JP6362853B2 - Blood vessel analyzer and method for operating blood vessel analyzer - Google Patents

Description

Embodiments described herein relate generally to a blood vessel analysis device and a method for operating the blood vessel analysis device .

  A technique for non-invasive prevention and diagnosis of stenosis due to coronary artery stenosis, cerebral aneurysm, or carotid artery plaque, which is one of the three major diseases, is desired. Yes. Coronary stenosis is a serious lesion leading to ischemic heart disease. For heart disease, it is necessary to determine whether or not stent treatment should be performed, but coronary angiography (CAG) using a catheter is the mainstream for stenosis diagnosis of coronary arteries.

  A diagnostic index for coronary artery lesions is the fractional flow reserve (FFR), defined as the ratio of maximum coronary blood flow in the presence of stenosis to maximum coronary blood flow in the absence of stenosis. . This blood flow ratio is substantially the same as the ratio of the stenosis distal coronary pressure to the stenosis proximal coronary pressure, and the pressure ratio is measured by a pressure sensor at the catheter tip.

  Here, if the stenosis analysis of the coronary artery can be performed with an X-ray CT image (CT: Computed Tomography, hereinafter simply referred to as CT image) of the heart, catheter surgery is not required, and the burden on the patient can be reduced with minimal invasiveness. Connected. At present, there is only a non-invasive judgment material based on an index based on the size of plaque or thrombus by CT image.

  If the pressure difference before and after the thrombus can be analyzed from the CT image by the structural fluid simulation, the influence of the thrombus (or plaque) can be quantified. Then, from the obtained analysis results, if a blood vessel abnormality such as the degree of stenosis (the degree of stenosis) can be diagnosed, it will be a great merit for doctors and patients.

  However, the analysis result includes a large amount of data such as flow velocity / flow rate / pressure data at each time. For this reason, in order to obtain an image for diagnosing a blood vessel abnormality, enormous calculation resources and enormous calculation time are required. For this reason, the above analysis result may be difficult to use for diagnosing abnormal blood vessels.

JP 2008-241432 A

Kadooka et al. (ITU Journal, Tailor-Made Medicine Developed by Heart Simulator: Introduction of the World's Most Advanced Heart Simulator and its Applications, Vol.41, No.6) James K. Min. , Et. al. , “Relational and design of the DeFACTO (Determination of Fractional Flow Reservoir by Anatomic Computed TomoGraphicAnalogiology) tudy”, Journal

The problem to be solved by the present invention is to provide a blood vessel analyzing apparatus and an operating method of the blood vessel analyzing apparatus that can easily apply the analysis result by the structural fluid simulation to the abnormality diagnosis of blood vessels.

According to the embodiment, the injection profile acquisition unit acquires the injection profile of the contrast agent injected into the blood vessel of the subject, and the CT value acquisition unit acquires the CT value in the blood vessel lumen after the contrast agent injection. . Further, the structural fluid analysis unit performs a structural fluid analysis simulation with the injection profile and the CT value as inputs. The first calculation unit obtains the relationship between the input and the output corresponding to the structural fluid analysis simulation from the result of the structural fluid analysis simulation using the injection profile and the CT value acquired in advance and the CT value in which the presence or absence of the blood vessel is known. A model to be represented is calculated and stored in the storage device. Then, the difference model calculation unit calculates a difference model indicating the degree of abnormality of the blood vessel by detecting a difference between the CT value of the subject performing the abnormality diagnosis of the blood vessel and the model stored in the storage device, The diagnosis unit evaluates the difference model to diagnose a blood vessel abnormality.

FIG. 1 is a schematic hardware configuration diagram of a medical image diagnostic apparatus (X-ray computed tomography apparatus) according to an embodiment. FIG. 2 is a diagram illustrating an example of a dynamic model related to a target region for structural fluid analysis of the medical image diagnostic apparatus according to the embodiment. FIG. 3 is a diagram illustrating a flow of structural fluid analysis processing of the medical image diagnostic apparatus according to the embodiment. FIG. 4 is a functional block diagram of the image processing apparatus of the medical image diagnostic apparatus according to the embodiment. FIG. 5 is a diagram schematically showing a cross section orthogonal to the core line of the blood vessel. FIG. 6 is a diagram showing temporal changes in the vascular core form used in the image tracking process. FIG. 7 is a diagram illustrating an example of the tracking process between time t and time t + Δt. FIG. 8 is a diagram illustrating a calculation example of bending deformation and rotational displacement of a blood vessel core line by image analysis / tracking processing. FIG. 9 is a diagram illustrating a cross section orthogonal to the core line of the shape model constructed by the dynamic model construction unit of the medical image diagnostic apparatus according to the embodiment. FIG. 10 is a diagram for explaining assignment of the forced displacement history to the shape model by the dynamic model construction unit of the medical image diagnostic apparatus according to the embodiment. FIG. 11 is a diagram illustrating another method for assigning the forced displacement history to the shape model by the dynamic model construction unit of the medical image diagnostic apparatus according to the embodiment, and illustrating an example of assignment when the blood vessel shape deformation index is twisted. It is. FIG. 12 is a diagram showing another assignment method of the forced displacement history to the shape model by the dynamic model construction unit of the medical image diagnostic apparatus of the embodiment, showing an assignment example when the blood vessel shape deformation index is bending. is there. FIG. 13 is a diagram for explaining the calculation of the posterior distribution and the identification of the average internal pressure related to the load condition (average pressure in the blood vessel) by the hierarchical Bayesian model and the Markov chain Monte Carlo method by the statistical identification unit of the medical image diagnostic apparatus of the embodiment. FIG. FIG. 14 shows the calculation of the posterior distribution and the identification of the material model parameter (equivalent elastic modulus of the blood vessel wall) regarding the material model parameter by the hierarchical Bayesian model and the Markov chain Monte Carlo method by the statistical identification unit of the medical image diagnostic apparatus of the embodiment. It is a figure for demonstrating. FIG. 15 is a diagram illustrating a display example of a spatial distribution of internal pressure that is one of the mechanical indices in the medical image diagnostic apparatus according to the embodiment. FIG. 16 is a diagram illustrating a display example of a spatial distribution of a flow velocity value that is one of blood flow indexes in the medical image diagnostic apparatus according to the embodiment. FIG. 17 is a diagram illustrating a display example of a graph regarding the blood pressure of the left coronary artery origin in the medical image diagnostic apparatus according to the embodiment. FIG. 18 is a diagram illustrating a display example of a graph relating to blood pressure in the vicinity of a branch point between LCX and LDA in the medical image diagnostic apparatus according to the embodiment. FIG. 19 is a diagram illustrating a display example of a graph relating to a blood pressure change in the core line direction in the medical image diagnostic apparatus according to the embodiment. FIG. 20 is a diagram illustrating another assignment example of the forced displacement history by the dynamic model construction unit of the medical image diagnostic apparatus according to the embodiment. FIG. 21 is a diagram illustrating another allocation example of the forced displacement history by the dynamic model construction unit of the medical image diagnostic apparatus according to the embodiment. FIG. 22 is a diagram illustrating another allocation example of the forced displacement history by the dynamic model construction unit of the medical image diagnostic apparatus according to the embodiment. FIG. 23 is a diagram illustrating deformation of the fiber group in the medical image diagnostic apparatus according to the embodiment. FIG. 24 is a diagram illustrating an orthogonal cross section of a dynamic model of a thick cylinder of the medical image diagnostic apparatus according to the embodiment. FIG. 25 is an enlarged view of a minute sector element in the medical image diagnostic apparatus of the embodiment. FIG. 26 is a diagram for explaining stenosis model identification in the medical image diagnostic apparatus according to the embodiment. FIG. 27 is a diagram for comparing the direct estimation result and the estimation result based on the stenosis model in the medical image diagnostic apparatus according to the embodiment. FIG. 28 is a functional block diagram when a blood vessel abnormality diagnosis is performed by the medical image diagnostic apparatus of the embodiment. FIG. 29 is a diagram illustrating an injection pattern of a contrast agent in the medical image diagnostic apparatus according to the embodiment. FIG. 30 is a graph showing changes in contrast agent concentration according to the presence or absence of stenosis. FIG. 31 is a graph for explaining a contrast medium concentration gradient.

  The blood vessel analysis device and the blood vessel analysis method of the embodiment can be applied to a computer device for structural fluid analysis of a blood vessel region included in a medical image generated by a medical image diagnostic device. This computer apparatus may be incorporated in the medical image diagnostic apparatus, or may be a workstation or the like separate from the medical image diagnostic apparatus. Hereinafter, a medical image diagnostic apparatus in which a computer device to which a blood vessel analysis device and a blood vessel analysis method according to an embodiment are applied will be described in detail with reference to the drawings.

(Overview)
In fluid analysis of arteries such as coronary arteries, it is necessary to appropriately consider the influence of blood vessel or plaque hardness, the influence of cardiac pulsation on arterial flow, and the influence of boundary conditions at the arterial inlet and outlet.

  For this reason, the medical image diagnostic apparatus according to the embodiment performs the following structural fluid analysis. That is, the medical image diagnostic apparatus uses the three-dimensional deformation history information obtained by image analysis of a medical four-dimensional image obtained by adding a time axis to a medical three-dimensional image, and the displacement history information obtained from image tracking to determine the blood vessel wall of the artery. The shape deformation index and the displacement constraint condition are calculated. Further, in the hardness calculation region, the medical image diagnostic apparatus gives a displacement freedom degree constraint condition based on a forced displacement history only to the surface of the blood vessel wall and to a node of the mechanical model of the blood vessel wall.

  The medical diagnostic imaging apparatus gives a forced displacement history to the surface and internal nodes in the vascular wall dynamic model in a region other than the hardness calculation region, and performs structural fluid analysis with a displacement constraint condition. Thereby, the structural fluid analysis for identifying the material model of the hardness calculation region and the boundary conditions of the inlet and outlet can be performed, and the blood vessel and hemodynamic analysis can be performed with high accuracy. Then, the medical image diagnostic apparatus creates a normal model at normal time and stenosis obtained by such a highly accurate structural fluid analysis, and diagnoses the degree of stenosis from the model identification with the diagnosis target data.

  Specifically, the medical image diagnostic apparatus according to the embodiment acquires a four-dimensional image of a blood vessel from a CT (Computed Tomography) diagnostic apparatus, and forces a movement vector of each point of the blood vessel part obtained from the image tracking of the blood vessel. Calculated as displacement history. Further, the medical image diagnostic apparatus calculates a physical quantity such as a volume change, a diameter change, a shape change, or a phase change of the blood vessel lumen from the image tracking information. In addition, the medical image diagnostic apparatus identifies the material model or boundary condition of the designated region using the calculated physical quantity as an index.

  The medical diagnostic imaging system uses vascular analysis based on such image tracking, structural fluid analysis, and inverse analysis (statistical identification method) to detect vascular lesions such as stenosis such as blood pressure, blood flow, stress, or strain in hemodynamics. Analyzes indicators necessary for prevention and diagnosis with high accuracy. Then, the medical image diagnostic apparatus analyzes a typical example in which the presence or absence of stenosis is known, creates a low-order standard model (nominal model) in advance, and performs stenosis based on model identification with the diagnosis target data. Diagnose the degree.

  The medical image diagnostic apparatus of such an embodiment can be applied to any kind of image diagnostic apparatus equipped with an imaging mechanism for scanning a subject. Examples of the medical image diagnostic apparatus according to the embodiment include an X-ray computed tomography apparatus (X-ray CT apparatus), a magnetic resonance diagnostic apparatus, an ultrasonic diagnostic apparatus, a SPECT (Single Photon Emission CT) apparatus, and a PET (Positron Emission Tomography). The present invention can be appropriately used for an apparatus, a radiotherapy apparatus, and the like. Hereinafter, the medical image diagnostic apparatus according to the embodiment is assumed to be an X-ray computed tomography apparatus for specific description.

  FIG. 1 is a schematic hardware configuration diagram of a medical image diagnostic apparatus (X-ray computed tomography apparatus) according to an embodiment. As shown in FIG. 1, the X-ray computed tomography apparatus includes a CT mount 10 and a console 20. The CT gantry 10 scans the imaging region of the subject with X-rays under the control of the gantry controller 23 of the console 20. The imaging site is, for example, the heart.

  The CT mount 10 includes an X-ray tube 11, an X-ray detector 13, and a data acquisition device 15. The X-ray tube 11 and the X-ray detector 13 are mounted on the CT mount 10 so as to be rotatable around the rotation axis Z. The X-ray tube 11 irradiates a subject into which a contrast medium has been injected with X-rays. The X-ray detector 13 detects X-rays generated from the X-ray tube 11 and transmitted through the subject, and generates an electric signal corresponding to the detected X-ray intensity.

  The data collection device 15 reads an electrical signal from the X-ray detector 13 and converts it into digital data. A set of digital data for each view is called a raw data set. A time-series raw data set relating to a plurality of scan times is transmitted to the console 20 by a non-contact data transmission device (not shown).

  The console 20 includes a gantry control unit 23, a reconstruction device 25, an image processing device 27, an input device 29, a display device 31, and a storage device 33 with the system control unit 21 as a center.

  The gantry control unit 23 controls each device in the console 20 in accordance with the scan condition set by the user via the input device 29.

