Determination of values of blood flow parameters

Abstract

A computer-implemented method for determining a value of a blood flow parameter for a blood vessel is provided, the method including calculating from spectral attenuation data, calculating the value of the blood flow parameter for the blood vessel using a plurality of different techniques, and providing the value of the blood flow parameter for the blood vessel based on the calculated value of the blood flow parameter.

Description

[Technical Field]

[0001] SUMMARY A computer-implemented method, a computer program product, and a system are disclosed that relate to determining values of blood flow parameters for blood vessels. [Background technology]

[0002] Various clinical studies involve the assessment of blood flow in the vascular system. For example, investigations of coronary artery disease "CAD" often involve the assessment of blood flow to assess the amount of blood supplied to regions of the heart. In this regard, various blood flow parameters have been investigated, including blood flow velocity, blood pressure, fractional flow reserve "FFR," instantaneous wave-free ratio "iFR," coronary flow reserve "CFR," Thrombolysis in Myocardial Infarction "TIMI" flow grade, index of microvascular resistance "IMR," and index of ultramicrovascular resistance "HMR."

[0003] Historically, values of the above-mentioned blood flow parameters have been measured using invasive devices such as pressure wires. For example, the value of fractional flow reserve (FFR) is often determined in CAD studies to assess the effect of stenosis on the delivery of oxygen to the myocardium. FFR is a ratio P d / P a where P d represents the distal pressure at the location distal to the stenosis, and P a represents the pressure proximal to the stenosis. FFR is typically calculated as P d and P a FFR values above 0.8 are typically considered clinically insignificant, while values below 0.8 are typically considered to represent increased clinical significance. Historically, these pressure values have been determined by placing invasive devices, such as pressure wires, at respective locations within the vasculature.

[0004] More recently, angiographic techniques have been developed to determine blood flow parameters, including FFR. According to fluid flow theory, pressure changes are, among other things, effects related to changes in fluid velocity. In the case of FFR, angiographic images of injected contrast media are analyzed to determine blood flow velocity. FFR is then calculated by using a hemodynamic model to estimate intravascular pressure values from blood flow velocity. Therefore, FFR, as well as other blood flow parameters, can be determined by angiography.

[0005] A known angiographic technique for measuring blood flow velocity is to sample the intensity of injected contrast in computed tomography (CT) angiographic image data over time and apply a mathematical model to the sampled data. In this regard, a technique for sampling reconstructed CT images to determine blood flow velocity is disclosed in Barfett, JJ et al., "Intra-vascular blood velocity and volumetric flow rate calculated from dynamic 4D CT angiography using a time-of-flight technique," Int J Cardiovasc Imaging, 2014, 30:1383-1392. A technique for sampling raw, or "projection," CT data to determine blood flow velocity is disclosed in Prevrhal, S. et al., "CT Angiographic Measurement of Vascular Blood Flow Velocity by Using Projection Data," Radiology, Vol. 261: Number 3 - December 2011, pages 923-929.

[0006] Some known angiography techniques for determining values of blood flow parameters use computational fluid dynamics (CFD) models. In this regard, a technique for determining FFR values is disclosed in the article by Taylor, "Computational Fluid Dynamics Applied to Cardiac Computed Tomography for Noninvasive Quantification of Fractional Flow Reserve: Scientific Basis," JACC, Vol. 61, No. 22, 2013. Another publication describing an angiography technique for measuring blood flow parameters using complex fluid dynamics (CFD) models is disclosed in Koo, BK et al., "Diagnosis of Ischemia-Causing Coronary Stenoses by Noninvasive Fractional Flow Reserve Computed From Coronary Computed Tomographic Angiograms: Results From the Prospective Multicenter DISCOVER-FLOW (Diagnosis of Ischemia-Causing Stenoses Obtained Via Noninvasive Fractional Flow Reserve) Study," JACC, Vol. 58, No. 19, 2011. Another technique is disclosed in Kim, HJ, et al., "Patient-Specific Modeling of Blood Flow and Pressure in Human Coronary Arteries," Annals of Biomedical Engineering, Vol. 38, No. 10, October 2010, pp. 3195-3209.

[0007] Another angiographic technique for measuring blood flow parameters involves the use of lumped parameter models. Lumped parameter models can be used to represent blood flow within a vascular region as an electrical circuit, where blood volumetric flow is represented as a current, blood pressure is represented as a voltage, and blood volume is represented by an electric charge. In the lumped parameter model, resistance to blood flow is represented by a linear or nonlinear electrical resistance, vessel wall compliance is represented by a capacitance, and blood inertia is represented by an inductance. Machine learning-based approaches for setting the values of linear and nonlinear resistors in static lumped parameter models based on angiographic geometric measurements of blood vessels in a vascular region are disclosed in the literature by Nickisch, H. et al., "Learning Patient-Specific Lumped Models for Interactive Coronary Blood Flow Simulations," MICCAI 2015, Part II, LNCS 9350, pp. 433-441, 2015, and WO 2016 / 001017. The values of blood flow and blood pressure in the model are then solved for to determine the value of the vascular FFR. Compared to computational fluid dynamics (CFD) models, the use of lumped parameter models to calculate the values of blood flow parameters offers the advantage of reducing both the computational load and model complexity. Summary of the Invention [Problem to be solved by the invention]

[0008] Generally, the basis for providing accurate values of blood flow parameters from angiographic image data is accurate anatomical delineation or segmentation of the vessel lumen. This image processing task may be performed manually, which is time-consuming and risks introducing user bias, or it may be performed automatically, in which case it depends on the technique used to segment the angiographic data. As a result, there may be variability in the values of angiographically derived blood flow parameters.

[0009] Therefore, there remains room for improvement in the reliability of angiographic measurements of blood flow parameters such as blood flow velocity, blood pressure, FFR, iFR, CFR, TIMI flow grade, IMR, and HMR. [Means for solving the problem]

[0010] According to one aspect of the present disclosure, there is provided a computer-implemented method for determining a value of a blood flow parameter for a blood vessel, the method comprising: receiving spectral attenuation data representative of injected contrast agent in a vascular region including the blood vessel, the spectral attenuation data being included at a plurality of different energy intervals DE 1..m defining x-ray attenuation within a blood vessel region in calculating a value of a blood flow parameter for the blood vessel from the spectral attenuation data using a number of different techniques; providing a value of a blood flow parameter for the blood vessel based on the calculated value of the blood flow parameter; It has.

[0011] Compared to the use of conventional X-ray attenuation data, the use of spectral attenuation data in the above-described method for calculating blood flow parameter values facilitates improved separation between attenuation resulting from the injected contrast agent and attenuation resulting from other media that may be present in the vascular region. Attenuation resulting from other media, such as calcium lesions within blood vessels, could otherwise be interpreted as contrast agents, thereby reducing the accuracy of the calculated blood flow parameters. The use of spectral attenuation data also helps reduce image artifacts resulting from media other than the contrast agent, which could similarly be mistaken for contrast agents, thereby reducing the accuracy of the calculated blood flow parameters. For example, calcium lesions can also cause "calcium blooming" in conventional CT data, resulting in image intensity values that are at risk of being misinterpreted as attenuation resulting from the contrast agent. Therefore, the use of spectral attenuation data facilitates more accurate calculation of blood flow parameter values. However, blood flow parameter values calculated using spectral attenuation data can still be subject to variation. For example, the choice of energy interval used to reconstruct the images from which blood flow parameter values are calculated can also affect the calculated blood flow parameter values. The values of model parameters, such as boundary conditions of the blood flow calculation algorithm used to calculate the blood flow parameters, may also affect the values. Similarly, the selection of the segmentation algorithm used to segment the reconstructed image data and to calculate the values of the blood flow parameters, and the selection of the blood flow calculation algorithm used to calculate the values of the blood flow parameters may also affect the values. The method calculates the values of the blood flow parameters for the blood vessels using multiple different techniques and provides the values of the blood flow parameters based on the calculated values, thereby improving the reliability of the provided values of the blood flow parameters.

