medical image processing system 100 shown. It should be noted that the enumeration of the operations is for clarity purposes, and these operations do not need to be performed in the order implied by the enumeration.

[0033] As described above, method 200 is performed on medical image data 140. Medical image data 140 can be obtained using standard coronary CTA as illustrated in the following example. Standard coronary CTA employs a 64-slice multi-detector CTA scanner (e.g., a Toshiba Aquilion 64 with 0.5 mm × 64 detector rows) or a 320-slice multi-detector CTA scanner (Toshiba Aquilion ONE). TM The scan was performed using a 0.5mm × 320 detector row, following current clinical guidelines. The scan extent was planned according to individual differences to cover most of the coronary arteries and aorta. Some patients were given a beta-blocker to slow the heart rate. Iodophenylhexol nonionic control was typically administered before image acquisition to highlight regions of interest in the CT images. ECG triggering was used for intended control of the control dose. Image acquisition was performed with breath-holding during inspiration. Scan parameters were: gantry rotation time 350–400 ms, tube potential 100–120 kV, field of view (FOV) 161–230 mm. Depending on the size and spacing of the heart, the scan time in the head-to-tail direction was 5.7–8.4 seconds per breath-hold. All CT images were recorded in 0.25 mm increments (i.e., 0.5 mm slice thickness) and saved in DICOM format for image processing.

[0034] In step 202, the 3D reconstruction module 131 of the medical image processing system 100 reconstructs a three-dimensional coronary artery tree model of the patient based on medical image data 140. In step 202, image processing can be performed to isolate arterial structures from the background CT image for reconstructing the 3D coronary artery tree. First, a Hessian matrix-based filter is applied to each individual transverse CT image to obtain higher-order geometric features; that is, the principal curvature of the image intensity. Eigenvector analysis is performed on the obtained Hessian matrix to determine whether the analyzed voxels belong to vascular structures. Finally, image-based graphic segmentation techniques are applied to the enhanced 3D image, and then a triangular surface mesh is built from the segmented voxel data using a traveling cubes algorithm. We found that some models require manual editing to remove artifacts and small vessels with diameters <1 mm.

[0035] In step 204, the stenosis lesion size determination module 132 of the medical image processing system 100 determines the size of the stenosis lesion based on the three-dimensional arterial tree model reconstructed in step 202. The reconstructed three-dimensional model allows for the measurement of the stenosis lesion's size, such as proximal, distal, and minimum luminal area, as well as its length.

[0036] In step 206, the blood flow simulation module 133 of the medical image processing system 100 simulates blood flow in a three-dimensional model. In step 206, computational fluid dynamics (CFD) is performed on the 3D model to obtain the flow rate through the congested coronary arteries of the lesion. To obtain flow information of the coronary artery lesion in a non-invasive manner, CFD simulation is performed. First, the 3D coronary artery model is discretized to divide the computational domain into small control volumes. Then, morphological and physiological boundary conditions are set for the boundaries. Afterward, the mass conservation equation and momentum conservation equation are solved iteratively. Finally, the flow information is obtained from the solution.

[0037] In step 206, the computational domain can be discretized into tetrahedral elements using the ANSYS workbench (a commercial software package). After mesh dependency testing, the coronary tree model is discretized to have a total of approximately 800,000 volumetric elements. Further mesh refinement results in a relative error of <1% at the maximum velocity.

[0038] In one embodiment, FLUENT is used. TMSolve the continuity equation and the Navier-Stokes equation. Model blood as a Newtonian fluid to simulate blood flow in a patient-specific coronary artery tree model. Specify appropriate boundary conditions at the inlet and outlet of each model to simulate physiological conditions and satisfy three key principles. First, it is assumed that the coronary artery supply meets the myocardial demand at rest. Therefore, the total resting coronary blood flow can be calculated by assessing myocardial mass based on CT images. Second, the resting resistance of each coronary artery branch is proportional to the size of the parent and child branch vessels. Third, it is assumed that coronary resistance decreases as expected during congestion, falling within the expected physiological range.

[0039] To simulate blood flow in a patient-specific coronary tree model, this study used FLUENT. TM Solve the continuity equation and the Navier-Stokes equation, as shown in equations (1) and (2) respectively: (1) (2) Where, x j It is the position in the Cartesian coordinate system, u j (or u) i ) represents the Cartesian component of velocity, P represents static pressure; ρ and μ are set to 1060 kg / m³. 3 and 4.5×10 -3 Pa•s represents the density and dynamic viscosity of blood in the epicardial arteries.