  The reconstruction device 25 generates CT image data related to the subject based on the raw data set. Specifically, first, the reconstruction device 25 generates a projection data set by pre-processing the raw data set. Pre-processing includes logarithmic conversion, nonuniformity correction, calibration correction, and the like. Next, the reconstruction device 25 performs image reconstruction processing on the projection data set to generate CT image data. Image reconstruction algorithms include analytical image reconstruction methods such as FBP (filtered backprojection) method, successive approximation images such as ML-EM (maximum likelihood expectation maximization) method and OS-EM (ordered subset expectation maximization) method. Existing algorithms such as reconstruction can be employed.

  In the embodiment, the reconstruction device 25 generates time-series CT image data based on the time-series projection data set. The CT image includes a pixel region related to a blood vessel contrasted with a contrast agent (hereinafter referred to as a blood vessel region). Note that the CT image may be slice data representing a two-dimensional spatial distribution of CT values, or volume data representing a three-dimensional spatial distribution of CT values. Hereinafter, it is assumed that the CT image is volume data. Time-series CT image data is stored in the storage device 33.

  The image processing device 27 constructs a dynamic model based on the time-series CT images and executes structural fluid analysis. Details of the processing of the image processing device 27 will be described later.

  The input device 29 receives various commands and information input from the user. As the input device 29, a keyboard, a mouse, a switch, or the like can be used.

  The display device 31 displays various information such as CT images and structural fluid analysis results. As the display device 31, for example, a CRT display, a liquid crystal display, an organic EL display, a plasma display, or the like can be used as appropriate.

  The storage device 33 is composed of various storage media such as a hard disk device. The storage device 33 stores various data such as time-series projection data and time-series CT image data. For example, the storage device 33 stores time-series CT image data in a medical image file format conforming to the DICOM (digital imaging and communications in medicine) standard. The storage device 33 may store medical data collected by an external device in association with time-series CT image data in a medical image file.

  The system control unit 21 includes a central processing unit (CPU), a read only memory (ROM), and a random access memory (RAM). The system control unit 21 functions as the center of the X-ray computed tomography apparatus. The system control unit 21 executes the blood vessel analysis program stored in the ROM or RAM to execute the blood vessel structure analysis processing of the embodiment.

  The system control unit 21, the image processing device 27, the input device 29, the display device 31, and the storage device 33 constitute a blood vessel analysis device 50. As in the embodiment, the blood vessel analyzing apparatus 50 may be incorporated in a medical image diagnostic apparatus (X-ray computed tomography apparatus), or may be a computer apparatus separate from the medical image diagnostic apparatus. When the blood vessel analyzing device 50 is separate from the medical image diagnostic device, the blood vessel analyzing device 50 collects medical data such as time-series CT images from the medical image diagnostic device or PACS (picturearchiving and communication systems) via the network. Just do it.

  Next, an operation example of the embodiment will be described in detail. Note that the medical image diagnostic apparatus according to the embodiment can analyze blood vessels in any part of the human body such as a cardiovascular vessel, a carotid artery, or a cerebral artery. Hereinafter, as an example, description will be made assuming that the blood vessel of the heart is the analysis target.

  Examples of the blood vessel of the heart include a coronary artery and an aorta. The coronary artery starts from the coronary artery origin of the aorta, travels on the myocardial surface, and enters the intima side from the epicardium side. The coronary arteries branch into myriad capillaries in the intima of the myocardium. After bifurcation, countless capillaries reintegrate to form the great cardiac vein and connect to the coronary sinus. Unlike other organs, the coronary vasculature is unique in that perfusion must be ensured in the course of mechanical changes such as myocardial contraction and relaxation.

  A characteristic of coronary blood flow is that it flows more when the perfusion pressure decreases in the left ventricular diastole than in the systole where the internal pressure at the coronary artery origin increases due to the mechanical blood flow inhibition effect due to myocardial contraction. Therefore, the normal coronary blood flow velocity waveform is bimodal in the systolic and diastolic phases, and the diastolic blood flow is dominant. In hypertrophic cardiomyopathy and aortic stenosis, retrograde waves are observed in the systole, and in aortic reflux, the systolic antegrade wave is increased, resulting in a specific blood flow waveform depending on the disease. Yes. In addition, the antegrade waveform in the diastole is closely related to the left ventricular dilation function, particularly the left ventricular relaxation. In the left ventricular relaxation delay example, the peak of the diastolic waveform shifts backward, and the deceleration leg tends to be gentle. In such cases, coronary blood flow during diastole cannot be sufficiently increased during tachycardia, which is thought to promote myocardial ischemia.

  Coronary blood flow is generated by applying coronary perfusion pressure equal to the aortic pressure (that is, the pressure at the aortic root where the coronary artery branches) to the left and right coronary arteries anatomically branching from the aortic root. Coronary blood flow resistance as well as driving pressure, which is aortic pressure, is important for determining coronary blood flow. A thick coronary vessel of 140 to 180 μm or more has about 20% of the resistance to coronary vessels, whereas a microvessel of 100 to 150 μm or less is said to have a lot of remaining resistance components. Therefore, in the absence of so-called coronary stenosis, the resistance value depends on the tonicity of coronary microvessels.

  Vascular resistance factors include vascular properties, arteriosclerosis, vascular stenosis, blood viscosity, and mechanical factors. Coronary microvessel tonus is defined by vascular properties, myocardial metabolism (myocardial oxygen consumption), neurohumoral factors, mechanical factors, various vasoactive substances as humoral factors, blood viscosity, cardiac hypertrophy, coronary artery Coronary circulatory disturbance is also affected by various lesions including sclerosis.

  Coronary blood flow pulsation is affected by the pulsation pattern of coronary blood flow, the control of intramyocardial blood flow by myocardial contraction, and the intramyocardial vascular response to mechanical stimulation. The mechanism by which myocardial contraction inhibits blood flow includes an increase in intramyocardial pressure, changes in intramyocardial vascular volume, and compression of intramyocardial blood vessels. The blood flow regulating factor in the myocardial diastole includes diastole coronary artery pressure, diastole extravascular force, heart rate, ratio of diastole to cardiac cycle, and myocardial relaxation.

  The blood vessel analysis device 50 constructs a dynamic model based on the time-series CT images, performs structural fluid analysis on the blood vessels of the heart using the dynamic model, and calculates the mechanical index and the blood flow rate index in the blood vessel. Calculate with high accuracy. In order to calculate a highly accurate mechanical index or vascular flow index, it is necessary to assign a highly accurate latent variable to the dynamic model. When constructing a dynamic model, the blood vessel analyzing apparatus 50 performs inverse analysis on the initial dynamic model and statistically identifies latent variables. Thereby, the blood vessel analysis device 50 can determine a highly accurate latent variable.

  The mechanical index means a mechanical index related to a blood vessel wall or blood. Examples of the mechanical index related to the blood vessel wall include an index related to the displacement of the blood vessel wall, an index related to stress and strain generated in the blood vessel wall, an index related to the distribution of internal pressure applied to the blood vessel lumen, and an index related to material properties representing the hardness of the blood vessel. And so on. Examples of the index relating to the material property representing the hardness of the blood vessel include an average slope of a curve indicating the relationship between stress and strain of the blood vessel tissue.

  The blood vessel flow rate index in the mechanical index related to blood means a hemodynamic index related to blood flowing through the blood vessel. Examples of the blood vessel flow rate index include blood flow rate, blood flow rate, blood viscosity, and the like. The latent variable is, for example, a material model parameter such as a blood vessel material constitutive equation or a blood material constitutive equation (eg, Young's modulus or Poisson's ratio), a load condition parameter such as an internal pressure distribution applied to the vascular lumen, It includes at least one of boundary condition parameters for structural analysis and fluid analysis, and variation distribution parameters related to uncertainties of time-series form indices and shape deformation indices.

  Here, the variation distribution parameter related to the uncertainty of the time-series shape index and the shape deformation index means that the medical image data includes a variation distribution caused by noise of each CT value and an ambiguity of a boundary threshold value of a living tissue. Uncertainty in spatial coordinates of boundary coordinates and feature points (blood vessel bifurcations, contrast agent dispersion, etc.) of vascular tissue and blood, or geometric structure The uncertainty of a parameter (such as a lumen radius of a cross section perpendicular to the core line) or the uncertainty of medical image data itself (such as a CT value or a boundary threshold value) is expressed as a probability distribution.

  The dynamic model is a numerical model for expressing the behavior of blood vessels and blood. The mechanical model has different types depending on the method of structural fluid analysis. For example, the dynamic model is classified into a continuum dynamic model and a simple dynamic model. The continuum mechanical model is used, for example, in a finite element method (FEM) or a boundary element method. The simple dynamic model is classified into, for example, a material dynamic model based on material dynamics and a hydrodynamic model based on rheology.

  Note that the type of the dynamic model is not particularly limited unless otherwise specified in the following description. The initial dynamic model means a dynamic model to which a sampling set (a set of combinations of parameters) related to the parameters of the latent variable obtained from the probability distribution of the latent variable and the variable range is assigned.

  FIG. 2 is a diagram illustrating an example of a dynamic model M1 related to a target region for structural fluid analysis (hereinafter referred to as an analysis target region). As shown in FIG. 2, the dynamic model M1 has an aorta region R1, a right coronary artery region R2, and a left coronary artery region R3. Blood flows from the aorta to the right or left coronary artery.

  As shown in FIG. 2, the end of the dynamic model M1 on the aortic origin side is set as the blood flow inlet, and the right coronary artery region end and the left coronary artery region end are set as the blood flow outlet. A boundary condition is set for each of the inlet and the outlet. The boundary conditions regarding the inlet include, for example, the blood flow velocity at the inlet or the pressure due to the blood flow, or the rate of change thereof. Boundary conditions for the outlet include, for example, the blood flow velocity at the outlet or the pressure due to the blood flow, or the rate of change thereof.

  Deformation of the aorta, right coronary artery, and left coronary artery is due to mechanical action on the blood vessel wall due to blood flow, mechanical action on the blood vessel wall due to heartbeat (external force), load condition at the blood vessel cross-sectional boundary, blood vessel wall Depending on various factors such as the material model, the unstressed state of the vessel, and the geometry of the vessel wall.

  Here, the mechanical action on the blood vessel wall caused by the blood flow includes, for example, an internal pressure caused by the blood flow and a shearing force caused by the blood flow. Due to the internal pressure caused by the blood flow, the blood vessel deforms in the radial direction of the blood vessel or in the direction perpendicular to the surface of the blood vessel lumen. The mechanical action on the blood vessel wall due to the pulsation of the heart and the shearing force due to the blood flow cause deformation due to the mechanical action on the blood vessel wall such as expansion / contraction, twisting and bending in the blood vessel core direction.

  The overall and local deformation behavior of the blood vessel, such as stretching, twisting, and bending with respect to the blood vessel core direction, is forced displacement in the aorta region R1, the right coronary artery region R2, and the left coronary artery region R3, or a part thereof, or the surrounding tissue. A load condition is set by giving a temporal change in (movement vector or rotational displacement) or load vector. Further, the deformation in the radial direction of the blood vessel based on the internal pressure caused by the blood flow or the direction perpendicular to the lumen surface is given as a temporal change in the pressure distribution to the blood vessel lumen, thereby setting the load condition.

  In the structural fluid analysis, a displacement constraint condition by forced displacement is assigned to the aorta region R1, the right coronary artery region R2, and the left coronary artery region R3. As a result, the degree of freedom of deformation of the blood vessel wall in the structural fluid analysis can be reduced, and the calculation convergence can be stabilized and the analysis time can be shortened.

  Further, for example, the degree of deformation of the blood vessel shape depends on the material of the blood vessel wall. Therefore, material models are assigned to the aorta region R1, the right coronary artery region R2, and the left coronary artery region R3. Also, for example, the degree of deformation of the blood vessel shape depends on the unstressed state of the blood vessel. As an initial value of the load condition, a residual stress distribution of the blood vessel may be given as the load condition.

  Here, in the element set composed of the node set and nodes that discretize the space of the numerical model for blood vessel analysis, identify the region that does not identify the analysis conditions such as the material model, boundary condition, load condition, etc. Displacement constraint of forced displacement history is given to the nodes in the region where analysis conditions are not identified, and the displacement constraint condition of forced displacement history is given only to the nodes on the blood vessel wall surface (outer surface) in the region where the material model is identified. The inside of the blood vessel wall secures a degree of freedom of displacement and does not give displacement restraint. Thereby, the deformation | transformation freedom degree of structural fluid analysis can be suppressed, and it can analyze stably and efficiently.

  However, a dummy dummy set for buffering may be provided on the surface of the blood vessel wall, and a forced displacement may be applied to a node on the surface. As a result, when there is a projection due to plaque or the like in the blood vessel lumen, or when a load due to internal pressure such as a blood vessel bifurcation affects the deformation outside the cross section in the core line direction, not only the blood vessel lumen but also the blood vessel wall The load vector applied to the blood vessel and the internal pressure can be separated and identified with reference to the shape index. Physiologically, it simulates the fat layer on the wall surface. On the other hand, in the numerical calculation, forced displacement is applied to the surface of the blood vessel wall, and high stresses that are different from the actual locality are generated inside the blood vessel wall. There is also an effect to avoid that.

  Parameters related to latent variables such as the material model, boundary condition, and load condition are identified by inverse analysis (statistical identification processing) based on a dynamic model described later. Accurate latent variables identified by inverse analysis are assigned to the dynamic model. For structural fluid analysis or fluid analysis or structural analysis or image analysis that takes into account the influence of external factors such as blood vessels and heart outside the analysis target blood vessel region to the analysis target blood vessel region using a dynamic model to which precise latent variables are assigned Based on hemodynamic analysis can be performed.