[0012] Further aspects, features, and advantages of the present disclosure will become apparent from the following description of exemplary embodiments that proceeds with reference to the accompanying drawings. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a flowchart illustrating an example of a computer-implemented method for determining a value of a blood flow parameter for a blood vessel, according to some aspects of the present disclosure. [Figure 2] 2 is a schematic diagram illustrating an example of a system 200 for determining a value of a blood flow parameter for a blood vessel, according to some embodiments of the present disclosure. FIG. [Figure 3] 1 is a schematic diagram illustrating a heart including an example of a blood vessel 110, according to some embodiments of the present disclosure. [Figure 4] FIG. 1 is a schematic diagram illustrating an example of a computer-implemented method for determining a value of a blood flow parameter for a blood vessel, according to some aspects of the present disclosure. [Figure 5] 1 is a graph showing the dependence of the mass attenuation coefficient on X-ray energy for two materials, iodine and water. [Figure 6] 15 is a schematic diagram illustrating an example of a lumped parameter model 140 including a set of model parameters 1501, according to some embodiments of the present invention. [Figure 7] FIG. 10 is a schematic diagram illustrating an example of an operation S140 for generating a contrast-adjusted reconstructed image according to some aspects of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0014] Examples of the present disclosure are provided with reference to the following description and drawings. In this description, for purposes of explanation, many specific details of several examples are set forth. Reference herein to an "example," "implementation," or similar language means that a feature, structure, or characteristic described in connection with an example is included in at least one of the examples. It should also be appreciated that features described in connection with one example may also be used in other examples, and that for purposes of brevity, not all features are necessarily replicated in each example. For example, features described in connection with a computer-implemented method may be implemented in a corresponding manner in a computer program product and in a system.

[0015] In the following description, reference is made to a method for determining a value of a blood flow parameter for a blood vessel. The blood vessel is located within a vascular region. In some examples, the vascular region is the heart. For example, in some examples, the blood vessel is a coronary artery. However, it should be understood that the blood vessel may generally be an artery or a vein. More generally, the method may be used to measure blood flow parameters for blood vessels in other parts of the body other than the heart. For example, the region of interest may alternatively be a blood vessel located in a peripheral region, such as a leg, arm, or brain.

[0016] Reference is also made herein to an example in which the calculated blood flow parameter is a vascular FFR value, however, it should be appreciated that this serves as an example only, and that the methods and systems disclosed herein may alternatively be used to calculate values for other blood flow parameters, such as, but not limited to, blood flow velocity, blood pressure, blood flow transit time, volumetric blood flow values, iFR values, CFR values, TIMI blood flow grades, IMR and HMR values, hypertensive stenosis resistance "HSR" values, zero flow pressure "ZFP" values, and instantaneous hypertensive diastolic velocity-pressure slope "IVDS" values.

[0017] It should be noted that the computer-implemented methods disclosed herein may be provided as a non-transitory computer-readable storage medium storing computer-readable instructions that, when executed by at least one processor, cause the at least one processor to perform the method. In other words, the computer-implemented methods may be implemented in a computer program product. The computer program product may be provided by dedicated hardware or hardware capable of executing software in association with appropriate software. When provided by a processor, the functionality of the method features may be provided by a single dedicated processor, by a single shared processor, or by multiple individual processors, some of which may be shared. One or more functions of the method features may be provided by a processor shared within a networked processing architecture, such as, for example, a client / server architecture, a peer-to-peer architecture, the Internet, or the cloud.

[0018] Explicit use of the terms "processor" or "controller" should not be construed as exclusively referring to hardware capable of executing software, but can implicitly include, but is not limited to, digital signal processor "DSP" hardware, read-only memory "ROM" for storing software, random access memory "RAM," non-volatile storage devices, and the like. Furthermore, examples of the present disclosure can take the form of a computer-usable storage medium, or a computer program product accessible from a computer-readable storage medium, the computer program product providing program code for use by or in connection with a computer or any instruction execution system. For purposes of this description, a computer-usable storage medium or computer-readable storage medium can be any apparatus that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system or device or propagation medium. Examples of computer-readable media include semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory "RAM," a read-only memory "ROM," a rigid magnetic disk, and an optical disk, current examples of which include compact disk-read only memory "CD-ROM," compact disk-read / write "CD-R / W," Blu-Ray™, and DVD.

[0019] As discussed above, there remains room for improvement in the reliability of angiographic measurements of blood flow parameters such as blood flow velocity, blood pressure, FFR, iFR, CFR, TIMI flow grade, IMR, and HMR.

[0020]

[0023] Figure 1 is a flowchart illustrating an example of a computer-implemented method for determining a value of a blood flow parameter for a blood vessel according to some aspects of the present disclosure. Figure 2 is a schematic diagram illustrating an example of a system 200 for determining a value of a blood flow parameter for a blood vessel according to some aspects of the present disclosure. Operations described in connection with the method illustrated in Figure 1 may be performed by the system 200 illustrated in Figure 2, and vice versa. Referring to Figure 1, the computer-implemented method for determining a value of a blood flow parameter for a blood vessel 110 includes: At S110, receiving spectral attenuation data 120 representing an injected contrast agent in a vascular region including a blood vessel 110, the spectral attenuation data being at a plurality of different energy intervals DE 1..m defining x-ray attenuation within a blood vessel region in At S120, calculating values of blood flow parameters of the blood vessel 110 from the spectral attenuation data 120 using a number of different techniques; In S130, providing a value of the blood flow parameter of the blood vessel 110 based on the calculated value of the blood flow parameter; Includes.

[0021] Compared to the use of conventional X-ray attenuation data, the use of spectral attenuation data in the above-described method for calculating blood flow parameter values facilitates improved separation between attenuation resulting from the injected contrast agent and attenuation resulting from other media that may be present in the vascular region. Attenuation resulting from other media, such as calcium lesions within blood vessels, could otherwise be interpreted as contrast agents, thereby reducing the accuracy of the calculated blood flow parameters. The use of spectral attenuation data also helps reduce image artifacts resulting from media other than the contrast agent, which could similarly be mistaken for contrast agents, thereby reducing the accuracy of the calculated blood flow parameters. For example, calcium lesions can also cause "calcium blooming" in conventional CT data, resulting in image intensity values that are at risk of being misinterpreted as attenuation resulting from the contrast agent. Therefore, the use of spectral attenuation data facilitates more accurate calculation of blood flow parameter values. However, blood flow parameters calculated using spectral attenuation data can still be subject to variation. For example, the selection of the energy interval for reconstructing the image and the energy interval from which the blood flow parameter values are calculated can also affect the calculated blood flow parameter values. The values of model parameters, such as boundary conditions of the blood flow calculation algorithm used to calculate the blood flow parameters, may also affect the values. Similarly, the selection of the segmentation algorithm used to segment the reconstructed image data and to calculate the values of the blood flow parameters, and the selection of the blood flow calculation algorithm used to calculate the values of the blood flow parameters may also affect the values. The method calculates the values of the blood flow parameters for the blood vessels using multiple different techniques and provides the values of the blood flow parameters based on the calculated values, thereby improving the reliability of the provided values of the blood flow parameters.

[0022] The above method will also be described with reference to FIG. 3, which is a schematic diagram illustrating a heart including an example of a blood vessel 110, according to some embodiments of the present disclosure. The heart shown in FIG. 3 is labeled with the left coronary artery, LCA, right coronary artery, RCA, as well as the LCA ostium and RCA ostium, which respectively define the openings of these arteries in the aorta. The operations described above with reference to FIG. 1 can be performed to determine values of blood flow parameters in cardiac vascular regions. For example, they can be used to determine values of blood flow parameters in the cardiac blood vessel 110 shown in FIG. 3. The blood flow parameter may be, for example, an FFR value. FFR can be calculated using the following formula: FFR=Pd / Pa Equation 1 Using the left coronary artery distal position Pos d can be calculated.

[0023] The FFR value is calculated using Equation 1 at the distal position Pos d Blood pressure P d and the proximal position Pos p Blood pressure P a , which in this example is taken as the ostium of the left coronary artery.

[0024] 1, in act S110, spectral attenuation data 120 is received. The spectral attenuation data 120 represents a contrast agent injected into a vascular region including a blood vessel 110 and measures x-ray attenuation within the vascular region at a plurality of different energy intervals DE 1..m Defined as:

[0025] The spectral attenuation data 120 received in act S110 may generally be projection X-ray data, i.e., 2D data that can be used to generate a 2D image, or volumetric X-ray data, i.e., 3D data that can be used to generate a 3D image. In the latter case, the data may be raw data, i.e., data that has not yet been reconstructed into a 3D image, or image data, i.e., data that represents a reconstructed 3D image.

