JP6713690B2 - Blood vessel shape constructing apparatus, method thereof, and computer software program - Google Patents

Description

The present invention relates to a blood vessel shape construction device for blood flow analysis by computational fluid dynamics, a method thereof, and a computer software program.

Cardiovascular diseases include vascular aneurysm, hardening, and stenosis. These diseases involve lesions in normal sites due to the influence of blood flow, and although many of these diseases die due to subsequent progress, their treatment is often life-threatening and difficult. In recent years, in order to elucidate and diagnose such refractory circulatory system diseases, and to predict the therapeutic effect, a technique for performing blood flow analysis of blood vessels including a lesion site has been demanded.

One method of performing this blood flow analysis is a method of performing computer processing by using Computational Fluid Dynamics (CFD). Computational fluid dynamics is a technique for acquiring the flow of a fluid by computational analysis by a computer. For example, in Japanese Patent No. 5596866, three-dimensional blood vessel shape data indicating a blood vessel shape is constructed from medical image data captured by a medical image capturing apparatus or the like, and blood flow analysis by computational fluid dynamics is performed based on the blood vessel shape. The technology is disclosed. Japanese Patent No. 559866 describes that the blood vessel treatment effect can be simulated by selecting information on the treatment method, correcting the blood vessel shape according to the information, and performing blood flow analysis by computational fluid dynamics.

In such blood flow analysis, the quality of the blood vessel shape to be calculated, that is, the quality of the three-dimensional shape data is important. This is because the quality of the blood vessel shape directly affects the accuracy of the result of the blood flow analysis described above.

However, in the conventional blood vessel shape construction, each processing such as selection of medical image data used for blood vessel shape construction and identification/extraction of a blood vessel region and a lesion area is performed using various kinds of independent devices and software, In addition, since the user manually modified and confirmed the medical image data and adjusted various parameters of the construction process, the blood vessel shape that is output may differ depending on the combination of the device and software and the degree of knowledge and experience of the user. There was a problem. That is, in the conventional blood vessel shape construction method, the blood vessel shape to be constructed has user dependence, and the quality of the blood vessel shape tends to vary. And, the variation in the quality of the blood vessel shape by such a conventional blood vessel shape construction method may affect the accuracy of the result of the blood flow analysis, and is a problem especially in a large-scale clinical experiment for comparing a huge amount of analysis results. It has been said.

In order for blood flow analysis to spread to medical practice, it is essential to ensure the accuracy of blood flow analysis results and to standardize and share the results. Therefore, there is no user dependency and the accuracy of blood flow analysis is high. It is important to have a technique for constructing a blood vessel shape of sufficient quality to guarantee

Patent No. 5596866

The present invention has been made in view of such a situation, and an object thereof is to eliminate the user dependency by fully processing each process of constructing a blood vessel shape and construct a blood vessel shape for blood flow simulation. To provide a blood vessel shape constructing apparatus, a method thereof, and a computer software program capable of performing quality control of the blood vessel.

According to the first main aspect of the present invention, a medical image is received, the image attribute of the medical image and the adequacy as the medical image are discriminated, and it is judged whether or not the medical image can be processed as an input image based on them. An input permission/inhibition determination unit, and a medical image determined to be processable as the input image is corrected/corrected based on the determined image attribute and adequacy, and the image quality of the medical image is standardized, and the correction is performed. An image standardization unit that outputs a correction degree indicating the degree of correction, a blood vessel shape construction unit that constructs a three-dimensional blood vessel shape based on the standardized medical image, and a blood vessel shape of the constructed three-dimensional blood vessel shape. A blood vessel shape quality evaluation unit that determines the construction accuracy and that calculates and outputs a comprehensive quality evaluation of the constructed three-dimensional blood vessel shape based on the blood vessel shape construction accuracy, the adequacy, and the correction degree, and the three-dimensional blood vessel. A blood vessel shape data output unit that outputs blood vessel shape data indicating a shape together with the shape evaluation is provided.

Here, according to an embodiment of the present invention, the image attribute of the medical image is information including information on a maker and modality of an imaging device that has captured the medical image, and an imaging condition, and as the medical image. The adequacy is information including the noise level of the medical image, and the input propriety determination unit executes the input propriety determination by referring to the image attribute and the adequacy with a predetermined template. In this case, it is preferable that the input availability determination unit determines the noise level by analyzing the medical image. Further, in this case, it is preferable that the input availability determination unit calculates different types of noise components according to the image attributes by image analysis and determines the noise level of each component based on the calculated noise components.

Further, according to another embodiment of the present invention, the image standardization unit performs correction/correction for removing a specific noise component from the medical image based on the image attribute and the adequacy. is there.