  The blood vessel analysis device 50 can solve the following four difficulties by identifying latent variables by inverse analysis regarding the construction of a dynamic model. Difficult Identification method of coronary artery material model. Difficult 2 Incorporation of coronary effects of heart shape deformation. Difficult 3 A method for identifying boundary conditions of coronary arteries. Difficult 4 Image analysis and structural fluid analysis using blood vessel shapes with variations due to the uncertainty of medical image data. By overcoming these four difficulties, the blood vessel analysis device 50 achieves improved analysis accuracy compared to conventional blood vessel structure fluid analysis in which latent variables are not identified by inverse analysis.

  Next, details of the structural fluid analysis process in the medical image diagnostic apparatus of the embodiment will be described. FIG. 3 is a diagram showing a typical flow of the structural fluid analysis process performed under the control of the system control unit 21. FIG. 4 is a functional block diagram of the image processing device 27.

  As shown in FIG. 3, in the structural fluid analysis process, first, the system control unit 21 reads a medical image file to be processed from the storage device 33 and supplies it to the image processing device 27. The medical image file includes, in addition to time-series CT image data, pressure value data related to the subject's blood vessel lumen, blood flow index observed value data, and plaque index. Other than the CT image, an MRI image or an ultrasonic echo image may be used. The time-series CT image data is data representing a three-dimensional spatial distribution of time-series CT values. The time-series CT images include, for example, about 20 CT images for one heartbeat, that is, about 20 heart phases.

  As shown in FIG. 3, the system control unit 21 causes the image processing device 27 to perform a region setting process (step S1). In step S <b> 1, the region setting unit 51 of the image processing device 27 sets an analysis target region for structural fluid analysis in a blood vessel region included in a time-series CT image. The analysis target region is set to an arbitrary part of the blood vessel region related to the coronary artery. For example, the region setting unit 51 sets the analysis target region and the identification target region in the blood vessel region by an instruction from the user via the input device 29 or image processing.

  Here, the structure of the blood vessel will be described with reference to FIG. FIG. 5 is a diagram schematically showing a cross section perpendicular to the blood vessel core line (hereinafter referred to as a blood vessel cross section). As shown in FIG. 5, the blood vessel has a tubular blood vessel wall. The central axis of the blood vessel wall is called the core wire. The inside of the vessel wall is called the lumen. Blood flows into the lumen. The boundary between the lumen and the blood vessel wall is called the blood vessel inner wall. A tissue around a blood vessel such as a myocardium is distributed outside the blood vessel wall. The boundary between the blood vessel wall and the tissue surrounding the blood vessel is called the blood vessel outer wall. Plaques may develop inside the vessel wall.

  As shown in FIG. 5, plaques are classified into, for example, calcified calcified plaques, rod-shaped plaques, and the like. Spider-like plaque is soft and has a risk of rupturing the inner wall of the blood vessel and oozing out into the blood vessel as a thrombus, and is sometimes called unstable plaque. Therefore, it is clinically useful to grasp the characteristics of plaque. Plaque properties and existing areas can be identified by a plaque index included in the medical image file. The plaque index can be relatively discriminated by, for example, the magnitude of the CT value normalized with reference to the CT value of bone. However, it is not easy to analyze the deformation characteristics and hardness of the plaque inside the blood vessel.

  When step S1 is performed, the system control unit 21 causes the image processing device 27 to perform image analysis / tracking processing (step S2). In step S2, the image analysis / tracking processing unit 53 of the image processing device 27 performs image processing on the time-series CT image to calculate a time-series blood vessel shape index and a time-series blood vessel shape deformation index. Specifically, the image analysis / tracking processing unit 53 calculates a time-series blood vessel shape index by performing an image analysis process on the time-series CT image, and performs a tracking process on the time-series CT image. A series of blood vessel shape deformation indices is calculated.

  More specifically, the image analysis / tracking processing unit 53 extracts a blood vessel region from each CT image in the image analysis processing, and relates to a pixel region relating to a blood vessel lumen (hereinafter referred to as a blood vessel lumen region) and a blood vessel wall. A pixel region (hereinafter referred to as a blood vessel wall region) is specified. The image analysis / tracking processing unit 53 uses, as a blood vessel shape index, a plurality of pixels on a region where a cross section perpendicular to a blood vessel core line or a surface perpendicular to a blood vessel lumen surface intersects a blood vessel lumen, a blood vessel wall, and a plaque region. The three-dimensional coordinates of are specified.

  The blood vessel shape index is not only a three-dimensional coordinate, but also a radius and diameter of a blood vessel lumen at a certain angle in a cross section perpendicular to the core line and a direction vector of 0 °, or an average area and average radius for all angles in the cross section, Alternatively, a blood vessel lumen volume surrounded by a plurality of cross sections perpendicular to the core line direction, or a geometric index such as a blood vessel wall volume or a plaque volume surrounded by a plurality of cross sections perpendicular to the lumen surface may be used.

  In the tracking process, the image analysis / tracking processing unit 53 displays a feature point, a feature shape, a representative point, a pixel, or the like in a blood vessel region, blood, a contrast agent, or a proton by an indication or image processing via the input device 29 from the user. Set multiple tracking points. For example, the image analysis / tracking processing unit 53 sets a tracking point set such as a blood vessel bifurcation or a surface feature shape. From the displacement data of the tracking point set obtained by the tracking processing of the image analysis / tracking processing unit 53 at each time (each cardiac phase), the displacement time of the node on the blood vessel wall surface, the blood vessel wall, or the blood vessel lumen of the dynamic model The change is calculated by interpolation processing or the like and given as a forced displacement.

  For example, the image analysis / tracking processing unit 53 defines a node on the blood vessel core line in the dynamic model. The image analysis / tracking processing unit 53 extracts deformation related to expansion / contraction, torsion, and bending in the core direction of the blood vessel from temporal changes in the displacement of the nodes on the surface of the blood vessel wall in the mechanical model, in the blood vessel wall, or in the blood vessel lumen. You may express by giving as a forced displacement of the node in the cross section perpendicular | vertical to a core wire and a core wire. Thus, the forced displacement data (forced displacement history) at the nodes at each time in the dynamic model is specified as the blood vessel shape deformation index.

  The image analysis / tracking process will be described below with reference to FIGS. FIG. 6 is a diagram showing temporal changes in the shape of the blood vessel core line. As shown in FIG. 6, for example, a time-series medical image is assumed to include 20 CT images per heartbeat. That is, CT images are obtained every 5% from 0% to 95% of the cardiac phase. The image analysis / tracking processing unit 53 extracts the core line of the blood vessel region. As shown in FIG. 6, the form of the core wire changes as the cardiac phase progresses.

  FIG. 7 is a diagram illustrating an example of the tracking process between time t and time t + Δt. As shown in FIG. 7, nodes in the dynamic model P1 to P10 are set on the blood vessel core line, and are dynamically connected to the nodes of the mechanical model of the blood vessel on each cross section. However, it is independent of the nodes of the blood dynamic model. It is assumed that the displacement data of the nodes P1 to P20 on the blood vessel core line is calculated based on the displacement data of the blood vessel tracking points by a process such as interpolation, and the forced displacement is set at each node. In order to describe the blood vessel shape deformation index and the blood vessel shape index, a local blood vessel region RA defined by the node P13 and the node P14 is considered.

  It is assumed that the distance between the node P13 and the node P14 in the core line direction is L and the radius of the blood vessel region is r at time t. From the image analysis / tracking processing unit 53, a forced displacement such as expansion, contraction, torsion, or bending in the blood vessel core direction between the node P13 and the node P14 is extracted, and the forced displacement of the node P13 (the movement displacement in the three-dimensional space and the core line direction). Rotational displacement) and forced displacement of the node P14 (moving displacement in the three-dimensional space and rotational displacement in the core line direction) are calculated.

  FIG. 8 is a diagram illustrating a calculation example of bending deformation and rotational displacement of the blood vessel core wire. As shown in FIG. 8, for example, the torsion angle may be calculated from a change in the normal direction vector c of the surface composed of the line a and the line b.

  As shown in FIGS. 7 and 8, the image analysis / tracking processing unit 53 performs forced displacement of each node on the core line (movement displacement in the three-dimensional space and the core line direction based on the coordinates of the tracking point and the movement vector. ), And a blood vessel shape deformation index is calculated. For example, the image analysis / tracking processing unit 53 calculates the temporal change in the coordinate difference between two adjacent nodes as the expansion / contraction distance ΔL in the core line direction.

  Further, the image analysis / tracking processing unit 53, for each node on each core line, between the node and another node on the blood vessel region cross section including the node (node in the blood vessel lumen, blood vessel wall, or plaque region). The time change of the distance is calculated as the expansion / contraction distance Δr in the radial direction. Further, the image analysis / tracking processing unit 53 calculates, for each tracking point, a twist angle Δθ in the core line direction of the node on the core line based on the coordinates and movement vectors of a plurality of tracking points in the vicinity of the tracking point. To do.

  Further, the image analysis / tracking processing unit 53 may calculate a flow velocity or an average flow velocity or average flow rate in the cross section in the core line direction as a blood flow index by image tracking of a contrast agent or proton in the blood region.

  The blood vessel shape deformation index is used as a forced displacement in the dynamic model. Hereinafter, the time-series blood vessel shape index is referred to as “shape history”, and the time-series blood vessel shape deformation index is referred to as “forced displacement history”.

  When step S2 is performed, the system control unit 21 causes the image processing device 27 to perform a construction process (step S3). In step S <b> 3, the dynamic model construction unit 55 of the image processing device 27 performs shape history (time-series blood vessel shape index), forced displacement history (time-series blood vessel shape index), and time-series medical image (CT image, Based on the DICOM data (MRI image, ultrasonic echo image, etc.), a dynamic model for the analysis target region is provisionally constructed. The dynamic model is a numerical model related to an analysis target region for performing structural fluid analysis.

  Hereinafter, step S3 will be described in detail. The dynamic model construction unit 55 first constructs a shape model for solving a dynamic model (mathematical model) based on the medical image and the shape history. The shape model schematically represents the geometric structure of the blood vessel region at each time. The shape model is divided into a plurality of discretized regions, for example. The vertices of each discretized region are called nodes.

  The dynamic model construction unit 55 may construct a shape model for each time based on a blood vessel region and a blood vessel shape index included in the medical image for each time, or a blood vessel included in the medical image for a specific time phase. A shape model for each time may be constructed based on the region and the blood vessel morphology index. Further, for example, when it is assumed that there is no residual stress in the blood vessel corresponding to the analysis target region as the initial load state, the time phase in which the blood vessel corresponding to the analysis target region is most contracted is set as the time phase in the no stress state. Assume no stress.

  FIG. 9 is a diagram showing a cross section orthogonal to the core wire of the shape model. As shown in FIG. 9, the shape model has a blood vessel lumen region and a blood vessel wall region from the core line toward the outside. When plaque exists, a plaque region may be provided in the blood vessel wall region. Further, when considering the influence on the blood vessel by the blood vessel peripheral tissue, a dummy element of the blood vessel peripheral tissue may be provided outside the blood vessel wall region.

  When the shape model is constructed, the dynamic model construction unit 55 obtains a sampling value relating to the parameter of the latent variable obtained from the probability distribution and the variable range of each latent variable (for example, from a set of combinations of each parameter by Markov chain Monte Carlo method or the like). Set Sampling to the dynamic model.

  For example, as shown in FIG. 2, the dynamic model construction unit 55 sets a boundary condition identification target region (hereinafter referred to as a boundary condition identification region) related to the entrance to the end of the aortic region R1 on the aortic origin side. A boundary condition identification region relating to the exit is set at the end of the coronary artery region R2 and the end of the left coronary artery region R3. The dynamic model construction unit 55 assigns a sampling value related to a boundary condition parameter obtained from the probability distribution of the boundary condition or the variable range to each boundary condition identification region.

  The dynamic model construction unit 55 also includes a material model identification target region (hereinafter referred to as a material model identification region) and a load condition identification target region (hereinafter referred to as a material model identification region) in the aorta region R1, the right coronary artery region R2, and the left coronary artery region R3. Hereinafter, this is referred to as a load condition identification region). The mechanical model construction unit 55 assigns a sampling value related to a material model parameter obtained from the probability distribution or variable range of the material model to each material model identification region, and obtains from each of the load condition identification regions from the probability distribution of the load condition and the variable range. Assign a sampling value for the load condition parameter

  It is said that blood vessels have residual stress even when the flow rate is zero. For example, the dynamic model construction unit 55 may assign the residual stress when the flow rate is 0 to the analysis target region as the initial value of the load condition. In addition, the dynamic model construction unit 55 may set a region for which the geometric structure is to be identified (hereinafter referred to as a geometric structure identification region) in a region where the geometric structure is uncertain. The geometric structure parameter is a variation distribution parameter related to the uncertainty of the geometric structure or a variation distribution parameter inherent in the medical image data. The variation distribution of each CT value or the boundary threshold value of the living tissue Or the like. Although details will be described later, the dynamic model construction unit 55 may set a material model in the plaque region. Details of the material model will be described later.

  When the shape model is constructed, the dynamic model construction unit 55 assigns the time-series blood vessel shape deformation index calculated in step S2, that is, the forced displacement history, to the shape model. A shape model to which a latent variable and a forced displacement history are assigned is called a “mechanical model”.