[0026] The spectral attenuation data 120 may be received by one or more processors 210 shown in FIG. 2. The spectral attenuation data 120 may be received via any form of data communication, including wired communication, optical communication, and wireless communication. As some examples, when wired or optical communication is used, the communication may be via signals transmitted over electrical or optical cables, and when wireless communication is used, the communication may be via RF or optical signals. In general, the one or more processors 210 may receive the spectral attenuation data 120 from a spectral X-ray projection imaging system, a spectral CT imaging system, or from another source, such as a computer-readable storage medium, the internet, or the cloud.

[0027] 1 , the spectral attenuation data 120 received in act S110 represents an injected contrast agent within a vascular region that includes blood vessel 110. As discussed above with reference to FIG. 3 , the spectral attenuation data 120 may represent, for example, a contrast agent injected into a cardiac region. In this regard, the spectral attenuation data received in act S110 may be generated after injection of a contrast agent into the vascular system. The contrast agent may include a substance such as iodine, or a lanthanide such as gadolinium, or another substance that provides visibility of the blood flow into which the contrast agent is injected.

[0028] Continuing to refer to FIG. 1, the spectral attenuation data 120 received in act S110 may generally be generated by a spectral X-ray imaging system, i.e., a spectral CT imaging system or a spectral X-ray projection imaging system.

[0029] Spectral CT imaging systems generate spectral attenuation data while rotating or stepping an x-ray source-detector arrangement around an imaging region. Examples of spectral CT imaging systems include cone-beam spectral CT imaging systems, photon-counting spectral CT imaging systems, dark-field spectral CT imaging systems, and phase-contrast spectral CT imaging systems. As an example, the spectral attenuation data 120 may be generated by a Spectral CT 7500 sold by Philips Healthcare, Best, The Netherlands.

[0030] An example of a spectral CT imaging system 220 that may be used to generate the spectral attenuation data 120 received in operation S110 is shown in FIG. 2. The spectral CT imaging system 220 shown in FIG. 2 includes an X-ray source 230 and an X-ray detector 240. The X-ray source 230 and the X-ray detector 240 are mechanically coupled to a gantry (not shown in FIG. 2). During operation, the X-ray source 230 and the X-ray detector 240 are rotated by the gantry about a rotation axis 250 while acquiring spectral attenuation data 120 that defines X-ray attenuation in a region of interest located in an imaging region of the imaging system 220. The spectral attenuation data, or more specifically, volumetric X-ray data, i.e., 3D data, obtained from multiple rotation angles about the rotation axis 250 can then be reconstructed into a volumetric image.

[0031] As described above, the spectral attenuation data received in operation S110 may alternatively be generated by a spectral X-ray projection imaging system. A spectral X-ray projection imaging system typically includes a support arm, such as a so-called “C-arm,” that supports an X-ray source and an X-ray detector. Alternatively, a spectral X-ray projection imaging system may include a support arm having a different shape from this example, such as an O-arm. A spectral X-ray projection imaging system typically generates projection data using a support arm that is held in a stationary position relative to the imaging region during image data acquisition. Therefore, a spectral X-ray projection imaging system may be used to acquire projection X-ray data, i.e., 2D data. However, a spectral X-ray projection imaging system may also acquire volumetric X-ray data, i.e., 3D data, by acquiring projection X-ray data while rotating the support arm around a rotation axis. Then, using an image reconstruction method, the projection X-ray data obtained from multiple rotation angles around the rotation axis can be reconstructed into a volumetric image in a manner similar to that of a spectral CT imaging system. Therefore, the spectral attenuation data received in operation S110 may alternatively be generated by a spectral X-ray projection imaging system.

[0032] The spectral attenuation 120 received at S110 is divided into a plurality of energy intervals DE 1..m Generally, there may be two or more energy intervals, i.e., m is an integer and m≧2, and a plurality of different energy intervals DE 1..m The ability to generate data defining X-ray attenuation data distinguishes spectral X-ray imaging systems from conventional X-ray imaging systems. By processing data from multiple different energy intervals, it is possible to distinguish between media that have similar X-ray attenuation values when measured within a single energy interval and that are indistinguishable from attenuation data generated by conventional X-ray imaging systems. In this regard, a variety of different configurations of the spectral X-ray imaging system may be used to generate the spectral attenuation data received in operation S110, some of which are described with reference to FIG. 2.

[0033] Referring to the exemplary spectral CT imaging system 220 shown in FIG. 2, in general, the X-ray source 230 may be provided by multiple monochromatic sources or by one or more polychromatic sources, and the X-ray detector 240 may be a common detector for detecting multiple different X-ray energy intervals, or each detector may be provided by a different X-ray energy interval DE. 1..m The direct conversion material may include a plurality of detectors that detect X-rays having energies within different X-ray energy intervals, a multi-layer detector in which X-rays having energies within different X-ray energy intervals are detected by corresponding layers, or a photon-counting detector that classifies detected X-ray photons into one of a plurality of energy intervals based on their individual energies. In a photon-counting detector, the associated energy interval may be determined for each received X-ray photon by detecting the pulse height induced by electron-hole pairs generated in response to absorption of the X-ray photon in the direct conversion material.

[0034] The various configurations of the X-ray source and X-ray detector described above can be used to generate a number of different energy intervals DE 1..m The X-ray detector 240 may be used in a spectral X-ray imaging system to generate spectral attenuation data defining the X-ray attenuation in a given X-ray energy interval. Generally, by switching the X-ray tube potential of a single X-ray source 230 in time, i.e., by "rapid kVp switching," or by switching or filtering the X-ray emission from multiple X-ray sources 230 in time, discrimination between different X-ray energy intervals can be provided in the X-ray source 230. In such a setup, a common X-ray detector 240 can be used to detect X-rays over a number of different energy intervals, and the energy interval DE 1..m Alternatively, the X-ray detector 240 may use a multi-layer detector or a photon-counting detector to generate attenuation data for each of the X-ray energy intervals DE 1..m Such a detector may distinguish between X-ray energy intervals DE 1..mSince X-rays from the X-ray source 230 can be detected almost simultaneously, there is no need for time switching in the X-ray source 230. Therefore, a multi-layer X-ray detector 240 or a count-and-count X-ray detector 240 can be used in combination with the polychromatic X-ray source 230 to detect X-rays from multiple different energy intervals DE 1..m The spectral attenuation data can be generated by

[0035] Other combinations of the X-ray source 230 and X-ray detector 240 described above may alternatively be used to provide a plurality of different energy intervals DE 1..m For example, in another setup, the need to sequentially switch between different x-ray sources 230 emitting x-rays at different energy intervals can be avoided by mounting the x-ray source-detector pairs on the gantry at rotationally offset positions about the axis of rotation 250. In this setup, each source-detector pair operates independently and detects x-rays at various energy intervals DE 1..m The separation between the spectral attenuation data of the various energy intervals DE is facilitated by the rotational offset of the source-detector pair. In this setup, to reduce the effects of X-ray scattering, an energy selection filter is applied to the X-ray detector 240 to separate the spectral attenuation data of the various energy intervals DE. 1..m Improved separation between the spectral attenuation data of

[0036] In yet another configuration, multiple different energy intervals DE 1..mSpectral attenuation data for energy intervals DE may be provided by blocking the X-ray beam in a conventional X-ray imaging device with one or more spectral filters. For example, a single filter can be mechanically moved in and out of the X-ray beam generated by the polychromatic X-ray source 230 to temporally control the spectrum of X-rays detected by the X-ray detector 240. The filter can, for example, transmit only a portion of the spectrum of X-rays emitted by the polychromatic X-ray source 230, thereby providing spectral attenuation data for a first energy interval when the filter blocks the beam. When the filter is removed from the beam, spectral attenuation data is provided for a second energy interval, e.g., the complete spectrum emitted by the polychromatic X-ray source. This allows for the time-series analysis of energy intervals DE. 1..m Multiple filters, each with a different X-ray transmission spectrum, can be sequentially switched in and out of the X-ray beam to increase the number of energy intervals.

[0037] Returning to FIG. 1 , in act S120, a number of different techniques are used to calculate the value of a blood flow parameter of the blood vessel 110. As some examples, the blood flow parameter may be a blood flow velocity, a blood pressure, an FFR value, or an iFR value. The different techniques used in act S120 may include, for example, using different techniques for reconstructing the received spectral attenuation data, different segmentation algorithms for segmenting the image reconstructed from the spectral attenuation data, using different blood flow calculation algorithms for calculating the value of the blood flow parameter, and using different sets of model parameters for the blood flow calculation algorithm. Further details of these techniques are provided below. Combinations of these and other techniques may also be used to provide different values for the blood flow parameter.