According to still another embodiment, the image normalization unit isotropic by equalizing spatial resolutions in the X-direction, Y-direction, and Z-direction by interpolation in an XYZ coordinate system of a medical image that is orthogonal to each other. The correction/correction is performed.

Further, according to still another embodiment, the image standardization unit executes correction/correction for increasing the resolution by increasing the spatial resolution of the medical image by image interpolation.

Further, according to still another embodiment, the image standardization unit performs correction of image distortion derived from imaging and correction of shape distortion of a local blood vessel shape to standardize the image quality. Is. In this case, the image standardization unit performs the correction of the local shape distortion of the blood vessel shape on the medical image determined as the non-contrast MRA by the determination of the image attribute, and the image standardization unit Luminance information of a region corresponding to the inside and outside of the blood vessel on the medical image so as to eliminate the flow velocity dependency between the inside and outside of the blood vessel, which is a characteristic of non-contrast MRA, in correcting the shape distortion. It is preferable to carry out based on. Further, in this case, the image standardization unit detects a portion of the medical image where the flow velocity dependency is stronger than a predetermined value based on the luminance information of the portion corresponding to the vicinity of the blood vessel wall on the medical image, and morphological information or blood It is preferable to correct the flow velocity dependency of the detected location based on the flow information. Further, in this case, it is preferable that the image standardization unit detects a portion where the brightness gradient value in each cross section orthogonal to the blood vessel center line in the vicinity of the blood vessel wall is equal to or less than a threshold value as a portion having a strong flow velocity dependency. .. When correcting the flow velocity dependency based on the morphological information of the blood vessel, the image standardization unit corrects the flow velocity dependency of the image by applying the statistical model of the blood vessel at the detected location. It is preferably one. In this case, it is preferable that the statistical model of the blood vessel is a mathematical model of a change in the blood vessel cross section in the blood vessel running direction obtained from the blood vessel shape database. When correcting the flow velocity dependency based on the blood flow information of the blood vessel, the image standardization unit corrects the flow velocity dependency of the image by applying a blood flow model of the blood vessel at the detected location. It is preferable that In this case, the blood flow model of the blood vessel is a mathematical model of blood flow information obtained by analyzing the actual blood flow information measured by the phase contrast MRI method or the spatial distribution depending on the flow velocity. Is preferred.

Further, according to another embodiment, the image standardization unit identifies a blood vessel as a high-intensity signal with respect to a medical image determined to be CTA (X-ray computed tomography image) by the determination of the image attribute. The correction/correction is performed to remove bone and/or calcification that are easily misrecognized. In this case, the image standardization unit removes bone by using anatomical features of the bone or compositely uses statistical models of blood vessels, and the statistical model of blood vessels is It is preferable that the change in the blood vessel cross section in the blood vessel running direction obtained from the shape database is mathematically modeled. Further, it is preferable that the image standardization unit removes calcification based on a brightness characteristic specific to calcification that is not present in bones or blood vessels.

Further, according to still another embodiment, the blood vessel shape constructing unit performs a structural analysis on the coarsely dividing unit that roughly divides the blood vessel region by image analysis of the medical image, and the roughly divided blood vessel region. A fine division unit that performs a fine division of a blood vessel region based on the result of the structural analysis, and a detection unit that detects the presence or absence of adhesion of a specific blood vessel based on the result of the structural analysis. In this case, it is preferable that the blood vessel shape construction unit re-executes the precision division of the blood vessel based on the determination of the presence or absence of the adhesion.

Further, according to still another embodiment, the blood vessel shape constructing unit performs a structural analysis on the coarsely dividing unit that roughly divides the blood vessel region by image analysis of the medical image, and the roughly divided blood vessel region. , And a precision division unit that performs a precision division of the blood vessel region based on the result of the structural analysis, and the structural analysis acquires the center line of the blood vessel, and when the center line of the blood vessel is obtained, Obtain an orthogonal image at each point, perform a luminance value analysis of the orthogonal image in each blood vessel, to determine the blood vessel site-specific inside and outside the blood vessel from the luminance value analysis, the precision dividing unit, the inside and outside of the blood vessel determination of The blood vessel region is divided based on the result. In this case, it is preferable that the inside/outside determination by the luminance value analysis is based on a half value of the maximum luminance value in the centerline orthogonal image. Further, it is preferable that the blood vessel shape constructing unit enhances the accuracy of the blood vessel region division/extraction by recursively performing the blood vessel inside/outside determination in the blood vessel region division/extraction.