  FIG. 10 shows a part of a mechanical model of blood vessels and blood as a shape model M2, and is a diagram for explaining assignment of forced displacement history to nodes of the dynamic model. FIG. 10 shows a part of the shape model M2. However, although FIG. 10 shows the case where the core wire is located within M2, the core wire may be located outside M2. As shown in FIG. 10, a plurality of nodes PN (PN1, PN2) are set in the shape model M2. A node on the core line is referred to as PN1, and a node in a blood vessel or blood dynamic model is referred to as PN2. The shape model M2 is set on the dummy element surface, blood vessel outer wall, blood vessel inner wall, plaque region surface, plaque region inside, or blood part. The dynamic model construction unit 55 assigns a forced displacement, that is, a blood vessel shape change index to each node PN1 of the shape model M2 for each time.

  Specifically, the dynamic model construction unit 55 connects adjacent nodes PN1 and PN1 on the core line with beam elements (or rigid elements) EB. Further, the dynamic model construction unit 55 connects the node PN1 and the other node PN2 included in the orthogonal cross section passing through the node PN1 by the beam element EB. The dynamic model construction unit 55 assigns a constraint condition related to the shape displacement direction of each blood vessel shape deformation index to the node PN1 and the beam element EB.

  In the region that identifies the internal pressure of the material model or the blood vessel lumen, the forced displacement includes expansion / contraction of the surface of the blood vessel wall (or dummy element) with respect to the core line direction, twist of the surface of the blood vessel wall (or dummy element), blood vessel wall (or dummy element) ) Surface bending deformation. For example, in a region where the material model and the internal pressure of the blood vessel lumen are not identified, not only forced displacement in the core line direction but also time-series expansion / contraction (displacement) of the blood vessel wall in the radial direction is assigned as the forced displacement history.

  Also, when the internal pressure contributes to deformation outside the cross section in the core line direction, such as when there is a protrusion in the blood vessel lumen or in a blood vessel bifurcation, the area around that area is not forcedly displaced. A forced displacement history is assigned only to a portion (for example, a surface node of a dummy element). In addition, a time-series blood vessel shape deformation index is assigned to the dynamic model construction unit 55, the node PN1, and the beam element EB as a forced displacement history. As a result, expansion deformation, torsion deformation, and bending deformation related to the whole and local blood vessels are expressed.

  The assignment target of the forced displacement history is not limited to the nodes and beam elements on the core line. FIG. 11 and FIG. 12 are diagrams showing another method for assigning the forced displacement history to the shape model. FIG. 11 shows an example of assignment when the blood vessel shape deformation index is twisted. FIG. 12 shows an example of assignment when the blood vessel shape deformation index is bending. As shown in FIGS. 11 and 12, the dynamic model construction unit 55 may directly assign a forced displacement history to the surface or internal node of the shape model. For example, in step S <b> 2, the image analysis / tracking processing unit 53 calculates a blood vessel shape deformation index such as an expansion / contraction amount, a torsion amount, and a bending amount for the feature point. The dynamic model construction unit 55 directly assigns the calculated blood vessel shape deformation index by interpolating (interpolating or extrapolating) the nodes around the feature points.

  The image processing apparatus 27 of the embodiment performs inverse analysis using the dynamic model provisionally constructed in step S3, and statistically identifies latent variables set in the dynamic model. The statistical identification process is performed in step S6 described later. Steps S4 and S5 are provided for calculating a blood vessel shape index and a blood flow index used for the statistical identification process, respectively.

  When step 3 is performed, the system control unit 21 causes the image processing apparatus to perform blood vessel stress analysis processing on 27 (step S4). In step S4, the vascular stress analysis unit 57 of the image processing device 27 performs vascular stress analysis on the current dynamic model to calculate a predicted value of a time-series vascular shape index. The blood vessel shape index may be any of the above-described blood vessel shape fingers. For example, a cross-sectional shape index of a lumen region or a cross-sectional shape index of a blood vessel wall in the blood vessel core direction may be used.

  Specifically, the cross-sectional shape index of the lumen region is at least one of the coordinate value of the pixel of interest in the lumen region and the geometric structure parameters of the lumen region (the radius of the lumen region, the diameter of the lumen region, etc.). One. In addition, the cross-sectional shape index of the blood vessel wall region specifically includes at least the coordinate value of the target pixel of the blood vessel wall region, the geometric structure parameters of the blood vessel wall region (the radius of the blood vessel wall region, the diameter of the wall region, etc.) One. Note that the predicted value means a calculated value of a blood vessel shape index calculated by performing vascular stress analysis on the dynamic model.

  When step 3 is performed, the system control unit 21 causes the image processing device 27 to perform blood fluid analysis processing (step S5). In step S <b> 5, the blood fluid analysis unit 59 of the image processing device 27 performs blood fluid analysis on the tentatively constructed dynamic model to calculate a predicted value of the time-series blood flow index. The blood flow index is at least one of a blood flow rate or a flow velocity, or a spatial and temporal average value thereof. Note that the predicted value means a calculated value of the blood fluid index calculated by performing blood fluid analysis on the dynamic model.

  When steps S4 and S5 are performed, the system control unit 21 causes the image processing device 27 to perform identification processing (step S6). In step S6, the statistical identification unit 61 of the image processing device 27 acquires the blood vessel morphology in which the predicted value of the blood vessel shape index calculated in step S4 and the predicted value of the blood flow index calculated in step S5 are collected in advance. The parameters of the latent variables of the mechanical model are statistically identified so as to match the observed value of the index and the observed value of the blood flow index.

  As illustrated in FIG. 4, the statistical identification unit 61 includes a first statistical identification unit 61-1 and a second statistical identification unit 61-2. The first statistical identification unit 61-1 statistically identifies the parameter of the latent variable so that the predicted value of the blood vessel shape index matches the observed value of the blood vessel shape index. The second statistical identification unit 61-2 statistically identifies the parameter of the latent variable so that the predicted value of the blood flow index matches the observed value of the blood flow index. Hereinafter, the first statistical identification unit 61-1 and the second statistical identification unit 61-2 will be described in order.

  Specifically, in step S6, the first statistical identification unit 61-1 sets a data distribution based on the predicted value and the observed value of the blood vessel shape index calculated in step S4. The data distribution indicates, for example, a multivariate normal distribution function related to an error between the predicted value and the observed value of the blood vessel shape index.

  A normal distribution function value regarding the error between the predicted value and the observed value at each node or each element in the dynamic model is calculated, and the product of these is set as a data distribution in the statistical identification processing unit. The data distribution may be set individually for each time or may be set collectively for a plurality of times.

  Next, the first statistical identification unit 61-1 assigns a prior distribution (prior probability distribution) to the latent variable of the dynamic model. Specifically, a prior distribution is assigned to each of the material model, the boundary condition, the load condition, and the parameters related to the uncertainty of the time-series shape index and shape deformation index. For example, a prior distribution relating to a pressure value related to a blood vessel lumen, which is one of parameters of a load condition, is assigned. The range of values that can be taken by the pressure value (assumed range) can be empirically limited in advance. The first statistical identification unit 61-1 calculates a probability distribution of internal pressure values, that is, a prior distribution, for each discretized region by executing a Monte Carlo simulation related to the internal pressure values within the assumed range.

  In addition, as a prior distribution, the dynamic model construction unit 55 has observed that the pressure distribution in the core line direction is smooth, that the pressure change with the passage of time is smooth, and that there is no backflow of blood flow. In this case, for example, a probability distribution expressed mathematically by a multivariate normal distribution function that the slope of the average pressure change in the core line direction is negative may be set as the prior distribution.

  It is possible to execute a Monte Carlo simulation regarding the parameters of the load condition according to the assumed probability distribution, and to obtain a sampling value of the load condition (latent variable) to be set in the dynamic model, within the assumed range. .

  Next, the first statistical identification unit 61-1 calculates a posterior distribution (a posteriori probability distribution) by performing a statistical identification process on the prior distribution and the data distribution for each latent variable. Examples of the statistical identification process include a hierarchical Bayes model and a Markov chain model. And the 1st statistical identification part 61-1 identifies the parameter of each latent variable from statistical values, such as the mode value and average value of posterior distribution, about each latent variable. For example, in the case of the above-described example, the posterior distribution related to the vascular lumen pressure value is calculated, and the identification value of the vascular lumen pressure value is calculated from the posterior distribution.

  FIG. 13 is a diagram for explaining the calculation of the posterior distribution and the identification of the average internal pressure regarding the load condition (average pressure in the blood vessel) by the hierarchical Bayes model and the Markov chain Monte Carlo method. As shown in FIG. 13, it is assumed that a calcified plaque region and a hook-like plaque region are set in a blood vessel extending from the blood vessel starting portion. The calcified plaque region is set as the material model identification region, and the calcified plaque region is set as the material model identification region. The blood pressure in the blood vessel decreases as the blood vessel progresses along the blood vessel core direction from the blood vessel starting portion. A plurality of nodes are set along the blood vessel core line. The posterior distribution of the intraluminal pressure is calculated in the orthogonal section (node section) including each node, and the mode value of the posterior distribution is identified.

  For example, the blood vessel shape index calculated in step S2 is used as the observed value of the blood vessel shape index.

  The process by the second statistical identification unit 61-2 is the same as the process by the first statistical identification unit 61-1, except that the index used for calculating the data distribution is different. That is, the second statistical identification unit 61-2 first sets a data distribution based on the predicted value and the observed value of the blood flow index calculated in step S5.

  Next, the second statistical identification unit 61-2 assigns a prior distribution to the latent variables of the dynamic model. For example, a prior distribution relating to a material model parameter relating to blood vessels, a material model parameter relating to blood, and a material model parameter relating to plaque is assigned. Examples of parameters of these material models include material model parameters such as elastic modulus, and parameters related to viscosity in blood constitutive equations. The assumed range and probability distribution of the parameters of the material model can be set empirically in advance.

  The second statistical identification unit 61-2 sets the probability distribution of the parameter of the material model for each discretized region, that is, sets a prior distribution, limits the range within these assumed ranges, and follows the assumed probability distribution of the material model. A Monte Carlo simulation on the parameters can be executed, and a sampling value of the material model parameter (latent variable) for setting to the dynamic model can be obtained.

  Next, the second statistical identification unit 61-2 calculates a posterior distribution by applying a statistical identification process to the prior distribution and the data distribution for each latent variable, and calculates each statistic value of the calculated posterior distribution. Identify the parameters of the latent variable. For example, in the case of the above-described example, the posterior distribution relating to the parameter of the material model is calculated, and the identification value of the parameter of the material model is calculated from this posterior distribution.

  FIG. 14 is a diagram for explaining a posteriori distribution calculation and identification of a material model parameter (equivalent elastic modulus of a blood vessel wall) regarding a material model parameter by a hierarchical Bayes model and a Markov chain Monte Carlo method. As shown in FIG. 14, the blood vessel model is assumed to be the same as that in FIG. By limiting to the material model identification region, the posterior distribution of the parameter (for example, equivalent elastic modulus) of the material model of the blood vessel wall is calculated, and the mode value of the posterior distribution is identified.

  Note that the observed value of the blood flow index is assumed to be, for example, a change in the blood flow sent to the aorta, and the observed value of the blood vessel shape index is changed from the time-series CT image by image processing. It can be used as a value (CFA). The flow rate and flow rate may be calculated by calculating the temporal change in the amount of movement of the feature points by image tracking of the contrast agent after injection of the contrast agent into the coronary artery. In addition, the concentration change amount of a specific region in the blood vessel core direction or temporal direction of the contrast agent is acquired, and the flow rate and the value of the concentration change divided by the distance in the core direction direction of each region and the temporal change rate of the concentration change are calculated. The flow rate may be calculated. In the case of MRI, proton image tracking is used, and in the case of ultrasonic echoes, the flow rate is calculated by contrast echography or the like.

  In addition, if it is not assumed that the coordinate value of each pixel in the analysis target area is a definite value, that is, if it is assumed that the geometric structure of the analysis target area is uncertain, the geometrical A structure may be included. In this case, the statistical identification unit 61 pre-distributes a probability distribution such as a normal distribution expressing a variation within a predetermined range of the coordinate value of each node in the core line direction or a variation within a predetermined range of the diameter of the analysis target region. It is good to set as. In this case, a restriction that the shape of the analysis target area is smooth and that the order of the nodes in the core line is unchanged may be set as a prior distribution. It is possible to perform Monte Carlo simulations on geometric parameters according to the assumed probability distribution, and limit the uncertainty parameters of the geometric structure (latent variables) to be set to the dynamic model. ) Sampling value can be obtained.

  In each step S6, both the statistical identification process by the first statistical identification unit 61-1 and the statistical identification process by the second statistical identification unit 61-2 may not be performed. That is, in each step S6, either the statistical identification process by the first statistical identification unit 61-1 or the statistical identification process by the second statistical identification unit 61-2 may be performed.

  In the above example, the first statistical identification unit 61-1 statistically identifies the parameters of the latent variable so that the predicted value of the blood vessel shape index matches the observed value of the blood vessel shape index, and the second The statistical identification unit 61-2 statistically identifies the parameters of the latent variable so that the predicted value of the blood flow index matches the observed value of the blood flow index. However, based on the structure-fluid coupling analysis, the statistical identification unit 61 matches the predicted value of the blood vessel shape index and the predicted value of the blood flow index with the observed value of the blood vessel shape index and the observed value of the blood flow index. Thus, the parameters of the latent variable may be identified statistically. Further details of the statistical identification process by the statistical identification unit 61 will be described later.