[0038] Continuing to refer to FIG. 1 , in act S130, a value of the blood flow parameter for the blood vessel 110 is provided based on the calculated values of the blood flow parameter. In one example, act S130 can include providing the value of the blood flow parameter for the blood vessel 110 as a weighted average of the individual calculated values. By providing the value of the blood flow parameter as a weighted average, reliability of the measurement is improved. In another example, a weighted average is calculated and outliers are removed before calculating the weighted average. In this example, the act of providing the value of the blood flow parameter for the blood vessel 110 based on the calculated values of the blood flow parameter in S130 can include: analyzing the calculated values of the blood flow parameters to identify outliers; providing a value of the blood flow parameter of the blood vessel 110 as a weighted average of the calculated values that are not identified as outliers; It has.

[0039] Excluding outliers from the weighted average calculated in this example has the effect of reducing the influence of potentially inaccurate values of the blood flow parameter on the provided value of the blood flow parameter, thus reducing the amount of variability in the calculated value of the blood flow parameter and, consequently, improving the reliability of the provided value.

[0040] In this example, analyzing the calculated values of the blood flow parameter to identify outliers may include performing a statistical analysis on the calculated values of the blood flow parameter. For example, a mean of the calculated values may be determined, and values that deviate from the mean by more than a predetermined amount may be identified as outliers. As another example, a standard deviation may be determined for the calculated values, and values that fall outside, for example, two standard deviations from the mean may be identified as outliers. As another example, a RANSAC method may be used to identify and remove outliers from a weighted mean.

[0041] The calculated values of the blood flow parameters of the blood vessel 110 provided in operation S130 are determined based on the values of the blood flow parameters and are calculated using different techniques, thereby improving the reliability of the provided values of the blood flow parameters for the vessel.

[0042] The computer-implemented methods described above may include one or more additional operations.

[0043] In one example, the method described with reference to Figure 1 may also include outputting a value representing the change in the calculated value of the blood flow parameter. Outputting the value representing the variability provides an indication of the level of confidence in the result, with a small amount of variability indicating a high level of confidence, and vice versa. The measure of variability may be output numerically or graphically, for example, via error bars, etc.

[0044] In another example, the operations of calculating (S120) a value of a blood flow parameter of the blood vessel 110 and providing (S130) a value of the blood flow parameter of the blood vessel 110 are performed at multiple locations along the blood vessel.

[0045] As an example, FFR values of a blood vessel may be provided at multiple locations along the vessel and then used by a physician to assess the severity of a stenosis along the vessel.

[0046] As mentioned above, the act of calculating the value of the blood flow parameter of the blood vessel 110 using a number of different techniques in S120 can be performed in a variety of ways.

[0047] In one embodiment, in S120, the step of calculating the motion comprises calculating a plurality of corresponding reconstructed images 130 1..n and calculating a value of the blood flow parameter from each reconstructed image.

[0048] This example will be described with reference to FIG. 4, which is a schematic diagram illustrating an example of a computer-implemented method for determining values of a blood flow parameter for a blood vessel, according to some aspects of the present disclosure. The spectral attenuation data 120 received in operation S110 described above is shown on the left side of FIG. 4. The spectral attenuation data 120 in this example is volumetric X-ray data generated by a spectral CT imaging system. In operation S120, the spectral attenuation data 120 is reconstructed using a plurality of different image reconstruction methods to generate a plurality of corresponding reconstructed images 130, as shown in FIG. 1..n In this example, the reconstructed image is applying different material decomposition algorithms to the spectral attenuation data 120; and / or Energy interval DE 1..m Reconstructing the spectral attenuation data 120 including different selections of is reconstructed by

[0049] The resulting reconstructed image represents the injected contrast agent within the vascular region. The injected contrast agent may include materials such as iodine or gadolinium. Such materials are often used as contrast agents due to their attenuation at x-ray energies used in diagnostic x-ray imaging systems. By reconstructing an image representing the injected contrast agent from spectral attenuation data, more reliable data regarding the flow of the injected contrast agent can be obtained compared to the use of traditional x-ray attenuation data.

[0050] In addition to attenuation resulting from the contrast agent, the spectral attenuation data may also represent attenuation resulting from one or more background materials, such as fat, water, bone, soft tissue, vascular calcification, air, and metals such as gold, titanium, tungsten, and platinum. Such materials are also often present near blood vessels, and therefore attenuation resulting from these materials may also be represented in the spectral attenuation data. For example, when imaging the cardiovascular system, bone in the form of portions of the vertebrae or ribs is often within the field of view of the spectral CT imaging system. Similarly, fiducial markers, implanted medical devices, and interventional devices are typically formed from metals such as those mentioned above, and attenuation resulting from these materials may also be captured in the spectral attenuation data. Thus, the reconstructed image generated in operation S120 may also represent one or more materials other than the contrast agent, such as:

[0051] An example of a material decomposition technique that can be used in operation S120 to reconstruct an image representing the injected contrast agent from the spectral attenuation data is disclosed in Brendel, B. et al., "Empirical, projection-based basis-component decomposition method," Medical Imaging 2009, Physics of Medical Imaging, edited by Ehsan Samei and Jiang Hsieh, Proc. of SPIE Vol. 7258, 72583Y. Another suitable material decomposition technique is disclosed in a document by Roessl, E. and Proksa, R., "K-edge imaging in X-ray computed tomography using multi-bin photon counting detectors," Phys Med Biol. 2007 Aug 7, 52(15):4679-96. Another suitable material decomposition technique is disclosed in published PCT patent application WO / 2007 / 034359 A2. Another suitable material decomposition technique is disclosed in Silva, AC et al., "Dual-energy (spectral) CT: applications in abdominal imaging", RadioGraphics 2011; 31(4):1031-1046.

[0052] Thus, in act S120, a number of different material decomposition algorithms such as these may be applied to the spectral attenuation data 120 to reconstruct an image representative of the injected contrast agent.

[0053] Generally, the X-ray attenuation spectrum of a material contains contributions from Compton scattering and the photoelectric effect. While the attenuation from Compton scattering is relatively similar for different materials, the attenuation from the photoelectric effect is strongly material-dependent. Both Compton scattering and the photoelectric effect exhibit energy dependence, and this effect is exploited by material decomposition techniques to distinguish between different materials.

[0054] Generally, material decomposition algorithms operate by decomposing the attenuation spectrum of an absorbing medium into contributions from a set of assumed "basis" materials. The energy-dependent X-ray attenuation of the assumed basis materials is typically modeled as a combination of absorption from Compton scattering and the photoelectric effect. Some materials also have k-edge (k-edge) energies that lie within the energy range used by diagnostic X-ray imaging systems, and this effect can also be exploited to distinguish between different materials. The spectral decomposition algorithm then seeks to estimate the amount of each basis material required to produce the measured X-ray attenuation at two or more energy intervals. Water and iodine are examples of basis materials that are often separated in clinical practice using so-called two-material decomposition algorithms. Non-fatty soft tissue, fat, and iodine are examples of basis materials that are often separated in clinical practice using three-material decomposition algorithms.

[0055] As mentioned above, if the k-absorption edge ("k-edge") energy of any basic material falls within the energy range used by diagnostic X-ray imaging systems, i.e., between approximately 30 and 120 keV, this can be used to help identify the basic material's contribution to X-ray attenuation. The k-edge energy of a material is defined as the minimum energy required for a photoelectric event to occur in the k-shell electron. The k-edge occurs at a characteristic energy for each material. The k-edge energy of a material is characterized by a sharp increase in its X-ray attenuation spectrum at the X-ray energy corresponding to the k-edge energy value. The k-edge energies of many materials present in the human body are too low to be detected by diagnostic X-ray imaging systems. For example, the k-edge energies of hydrogen, carbon, oxygen, and nitrogen are less than 1 keV. However, materials such as iodine (k-edge = 33.2 keV), gadolinium (50.2 keV), gold (80.7 keV), platinum (78.4 keV), tantalum (67.4 keV), holmium (55.6 keV), and molybdenum (k-edge = 20.0 keV) have k-edge energy values that allow their differentiation in spectral CT projection data acquired from a diagnostic x-ray imaging system.

[0056] As an example, FIG. 5 is a graph showing the dependence of mass attenuation coefficients on X-ray energy for two example materials, iodine and water. The mass attenuation coefficients shown in FIG. 5 represent X-ray attenuation. The sharp increase in iodine X-ray attenuation at 33.2 keV facilitates separation between the contributions of each of these materials in a composite attenuation spectrum that includes attenuation resulting from both of these materials. For example, X-ray attenuation resulting from iodine and water can be separated by using a first selection of energy intervals ES1 in the spectral attenuation data, where one energy interval DE1 is close to the k-edge energy of 33.2 keV and another energy interval DE2 is significantly above the k-edge energy.