According to a second main aspect of the present invention, a computer software program executed by a computer, each instruction stored in the following storage medium: a computer receives a medical image, and an image of the medical image is received. An input acceptability determination unit that determines the attribute and the adequacy as a medical image and determines whether or not it can be processed as an input image based on them, and the computer determines the medical image that can be processed as the input image. The image standardization unit that corrects/corrects based on the determined image attributes and the adequacy, standardizes the image quality of the medical image, and outputs the correction degree indicating the degree of the correction/correction, and the computer, A blood vessel shape constructing unit that constructs a three-dimensional blood vessel shape based on a standardized medical image, and a computer determines the blood vessel shape construction accuracy of the constructed three-dimensional blood vessel shape, and this blood vessel shape construction accuracy, the Adequacy, a blood vessel shape quality evaluation unit that calculates and outputs a comprehensive quality evaluation of the constructed three-dimensional blood vessel shape based on the correction degree, and a computer, together with the shape evaluation, the blood vessel shape data indicating the three-dimensional blood vessel shape. A computer software program is provided which has a blood vessel shape data output unit for outputting.

According to a third main aspect of the present invention, in a method executed by a computer, the computer receives a medical image and discriminates the image attribute of the medical image and the adequacy as the medical image. An input availability determination step of determining whether the image can be processed as an input image based on them, and a medical image determined to be processable as the input image by a computer based on the determined image attribute and adequacy. An image standardization step of correcting/correcting and standardizing the image quality of the medical image, and outputting a correction degree indicating the degree of the correction/correction, and a computer based on the standardized medical image And the computer determines the blood vessel shape construction accuracy of the constructed three-dimensional blood vessel shape, and constructs the blood vessel shape based on the blood vessel shape construction accuracy, the adequacy, and the correction degree. A blood vessel shape quality evaluation step of calculating and outputting a comprehensive quality evaluation of the three-dimensional blood vessel shape; and a computer, a blood vessel shape data output step of outputting blood vessel shape data indicating the three-dimensional blood vessel shape together with the shape evaluation. A characterizing method is provided.

It should be noted that other features of the present invention other than those described above can be easily understood by those skilled in the art by referring to the following "Description of Embodiments" and the drawings.

FIG. 1 is a diagram for explaining computational fluid dynamics. FIG. 2 is a diagram showing a flow of blood flow analysis by computational fluid dynamics. FIG. 3 is a diagram showing a blood vessel shape construction process. FIG. 4 is a schematic configuration diagram showing an embodiment of the present invention. FIG. 5 is a diagram showing a flow of blood vessel shape construction showing an embodiment of the present invention. FIG. 6 is a diagram for explaining factors that cause qualitative differences in medical images. FIG. 7 is a block diagram showing the image discrimination unit. FIG. 8 is a block diagram showing the image standardization unit. FIG. 9 is a configuration diagram showing a blood vessel shape construction unit. FIG. 10: is a figure explaining the phenomenon which seems to have adhered. FIG. 11 is a diagram for explaining a graph of a blood vessel structure and a depth-first search of the graph. FIG. 12 is a diagram for explaining the adhesion separation process. FIG. 13 is a diagram for explaining the operations of the shape measuring unit and the quality evaluation unit. FIG. 14 is a diagram for explaining quality determination of blood vessel shape data.

First, the specific features of the present invention will be described.