[0040] Pressure and resistance boundary conditions were specified at the inlet and outlet of each model to simulate physiological conditions. To allocate the total inlet pressure, the mean brachial artery pressure was calculated using patient-specific systolic and diastolic pressures and matched to the mean aortic pressure.

[0041] A key element in determining the value of coronary artery resistance at the outlet is the specified reference pressure (P0). Physiological studies report that at low pressures, the coronary pressure streamlines are concave towards the flow axis, but straight at physiological pressures. The zero-flow pressure intercept within the physiological pressure range (i.e., P0 in this application) exceeds five to ten times the diastolic pressure of the coronary vein or left ventricle. Therefore, in this study, we propose a novel method to determine the reference pressure P0 and thus determine the outlet resistance via an iterative process. Specific details are as follows. In short, the total resistance R at rest... inlet Defined by equation (3).

[0042] (3) Among them, P inletThis represents the mean aortic pressure estimated based on mean branchial artery pressure. Total coronary flow at rest can be estimated based on myocardial mass. Therefore, Q inlet It can be determined from CT images.

[0043] At the exit point of each coronary artery, (4) The resistance R of the downstream vascular system of each coronary artery branch i Resistance R of the coronary tree inlet The correlation between them can be estimated according to the similarity law of equation (5) below.

[0044] (5) Where, N i This represents the ratio of the flow rate at the inlet to the flow rate at the i-th outlet.

[0045] According to (3) and (5), R i It can take the following forms: (6) According to the law of conservation of mass, the total outflow is equal to the inflow (Equation (7)). (7) According to equations (4), (5) and (7), P0 is given by the following formula: (8) During congestion, it is assumed that the coronary microcirculation has a predictable response to adenosine, and that the resistance decreases to K times the resting value (where K = 0.21), which is within the expected physiological range.

[0046] According to equation (6), the resistance to congestion is given by the following formula: (9) To ensure smooth convergence during the CFD numerical iteration process, P0 and R i Initialized to 20 mmHg and 1.0e+8 Pa•s / m respectively. 3 Then, the process is updated using the under-relaxation scheme formulated by equations (10) and (11) until the total outflow from all outlets matches the inflow during engorgement.

[0047] To ensure smooth convergence during the CFD numerical iteration process, P0 and Ri were initialized to 20 mmHg and 1.0e+8 Pa•s / m, respectively. 3 Then, the process is updated using the under-relaxation scheme formulated as in equations (10) and (11) until the total outflow from all outlets matches the inflow rate during engorgement.

[0048] (10) (11) Where α is the underrelaxation factor. P 0,old and R i,hyperemia,old This represents the reference pressure and resistance in the last iteration, while P... 0,new and R i,hyperemia,new This indicates the reference pressure and resistance values ​​in the next iteration. Since it has been found that minimum microvascular resistance is not related to the severity of epicardial stenosis, it is assumed that the congestive microcirculatory resistance distal to the stenosis is the same as the resistance of the coronary artery without stenosis, as indicated in equation (11).

[0049] By utilizing these novel iterative boundary conditions, the computational burden can be reduced compared to solving complex lumped parameter cardiac and coronary artery models.

[0050] For the 21 patients investigated in this study, the calculated value of P0 was 38.6 ± 9.5 mmHg, which is close to the 37.9 ± 9.8 mmHg measured by Dole et al. for 10 patients (Dole WP, Richards KL, Hartley CJ, Alexander GM, Campbell AB, Bishop VS. "Diastolic coronary artery pressure-flow velocity relations in conscious man", Cardiovasc Res. 1984;18(9):548-554), and the 36 ± 9 mmHg measured by Nanto et al. for 15 subjects in a vasodilated state (Nanto S, Masuyama T, Takano Y. "Determination of coronary zero flow pressure by analysis of The baseline pressure–flow in humans"). Jpn Circ J.2001;65(9):793-796). Apply no-slip boundary conditions to the vessel wall.