[0057] In operation S120, a plurality of different material decomposition algorithms may be used to reconstruct separate images representing the injected contrast agent, and a value of the blood flow parameter is then calculated from each of the reconstructed images to determine a provided value of the blood flow parameter. In this regard, the plurality of different material decomposition algorithms used may be provided by the material decomposition algorithms cited above. Alternatively, or additionally, in operation S120, the energy interval DE 1..m Various selections of energy intervals DE1 and DE2 may be used to reconstruct separate images representing the injected contrast agent, and the numerical value of the blood flow parameter is calculated from each of these reconstructed images. In this regard, FIG. 5 shows an example of a selection of energy interval ES1 for generating a reconstructed image by separating iodine from water. Alternative selections of energy intervals for separating iodine from water may be provided by varying the width or center energy of energy intervals DE1 and / or DE2. As shown by the second selection of energy interval ES2 in FIG. 5, different selections of energy intervals may also be provided, in which case different numbers of energy intervals and different ranges of energy intervals, i.e., energy interval DE 1..5 is used to separate the iodine from the water. Once the energy interval is selected, a material decomposition algorithm, such as one of those described above, may be used in operation S120 to reconstruct each of the images representing the injected contrast agent.

[0058] Continuing with the example shown in Figure 4, in operation S120, after reconstructing a plurality of images each representing an injected contrast agent, a value of a blood flow parameter is determined for each of the reconstructed images. This operation is shown in the center portion of Figure 4, where a blood flow calculation algorithm 140 calculates the blood flow parameter for each of the reconstructed images 130. 1..34, the blood flow computation algorithm 140 includes a lumped parameter model. Using the lumped parameter model, blood flow within a vascular region can be represented as an electrical circuit, with the volumetric flow rate of blood within the vascular region represented by a current and the pressure of blood within the vascular region represented by a voltage. Lumped parameter models are described in more detail below.

[0059] Instead of using a lumped parameter model to calculate values of the blood flow parameters from each of the reconstructed images in act S120, other techniques may alternatively be used, including the time-of-flight approach for sampling reconstructed CT images to determine blood fluid velocities as disclosed in the above-cited Barfett, JJ et al., the technique for sampling raw or "projection" CT data to determine blood fluid velocities as disclosed in the above-cited Prevrhal, S. et al., and computational fluid dynamics ("CFD") techniques, such as the techniques for determining FFR values as disclosed in Taylor, CA et al., the above-cited Koo, BK et al., and the above-cited Kim, HJ et al.

[0060] Continuing with reference to Figure 4, once the value of the blood flow parameter is calculated from each of the reconstructed images using the blood flow calculation algorithm in act S120, the method shown in Figure 4 continues to act S130, where a value of the blood flow parameter for the blood vessel 110 is provided based on the calculated value of the blood flow parameter. This may be performed as described above, for example, by calculating the provided value of the blood flow parameter as a weighted average of the individual values from the reconstructed images. This may also include identifying outliers and excluding them from the weighted average, as described above.

[0061] Alternative approaches can also be used to calculate values of blood flow parameters of blood vessel 110 from spectral attenuation data 120 in act S120. As described in the approaches above, these can be used to generate multiple corresponding reconstructed images 130 using multiple different image reconstruction methods. 1..n and calculating the value of the blood flow parameter from each reconstructed image.

[0062] In one exemplary approach, in act S120, the received spectral attenuation data 120 is reconstructed to provide a reconstructed image 1301, and values for blood flow parameters are calculated from the image using a blood flow calculation algorithm 140. The values of the blood flow parameters are calculated based on a set of model parameters 150 of the blood flow calculation algorithm. 1..i is calculated using various values for

[0063] In this example, the model parameters may include geometric parameters of the model and / or boundary conditions of the model. The blood flow computation algorithm may include a lumped parameter model or a CFD model, as described above. For example, if the model is a lumped parameter model, values of the blood flow parameters may be determined for the model using different values for a set of geometric parameters of the model or using different values for a set of boundary conditions of the model. With reference to FIG. 4 , the geometric parameters of the model may be calculated from the reconstructed image, for example, from the reconstructed image 1301. The set of phantom parameters 150 1..i For example, the set of model parameters 150 may represent the diameter of the vessel lumen at different locations along its length. By applying different thresholds to the image intensity values in the reconstructed image, which are used to distinguish contrast agents within the vessel, different values of diameter can be determined from the vessel wall, whose cross-sectional area represents the vessel lumen diameter. Alternatively, different segmentation algorithms can be used. 1..iAlternatively, σ can represent boundary conditions of the model, such as pressure values at the inlet and outlet of the lumen of a vascular region. Different values for the boundary conditions can be provided by making different assumptions within the model regarding the actual pressure values. Such assumptions are detailed in the cited literature for each model. As an example, the overall global pressure drop boundary condition can be varied to estimate the sensitivity of the model.

[0064] In another exemplary approach, in act S120, the received spectral attenuation data 120 is reconstructed to provide a reconstructed image, a segmentation algorithm is applied to the reconstructed image, and multiple values for blood flow parameters are calculated from the segmented image using the blood flow calculation algorithm. The values of the blood flow parameters are calculated from the segmented image obtained by applying multiple different segmentation algorithms to the reconstructed image. For this purpose, various different segmentation algorithms can be used, such as model-based segmentation, watershed-based segmentation, region growing, level setting, or graph cut. Alternatively, different parameter settings of a single algorithm can be used.

[0065] In another exemplary approach, in act S120, values of blood flow parameters are calculated from the received spectral attenuation data 120 using each of a plurality of different blood flow calculation algorithms.

[0066] In this example, different blood flow calculation algorithms may be provided by, for example, different lumped parameter models or different CFD models.

[0067] The above-described exemplary approaches may be used alone or in combination to calculate multiple values of blood flow parameters of blood vessel 110 in act S120 using different techniques.

[0068] As mentioned above, in some examples, such as the example shown in Figure 4, operation S120 of calculating the value of the blood flow parameter of blood vessel 110 is performed using blood flow computation algorithm 140, which includes a lumped parameter model. The lumped parameter model includes a set of equations that describe blood flow within the model, and the equations are solved to calculate the value of the blood flow parameter.

[0069] Generally, a lumped parameter model can be used to represent blood flow within a vascular region as an electrical circuit. The volumetric flow rate of blood within the vascular region is represented by a current, and the pressure of blood within the vascular region is represented by a voltage. The volume of blood within the vascular region is represented by an electric charge. As described above, in a lumped parameter model, resistance to blood flow is represented by an electrical resistance, vessel wall compliance is represented by a capacitance, and blood inertia is represented by an inductance. Machine learning-based approaches for setting values of resistors, capacitors, and indices in such lumped parameter models based on angiographic geometric measurements of blood vessels within the vascular region are also disclosed in the above-cited publication by Nickisch, H. et al. and in publication WO 2016 / 001017. The values of blood flow and blood pressure in the model are then solved to determine the value of the FFR of the vessel. Another technique for providing and determining parameters of a lumped parameter model is disclosed in the above-cited publication by Kim, H. et al.

[0070] More specifically, as described in the above-cited article by Nickisch, H. et al., geometric measurements from the reconstructed angiographic images can be used to set values of model parameters in a lumped parameter model. These model parameters include values of electrical components in an electrical circuit and values of voltages or currents at nodes in the electrical circuit, i.e., boundary conditions. Such model parameters are determined from geometric measurements in the reconstructed images. Such measurements include, for example, the length, diameter, and curvature of the lumen in the vascular region.

[0071] FIG. 6 is a schematic diagram illustrating an example of a lumped parameter model 140 including a set of model parameters 1501 according to some embodiments of the present invention. The lumped parameter model shown in FIG. 6 represents branches in a vascular region as vessel segments and branches in a vascular region as nodes. As described in the above-cited paper by Nickisch, H. et al., a lumped parameter model generally includes two macroscopic component types: nonlinear vessel segment resistors (white tubes) and boundary conditions (black boxes). FIG. 6 shows boundary conditions for vessel segments such as "LCX" (left circumflex artery), "M1" (marginal artery branching from the left circumflex artery), "LAD" (left anterior descending artery), and "D1" (first diagonal artery). While the boundary conditions may be pressure or flow sources driving the network, any (lumped) boundary condition that drives a conventional FE model can be used. The challenge is to translate the local geometry of the vessel (radius, circumference, cross-sectional area) into nonlinear resistor parameters.