(Characteristics of the Invention 1: Blood Vessel Shape Constructing Device with Fully Automatic Control Function)
1-1. INDUSTRIAL APPLICABILITY The present invention is a blood vessel shape construction apparatus for blood flow analysis by computational fluid dynamics, which eliminates user dependence by fully automating the process of inputting a medical image and outputting a blood vessel shape and quality evaluation. It is a blood vessel shape construction device characterized by the above.
1-2. The blood vessel shape constructing means of the feature 1-1 is a means for inputting a captured medical image, a means for discriminating a medical image, a means for standardizing a medical image, a means for constructing a blood vessel shape, a means for measuring a blood vessel shape, a blood vessel shape. And a means for outputting a blood vessel shape and a quality score.
1-3. The means for discriminating the medical image of the characteristic 1-2 is a maker, a means for discriminating modality (device discrimination), a means for discriminating imaging conditions (image discrimination), and discriminating whether the image is sufficient for blood flow analysis. Means (determination of input propriety).
1-4. The feature 1-3 is a means for determining whether or not the image is sufficient for blood flow analysis (input permission/inhibition determination), and determines whether or not it is possible to compare conditions satisfying a medical image that can withstand blood flow analysis with a listed template.
1-5. Characteristic 1-2 standardizing means for medical images includes means for removing noise (noise removal), means for making an image isotropic by interpolation (isotropic), means for increasing image resolution by interpolation (higher resolution). ), means for correcting an image (image correction).
1-6. The means 1-2 for constructing the blood vessel shape includes means for dividing the blood vessel into areas (area division), means for structurally analyzing the blood vessels (vascular structure analysis), means for detecting the presence or absence of adhesion (adhesion detection), and blood vessels. It has means for setting an area (area setting) and means for constructing a blood vessel shape (blood vessel shape construction).
1-7. The feature 1-6 that divides the blood vessel into regions is a means that extracts the blood vessel by performing the inside/outside determination of the blood vessel based on the characteristic of the brightness value of the portion corresponding to the blood vessel on the image.
1-8. The means for structurally analyzing the blood vessel having the feature 1-6 is a means for performing labeling by dividing the blood vessel extracted in the feature 1-7 by centerline extraction to divide the blood vessel into individual blood vessels and lesions. 1-9. Precisely correct the result by repeating the features 1-7 and 1-8 based on the result of the rough extraction by performing the features 1-7 and 1-8 first as rough extraction of the blood vessel. Extract. In the precision extraction, adaptive blood vessel region segmentation in which the threshold value is adaptively changed for each brightness value of each blood vessel, and the blood vessel structure analysis is performed again based on the result, or region segmentation is performed until the threshold value is equal to or less than a certain threshold value. It is possible to perform recursive blood vessel segmentation by repeatedly performing blood vessel structure analysis.
1-10. The feature 1-6 for detecting the presence or absence of adhesion is a means for confirming the presence or absence of a loop of a blood vessel that cannot normally exist when the blood vessel is viewed as a whole. If a loop is confirmed, it is determined that an artifact due to adhesion between blood vessels has occurred, and the adhesion is separated by performing the precise extraction of the feature 1-9 again.
1-11. The means for setting the region of the blood vessel of the feature 1-6 includes in the blood flow analysis by comparing the conditions to be satisfied as the region setting for the blood vessel/lesion labeled in the feature 1-8 with the listed template. This is a means for extracting only the region.
1-12. The means 1-6 for constructing the shape of the blood vessel is means for forming the surface of the blood vessel from the minute triangular elements by means such as the marching cube method for the blood vessel whose region is set in the characteristic 1-9.
1-13. The feature 1-2 measuring the blood vessel shape is a means for creating a cross-section orthogonal to the blood vessel at each point on the center line obtained in the feature 1-8 and measuring the shape.
1-14. The means for evaluating the quality of the blood vessel shape of the feature 1-2 calculates the quality score based on the brightness gradient in the vicinity of the blood vessel wall in the obtained blood vessel shape.

(Feature 2 of the invention: blood vessel shape construction device having image standardization function)
2-1. The present invention is a blood vessel shape constructing apparatus for blood flow analysis by computational fluid dynamics, which has an image standardization function that corrects the difference in quality of medical images.
2-2. The means for standardizing the image of the characteristic 2-1 is a means for removing noise (noise removal), a means for making an image isotropic by interpolation (isotropic), and a means for increasing image resolution by interpolation (higher resolution). , Means for correcting an image (image correction).
2-3. The means for removing noise of feature 2-2 is a means for removing different image noise depending on the manufacturer, modality, imaging condition, and imaging target.
2-4. The means for making the image isotropic by the interpolation of the feature 2-2 is a means for making the spatial resolution of the medical image, which may be different in the in-plane direction and the out-of-plane direction, into a cube having no direction dependency.
2-5. The means for increasing the image resolution by the interpolation of the feature 2-2 is a means for artificially doubling the original spatial resolution of the image by the image interpolation.
2-6. The means for correcting the image of the characteristic 2-2 is a means for correcting the difference in the characteristics of the image which differs depending on the manufacturer, modality, and imaging condition.
2-7. In Feature 2-6, in the case of non-contrast MRA, a means for correcting the flow velocity dependence of the brightness value, which is the property of the MRA image, is provided.
2-8. In Feature 2-7, the means for correcting the flow velocity dependency has a means for detecting a portion having a strong flow velocity dependency, and a means for correcting the flow velocity dependency based on blood vessel morphology information or blood flow information.
2-9. In Feature 2-8, the means for detecting a place having a strong flow velocity dependency is based on a brightness gradient value in the vicinity of the blood vessel wall on each cross section orthogonal to the blood vessel center line. Since the location where the brightness gradient value is low is the location where the flow velocity is reduced, it becomes a factor that distorts the shape of the blood vessel wall.
2-10. In Feature 2-8, when the blood vessel morphology information is used to correct the flow velocity dependency, the flow velocity dependency of the image is corrected by applying the statistical model of the blood vessel at the location where the flow velocity dependency is detected. It is a thing.
2-11. In Feature 2-10, the blood vessel statistical model is a mathematical model of how the blood vessel cross-section changes in the blood vessel running direction obtained from the blood vessel shape database.
2-12. In Feature 2-8, when the blood flow information of the blood vessel is used to correct the flow velocity dependence, the blood flow model of the blood vessel is applied at the location where the flow velocity dependence is detected to determine the flow velocity dependence of the image. To correct.
2-13. In Feature 2-12, the blood flow model of a blood vessel may use actual blood flow information measured by a phase contrast MRI method or the like, or a blood flow obtained by analyzing a flow velocity-dependent spatial distribution. You may correct from information.
2-14. In Feature 2-6, in the case of CTA, a means for removing bone and calcification that are erroneously recognized as blood vessels as a high-brightness signal is provided.
2-15. In the feature 2-14, as a means for removing the bone, a means utilizing the anatomical feature of the bone may be used, or the statistical model of the blood vessel shown in the feature 2-11 may be used in combination. Good.
2-16. In Feature 2-14, as a means for removing calcification, a means which does not exist in bones or blood vessels and which utilizes luminance characteristics specific to calcification may be used. For example, since calcification is a phenomenon that occurs locally, high-brightness signals are isolated in islands. Calcification can be detected and removed by extracting an isolated high-brightness signal.