[0051] In step 208, the analysis and modeling module 134 of the medical image processing system 100 uses an analysis model to predict the pressure drop across the stenotic lesion. Based on the size of the stenotic lesion, such as the proximal, distal, and minimum lumen areas and length measured from the reconstructed 3D model, and the flow information obtained from CFD simulations, an analytical analysis is performed according to the following method.

[0052] We use an analytical model to predict the pressure drop (ΔP1) across the stenotic lesion. Since the pressure drop is greater with longer stenotic lesions due to the inlet effect, we first use Equation 12 to solve for the dimensionless radius (γ) of the inviscid fluid core to determine whether the inlet effect is included. (12) L and Q represent the length and flow rate of the stenotic lesion, respectively. μ and ρ are the viscosity and density of the blood.

[0053] If γ ≥ 0.05, the length of the stenotic lesion is short, and the inlet effect can be ignored. Therefore, the pressure drop in a vessel with a single stenotic lesion can be calculated using the following formula: (13) A distal A proximal and A stenosis These represent the lumen area at the distal, proximal, and stenotic lesion locations, respectively.

[0054] For long stenotic lesions (γ<0.05), considering the inlet effect, the total pressure drop across a single stenotic lesion is calculated as follows: (14) As described above, this analytical model is derived from an energy-saving perspective, taking into account energy losses from convection and diffusion, as well as energy losses caused by sudden contraction and expansion of the lumen area. In this application, data obtained from simulations are used to indicate the flow rate of congested coronary arteries through coronary artery branches. Combining the dimensions of the stenotic lesion (proximal, distal, and minimum lumen area, as well as length), the pressure drop (ΔP1) across the stenotic lesion is derived from the analytical model. The pressure drop ΔP2 across the normal vessel segment (from the inlet to the segment before the stenotic lesion) is calculated according to the Poiseuille equation.

[0055] In step 210, the FFR determination module 135 of the medical image processing system 100 determines the FFR using the following equation:

[0056] Where P represents patient-specific mean aortic pressure. In this application, the FFR calculated by this method is denoted as FFR. BThis indicates that the FFR is calculated using the modified Bernoulli equations (13) and / or (14) described above.

[0057] In step 212, the threshold comparison module 136 of the medical image processing system 100 compares the calculated FFR with a threshold to classify the stenotic lesion. If the lesion causes FFR... B If the stenosis is ≤0.8, the lesion is classified as ischemic. Conversely, FFR... B Lesions with a value >0.8 are classified as non-ischemic lesions.

[0058] Figure 3 This is a flowchart illustrating a method for assessing the functional significance of coronary artery disease according to an embodiment of the present invention.

[0059] As described above, method 300 relates to computer-aided techniques and methods for analyzing CTA images and assessing the functional significance of coronary artery disease. Method 300 includes CTA image acquisition 310, CTA image processing 320, CFD simulation 330, and analytical analysis 340. After acquiring CTA images via a CTA scanner, image processing 330 is performed to reconstruct a 3D coronary tree. For processing the CTA images, techniques such as Hessian matrix-based filtering 322, eigenvector analysis 324, and image segmentation based on graphic cutting 326 are applied. The dimensions of the stenotic lesions, such as proximal, distal, and minimum luminal area, as well as their length, can be measured using the reconstructed 3D model.

[0060] For the 3D model, steady-flow CFD was performed to obtain the flow rate through the congested coronary arteries of the lesion. A typical CFD simulation 330 includes meshing 332 (discretizing the computational domain into small control volumes), setting morphological and physiological boundary conditions 334, solving the mass conservation equation and momentum conservation equation 336, and post-processing.

[0061] Finally, analytical analysis 340 was performed to calculate FFR using the stenotic lesion size (proximal, distal, and minimum luminal area, as well as length) and the flow rate through the congested coronary artery of the lesion. B .

[0062] As described above, in one embodiment, four steps may be considered: image acquisition 310, image processing 320, CFD simulation 330, and analytical analysis 340. Preferred practices of the present invention can be implemented using many high-level programming languages, including Matlab and C / C++.

[0063] In image acquisition step 310, CT images are acquired in clinical practice. These images conform to the DICOM protocol. Meta-information of all images is examined and recorded.