[0072] The hydraulic analogue of Ohm's law is TIFF2025527149000002.tif20158: The pressure drop p and flow f through a slender pipe are linearly transferred by a function p F The resistance constant w P is Poiseuille's law, which states that depends on the length of the pipe and its cross-sectional area. Poiseuille friction is the only hydraulic effect that causes pressure drop (or energy loss) in a hydraulic network. Friction between the vessel wall and the vessel is caused by changes in cross-sectional area, branching, and blood and curvature. In the adopted framework, shown in Table 1, a piecewise polynomial (invertible and point-symmetric) effect transfer function TIFF2025527149000003.tif15153, depending only on the local vascular geometry and material constants such as blood concentration r and viscosity m. e and geometric specific weighting Model hydraulic specific effects with TIFF2025527149000004.tif12153.d e If = 1, the resistance is restored and de For ≠ 1, we speak of varistors. The functional forms of the effect transfer functions shown in Table 1 below were taken from the fluid mechanics literature where individual hydraulic effects were studied experimentally and analytically in isolation. To combine multiple interdependent effects e = 1..E, local effect superposition TIFF2025527149000005.tif19146 is the effect-specific coefficient α e The bottom part of Table 1 shows a description of the local effect superposition and tree segment compression. A total of six different hydraulic effects are included, the first five of which have a single coefficient each, and the last one has three coefficients, resulting in a total of eight coefficients. To compress the size of the hydraulic network, the transfer function TIFF2025527149000006.tif14161 is a single tree segment transfer function along the center line of the tree segment c = 1..C This sums to TIFF2025527149000007.tif18154. This is possible because the flow through a tree segment is constant within the segment. Once the simulation is performed and the value of f is known, the compression behavior can be reversed by expansion. The hydraulic effects are also location-specific, in the sense that the first five effects are active everywhere except at bifurcations, and the last effect is active only at bifurcations. This representation of the hydraulic network as an assembly of transfer functions encoding the hydraulic effects is then used to simulate blood flow. TIFF2025527149000008.tif105124

[0073] Table 1 shows the transfer function and hydraulic effects. The geometric weights are obtained from the local cross-sectional area A, perimeter P, length l, radius r, and curvature κ of the vessel. Blood density is denoted by r and viscosity by m. Note that the cross section is not circular. Local effect specific transfer function TIFF2025527149000009.tif21126 is a local transfer function These are linearly combined to obtain the local transfer function TIFF2025527149000010.tif23120. TIFF2025527149000011.tif20128 is summed along the centerline c = 1..C and the tree segment transfer function Generates TIFF2025527149000012.tif24142.

[0074] Next, using a modified nodal analysis technique described in more detail in the above-cited paper by Nickisch, H. et al., the circuit graph representation of the model is solved to determine the values of pressure and volumetric flow at each node in the model. The model parameters a in Table 1 are then learned by comparing the lumped model predictions with the CFD reference simulation. This yields the values of pressure and flow at locations along the centerline of the vessel.

[0075] Thus, in an example where the operation of calculating the value of the blood flow parameter for the blood vessel 110 in S120 is performed using a lumped parameter model, the method described above with reference to FIG. 1 can be extracting, from each of the reconstructed images, model parameters 150 of a corresponding lumped parameter model representing blood flow within the vascular region; In S120, calculating the value of the blood flow parameter of the blood vessel 110 is performed using the corresponding lumped parameter model and model parameters 150. It may also include.

[0076] In these examples, the act of extracting the model parameters 150 is segmenting the reconstructed image to provide a three-dimensional model representing the vascular region; determining geometric data of the vascular region from the three-dimensional model; determining values of model parameters 150 from the geometric data; It has.

[0077] In these examples, various known segmentation algorithms can be used to segment the reconstructed image. The determined geometric data can include, for example, lumen dimensions in the vascular region, measurements of lumen curvature in the vascular region, cross-sectional area, and lumen shape, etc., as described above with reference to Table 1. Image intensity values in the segmented image can also be used to provide values for model parameters. For example, flow boundary conditions can be determined based on image intensities within the segmented vessel.

[0078] Returning to FIG. 1 , the value of the blood flow parameter provided in act S130 may be provided in various ways. Generally, the value of the blood flow parameter is output. For example, the value of the blood flow parameter may be provided by outputting the value to a display device, such as monitor 260 of system 200 shown in FIG. 2 . Alternatively, the value of the blood flow parameter may be output, for example, to a computer-readable storage medium or a printer device. In some examples, the value of the blood flow parameter is output numerically, while in other examples, the value of the blood flow parameter may be output graphically. For example, the value of the blood flow parameter may be output as a color-coded icon. The value of the blood flow parameter may also be calculated at multiple locations along the length of the blood vessel and output graphically. For example, an image showing a blood vessel or a region of the blood vessel may be output as a color-coded overlay on the image, with the value of the blood flow parameter displayed at corresponding locations along the length of the blood vessel. A physician can then use the provided value of the blood flow parameter to inform a clinical diagnosis regarding the subject.

[0079] The inventors have also observed that despite using reconstructed spectral attenuation data to visualize stenosis in vascular regions, it can be difficult to convey to a physician some of the underlying anatomical causes of the stenosis in the reconstructed image. For example, when a blood vessel contains a stenosis, it is useful to convey to the physician the relative amounts of calcification, soft tissue, plaque, blood vessels, and lumen to help the physician make a diagnosis of the vessel. However, these amounts can represent relatively small portions of the reconstructed image, making them difficult to identify. In response to this observation, in some instances, contrast in the reconstructed image is adjusted based on the value of a blood flow parameter provided to the blood vessel. This has the effect of drawing the physician's attention to the anatomical causes of clinically significant values of the blood flow parameter.

[0080] In one example, the method described above with reference to FIG. generating a contrast-adjusted reconstructed image from the provided values of the blood flow parameters for the blood vessel 110 and an angiographic image reconstructed from the spectral attenuation data 120 or from conventional X-ray attenuation data representing the blood vessel 110, in S140, wherein the contrast in the reconstructed image is adjusted based on the provided values of the blood flow parameters for the blood vessel 110; Includes.

[0081] In this example, the contrast may be adjusted throughout the entire contrast-adjusted reconstructed image, or only in a predetermined region 160. In the latter case, contrast may be enhanced in a predetermined region surrounding a location in the reconstructed angiographic image where a given value of the blood flow parameter meets a predetermined criterion. For example, if the value of the blood flow parameter is below a threshold value at that location, contrast may be adjusted only within a circular region surrounding the location within the blood vessel. By locally adjusting contrast in this manner, the physician's attention is focused on features within the anatomical region surrounding the clinically significant finding where the value of the blood flow parameter exceeds the threshold value, while the appearance of the remainder of the reconstructed image is not altered. For example, contrast may be increased in the contrast-adjusted reconstructed image.

[0082] Contrast can be adjusted in various ways. For example, gray-level mapping, contrast stretching, histogram correction, or a "windowing" technique, also known as contrast enhancement, can be used. As an example, a so-called "level and window" approach can be used, in which values are set for the window width, i.e., the range of X-ray attenuation values mapped between grayscale black and grayscale white values, and the window level, i.e., the center of the range of X-ray attenuation values. By setting the window width and window level values, image intensity values in the reconstructed image can be mapped to new values in the contrast-adjusted reconstructed image based on the value of the blood flow parameter. For example, decreasing the window level can increase image brightness in the contrast-adjusted reconstructed image, thereby improving the visibility of anatomical features.

[0083] The angiographic image used to generate the contrast-adjusted angiographic image may be generated from the spectral attenuation data 120 received in act S110 or from conventional X-ray attenuation data representing the blood vessel 110. In the former case, the angiographic image may be the reconstructed image used to calculate the value of the blood flow parameter of the blood vessel 110 in act S120, or may be a separate image reconstructed from the spectral attenuation data. In either case, the decision of which of multiple image reconstruction methods to use to generate the angiographic image or which of multiple windowing techniques to use to provide the contrast-adjusted image may be made using a classification algorithm such as a random forest or a support vector machine. The classification algorithm may determine which technique to use to adjust the contrast based on rules regarding, for example, the type of anatomical region represented in the image, the value of the blood flow parameter, etc. For example, a strong localized pressure drop in the reconstructed image is expected to indicate the presence of a lesion. As a result, plaque and narrow contrast-filled vessels are expected to be present. According to the rules, a predetermined localized region surrounding the pressure drop may be displayed purely as a contrast-adjusted image in the contrast-adjusted image. The remainder of the contrast-adjusted image may be displayed as classical Hounsfield unit x-ray attenuation values. Additional layers may be provided within a given localized region where plaque characteristics are identified. For example, a material image, such as a calcium image, may be provided to highlight the presence of plaque in the vicinity of the lesion. As another example, in an anatomical region with a relatively high perfusion level, the contrast between perfused muscle and contrast-filled blood vessels may be different. In this anatomical region, a rule may dictate that the contrast material is not displayed because there is too much contrast in the tissue.