(Feature 3 of the Invention: Blood vessel shape construction device having adaptive blood vessel region dividing function)
3-1. The present invention is a blood vessel shape constructing apparatus for blood flow analysis by computational fluid dynamics, which eliminates the influence of different brightness characteristics in a blood vessel region-specific manner by performing blood vessel region adaptive determination of the blood vessel region in blood vessel region extraction. It is a device for constructing a blood vessel shape.
3-2. Feature 3-1 performs region segmentation and blood vessel structure analysis as rough extraction of region segmentation of blood vessels. The region division is performed with a threshold value, and the center line of the blood vessel is acquired by the blood vessel structure analysis. When the center line of the blood vessel is obtained, the orthogonal image is acquired at each point on the center line, and the brightness value analysis of the orthogonal image is performed for each blood vessel. This makes it possible to quantify the variation in the brightness value among the blood vessels.
3-3. The means for performing the blood vessel region-specific inside/outside blood vessel determination based on the luminance value analysis of the feature 3-2 includes means for setting a half value of the maximum luminance value in the centerline orthogonal image as a reference.
3-4. The means for performing the blood vessel site-specific inside/outside determination based on the brightness value analysis of the feature 3-2 is a means for using the maximum point of the brightness gradient obtained by setting polar coordinates from the maximum brightness value in the centerline orthogonal image as a reference. Including.
3-5. The means for performing the blood vessel region-specific inside/outside blood vessel determination based on the luminance value analysis of feature 3-2 includes means for setting the maximum point of the luminance gradient obtained by setting polar coordinates from the center of gravity of the centerline orthogonal image as a reference.

(Feature 4 of the invention: blood vessel shape construction device having recursive blood vessel region dividing function)
4-1. The present invention is a blood vessel shape constructing apparatus for blood flow analysis by computational fluid dynamics, in which the accuracy of blood vessel area extraction is improved by recursively determining the inside and outside of the blood vessel in blood vessel area extraction. ..
4-2. The means for improving the accuracy of blood vessel region extraction of feature 4-1 includes means for performing rough extraction of blood vessels and means for performing precise extraction of blood vessels.
4-3. In the feature 4-2, the means for roughly extracting blood vessels includes means for extracting a rough blood vessel shape by rough inside/outside determination, and means for extracting a rough center line from the rough blood vessel shape.
4-4. In the feature 4-3, the means for extracting the rough blood vessel shape by the rough inside/outside determination is a means for extracting the rough blood vessel shape by binarizing an image using a predetermined value or a reference value calculated from a luminance value histogram. Is.
4-5. In feature 4-2, the means for precisely extracting blood vessels is a means for creating an orthogonal cross section at each point of the rough centerline, a means for creating a centerline orthogonal image by image interpolation, and a luminance value analysis of the centerline orthogonal image. It has a means, a means for making a precision blood vessel inside/outside determination based on luminance value analysis, and a means for creating a precision center line from the result of the precision blood vessel inside/outside determination.
4-6. The means for performing the precision blood vessel inside/outside determination based on the luminance value analysis of the feature 4-5 includes means for setting a half value of the maximum luminance value in the centerline orthogonal image as a reference.
4-7. The means for performing precision blood vessel inside/outside determination based on the luminance value analysis of feature 4-5 includes means for setting the maximum point of the luminance gradient obtained by setting polar coordinates from the maximum luminance value in the centerline orthogonal image.
4-8. The means for performing the precision blood vessel inside/outside determination based on the luminance value analysis of the feature 4-5 includes means for setting the maximum point of the luminance gradient obtained by setting polar coordinates from the center of gravity of the centerline orthogonal image as a reference.
4-9. The rough extraction and the fine extraction of the feature 4-2 may be performed in a single stage or repeatedly.