[0064] In image processing step 320, image processing is performed to isolate arterial structures from the background CT image in order to reconstruct the 3D coronary artery tree. First, a Hessian matrix-based filter 332 is applied to each individual transverse CT image to obtain high-order geometric features. Then, the eigenvectors of the obtained Hessian matrix are analyzed to determine whether the analyzed voxels belong to vascular structures. Finally, a 3D surface and volumetric mesh are generated using a graph-based image segmentation technique 326 and a traveling cubes algorithm.

[0065] On the reconstructed 3D model, the size of the stenotic lesion, such as the proximal, distal, and minimum luminal area and length, can be measured, and the DS of the stenotic lesion can be calculated.

[0066] In CFD simulation step 330, a steady-flow CFD simulation is performed. To obtain flow information of coronary artery lesions in a non-invasive manner, a CFD simulation is conducted. First, in meshing step 332, the 3D coronary artery model is discretized to divide the computational domain into small control volumes. Then, in setup step 334, morphological and physiological boundary conditions are set at the boundaries. Afterwards, the mass conservation equation and momentum conservation equation are solved iteratively. Finally, in step 336, flow information is obtained from the solution.

[0067] In the analytical analysis step 340, based on the dimensions of the stenotic lesion, such as the proximal, distal, and minimum luminal areas and lengths measured from the reconstructed 3D model, and the flow information obtained from the CFD simulation, the obtained FFR is analyzed. B Analytical analysis was performed on hemodynamic parameters.

[0068] In step 350, the FFR value is compared to a threshold. If the lesion causes FFR... B If the stenosis is ≤0.8, the lesion is classified as ischemic. Conversely, if the FFR is ≤0.8, the lesion is classified as ischemic. B Lesions with a value >0.8 are classified as non-ischemic lesions.

[0069] Figure 4a , 4b An example of applying the method described above to medical image data of two patients with ischemic coronary artery disease is shown.

[0070] like Figure 4a As shown, a patient 400 has an ischemic lesion 410 along the middle left anterior descending artery (LAD) 420.

[0071] like Figure 4b As shown, a patient 450 has an ischemic lesion 460 along the first oblique branch 480 of LAD 470.

[0072] Figure 4a and 4b The calculated anatomical parameters of the diameter stenosis (DS) are also shown; the fractional flow reserve (FFR) calculated using a method according to an embodiment of the present invention as described above. B ; Fractional flow reserve (FFR) measured via invasive coronary catheterization. For Figure 4a The patients shown had a DS rate of 69%; FFR B The calculated value is 0.67, and the measured FFR is 0.74. For Figure 4b The patients shown have a DS calculated as 52%; FFR B The calculated value is 0.72, and the measured FFR is 0.79.

[0073] Figure 5a , 5b An example of applying the method described above to medical image data of two patients with non-ischemic coronary artery disease is shown.

[0074] like Figure 5a As shown, a patient 500 has a non-ischemic lesion 510 along the middle LAD 520.

[0075] like Figure 5b As shown, a patient 550 has a non-ischemic lesion 460 along the proximal LAD 570.

[0076] Figure 5a and 5b The calculated anatomical parameters of the diameter stenosis (DS) are also shown; the fractional flow reserve (FFR) calculated using a method according to an embodiment of the present invention as described above. B ; Fractional flow reserve (FFR) measured via invasive coronary catheterization. For Figure 5a The patients shown had a DS rate of 61%; FFR B The calculated value is 0.89, and the measured FFR is 0.83. For Figure 5b The patients shown have a DS calculated as 51%; FFR B The calculated value is 0.97, and the FFR measurement is 0.97.

[0077] like Figure 4a , 4b As shown in 5a and 5b, all FFRs obtained non-invasively by the present invention B Hemodynamic parameters were close to those measured by invasive angiography for FFR. Two patients with non-ischemic lesions had DS values ​​greater than 50%, suggesting that DS may lead to false positive predictions. B Hemodynamic parameters are crucial for non-invasively assessing the functional significance of coronary artery disease.

[0078] To validate our invention, we conducted a pilot study; 21 patients were recruited and subsequently underwent CTA and invasive angiography for the diagnosis of CAD. FFR was measured in a total of 32 vessels.

[0079] Figure 6a This shows the measured FFR and the calculated FFR. B A graph showing the correlation between them. For example... Figure 6a As shown, the calculated FFR B It showed a good correlation with FFR (R=0.825, p<0.001).