[0084] In one example, a neural network is used to adjust the contrast in the contrast-adjusted image. The neural network can adjust the contrast of the reconstructed image directly, or the neural network can be used to identify which of multiple reconstruction methods to apply or which of multiple windowing techniques to apply to provide the contrast-adjusted image. This example is described with reference to FIG. 7, which is a schematic diagram illustrating an example of an operation S140 for generating a contrast-adjusted reconstructed image according to some aspects of the present disclosure. In this example, the operation for generating a contrast-adjusted reconstructed image includes: inputting the provided values of the blood flow parameters for the blood vessel 110 and the corresponding reconstructed image, or the corresponding angiographic image reconstructed from conventional X-ray attenuation data representing the blood vessel 110, into the neural network; and The neural network is trained to generate a contrast-adjusted reconstruction from input values of the blood flow parameters and the corresponding reconstruction.

[0085] 7, the blood flow parameter is an FFR value, and contrast is adjusted only within a predetermined region 160 surrounding the location of blood vessels whose FFR value is below a threshold of 0.8. The neural network may be implemented by various architectures, such as a convolutional neural network ("CNN"), a recurrent neural network ("RNN"), or a transformer.

[0086] The neural network NN illustrated in Figure 7 is receiving training data, the training data comprising a plurality of volumetric training images representing the vascular region and corresponding values of blood flow parameters for blood vessels within the vascular region;

[0087] receiving ground truth data, the ground truth data having, for each of the volumetric training images, a corresponding ground truth contrast-adjusted reconstructed image, the contrast in the reconstructed image being adjusted;

[0088] inputting the training data into the neural network;

[0089] For each of the plurality of input volumetric training images,

[0090] using the neural network to predict a corresponding contrast-adjusted reconstructed image;

[0091] adjusting parameters of the neural network based on a difference between the predicted contrast-adjusted reconstructed image and the ground truth contrast-adjusted reconstructed image;

[0092] repeating the predicting and adjusting steps until a stopping criterion is met; The contrast-adjusted reconstructed image is trained to generate the contrast-adjusted reconstructed image by

[0093] In this example, the volumetric training images may be, for example, CT images reconstructed from spectral CT data. Corresponding values of blood flow parameters may be determined using the techniques described above or by invasive measurements in the vasculature, for example, using blood flow sensing wires, pressure wires, or Doppler ultrasound catheters. Ground truth data may be provided by an expert who adjusts the contrast of the entire image, or for a predetermined region around a blood vessel where the blood flow parameter has clinical significance. If the neural network is trained to adjust the contrast directly, the ground truth data may include contrast-adjusted image intensities. If the neural network is trained to identify which of multiple reconstruction methods to apply to provide a contrast-adjusted image, the ground truth data may include an indication of the type of image reconstruction method to be used to generate the contrast-adjusted image, an indication of the energy interval to be used to generate the contrast-adjusted image, or an indication of the type of windowing technique to be used to generate the contrast-adjusted image.

[0094] Generally, training a neural network involves inputting a training dataset into the neural network and iteratively adjusting the neural network's parameters until the trained neural network provides accurate outputs. Training is often performed using dedicated neural processors such as graphics processing units (GPUs), neural processing units (NPUs), or tensor processing units (TPUs). Training often employs a centralized approach, in which cloud- or mainframe-based neural processors are used to train the neural network. Following its training with the training dataset, the trained neural network can be deployed to a device for analyzing new input data during inference. Processing requirements during inference are significantly lower than those needed during training, allowing neural networks to be deployed to a variety of systems, such as laptop computers, tablets, and mobile phones. Inference can be performed, for example, on a central processing unit (CPU), GPU, NPU, TPU, server, or in the cloud.

[0095] Therefore, the process of training the neural network NN shown in FIG. 7 includes adjusting its parameters. The parameters, more specifically, the weights and biases, control the behavior of the activation function in the neural network. In supervised learning, the training process automatically adjusts the weights and biases so that when input data is presented, the neural network accurately provides the corresponding expected output data. To do this, a loss function or error value is calculated based on the difference between the predicted output data and the expected output data. The value of the loss function may be calculated using functions such as negative log-likelihood loss, mean absolute error (or L1 norm), mean squared error, root mean squared error (or L2 norm), Huber loss, or (binary) cross-entropy loss. During training, the value of the loss function is typically minimized, and training is terminated when the value of the loss function meets a stopping criterion. In some cases, training is terminated when the value of the loss function meets one or more of several criteria.

[0096] Various methods are known for solving loss minimization problems, such as gradient descent and quasi-Newton methods. Various algorithms have been developed to implement these methods and their variations, including, but not limited to, stochastic gradient descent (SGD), batch gradient descent, mini-batch gradient descent, Gauss-Newton, Levenberg-Marquardt, Momentum, Adam, Nadam, Adagrad, Adadelta, RMSProp, and Adamax optimizers. This process is called backpropagation because derivatives are calculated starting from the last layer, or output layer, moving toward the first layer, or input layer. These derivatives inform the algorithm how to adjust the model parameters to minimize the error function. That is, adjustments to the model parameters are made starting from the output layer and working backward through the network until the input layer is reached. In the first training iteration, the initial weights and biases are often randomized. The neural network then predicts output data, which is also random. Backpropagation is then used to adjust the weights and biases. The training process is performed iteratively, adjusting the weights and biases at each iteration. Training is terminated when the error or difference between the predicted output data and the expected output data is within an acceptable range for the training data or some validation data. The neural network can then be deployed, and the trained neural network will make predictions for new input data using the trained values of its parameters. If the training process is successful, the trained neural network will accurately predict the expected output data from the new input data.

[0097] In one example, patient data may be input to the neural network NN described with reference to Figure 7, and a contrast-adjusted angiographic image may be generated based on the patient data. In this example, the operation of generating the contrast-adjusted reconstructed image may include: receiving patient data relating to a vessel; inputting patient data into a neural network; generating a contrast-adjusted angiographic image further based on the patient data; The patient data includes electronic health record data regarding the vessel and / or identification of a device implanted within the vessel.

[0098] The patient data may include information about the patient, such as body mass index, smoking history, age, gender, data identifying devices implanted within the blood vessels, etc. In this example, a neural network is trained to generate contrast-adjusted angiographic images further based on the patient data. In doing so, the neural network can provide a contrast-adjusted image that better matches the blood vessels in the reconstructed image.

[0099] The above-described method may be performed on multiple vessels within a vascular region, or on selected vessels. In the latter case, in one example, the method with reference to FIG. 1 further comprises: receiving an input identifying a blood vessel 110 in the spectral attenuation data 120; and S130, which provides values of blood flow parameters of the vessel 110, is performed for the identified vessel.

[0100] The operation of identifying the blood vessel 110 may be performed on an image reconstructed from the spectral attenuation data received in operation S110 or on an image reconstructed from conventional X-ray attenuation data. The blood vessel may be identified manually or automatically. Manual identification of the blood vessel 110 may be performed, for example, by a user manipulating a user input device in combination with a displayed reconstructed image. Automatic identification of the blood vessel 110 may be performed, for example, using a feature detector or a trained neural network. Segmentation may be performed on the reconstructed image to aid in identifying the blood vessel. Various segmentation algorithms are known for this purpose. The feature detector or trained neural network may automatically identify candidate blood vessels for determining the value of a blood flow parameter. The blood vessel may be identified based on the type of blood vessel or the presence of an abnormality, such as a stenosis, within the blood vessel. As an example, a feature detector or a trained neural network may be used to identify the left coronary artery and a stenosis therein in the reconstructed image, identifying the blood vessel as a candidate blood vessel for use in performing blood flow measurements, such as FFR. Blood flow measurements may then be provided for the blood vessel at the location of the stenosis.