(Characteristic 5 of the invention: Blood vessel shape construction device having automatic region setting function)
5-1. The present invention is a blood vessel shape building device for blood flow analysis by computational fluid dynamics, which has a function of automatically setting a blood flow analysis region for an extracted blood vessel.
5-2. The means for automatically setting the blood flow analysis region of the characteristic 5-1 has a means for analyzing the structure of the blood vessel and a means for determining whether or not to include it in the analysis area.
5-3. Characteristic 5-2 is a means for analyzing the structure of a blood vessel, a means for determining a main blood vessel, a means for tracking the running of each blood vessel from the center line of the main blood vessel and comparing it with an anatomical template to determine a subordinate blood vessel, a main blood vessel. And a means for labeling the blood vessel name to the subordinate blood vessel.
5-4. The means for determining whether or not to include the feature 5-2 in the analysis region is performed by referring to a template in which a set of a main blood vessel and a dependent blood vessel is listed in advance as a blood flow analysis region.

(Feature 6 of the invention: blood vessel shape construction device having adhesion detection and separation functions)
6-1. The present invention is a blood vessel shape constructing apparatus for blood flow analysis by computational fluid dynamics, which has a function of automatically detecting and separating the presence or absence of adhesion of blood vessels when evaluating the blood vessel shape obtained from a medical image. It is a shape building device.
6-2. The means for automatically detecting and separating the presence or absence of adhesion of a blood vessel of the feature 6-1 has a means for detecting the presence or absence of a blood vessel loop in a blood vessel structure analysis, and a means for reconstructing a blood vessel shape from region division when adhesion is detected. ..
6-3. In the feature 6-2, the means for detecting the presence or absence of the blood vessel loop is performed by preparing data in which the blood vessel center line and the center line end point are associated with each other and detecting whether or not there are a plurality of blood vessels sharing the center line end point. ..
6-4. In Feature 6-2, when adhesion is detected, the means for reconstructing the blood vessel shape from the area division is performed by increasing the accuracy of the blood vessel inside/outside determination reference at the relevant location.
6-5. In the feature 6-4, the means for improving the accuracy of the blood vessel inside/outside judgment is to apply the blood vessel inside/outside judgment method on the assumption that the blood vessels do not overlap with each other.

Next, an embodiment of the present invention will be specifically described with reference to the drawings.

As described above, the present invention relates to a blood vessel shape constructing apparatus for blood flow analysis by computational fluid dynamics (CFD), its method, and a computer software program, and in particular, a medical image as an input. An apparatus for generating a blood vessel shape having no user dependence by fully automating the steps of constructing a blood vessel shape model and outputting a quality evaluation.

(Blood flow analysis using computational fluid dynamics)
To simplify the description of this embodiment, the concept of blood flow analysis using computational fluid dynamics will now be described.

In computational fluid dynamics, the flow of a fluid is calculated and output by calculation analysis by a computer. More specifically, the governing equation (continuity equation, Navier-Stokes equation) that describes the flow is replaced with an algebraic equation to obtain an approximate solution by sequential calculation.

In this embodiment, as shown in FIG. 1, four types of blood vessel shapes 1, blood physical properties 2, boundary conditions 3, and calculation conditions 4 are used as inputs. Based on these inputs, a computational fluid dynamics calculation (CFD) is executed, and a pressure field 5 and a velocity field (also called “velocity field”) 6 in space are output. In this example, the pressure field 5 and the velocity field 6 in space-time are calculated by solving the governing equation as a time evolution type.

FIG. 2 shows a flow of blood flow analysis by the above-mentioned computational fluid dynamics. In the flow of blood flow analysis, first, three-dimensional shape data (surface mesh) representing a blood vessel shape is constructed from a medical image obtained by X-ray CT or the like (FIGS. 2A to 2B). Then, for the constructed three-dimensional shape data of the blood vessel shape, blood physical properties 2, boundary conditions 3, calculation conditions 4 and the like are set, and numerical fluid analysis calculation is executed (FIGS. 2C to 2G). .. Here, the blood property 2 is specifically the density and viscosity of blood. The boundary condition 3 is specifically the flow rate at the end surface of each blood vessel, the constraint condition at the blood vessel wall surface, and the like. For example, when the blood vessel wall is fixed, it is common to apply a so-called non-slip condition that makes the velocity zero by ignoring the slip of the fluid on the wall, but this is also part of the boundary condition. is there. The calculation condition 4 includes calculation grid generation for the blood vessel shape of the calculation target, equation discretization related to the equation solution method, and simultaneous equation solution method, and these conditions may be hierarchically set to set each condition.