[0080] Figure 6b It is based on the FFR measured and calculated for each vessel. B The Bland-Altman plot. (See example.) Figure 6b As shown, FFR and FFR B The differences between them are negligible.

[0081] Figure 7 It is FFR B Area under the curve (AUC) curves for each vessel performance at ≤0.8. FFR at each vessel level. B The area under the receiver working characteristic (AUC) is 0.968.

[0082] FFR on the basis of each vessel B The diagnostic accuracy, positive predictive value (PPV), negative predictive value (NPV), sensitivity, and specificity were 87.5%, 80.0%, 90.9%, 80.0%, and 90.9%, respectively. In summary, FFR... B It is a promising indicator for diagnosing the hemodynamic significance of coronary artery stenosis.

[0083] While exemplary embodiments have been described in the foregoing description, those skilled in the art will understand that many variations may be made to the embodiments within the scope of the invention as defined by the appended claims.

Claims

1. A computer-readable medium carrying processor-executable instructions, which, when executed on a processor, cause the processor to perform a medical image processing method, the method being configured to determine, based on medical image data, a fractional flow reserve through a stenotic lesion in a coronary artery, the medical image data comprising a set of images of a coronary artery region of a patient, the coronary artery region including the stenotic lesion, the method comprising: A three-dimensional coronary artery tree model of the patient was reconstructed based on the medical image data; The size of the stenosis is determined based on the patient's three-dimensional coronary tree model, wherein the size of the stenosis includes the length of the stenosis; The blood flow in the patient's three-dimensional coronary tree model was simulated to determine the model flow rate; Based on the model flow rate, and depending on the size of the stenotic lesion, the analysis model is used to predict the model pressure drop across the stenotic lesion, including determining the radius of the non-viscous fluid core based on the length of the stenotic lesion and the model flow rate, comparing the radius of the non-viscous fluid core with a radius threshold to determine whether an inlet effect is included in the analysis model, wherein if the radius of the non-viscous fluid core is greater than or equal to the radius threshold, the inlet effect is ignored in the analysis model; and Based on the model pressure drop, the fractional blood flow reserve through the stenotic lesion is determined.

2. The computer-readable medium according to claim 1, wherein, The blood flow in the patient's three-dimensional coronary tree model is simulated to determine the model flow rate, including simulating steady-state blood flow.

3. The computer-readable medium according to claim 2, wherein, Simulating steady-state blood flow involves iteratively determining steady-state fluid resistance and pressure values.

4. The computer-readable medium according to claim 3, wherein, Determining steady-state fluid resistance and pressure values ​​iteratively involves updating the steady-state fluid resistance and pressure values ​​iteratively using an under-relaxation scheme.

5. The computer-readable medium according to any one of claims 3 to 4, wherein, The patient's three-dimensional coronary tree model includes at least one inlet and multiple outlets. The method further includes estimating the total inflow into the patient's three-dimensional coronary tree model based on the medical image data, and wherein determining the steady-state fluid resistance and pressure values ​​iteratively includes updating the steady-state fluid resistance and pressure values ​​iteratively until the total outflow from the multiple outlets matches the total inflow.

6. The computer-readable medium according to any one of claims 1-4, wherein, Simulate blood flow in the patient's three-dimensional coronary tree model to determine model flow, including simulating congested blood flow.

7. The computer-readable medium according to any one of claims 1-4, wherein, The analytical model is an improved version based on the Bernoulli equation.

8. A computer-readable medium carrying processor-executable instructions, which, when executed on a processor, cause the processor to perform a method for assessing the functional significance of coronary artery stenosis based on medical image data, the method comprising: A three-dimensional coronary artery tree model of the patient was reconstructed based on the medical image data; The size of the stenosis is determined based on the patient's three-dimensional coronary tree model, wherein the size of the stenosis includes the length of the stenosis; The blood flow in the patient's three-dimensional coronary tree model was simulated to determine the model flow rate; Based on the model flow rate, and depending on the size of the stenotic lesion, the analysis model is used to predict the model pressure drop across the stenotic lesion, including determining the radius of the non-viscous fluid core based on the length of the stenotic lesion and the model flow rate, comparing the radius of the non-viscous fluid core with a radius threshold to determine whether an inlet effect is included in the analysis model, wherein if the radius of the non-viscous fluid core is greater than or equal to the radius threshold, the inlet effect is ignored in the analysis model; and Based on the model pressure drop, the fractional flow reserve (FRR) through the stenotic lesion is determined; the FRR through the coronary stenotic lesion is compared with a threshold to assess the functional significance of the coronary stenotic lesion.