  When step S6 is performed, the system control unit 21 causes the image processing device 27 to perform setting processing (step S7). In step S7, the dynamic model construction unit 55 of the image processing device 27 sets the parameter of the latent variable calculated in step S6 to the dynamic model.

  When step S7 is performed, the system control unit 21 determines whether or not the identification end condition is satisfied (step S8). If it is determined in step S8 that the identification end condition is not satisfied (step S8: NO), the system control unit 21 repeats steps S4, S5, S6, S7, and S8.

  Here, the identification end condition is expressed by whether or not an index for determining the end of identification (hereinafter referred to as an identification end index) reaches a specified value. As an identification end index, for example, a difference value between a predicted value and an observed value of a blood vessel morphology index can be cited. In this case, the system control unit 21 determines that the identification end condition is not satisfied when the difference value is larger than the predetermined value, and determines that the identification end condition is satisfied when the difference value is smaller than the predetermined value. To do.

  Further, the identification end index may be, for example, the number of sampling points of the Monte Carlo method. In this case, when the number of sampling points is smaller than the predetermined value, the system control unit 21 determines that the identification end condition is not satisfied, and when the number of sampling points is larger than the predetermined value, the identification end condition is Determined to be satisfied. When the identification end condition is satisfied, the dynamic model construction unit 55 sets the latest dynamic model at that time as the final dynamic model.

  When the final mechanical model is constructed, the mechanical model construction unit 55 associates the observed value of the blood vessel shape deformation index, the load condition parameter in the final mechanical model, and the material model parameter (hereinafter referred to as the related model). Called a model). The related model is stored in the storage device 33. The related model may be stored in association with patient information, examination information, or the like for ease of search. Note that the observed values of the blood vessel shape index and blood flow index, the load condition parameter in the final dynamic model, and the material model parameter do not necessarily have to be associated with each other in the form of the model. Also good.

  The above steps S4, S5, S6, S7, and S8 may be repeated with the same identification method, or may be repeated with different identification methods. When iterating with different identification methods, for example, first, a latent variable may be tentatively identified using a simple dynamic model, and then a latent variable may be accurately identified using a continuum dynamic model. . Thus, by performing the statistical identification process in two steps using different methods, the parameters of the latent variable can be converged in a short time. As a method of using a simple dynamic model, there is a method of using internal pressure and external pressure for the material dynamics equation of the thick-walled cylinder, Hagen-Poiseuille flow and modified Bernoulli equation. As a method using the continuum mechanics model, FEM structural fluid analysis can be mentioned. Details of the identification method using a simple mechanical model and the identification method using a continuum mechanical model will be described later.

  If it is determined in step S8 that the identification completion receipt is satisfied (step S8: YES), the system control unit 21 may cause the image analysis / tracking processing unit 53 to perform correction processing (step S9). In step S9, the image analysis / tracking processing unit 53 performs the structural fluid analysis result (predicted value of the mechanical index and predicted value of the blood fluid index) performed based on the latent variable obtained by the inverse analysis by the statistical identification method. The shape of the blood vessel region included in the time-series medical image may be corrected so as to match the observed value (the observed value of the mechanical index and the observed value of the blood fluid index).

  The display device 31 displays a diagnosis result based on the corrected time-series medical image. Thereby, the blood vessel analysis device 50 can display the diagnosis result in consideration of the final dynamic model. Alternatively, the display device 31 may display on the screen a blood vessel location / region where the identification by the inverse analysis and the observation result do not match by the structural fluid analysis.

  For example, an image of a cardiac phase in which the behavior of a blood vessel is fast is often blurred, and there are locations and regions with large errors in the blood vessel shape observed by image analysis based on a medical image. The cardiac phase data in which the blood vessel behavior is relatively stable can be obtained from an image with little noise. Based on such blood vessel shape data with a small error distribution, the blood vessel shape in the cardiac phase with a large error can be correctly interpolated by using a dynamic model. Along with the correctly interpolated shape, it can be displayed with the variation distribution from the original data. Thereby, the stability of the blood vessel shape display can be ensured and the user can recognize the uncertainty of the shape.

  When step S9 is performed, the system control unit 21 causes the image processing device 27 to perform vascular stress analysis processing (step S10). In step S <b> 10, the vascular stress analysis unit 57 of the image processing device 27 performs vascular stress analysis on the final mechanical model to calculate a spatial distribution of predicted values of time-series mechanical indices. Specifically, the predicted value of the mechanical index is calculated for each discretized region.

  When step S9 is performed, the system control unit 21 causes the image processing device 27 to perform blood fluid analysis processing (step S11). In step S11, the blood fluid analysis unit 59 of the image processing device 27 performs blood fluid analysis on the tentatively constructed dynamic model to calculate a spatial distribution of predicted values of the time series blood flow index. Specifically, the predicted value of the blood flow index is calculated for each discretized region.

  Note that FFR may be calculated as a mechanical index or a blood flow index.

  When steps S9 and S10 are performed, the system control unit 21 causes the display device 31 to perform display processing (step S11). In step S11, the display device 31 displays the predicted value of the time-series dynamic index calculated in step S9 and the predicted value of the time-series blood flow index calculated in step S10.

  For example, the display device 31 displays a time-series dynamic index or a time-series blood vessel flow index as a moving image of a time-series dynamic model in a color corresponding to the predicted value. For this reason, the display device 31 holds a color table indicating the relationship between various predicted values and color values (for example, RGB). The display device 31 specifies a color value corresponding to the predicted value using a color table, and displays a discretization region corresponding to the predicted value with a color corresponding to the specified color value.

  FIG. 15 is a diagram illustrating a display example of a spatial distribution of internal pressure that is one of the mechanical indices. As shown in FIG. 15, the display device 31 displays each discretized area constituting the dynamic model in a moving image with a color corresponding to the internal pressure value related to the discretized area. By observing the dynamic model, the user can grasp the dynamic index that changes with time and space by color.

  FIG. 16 is a diagram illustrating a display example of a spatial distribution of a flow velocity value that is one of blood flow index. As shown in FIG. 16, the display device 31 displays each discretized area constituting the dynamic model in a moving image with a color corresponding to the flow velocity value related to the discretized area. By observing the dynamic model, the user can grasp the blood flow index that changes with time and space by color.

  For example, when the inside of the blood vessel is completely stenotic, the internal pressure at the stenotic site is smaller than the internal pressure at the non-stenotic site. When the internal pressure is designated as the mechanical index, the user can determine the presence or absence of stenosis based on a local color difference on the dynamic model. Further, the flow rate of the stenosis site is smaller than that of the non-stenosis site. When the flow rate is designated as the blood flow index, the user can determine the presence or absence of stenosis based on a local color difference on the dynamic model.

  Further, the vascular stress analysis unit 57 may calculate the spatial distribution of the hardness value as a mechanical index based on the identification result of the material model parameter of the plaque region. In this case, the display device 31 may display the spatial distribution of the hardness value related to the plaque region on the dynamic model. The display device 31 may display an internal pressure distribution, a stress distribution, and a strain distribution around the plaque region. The user can use these displays to estimate the nature of the plaque and the likelihood of failure.

  The predicted values of the mechanical index and the blood flow index are not limited only to the method of expressing with the color of the discretized region of the dynamic model. For example, as shown in FIG. 17, FIG. 18, and FIG. FIG. 17 is a graph relating to the blood pressure of the left coronary artery origin. The vertical axis of the graph of FIG. 17 is defined by normalized blood pressure, and the horizontal axis is defined by cardiac phase [%]. FIG. 18 is a graph relating to blood pressure near the branch point between LCX and LDA. The vertical axis of the graph of FIG. 18 is defined by normalized blood pressure, and the horizontal axis is defined by cardiac phase [%]. FIG. 19 is a graph relating to changes in blood pressure with respect to the core line direction. The vertical axis of the graph of FIG. 19 is defined by the blood pressure ratio, and the horizontal axis is defined by the distance [mm] from the aorta. The display device 31 allows the user to easily grasp these values by displaying the predicted values of the mechanical index and the blood flow index in a graph.

  When step S11 is performed, the structural fluid analysis process ends.

  In FIG. 10, the forced displacement history is set in the core line portion and the outer wall portion of the shape model, but the setting location of the forced displacement history is not limited to this. For example, the forced displacement history may be set in a blood vessel wall region between the core line portion and the outer wall portion.

  Further, the assignment target of the constraint condition of the forced displacement history may be divided according to whether the boundary condition and the material model are identified. FIG. 20 is a diagram illustrating another allocation example of the forced displacement history, and illustrates a cross section of the shape model. For example, as shown in FIG. 20A, when the boundary condition and the material model are identified, the forced displacement history is assigned only to the node PN2 on the outer wall portion OW of the shape model, and the node PN3 of the blood vessel wall region RV is assigned to the node PN3. There is no need to assign a forced displacement history.

  Further, as shown in FIG. 20B, if the boundary condition and the material model are not identified, the forced displacement history is assigned to both the node PN2 of the outer wall portion of the shape model and the node PN3 of the blood vessel wall region RV. Good. In this case, a forced displacement history is assigned to the node PN1 on the core line. Further, the node PN2 and the node PN1 of the outer wall portion OW may be connected by the beam element EB, and the forced displacement history may be assigned to the nodes PN2 and PN3 on the beam element EB. At this time, contraction and expansion in the radial direction are expressed by expansion and contraction of the beam element EB. Note that the forced displacement history need not be assigned to the blood vessel lumen region RI.

  FIG. 21 is a diagram showing another example of assignment of forced displacement history, and shows a cross section of a shape model including a dummy element RD of a tissue around a blood vessel. As shown in FIG. 21, the dummy element RD is set outside the blood vessel wall region RN. When the shape model includes the dummy element RD, the node PN4 is set in the dummy element RD in addition to the blood vessel wall region RN. A forced displacement history is also assigned to the node PN4.

  The mechanical model construction unit 55 assigns a forced displacement history to the node PN3 included in the blood vessel wall region RV when identifying the boundary condition and the material model, and does not assign a forced displacement history when the boundary condition and the material model are not identified. Also good. When the forced displacement history is assigned to the node PN3, the material model is identified with reference to the shape index related to the blood vessel wall region RV in addition to the shape index related to the lumen region RI.

  FIG. 22 is a diagram showing another example of assignment of the forced displacement history, and shows a cross section of the shape model including the plaque region RP. As shown in FIG. 22, the plaque region RP is included in the blood vessel wall region RV. The plaque region RP is set in the material model identification region. For the plaque region RP, a material model is identified in consideration of the lumen shape index, the blood vessel wall shape index, and the plaque index.

  As described above, the plaque index is, for example, data relating to the plaque property obtained by the tissue property diagnosis by the ultrasonic diagnostic apparatus. The mechanical model construction unit 55 divides the plaque region into a plurality of partial regions according to the properties, and individually sets the material model identification regions in the plurality of partial regions. It is preferable to set a parameter range in advance for each partial region according to the properties of the partial region. By the above-described statistical identification process, the material model parameters for each partial region are identified by the statistical identification unit 61. In step S10, the display device 31 displays an index related to the material characteristics of the blood vessel as a mechanical index, so that the user can accurately and easily grasp the property of the plaque.

  Next, a material model that is one of latent variables will be described in detail. As a blood vessel material model, an elastic model, a superelastic model, an anisotropic hyperelastic model, a superelastic model considering viscosity characteristics, and the like are applicable. Examples of anisotropic superelastic models include Y. C. A mathematical model proposed by Fun or a mathematical model called Holzapfel-Gasser constitutive equation can be applied.

  The strain energy per unit reference volume is expressed by the following equation (1). The first term of the formula (1) represents the energy related to the shear deformation of the isotropic basic material not containing collagen. Further, the second term of the formula (1) represents the energy related to the volume deformation of the isotropic basic material not containing collagen. The third term of the formula (1) represents the contribution of each group of collagen fibers (considering the dispersion in the fiber direction).

  FIG. 23 is a diagram illustrating deformation of a fiber group. As shown in FIG. 23, a cylindrical outer membrane is assumed. The deformation of the fiber group in the average direction A on the surface defined by the core line direction z and the circumferential direction θ is expressed by the following equation (2).

  Parameters relating to the material model in the equations (1) and (2) include a material parameter and a fiber dispersion parameter as shown in Table 1 below. C10, D, K1, K2, etc. are used as material parameters, and Kappa, γ, etc. are used as fiber dispersion parameters. The default values and constraint conditions for each parameter are as shown in Table 1.

  As the blood material model, a Casson constitutive expression such as the following expression (3) and an HB constitutive expression such as the expression (4) are suitable.

  The parameters of the material model are identified by the statistical identification unit 61 in the above-described step S6 by statistical identification processing using the blood vessel shape index and the blood fluid index.

  Next, details of the statistical identification processing performed by the statistical identification unit 61 will be described.

  Observation variables such as blood vessel shape indices and blood flow indices measured from time-series medical images have uncertainties. The statistical identification unit 61 uses a statistical method based on a hierarchical Bayesian model and a Markov chain Monte Carlo method as a statistical identification method of a latent variable in a situation where such uncertainty exists.