[0101] In another example, a computer program product is provided, the computer program product comprising instructions that, when executed by one or more processors, cause the one or more processors to perform a method for determining a value of a blood flow parameter of a blood vessel 110, the method comprising: At S110, receiving spectral attenuation data 120 representing an injected contrast agent in a vascular region including a blood vessel 110, the spectral attenuation data being at a plurality of different energy intervals DE 1..m defining x-ray attenuation within a blood vessel region in At S120, calculating values of blood flow parameters of the blood vessel 110 from the spectral attenuation data 120 using a number of different techniques; In S130, providing a value of the blood flow parameter of the blood vessel 110 based on the calculated value of the blood flow parameter; Includes.

[0102] In another example, a system 200 is provided for determining a value of a blood flow parameter of a blood vessel 110. The system includes: At S110, receiving spectral attenuation data 120 representing an injected contrast agent in a vascular region including a blood vessel 110, the spectral attenuation data being spaced apart from a plurality of different energy intervals DE 1..m defining x-ray attenuation within a vascular region within the At S120, calculating values of blood flow parameters of the blood vessel 110 from the spectral attenuation data 120 using a number of different techniques; In S130, providing a value of the blood flow parameter of the blood vessel 110 based on the calculated value of the blood flow parameter; The one or more processors 210 are configured to execute the

[0103] An example of system 200 is shown in Figure 2. It should be noted that system 200 may also include one or more of a spectral X-ray imaging system for generating the spectral attenuation data 120 received in operation S110, such as, for example, a spectral CT imaging system 220 shown in Figure 2, a monitor 260 for displaying provided values such as blood flow parameters, reconstructed images, etc., a patient bed 270, an injector (not shown in Figure 2) for injecting contrast agent into the vasculature, and a user input device configured to receive user input, such as a keyboard, mouse, touch screen, etc.

[0104] The above examples should be understood as illustrating, not limiting, the present disclosure. Further examples are contemplated. For example, examples described in connection with a computer-implemented method may also be provided in a corresponding manner by a computer program product, a computer-readable storage medium, or the system 200. It should be understood that features described with respect to any one embodiment may be used alone or in combination with other described features, and may be used in combination with one or more other features of the embodiment or with combinations of other embodiments. Furthermore, equivalents and modifications not described above may also be used without departing from the scope of the present invention as defined in the appended claims. In the claims, the word "comprising" does not exclude other elements or operations, and the indefinite articles "a" or "an" do not exclude a plurality. The mere fact that certain features are recited in mutually different dependent claims does not indicate that a combination of these features cannot be advantageously used. Any reference signs in the claims should not be construed as limiting their scope.

Claims

1. A computer implementation method for determining the values ​​of blood flow parameters for blood vessels, The steps include receiving spectral attenuation data representing an injected contrast agent within a vascular region including the blood vessel, wherein the spectral attenuation data defines X-ray attenuation within the vascular region at multiple different energy intervals; The steps include: calculating the values ​​of blood flow parameters for the blood vessel using multiple different techniques from the spectral attenuation data; A step of providing a blood flow parameter value for the blood vessel based on the calculated blood flow parameter value. It has, The step of calculating the blood flow parameter values ​​for the aforementioned blood vessel is: The steps include: reconstructing the received spectral attenuation data using multiple different image reconstruction methods to provide multiple corresponding reconstructed images; and calculating the blood flow parameter values ​​from each of the reconstructed images. A step of reconstructing the received spectral attenuation data to provide a reconstructed image, and calculating values ​​for the plurality of blood flow parameters from the image using a blood flow calculation algorithm, wherein each value of the blood flow parameter is calculated using different values ​​for a set of model parameters of the blood flow calculation algorithm. Steps include: reconstructing the received spectral attenuation data to provide a reconstructed image; applying a segmentation algorithm to the reconstructed image; and using a blood flow calculation algorithm to calculate values ​​for a plurality of blood flow parameters from the segmented image, wherein each value of the blood flow parameter is calculated from segmented images obtained by applying different segmentation algorithms to the reconstructed image. The steps include: calculating the value of the blood flow parameter from the received spectral attenuation data using each of several different blood flow calculation algorithms; Having one or more method.

2. The step of calculating the blood flow parameter value for the blood vessel is to reconstruct the received spectral attenuation data using a plurality of different image reconstruction methods to provide a plurality of corresponding reconstructed images, and to calculate the blood flow parameter value from each of the reconstructed images. It has, The reconstructed image is The steps of applying a different material decomposition algorithm to the spectral decay data, and / or Steps to reconstruct spectral decay data having different selections of energy intervals. Reconstructed by The computer implementation method according to claim 1.

3. The method further includes: Steps include extracting model parameters for a corresponding concentration parameter model representing blood flow within the vascular region from each of the reconstructed images. It has, The step of calculating the blood flow parameter values ​​for the blood vessel is performed using the corresponding centrifugation parameter model and the model parameters. The computer implementation method according to claim 2.

4. The step of extracting the model parameters is: The steps include segmenting the reconstructed image to provide a three-dimensional model representing the vascular region, A step of determining geometric data for the vascular region from the three-dimensional model, The steps include determining the values ​​of the model parameters from the geometric data and A computer implementation method according to claim 3, comprising:

5. The concentrated parameter model represents the blood flow within the vascular region as an electrical circuit, The computer implementation method according to claim 3, wherein the volumetric flow rate of blood within the vascular region is represented by an electric current in the electrical circuit, and the pressure of blood within the vascular region is represented by a voltage in the electrical circuit.

6. The step of providing a value for the blood flow parameter for the blood vessel based on the calculated value of the blood flow parameter is: The steps include analyzing the calculated values ​​of the blood flow parameters to identify outliers, A step of providing a value for the blood flow parameter for the blood vessel as a weighted average of the calculated values ​​that are not identified as outliers. A computer implementation method according to claim 1, comprising:

7. The computer implementation method according to claim 6, wherein the step of analyzing the calculated values ​​of the blood flow parameters comprises the step of performing a statistical analysis on the calculated values ​​of the blood flow parameters.

8. The computer implementation method according to claim 1, further comprising the step of outputting a value representing a variation in the calculated value for the blood flow parameter.

9. The computer implementation method according to claim 1, wherein the steps of calculating a value of a blood flow parameter for the blood vessel and providing a value of a blood flow parameter for the blood vessel are performed at a plurality of locations along the blood vessel.

10. The method further, A step of generating a contrast-adjusted reconstructed image from the blood flow parameter values ​​for the provided blood vessel and an angiographic image reconstructed from the spectral attenuation data, or an angiographic image reconstructed from conventional X-ray attenuation data representing the blood vessel, wherein the contrast in the reconstructed image is adjusted based on the blood flow parameter values ​​provided for the blood vessel. A computer implementation method according to claim 1, comprising:

11. The step of generating the contrast-adjusted reconstructed image is: The step of inputting the blood flow parameter values ​​for the provided blood vessels, and the corresponding reconstructed image, or the corresponding angiographic image reconstructed from conventional X-ray attenuation data representing the blood vessels, into a neural network. It has, The neural network is trained to generate the contrast-adjusted reconstructed image and the corresponding reconstructed image from the input values ​​of the blood flow parameters. The computer implementation method according to claim 10.

12. The neural network is A step of receiving training data, wherein the training data comprises a plurality of volumetric training images representing the vascular region and corresponding values ​​of blood flow parameters for the blood vessels within the vascular region. A step of receiving ground truth data, wherein the ground truth data has a corresponding ground truth contrast-adjusted reconstructed image for each of the volume training images, and the contrast in the reconstructed image is adjusted. The steps include inputting the training data into the neural network, For each of the aforementioned multiple input volume training images, The steps include using the aforementioned neural network to predict the corresponding contrast-adjusted reconstructed image, The steps include adjusting the parameters of the neural network based on the difference between the predicted contrast-adjusted reconstructed image and the ground-truth contrast-adjusted reconstructed image, The steps of predicting and adjusting are repeated until the stopping criteria are met. The computer implementation method according to claim 11, which is trained to generate the contrast-adjusted reconstructed image.

13. The computer implementation method according to claim 10, wherein the contrast is adjusted only in a predetermined area within the reconstructed image.

14. The method described above is: The step of receiving an input that identifies the blood vessel in the spectral decay data. It further possesses, The step of providing values ​​for blood flow parameters for the blood vessel is performed for the identified blood vessel. The computer implementation method according to any one of claims 1 to 13.

Images