9. The computer-readable medium according to claim 8, wherein, The functional significance of a coronary artery stenosis is assessed by comparing its fractional flow reserve (FRR) with a threshold, including: classifying the stenosis as ischemic if the FRR is less than or equal to the threshold; and classifying the stenosis as non-ischemic if the FRR is greater than the threshold.

10. The computer-readable medium according to any one of claims 8 to 9, wherein, The threshold is 0.

80.

11. A medical image processing system for determining a fractional flow reserve (FVR) of a stenotic lesion through a coronary artery based on medical image data, the medical image data comprising a set of images of a patient's coronary artery region including the stenotic lesion, the system comprising: The computer processor and data storage device include a three-dimensional model reconstruction module comprising non-transient instructions, a stenosis lesion size determination module, a blood flow simulation module, an analysis modeling module, and a fractional blood flow reserve determination module. The non-transient instruction can be operated by the processor to: A three-dimensional coronary artery tree model of the patient was reconstructed based on the medical image data; The size of the stenosis is determined based on the patient's three-dimensional coronary tree model, wherein the size of the stenosis includes the length of the stenosis; The blood flow in the patient's three-dimensional coronary tree model was simulated to determine the model flow rate; Based on the model flow rate, and depending on the size of the stenotic lesion, the analysis model is used to predict the model pressure drop across the stenotic lesion, including determining the radius of the non-viscous fluid core based on the length of the stenotic lesion and the model flow rate, comparing the radius of the non-viscous fluid core with a radius threshold to determine whether an inlet effect is included in the analysis model, wherein if the radius of the non-viscous fluid core is greater than or equal to the radius threshold, the inlet effect is ignored in the analysis model. Based on the model pressure drop, the fractional blood flow reserve through the stenotic lesion is determined.

12. The system according to claim 11, wherein, The blood flow simulation module includes non-transient instructions that can be operated by the processor to: simulate blood flow in the patient's three-dimensional coronary tree model by simulating steady-state blood flow to determine the model flow rate.

13. The system according to claim 12, wherein, The blood flow simulation module includes non-transient instructions that can be operated by the processor to simulate steady-state blood flow by iteratively determining steady-state fluid resistance and pressure values.

14. The system according to claim 13, wherein, The blood flow simulation module includes non-transient instructions that can be operated by the processor to: iteratively update the steady-state fluid resistance and pressure values ​​using an under-relaxation scheme, and iteratively determine the steady-state fluid resistance and pressure values.

15. The system according to any one of claims 13 to 14, wherein, The blood flow simulation module includes non-transient instructions operable by the processor to: estimate the total inflow into the patient's three-dimensional coronary tree model based on the medical image data, wherein determining the steady-state fluid resistance and pressure values ​​iteratively includes updating the steady-state fluid resistance and pressure values ​​iteratively until the total outflow from all outlets of the patient's three-dimensional coronary tree model matches the total inflow.

16. The system according to claim 15, wherein, The blood flow simulation module includes non-transient instructions that can be operated by the processor to: simulate blood flow in the patient's three-dimensional coronary tree model by simulating congested blood flow to determine the model flow rate.

17. The system according to any one of claims 11 to 14, wherein, The analytical model is an improved version based on the Bernoulli equation.

18. The system according to any one of claims 11 to 14, wherein the data storage device further comprises a threshold comparison module, the threshold comparison module comprising non-transient instructions operable by the processor to: The fractional flow reserve of the stenotic lesions through the coronary arteries is compared with a threshold to assess the functional significance of the stenotic lesions.

19. The system according to any one of claims 11 to 14, wherein, The threshold comparison module further includes non-transient instructions, which can be operated by the processor to: classify the stenotic lesion as ischemic when the fractional flow reserve of the stenotic lesion passing through the coronary artery is less than or equal to the threshold; and classify the stenotic lesion as non-ischemic when the fractional flow reserve of the stenotic lesion passing through the coronary artery is greater than the threshold.

20. The system according to claim 19, wherein, The threshold is 0.8.

Images