  As described above, in step S6, the statistical identification unit 61 sets a data distribution based on the predicted value and the observed value of the blood vessel shape index or blood flow index calculated in step S4. The data distribution indicates, for example, a multivariate normal distribution function related to an error between a predicted value and an observed value of a blood vessel shape index or a blood flow index. The data distribution may be set individually for each time or may be set collectively for a plurality of times.

  Next, the statistical identification unit 61 assigns a prior distribution to the forced displacement and the latent variable of the shape model. The prior distribution indicates a probability distribution of possible values. Next, the statistical identification unit 61 executes a parameter survey of a numerical simulation regarding the latent variable, and constructs a model that expresses the relationship between the latent variable and the blood vessel shape index or the blood flow index. For example, the relationship between the material model parameter, the internal pressure distribution parameter, and the blood vessel shape index or blood flow index is defined in the model. The relationship between the blood vessel shape index or blood flow index and the latent variable may be defined by a database or a table instead of a model. These models, databases, or tables are stored in the storage device 33.

  The statistical identification unit 61 calculates a probability distribution of the blood vessel shape index or the blood flow index from the prior distribution using a model, a database, or a table. The statistical identification unit 61 statistically identifies latent variables from the posterior distribution obtained by the hierarchical Bayesian model and the Markov chain Monte Carlo method.

  Specifically, this identification problem is a defect setting problem that does not satisfy the following three conditions. The three conditions are (1) the existence of a solution is guaranteed, (2) the solution is uniquely determined, (3) the solution continuously changes with respect to the data, and the solution is stable with respect to the measurement error, It is. The defect setting problem is easy to handle when viewed within the framework of normalization theory and its extensions. The standard normalization theory is insufficient to solve the defect setting problem. In order to solve the defect setting problem, a methodology for detecting discontinuities in the internal state and using the detected discontinuous points for estimation of the internal state is required. In this respect, Markov random field theory is effective in the statistical identification processing of the present embodiment.

  In an environment where uncertainty exists in the blood vessel shape index and the blood flow index, the statistical identification unit 61 identifies the probability distribution parameter of the latent variable under an appropriate constraint condition. In order to determine appropriate constraints, it is necessary to know the nature of the solution in advance. The statistical identification unit 61 generates a database related to the constraint condition of the solution space based on the simulation and the observation value.

  The statistical identification unit 61 uses the generated database to execute statistical identification processing based on Markov random field theory and a hierarchical Bayesian model for a super-multi-degree-of-freedom large-scale problem. In the setting of the prior distribution as a constraint condition, the probability distribution of parameters relating to these factors is individually configured in parallel based on many numerical experimental results. The statistical identification unit 61 identifies parameters of latent variables by integrating a plurality of probability distributions and interpolating data loss. For this processing, the statistical identification unit 61 performs estimation by a hierarchical Bayesian method based on a model using Markov random field theory. This is a mechanism that can estimate pressure and flow rate distribution under certain load conditions and boundary conditions based on the identified intermediate variables from the measurement results of the deformation state of the structure to be analyzed.

  The identification problem of material models, boundary conditions, and load conditions in structural fluid analysis of coronary arteries is positioned as nonlinear inverse analysis, and the uniqueness and stability of the solution are often not guaranteed. Since realistically possible ranges of material properties and blood pressure of living tissue can be assumed as a priori information, these can be set as a probability distribution of prior distribution. In addition, since pressure and displacement can be assumed to be spatially and temporally smooth, this information can also be set as a priori probability distribution as a priori information.

  Alternatively, when the fact that no back flow occurs in the blood flow can be taken into account, the fact that the slope of the overall pressure distribution in the blood vessel core direction is negative (there is a pressure drop) may be used as a constraint condition. For load conditions (internal pressure distribution, etc.), boundary conditions, and material model, square error distribution between observed value of blood vessel shape deformation index based on time-series CT image and predicted value of blood vessel shape deformation index based on dynamic model Can be set as the data distribution.

  A square error distribution related to the observable average flow rate may be added as a data distribution. Based on these prior distribution and data distribution, the posterior distribution can be calculated using the hierarchical Bayesian model and the Monte Carlo method. The identification value of the parameter of the latent variable can be obtained by the occurrence probability and variance of the posterior distribution. It can be said that the higher the probability of occurrence and the smaller the variance, the higher the confidence value.

  Even when the posterior distribution is a multimodal distribution, an identification value having a small variance among a plurality of identification values may be selected. Or, when multiple identification values can exist, perform structural fluid analysis under each identification condition, recognize each possibility, and use the identification values and analysis results as guide information for diagnosis and prevention be able to. Since an error is included in the time-series CT image, an error is also included in the blood vessel shape index related to each node of the dynamic model. For this reason, for example, each blood vessel shape index is treated as a normal distribution random variable having an average value of the predicted value of the blood vessel shape index measured from a time-series CT image, and includes a restriction that the spatial order of positions is maintained. In addition, a prior distribution may be set.

  In addition, there are cases where there is no uniqueness in identification of parameters of latent variables, and a plurality of candidates can be considered. In this case, by checking the fluctuation range of the sample set of the identification value of the latent variable with respect to the random sampling point according to the uncertainty of the blood vessel shape index measured from the time-series CT image, the robustness of each identification value candidate is checked. Judgment (stability). The final identification value may be determined based on the robustness of each identification value candidate.

  Next, details of the dynamic model will be described. As described above, the dynamic model construction unit 55 can construct different types of dynamic models depending on the type of the dynamic model. When FEM based on continuum mechanics is used, the mechanical model construction unit 55 includes both a shape model for stress analysis of the blood vessel wall (FEM model) and a shape model for fluid analysis of blood (FEM model). Build up.

  When a simple identification method based on material mechanics is used, the relationship between pressure, elastic modulus, and displacement is approximately obtained from the equation of a thick cylinder that receives internal pressure in material mechanics. In this case, thick cylindrical approximation is used as a shape model for each of a plurality of discretized regions arranged in the core line direction. Specifically, the dynamic model construction unit 55 identifies the blood vessel lumen shape, the blood vessel wall surface shape, and the cross-sectional center on the cross section passing through the nodes discretely arranged on the core line.

  Next, the dynamic model construction unit 55 calculates the average area, the average radius of the lumen, and the average wall thickness based on the blood vessel lumen shape and the blood vessel wall surface shape. Based on the average area, the average radius of the lumen, and the average wall thickness, the dynamic model construction unit 55 constructs a shape model by performing thick-walled cylinder approximation on the blood vessel region of each discretization region.

  When a simple identification method based on flow mechanics is used, the modified Bernoulli equation in fluid mechanics or the Hagen-Poiseille Flow is used to approximate the mean blood pressure and mean flow. Use the formula. In this case, a shape model for approximating the relationship between the blood pressure difference and the flow rate of each of the plurality of discretized regions is constructed. Here, the blood pressure difference is a pressure difference between the inlet blood pressure and the outlet pressure, and the flow rate means the inlet flow rate (or flow velocity) and the outlet flow rate (or flow velocity) per unit time. However, the dynamic model construction unit 55 associates the movement vector for each node with the rotational displacement for the node adjacent to the node at each time based on the expansion / contraction distance and the twist angle in the core line direction calculated in step S2. And good.

  Next, structural fluid analysis using a continuum mechanics model will be described.

  The kinematics for blood vessels and blood (substances that make up blood vessels and blood) are independent of the forces causing that motion. The basic concept of kinematics for blood vessels and blood is an abstraction of the intuitive notion of position, time, objects, motion, and a collection of deformable substances into mathematical terms. The basic kinematic tensors that govern the local analysis of deformation and motion for blood vessels and blood are the deformation gradient tensor F and the velocity gradient tensor L.

  The deformation gradient tensor F defines the size and shape changes that occur in the moving blood vessels and blood material elements. The deformation gradient tensor F is represented by the product of a rotation tensor (strict orthogonal tensor) R and a stretching tensor (positive symmetric tensor) U and V. The stretching tensors U, V are brought about by first imposing a stretch on the direction defined by the orthonormal vector R in the basic form and then imposing a rigid body rotation given by the orthonormal vector R. Note that the order of imposing stretch and the order of imposing rigid body rotation may be reversed.

  The velocity gradient tensor L does not depend on the reference form but depends only on the current form. The velocity gradient tensor L defines the rate at which the size and shape changes that occur in the moving blood vessels and blood material elements occur. The velocity gradient tensor L can be separated into a strain rate tensor D (symmetric tensor) and a spin tensor (antisymmetric tensor). The strain rate tensor D represents the rate of change of the stretch when the object just passes through its current form. The spin tensor represents the rate of change of rotation when an object just passes through its current form.

  Displacement constraint conditions (including temporal changes) are assigned to the deformation gradient tensor and velocity gradient tensor relating to a part of the outer surface node (or integration point) of the mechanical model of the blood vessel. Predicted values (deformation gradient tensor, velocity gradient tensor, or their function values (for example, displacement or area)) and observation values (deformation obtained from observation data) for the lumen node (or integration point) of the dynamic model The material model parameters, loading conditions (surface force vector in the lumen of the dynamic model), and boundary conditions (force vector at the vessel boundary) are matched so that the gradient tensor, velocity gradient tensor, or their function values match. Identify. Here, the initial state of the stress inside the blood vessel may be assumed in advance or may be identified.

  Dynamic models based on continuum mechanics are based on equations representing the balance of mass, kinematics, momentum, angular momentum, and energy in moving blood vessels and blood. The concepts of mass, force, heat, and internal energy are fundamental. The balance law is that the momentary total derivative of momentum is equal to the sum of object force and contact force, the momentary total derivative of angular momentum is equal to the sum of object torque and contact torque, and kinetic energy and internal It means that the change with time in energy is equal to the sum of power (mechanical energy), heat supply per unit time and heat flux.

  From balance laws, constitutive equations, and jump conditions, field equations such as deformation gradient tensors, velocity gradient tensors, and stress tensors can be derived to describe vascular and blood dynamic models. In addition, the strain fields in blood vessels and blood satisfy the fitting conditions.

  Here, the constitutive equation gives a relationship of a set of 10 scalar equations composed of density, internal energy, velocity vector, stress tensor, heat flux vector, and temperature. Of the 17 scalar fields in the field balance equation, ie, density, internal energy, velocity vector, stress tensor, and heat flux vector, the field balance gives 8 scalar relationships and added the temperature Give the remaining unknown relationships. However, the object force b and the heat source r are known. The parameters of the equations giving these scalar fields are material model parameters.

  The mechanics model based on continuum mechanics is a numerical analysis method using the finite element method or the boundary element method, and approximates the field such as displacement vector, stress tensor, strain tensor, and velocity vector, given boundary condition, load condition, And can be calculated based on the material model.

  In the structure-fluid coupled analysis, the method of solving the structure and fluid equations may be either a monolithic method or a partitioned method. Further, the coupling at the interface between the structure and the fluid may be weakly coupled or strongly coupled. Further, in fluid analysis, a moving boundary may be handled by a boundary surface tracking type technique represented by the ALE method, or “Immersed Boundary method”, “Immersed Finite Element Method”, or “Fittious Domain Method”. A boundary surface supplement type technique such as the above may be used.

  Next, as an example of a simple dynamic model, the material dynamics equation of the thick cylinder subjected to internal pressure and external pressure, the Hagen-Poiseuille flow and the modified Bernoulli equation will be described in detail.

  First, an equation of material dynamics of a thick cylinder will be described with reference to FIGS. FIG. 24 is a diagram showing an orthogonal cross section of a dynamic model of a thick cylinder. FIG. 25 is an enlarged view of the minute sector element of FIG. Expressions such as stress, strain, and displacement when the internal pressure pa and the external pressure pb are applied to a thick cylinder having the inner radius ra and the outer radius rb will be described. E and ν are material model parameters. E represents the elastic modulus, and ν represents the Poisson's ratio. For thick cylinders, it is necessary to consider the radial stress σr and the radial distribution of the circumferential stress σθ. Hereinafter, the axial strain εz is assumed to be uniform with respect to the position and orientation of the cross section.

  Moreover, although the case where a cylindrical cross section is axially symmetrical is demonstrated, arbitrary shapes may be sufficient. When the cylindrical cross section is axisymmetric, the equilibrium condition may be considered only in the radial direction of an arbitrary cross section. Consider a balance of forces in the radial direction on a concentric cylinder with radii r and r + dr on a given cross section and a small sector element with unit thickness 1 cut at a central angle dθ. Since the deformation is also axisymmetric, no shear stress is generated on the ab plane and the bc plane, so that only normal stress acts. Therefore, the force balance in the radial direction can be expressed as the following equation (5).

  σr rdθ + 2σθr sin (dθ / 2)-(σr + (dσr / dr) dr) (r + dr) dθ = 0 (5)

  Here, since dr is smaller than r and dσr is smaller than σr, if high-order minute terms included in equation (5) are omitted and sin (dθ / 2) ≈dθ / 2, then equation (5) Can be expressed as the following equation (6).

  rdσr / dr + σr-σθ = 0 (6)

  If the radial displacement at the radius r is u, the displacement in the same direction at u + dr is u + (du / dr) dr, so the radial strain εr at the radius r is εr = du / dr. Further, due to the radial displacement u, the circle with the radius r becomes a circle with the radius r + u. Therefore, the circumferential strain εθ can be expressed as the following equation (7).