Next, with reference to FIG. 3, the concept of the construction process of the three-dimensional shape data representing the blood vessel shape will be described. In the blood vessel shape construction, first, a medical image is acquired (FIG. 3(a)). Next, the inside/outside of the blood vessel is determined from the medical image, and the blood vessel and the other area are segmented (FIG. 3B). Next, an area to be subjected to blood flow analysis is set (FIG. 3(c)). After that, the surface shape of the blood vessel wall is constructed by the marching cube method or the like (FIG. 3D). That is, the voxel space of the image is transferred to the polygon space. Therefore, at this point the vessel wall is constructed from micro-polygonal elements such as micro-triangular elements.

In such a blood vessel shape construction, when each process shown in FIG. 3 is performed by an independent system or software, the constructed blood vessel shape may vary due to a difference in processing selected by the user. That is, user dependency cannot be eliminated. In order to construct a blood vessel shape in which user dependence is eliminated from various medical images, the following points are given as specific problems.
(1) The quality of the medical image to be input varies depending on the manufacturer, modality, imaging condition, and imaging target, but it is not clear what kind of medical image can achieve blood flow analysis that can guarantee accuracy. ..
(2) Although the brightness value in the blood vessel changes depending on the place, it has not been clarified which method for determining the blood vessel inside and outside can achieve the blood flow analysis with guaranteed accuracy.
(3) The result of blood flow analysis changes depending on how the region is set, but it is not clear what kind of blood vessel region setting method can achieve accuracy of blood flow analysis.
(4) Although the accuracy of shape construction itself changes depending on the size of the minute triangle, it has not been clarified what kind of shape construction method can achieve blood flow analysis that guarantees accuracy.

As described above, the quality of the blood vessel shape, that is, the shape reproducibility directly affects the blood flow analysis result. Therefore, it is important to construct a high-quality blood vessel shape with high reliability, and the blood vessel shape constructing apparatus in this embodiment guarantees the accuracy of the result of blood flow analysis having no user dependence by the processing described below. It is intended to construct a blood vessel shape with a certain quality.

(Outline of this embodiment)
In view of the above problems, the inventor of the present application has made an invention from the following viewpoints (I) to (VI).

(I) Blood vessel shape constructing device having fully automatic control function
A first feature of the present invention is a blood vessel shape constructing apparatus for blood flow analysis by computational fluid dynamics, which is a fully automated process of inputting a medical image and outputting a blood vessel shape and quality evaluation. A blood vessel shape construction device characterized by eliminating user dependence.
The present invention has solved the problem by taking a new viewpoint of performing absolute evaluation based on blood flow information obtained from the blood flow analysis result. More specifically, the wall shear stress acting on the blood vessel wall is an important index, and each process of vascular shape construction is optimized based on keeping the wall shear stress within a certain range to achieve quality control as a whole. We have developed a device for constructing blood vessel shapes. According to the present invention, it is possible to provide a blood flow analysis result with accuracy guaranteed, without depending on the experience and knowledge of the user.

(II) Blood vessel shape construction device having image standardization function
The second feature of the present invention is to provide an image standardization function in a blood vessel shape constructing apparatus that eliminates user dependency by fully processing each step of blood vessel shape constructing and enables quality control of blood vessel shape constructing. It is a thing.
It is known that the quality of medical images varies depending on the manufacturer, modality, imaging condition, and imaging target, but has not been standardized. In the conventional blood flow analysis, the blood vessel shape is constructed without correcting the input medical image. For example, in MRA, the signal intensity is an image in which the flow velocity dependence is emphasized, and in CTA and DSA, the contrast agent concentration dependence is emphasized. In the former case, there is a problem that the blood vessel shape is distorted according to the flow velocity distribution. .. The essential cause of this is that when images are compared as shape information, they are only relative evaluations, and accuracy evaluations based on true values cannot be achieved. The present invention has solved the problem by taking a new viewpoint of performing absolute evaluation based on blood flow information obtained from the blood flow analysis result. More specifically, the wall shear stress acting on the blood vessel wall is an important index, and the medical image correction method has been developed on the basis of keeping the wall shear stress within a certain range. According to the present invention, it is possible to provide a blood flow analysis result with accuracy guaranteed, without depending on the experience and knowledge of the user.