  εθ = (2π (r + u) -2πr) / 2πr = u / r (7)

  Moreover, the following (8) Formula or (9) Formula is obtained from the relational expression of stress and strain.

  d2u / dr2 + (1 / r) (du / dr) -u / r2 = 0 (8)

  d2u / dr2 + d (u / r) / dr = 0 (9)

  The following formula (10) can be obtained by integrating the formula (8) or the formula (9).

  u = c1r + c2 / r (10)

  As a result, the following equations (11), (12), and (13) are obtained.

  σr = (E / ((1 + ν) (1-2ν))) (c1- (1-2ν) (c2 / r2) + νεz) (11)

  σθ = (E / ((1 + ν) (1-2ν))) (c1 + (1-2ν) (c2 / r2) + νεz) (12)

  σz = (Eν / ((1 + ν) (1-2ν))) (2c1 + ((1-ν) / ν) εz) (13)

  The constants c1 and c2 in the equations (11), (12), and (13) are determined from the peripheral conditions, that is, σr = −pa at the inner circumference r = ra of the cylinder and σr = −pb at the outer circumference r = rb. be able to. Under these peripheral conditions, the following expressions (14), (15), and (16) can be obtained from the expressions (11), (12), and (13), respectively. Further, the displacement u can be expressed as the following equation (17).

  σr = (1 / (rb2-ra2)) (ra2 (1-rb2 / r2) pa-rb2 (1-ra2 / r2) pb) (14)

  σθ = (1 / (rb2-ra2)) (ra2 (1 + rb2 / r2) pa-rb2 (1 + ra2 / r2) pb) (15)

  σz = 2ν (ra2pa-rb2pb) / (rb2-ra2) + Eεz = ν (σr + σθ) + Eεz (16)

  u = ((1 + ν) (1-2ν) / E) ((ra2pa-rb2pb) / (rb2-ra2)) r + ((1 + ν) / E) ((ra2rb2) / ((rb2-ra2) r)) (pa-pb) -νεzr (17)

  Expressions (16) and (17) include the term εz. Therefore, equation (16) varies depending on the constraint condition at the boundary of the target cylinder. For example, when both ends of the target cylinder are constrained, εz = 0, both ends are opened, and σz = 0.

  The shear stress τr in the rθ plane can be expressed as the following equation (18).

  τr = (1/2) (σr−σθ) = ((ra2rb2) / ((rb2-ra2) r2)) pb (18)

  Since the shear stress τr ′ in the θz plane becomes maximum when σz = 0, when σz = 0, it can be expressed as the following equation (19).

  τr ′ = (1/2) | σθ | = (1/2) ((rb2) / (rb2-ra2)) (1 + ra2 / r2) pb (19)

  Next, the Hagen-Poiseuille flow and the modified Bernoulli equation will be described. When the fluid flows into the circular pipe, the pressure decreases as it proceeds downstream, and the flow velocity distribution also gradually changes. When the flow flows into the pipe, a boundary layer develops from the pipe wall, and as the flow proceeds downstream, the boundary layer increases in thickness, and finally the flow in the pipe is covered with the boundary layer. For this reason, the velocity distribution changes from a substantially flat distribution at the pipe inlet to a downstream parabolic distribution, and the velocity distribution thereafter does not change. This state is called a fully developed flow, and the pressure drop due to pipe friction loss is also a constant rate. The section where the flow reaches the flow developed from the pipe entrance is called the run-up section or the entrance section, and the length of the section is called the run-up distance or the entrance length. Since the velocity distribution of the fully developed flow does not change in the downstream direction, as shown in the following equation (20), the acting force of the pressure loss ΔP caused by the pipe friction loss and the friction force of the shear stress τ caused by the viscosity of the fluid Will be balanced. In the case of a flow in a circular pipe, it is said that the flow becomes laminar when the Reynolds number Re is about 2300 or less.

  In the case of laminar flow, the velocity distribution u is represented by an axisymmetric rotational paraboloid as shown in the following equation (21).

  The flow rate Q can be expressed as the following equation (22) by integrating the velocity distribution u over the entire pipe cross section.

  The cross-sectional average flow velocity v can be expressed as the following equation (23).

The pressure gradient is constant in the flow direction and the pressure decreases. The pressure gradient can be expressed as the following equation (24) using the pressure drop Δp of the tube length l. Substituting equation (22) into equation (24) yields equation (25). Equation (25) means that the flow rate Q is proportional to the pressure loss Δp. The flow that satisfies this relationship is called the Hagen-Poiseuille flow.

  The modified Bernoulli equation when there is a loss can be expressed as the following equations (26) and (27).

  Here, l is the tube length, d is the tube inner diameter, v is the tube average velocity, and λ is the tube friction coefficient. The pipe friction coefficient λ is a value determined by the Reynolds number Re when the flow is laminar and by the Reynolds number Re and the surface roughness when the flow is turbulent. Frictional resistance is always affected by the viscosity of the fluid. This frictional resistance is energy loss because it consumes the power or energy that drives the flow.

  When the Hagen-Poiseuille flow equation is transformed to the Darcy-Weissbach equation, the following equation (28) relating to the pressure difference, flow velocity, and pipe inner diameter is obtained.

  This completes the description of the Hagen-Poiseuille flow and the modified Bernoulli equation.

  As described above, the image processing device 27 can identify latent variables using the material dynamics equation and the Hagen-Poiseuille flow and modified Bernoulli equations. For example, as a dynamic model, consider a case in which the deformation of a blood vessel is expressed by an equation of material dynamics of a thick-walled tube, and the change in the tube diameter is expressed by an internal pressure change and an elastic modulus.

  Assuming that the stress-free state is the initial shape (for example, the state in which the blood vessel contracts the most), if the elastic modulus of the blood vessel wall and plaque is set to a certain value, the observed value of the blood vessel shape deformation index such as the average radius of the blood vessel lumen A relational expression between the amount of change in time and the amount of change in internal pressure is obtained. The observed value of the blood vessel shape deformation index is measured from a time-series CT image. The temporal change of the internal pressure distribution of the blood vessel is determined so as to match the temporal change amount of the observed value of the blood vessel shape deformation index. The predicted value of the blood vessel flow rate index is measured by performing blood fluid analysis under this internal pressure distribution. If the predicted value of the blood vessel flow rate index does not match the observed value, the image processing device 27 changes the initially determined elastic modulus of the blood vessel wall or plaque and performs the same analysis.

  By repeating this, the image processing device 27 can detect potentials such as the elastic modulus of the blood vessel wall and plaque that match the observed value of the blood vessel shape deformation index and the observed value of the blood flow index, the internal pressure distribution, and the pressure boundary condition of the fluid analysis. Variables can be determined. In order to perform this determination method more efficiently and stably, a statistical identification method using a hierarchical Bayesian model and a Markov chain Monte Carlo method may be used.

  As described above, the blood vessel analysis device 50 includes the storage device 33, the region setting unit 51, the image analysis / tracking processing unit 53, the dynamic model construction unit 55, and the statistical identification unit 61. The storage device 33 stores time-series medical image data related to the blood vessels of the subject. The region setting unit 51 sets an analysis target region in a blood vessel region included in a time-series medical image. The image analysis / tracking processing unit 53 performs image processing on the time-series medical image and calculates a time-series shape index and a time-series shape deformation index of the analysis target region.

  The mechanical model construction unit 55 provisionally constructs a mechanical model related to the structural fluid analysis of the analysis target region based on the time-series form index, the time-series shape deformation index, and the time-series medical image. The statistical identification unit 61 analyzes the region to be analyzed so that the predicted values of the blood vessel shape index and the blood flow rate index based on the tentatively constructed dynamic model match the pre-measured values of the blood vessel shape index and the blood flow rate index. To identify the latent variables of the dynamic model.

  With the above configuration, the blood vessel analysis device 50 according to the embodiment identifies latent variables such as a material model, boundary conditions, load conditions, and a geometric structure by inverse analysis using a blood vessel shape deformation index and a blood flow index. be able to.

  The blood vessel analysis device 50 repeatedly performs the inverse analysis while changing the latent variable, thereby achieving the above four points of difficulty, namely, 1. 1. Identification method of coronary artery material model 2. Incorporating the effects of heart shape deformation on coronary arteries. 3. Identification method of coronary artery boundary condition; It is possible to identify a latent variable in consideration of all of a material model using a blood vessel shape having uncertainty, a load condition, and a boundary condition identification method.

  Therefore, the blood vessel analysis device 50 can perform structural fluid analysis in consideration of the influence of external factors such as blood vessels and the heart that are not depicted in the CT image. Thus, according to the medical image diagnostic apparatus of the embodiment, the accuracy of the structural fluid analysis of blood vessels can be improved.

  Next, various indices (blood vessel cross-sectional area, change in stress, blood flow velocity, flow rate, pressure, etc.) obtained from the results of structural analysis, fluid analysis, or structural fluid coupled analysis using such a vascular dynamic model A method of creating a nominal model using a change etc. and diagnosing a blood vessel abnormality by model identification with the data to be diagnosed will be described. Hereinafter, an example of diagnosing the degree of stenosis of a blood vessel will be described as a blood vessel abnormality diagnosis.

When two sets of indices are selected and one is input and the other is output, if the input / output relationship has a causal relationship, an input / output model is assumed and the model can be identified. That is, as shown in FIG. 26, an index that does not depend on the presence or absence of stenosis is input u _1, and an index that is affected by the presence or absence of stenosis is output. Further, y ₁ output without _stenosis, when the output of there stenosis and y _2, for no stenosis input u _1, model G ₁ is is assumed that the output y _1. Similarly, in the case of there stenosis, the input _{u 1,} model _{G 2} is assumed that the output _{y 2.}

At this time, model G ₁ and the model G ₂ is represented by a finite number of parameters, if they can compare models with each other, it can be evaluated narrowing of the blood vessels. If the model identification accuracy is high, the output y ₁ hat output when the input u ₁ is input to the estimated model G ₁ hat has a small error from the observed y ₁ . “Hat” means an estimated value.

In general, in order to increase the estimation accuracy, it is necessary to increase the order of the model. Comparison between higher-order models tends to be unstable because the number of parameters to be evaluated increases. Therefore, a model identified in advance is set as a “nominal model”, and its output y ₁ hat is calculated. as input y ₁ hat, assuming a new input-output model _{G 12} to output the index _{y 2} to be compared to identify. Since this model expresses the presence or absence of stenosis, it is hereinafter referred to as “stenosis model”. y ₁ hat includes a higher-order model element. For this reason, a low-order model with reduced order can be used as the stenosis model.

An example is shown in FIG. In the graph of FIG. 27, the horizontal axis indicates time, and the vertical axis indicates CT value (contrast medium concentration). In the graph of FIG. 27, a graph indicated by a circle is a graph of output y _{2 with} stenosis. The solid line graph is a graph of the output y ₂ hat when the nominal model G ₂ is identified and the common input u ₁ is input. In contrast, the dotted line of the graph, the output y ₁ hat estimated at different nominal model G ₁ identified from the output y ₁ no constriction as input to identify stenosis model G ₁₂ which is the output of y _2, y It is a graph of the estimated value when ₁ hat is input. The nominal model _{G 1} 3/6 primary stenosis model _{G 12} is 2/2 next seen the high estimation accuracy, yet few parameters.

  FIG. 28 is a functional block diagram of the blood vessel analyzing device 50 when the medical image diagnostic device according to the embodiment diagnoses the degree of stenosis of blood vessels. In this case, the blood vessel analysis device 50 includes the above-described image processing device 27 that performs structural fluid analysis, and a storage device 33 that stores a nominal model (various parameters of an ARX model to be described later) and a contrast medium injection profile acquisition unit 71. Have. The blood vessel analyzing apparatus 50 includes a first input / output model calculation unit 72, a time series CT image acquisition unit 73, a blood vessel core line extraction unit 74, a time series CT value extraction unit 75, a second input / output model calculation unit 76, and a stenosis model. A calculation unit 77 and a stenosis degree diagnosis unit 78 are included.

  The contrast agent injection profile acquisition unit 71 is an example of an injection profile acquisition unit. The time series CT image acquisition unit 73 is an example of a time series CT value acquisition unit. The image processing device 27 is an example of a structural fluid analysis unit. The first input / output model calculation unit 72 is an example of a first calculation unit. The contrast agent injection profile acquisition unit 71, the first input / output model calculation unit 72, the time series CT image acquisition unit 73, the image processing device 27, and the storage device 33 are examples of a time series model generation unit. The storage device 33 is an example of a storage device. The time series CT value extraction unit 75 is an example of a time series CT value extraction unit. The stenosis model calculation unit 77 is an example of a stenosis model calculation unit. The stenosis degree diagnostic unit 78 is an example of a diagnostic unit. The second input / output model calculation unit 76 is an example of a second calculation unit.

  The contrast agent injection profile acquisition unit 71 to the stenosis degree diagnosis unit 78 may be all configured by hardware or all configured by software. Alternatively, a part may be configured by hardware and the rest may be configured by software.

  First, when diagnosing the degree of stenosis of a blood vessel, the above-described nominal model is prepared in advance. To explain in sequence, a contrast agent is often used for angiography for CT image acquisition. When a certain amount of contrast agent is injected from, for example, the vein of the upper arm, the CT value (HU value) of the region through which the contrast agent has passed rises, making it easy to distinguish from tissues other than blood vessels. The CT gantry 10 is controlled by the system control unit 21 and the gantry control unit 23 so as to capture CT images in time series. Thereby, the speed at which the CT value changes can be observed.

  As an example, FIG. 29 shows a contrast agent injection pattern. The solid line graph, the dotted line graph, and the alternate long and short dash line graph in FIG. 29 respectively show contrast agent injection patterns. The solid line graph in FIG. 29 shows an injection pattern in which the contrast medium is slowly injected over about 7 seconds. The dotted line graph in FIG. 29 shows an injection pattern in which a contrast medium is injected at a high speed over about 4.5 seconds. The dashed-dotted line graph in FIG. 29 shows an injection pattern in which the contrast medium is injected at a medium speed over about 6 seconds. The injection of the contrast agent may be performed manually by the doctor in charge or the like, or may be performed using an injector under the supervision of the doctor in charge.

  FIG. 30 is a graph of a change in CT value in a region of interest after contrast medium injection. Of the injection patterns shown in the solid line graph, dotted line graph, and alternate long and short dash line graph of FIG. 29, the same graph as in FIG. 30 can be obtained regardless of the injection pattern. In FIG. 30, the circled graph shows the change in CT value without stenosis. Further, in FIG. 30, a black circle indicates a change in CT value when stenosis occurs. In FIG. 30, when there is no stenosis in the region of interest, the change in the CT value rises quickly and the peak increases as shown by the circle. On the other hand, when a stenosis exists on the upstream side of the region of interest, the change in the CT value has a slow rise as shown in the graph ●, and the peak is also low. The medical image diagnostic apparatus according to the embodiment uses an input / output model in which a contrast agent injection pattern is input and a CT value in a region of interest is output.

  Returning to the description of FIG. 28, when generating a nominal model, the contrast agent injection profile acquisition unit 71 acquires the injection profile of the contrast agent described with reference to FIG. 29. The acquired contrast agent injection profile is supplied to the image processing device 27, the first input / output model calculation unit 72, and the second input / output model calculation unit 76.

  The above-described image processing device 27 performs structural fluid analysis on the time-series CT image (may be an MRI image or the like) acquired by the time-series CT image acquisition unit 73. The time-series CT image acquisition unit 73 may acquire a time-series CT image stored in the storage device 33, or may acquire a CT image captured in real time and time-series by the CT mount 10. Also good. By the analysis of the image processing device 27, it is possible to simulate an output (concentration change in contrast agent) corresponding to an injection profile of a contrast agent that is one of fluids. The structural fluid analysis unit simulates a change in the concentration of the contrast agent injected with the injection pattern described with reference to FIG. Then, information indicating the change in the concentration of the contrast agent shown in FIG. 30 as the simulation result is supplied to the first input / output model calculation unit 72.

  The first input / output model calculation unit 72 (and the second input / output model calculation unit 76) calculates a model expressed by a linear difference equation including a finite number of parameters. In other words, the first input / output model calculation unit 72 (and the second input / output model calculation unit 76) calculates a so-called ARX model as an example of an input / output model expressed by a finite number of parameters. ARX is an abbreviation of “Auto Regressive with eXenous inputs” or (Auto Regressive with eXternal inputs). The ARX model has u (t) as input, y (t) as output, and error ε (t). It is represented by the following equation (29).

  A (q) y (t) = B (q) u (t) + ε (t) (29)

  However, A (q) is the following (30) Formula, B (q) is as the following (31) Formula.

A (q) = 1 + a ₁ q ^-1 +… + a _m q ^-m … (30)

B (q) = b ₁ q ^-1 +… + b _n q ^-n … (31)

“T” in the equation (29) means discrete sampling data sampled at each time (see the CT values (sampling data) of ○ or ● in FIG. 30). Further, “q” in the expressions (30) and (31) means a shift parameter that shifts along the time axis. For example, “q ⁻¹ ” means that the previous value in time is used. In addition, the model in this example is expressed by a total of m + n parameters. The parameters A and B of this model are determined using the least square method so as to minimize the sum of squares of the formula error.

  Each parameter of the ARX model calculated in this way is stored in the storage device 33 as a nominal model. As an example, as the nominal model stored in the storage device 33, a plurality of nominal models such as a nominal model without stenosis, a nominal model with a stenosis degree of 30%, and a nominal model with a stenosis degree of 50% are stored. Is done. In this example, the ARX model is used as the input / output model, but any model may be used as long as it is a model that can be expressed by parameters.

  Next, when such a nominal model is stored in the storage device 33, the degree of stenosis of the blood vessel can be diagnosed. When the blood vessel stenosis degree is diagnosed, the time-series CT image acquisition unit 73 acquires a time-series CT image of the subject into which the contrast agent has been injected. In addition, the blood vessel core line extraction unit 74 extracts the blood vessel core line from the acquired time-series CT image as described above. The time series CT value extraction unit 75 extracts the CT value of the attention area of the blood vessel from which the core line has been extracted in time series. Each CT value extracted in this time series corresponds to a CT value indicated by a circle in the graph of FIG.

  Next, like the first input / output model calculation unit 72, the second input / output model calculation unit 76 calculates a so-called ARX model as an example of an input / output model expressed by a finite number of parameters. The second input / output model calculation unit 76 acquires, via the contrast agent injection profile acquisition unit 71, the injection profile of the contrast agent injected into the subject when each CT value is acquired. Then, the second input / output model calculation unit 76 calculates an ARX model that is a parameter set similar to the nominal model stored in the storage device 33 from the contrast agent injection profile and each CT value extracted in time series. Calculate as the second input / output model.

  Finally, the stenosis model calculation unit 77 calculates a stenosis model that is a difference between each parameter of the nominal model (parameter corresponding to the solid line graph in FIG. 27) stored in the storage device 33 and the second input / output model. Then, the stenosis degree diagnosis unit 78 diagnoses the calculated stenosis model parameter as the stenosis degree of the blood vessel.

  That is, the ARX model is used to calculate the ARX model using the output result when the contrast medium injection profile is input to the nominal model as the input and the output result when the contrast medium injection profile is input to the second input / output model. If there is no difference between the second input / output model and the nominal model, the transfer function of the obtained stenosis model is “1”, indicating that the second input / output model and the nominal model are the same. become.

  The stenosis degree diagnosis unit 78 may evaluate the stenosis model as it is (how far away from the transfer function 1), and displays the degree of stenosis of the blood vessel, for example, 30% stenosis, 10% stenosis, etc. May be. Evaluation may be performed using a plurality of nominal models created using data with different degrees of stenosis, and “which nominal model is closest” may be displayed. Furthermore, the stenosis degree diagnosis unit 78 may display the degree of stenosis on a scale with the normal value without stenosis being “0”. That is, as long as the degree of stenosis is known, any display format may be used.

  As is clear from the above description, the medical image diagnostic apparatus according to the embodiment performs blood pressure, blood flow, stress in hemodynamics by blood vessel analysis based on image tracking, structural fluid analysis, and inverse analysis (statistical identification method). Alternatively, the index necessary for prevention and diagnosis of vascular lesions such as stenosis such as strain is analyzed with high accuracy. Then, the medical image diagnostic apparatus according to the embodiment analyzes the typical example in which the presence or absence of stenosis is known, creates a low-order standard model (nominal model) in advance, Diagnose stenosis from model identification.

  This makes it possible to diagnose abnormalities of blood vessels by effectively using the analysis results calculated by the image processing apparatus. Further, by diagnosing the degree of stenosis based on the model identification of the low-order standard model (nominal model) and the diagnosis target data, it is possible to diagnose a blood vessel abnormality with a small amount of calculation resources and in an extremely short time.

  Further, an ARX model is calculated by the second input / output model calculation unit 76 from each CT value extracted by the time-series CT value extraction unit 75 and the contrast agent injection profile, and is stored in the storage device 33. The nominal model is compared by a stenosis model calculation unit 77. For this reason, since the same ARX model can be compared, the stenosis degree can be diagnosed with high accuracy.

(Modification 1)
In the description of the above-described embodiment, the ARX model is calculated by the second input / output model calculation unit 76 in order to compare the same ARX models, but each CT value extracted by the time series CT value extraction unit 75 The stenosis degree may be diagnosed by comparing the nominal model stored in the storage device 33 with the stenosis model calculation unit 77. In this case, since the second input / output model calculation unit 76 can be omitted, the configuration of the medical image diagnostic apparatus according to the embodiment can be simplified and the cost can be reduced.

(Modification 2)
As the time-series CT value, the time-series data of the gradient obtained by obtaining the CT value in the core line direction including before and after the stenosis of the blood vessel and calculating the gradient (contrast agent concentration gradient) in the blood vessel running direction may be used. FIG. 31 is an example of a contrast medium concentration gradient graph. In FIG. 31, a graph indicated by a solid line is a graph of a contrast medium concentration gradient when there is no stenosis. The contrast agent concentration gradient is a negative gradient, and when there is no stenosis, the gradient of the contrast agent falls at an early time and returns to 0 level at an early time. On the other hand, in FIG. 31, a graph indicated by a dotted line is a graph of a contrast medium concentration gradient in the case of stenosis. In the case of stenosis, the graph shows a gradient that slowly falls over time and gradually returns to level 0 over time. Even if time series data having such a gradient is used, the same effect as described above can be obtained.

  Although the embodiment and the modification have been described above, the embodiment and the modification are presented as examples, and are not intended to limit the scope of the invention. These novel embodiments and modifications can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. The embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 10 CT mount 11 X-ray tube 13 X-ray detector 15 Data acquisition apparatus 20 Console 21 System control part 23 Mount control part 25 Reconstruction apparatus 27 Image processing apparatus 29 Input apparatus 31 Display apparatus 33 Storage apparatus 50 Blood vessel analysis apparatus 51 Area setting Unit 53 image analysis / tracking processing unit 55 dynamic model construction unit 57 vascular stress analysis unit 59 blood fluid analysis unit 61 statistical identification unit 71 contrast medium injection profile acquisition unit 72 first input / output model calculation unit 73 time-series CT image acquisition unit 74 Blood vessel core line extraction unit 75 Time series CT value extraction unit 76 Second input / output model calculation unit 77 Stenosis model calculation unit 78 Stenosis degree diagnosis unit

Claims (8)

An injection profile acquisition unit for acquiring an injection profile of a contrast medium injected into a blood vessel of a subject;
A CT value acquisition unit for acquiring a CT value in the blood vessel lumen after the contrast agent injection;
A structural fluid analysis unit for performing a structural fluid analysis simulation using the injection profile and the CT value as inputs;
A model representing the relationship between the input and output corresponding to the structural fluid analysis simulation is obtained from the result of the structural fluid analysis simulation using the injection profile and the CT value acquired in advance and the presence or absence of a vascular abnormality is known. A first calculation unit for calculating;
A storage device for storing the model;
A difference model calculation unit that calculates a difference model indicating a degree of an abnormality of a blood vessel by detecting a difference between a CT value of a subject that performs an abnormality diagnosis of the blood vessel and the model stored in the storage device;
A blood vessel analyzing apparatus comprising: a diagnosis unit that evaluates the difference model and diagnoses a blood vessel abnormality. The first calculation unit calculates a nominal model representing the same input / output relationship as in the structural fluid analysis simulation as a model representing an input / output relationship corresponding to the structural fluid analysis simulation. Item 1. The blood vessel analysis device according to Item 1. A CT value extraction unit that extracts a CT value of a target region from a CT value of a subject that performs a blood vessel abnormality diagnosis acquired by the CT value acquisition unit;
From the CT value of the target region extracted by the CT value extraction unit, and the injection profile of the contrast agent injected into the subject that performs the abnormality diagnosis of the blood vessel acquired by the injection profile acquisition unit, A second calculator that calculates a second input / output model corresponding to the model;
The differential model calculation unit, vessel analysis apparatus according to claim 1 or claim 2, characterized in that for calculating said difference model representing the difference between the and the model second input model. The blood vessel according to any one of claims 1 to 3 , wherein the CT value acquisition unit acquires a CT value from a CT image photographed in a target region after the occurrence of a blood vessel abnormality. Analysis device. Obtaining a CT value in a blood vessel core direction including before and after the abnormal site of the blood vessel, and having a concentration gradient calculation unit for calculating a contrast agent concentration gradient in the blood vessel running direction;
The blood vessel analysis device according to any one of claims 1 to 4 , wherein a CT value of the calculated contrast medium concentration gradient is used as the CT value. 4. The blood vessel analysis device according to claim 3 , wherein at least one of the first calculation unit and the second calculation unit calculates a model expressed by a linear difference equation including a finite number of parameters. The blood vessel analysis device according to any one of claims 1 to 6 , wherein a time-series CT value is used as the CT value. The system controller
Causing the injection profile acquisition unit to execute an injection profile acquisition step of acquiring an injection profile of the contrast medium injected into the blood vessel of the subject;
The CT value acquisition unit, to execute a CT value acquisition step of acquiring CT values in the blood vessel lumen after the contrast medium injection,
The structural fluid analysis unit, to execute a structural fluid analysis step of performing a structural fluid analysis simulation of the injection profile and the CT value as an input,
The first calculation unit, said injection profile and advance acquired from the results whether the vascular abnormality has the structural fluid analysis simulation using the CT value has been found, an input corresponding to the structural fluid analysis simulation output A first calculation step of calculating a model representing the relationship and storing the model in a storage device;
The difference model calculation unit, by detecting the difference between the model stored in the storage device and CT value of the object performing abnormality diagnosis of the blood vessels, the difference calculating a difference model indicating the degree of vascular abnormalities Run the model calculation step,
The diagnosis unit to execute a display step of displaying the results of the evaluation by evaluating the difference model,
A method of operating a blood vessel analyzing apparatus .