(III) Blood vessel shape constructing device having adaptive blood vessel region dividing function
A third feature of the present invention is that an automatic blood vessel shape segmentation function in a blood vessel shape construction apparatus that eliminates user dependence by fully processing each step of blood vessel shape construction and enables quality control of blood vessel shape construction. Is provided.
When extracting a blood vessel from a medical image by region division, the inside/outside determination of the blood vessel is performed. It is known that the signal intensity in blood vessels varies depending on the site. More specifically, the signal intensity is high in a main blood vessel having a relatively large diameter. Therefore, if the internal and external criteria for main blood vessels are applied to the entire vascular system, subordinate blood vessels will be underestimated. Also, if the subordinate blood vessels are used, the main blood vessels will be overestimated. The present invention has solved the problem by taking a new viewpoint of performing absolute evaluation based on blood flow information obtained from the blood flow analysis result. More specifically, the wall shear stress acting on the blood vessel wall is an important index, and we have developed an adaptive blood vessel segmentation method based on keeping the wall shear stress within a certain range. According to the present invention, it is possible to provide a blood flow analysis result with accuracy guaranteed, without depending on the experience and knowledge of the user.

(IV) Blood vessel shape constructing device having recursive blood vessel region dividing function
A fourth feature of the present invention is that the recursive blood vessel region segmentation function in the blood vessel shape constructing apparatus enables the quality control of the blood vessel shape constructing by eliminating the user dependency by fully performing each step of the blood vessel shape constructing. Is provided.
When extracting a blood vessel from a medical image by region segmentation, it is necessary to determine whether the blood vessel is inside or outside. The inside/outside determination of the blood vessel is performed by setting some threshold value with respect to the change in the signal intensity from the inside to the outside of the blood vessel. Here, the running direction of the blood vessel is arbitrary, and a phenomenon such that a certain part is close to being perpendicular to the image capturing direction and another part is close to being parallel appears. Since the signal intensity changes depending on the blood vessel running direction, it is desirable to extract the shape as viewed from the blood vessel running direction. However, in the extraction of the blood vessel region so far, a highly accurate blood vessel inside/outside determination method considering the blood vessel running direction has not been reported. The present invention has solved the problem by taking a new viewpoint of performing absolute evaluation based on blood flow information obtained from the blood flow analysis result. More specifically, it became clear that the wall shear stress acting on the blood vessel wall can be kept within a certain range by performing the blood vessel region extraction stepwise from the coarse extraction to the fine extraction based on the center line. It was In other words, we have developed a recursive blood vessel region segmentation method that recursively determines the inside and outside of a blood vessel. According to the present invention, it is possible to provide a blood flow analysis result with accuracy guaranteed, without depending on the experience and knowledge of the user.

(V) Blood vessel shape construction device having automatic region setting function
The fifth feature of the present invention is to provide an automatic region setting function in a blood vessel shape construction apparatus that eliminates user dependence by fully processing each step of blood vessel shape construction and enables quality control of blood vessel shape construction. To do.
It is necessary to set an analysis area for the blood vessel shape extracted by the area division. Until now, the analysis area was arbitrarily determined by the user, and this was a factor that caused variations in analysis results. The present invention has solved the problem by taking a new viewpoint of performing absolute evaluation based on blood flow information obtained from the blood flow analysis result. More specifically, the wall shear stress acting on the blood vessel wall is an important index, and based on keeping the wall shear stress within a certain range, the main blood vessels to be included in the analysis region are We have created a template that shows dependent blood vessels, and have developed a method for automatically setting blood vessel regions that eliminates user dependence by automating region setting based on the template. According to the present invention, it is possible to provide a blood flow analysis result with accuracy guaranteed, without depending on the experience and knowledge of the user.

(VI) Blood vessel shape construction device having adhesion detection and separation functions
The sixth feature of the present invention is to perform adhesion detection and separation functions in a blood vessel shape construction device that eliminates user dependence by fully processing each step of blood vessel shape construction and enables quality control of blood vessel shape construction. Is provided.
Although the blood vessel shape extracted by the region division originally has no adhesion, blood vessel adhesion may occur due to an artifact due to image processing. In this case, the user visually confirms the presence or absence of adhesion, and manually images the adhesion site to separate blood vessels from each other. For this reason, the user's arbitrariness was generated, which was a cause of variation in the analysis results. Therefore, the present invention has developed a method for constructing a blood vessel shape that automates the detection and separation of blood vessel adhesions. According to the present invention, it is possible to provide a blood flow analysis result with accuracy guaranteed, without depending on the experience and knowledge of the user.

Hereinafter, one embodiment of the present invention will be described in more detail.

(Best embodiment)
FIG. 4 is a schematic configuration diagram showing a blood vessel shape construction device according to this embodiment.

The blood vessel shape construction device 10 includes a program storage unit 60 and data storage units 70 and 71 connected to a bus 50 to which a CPU 18, a RAM 19 and an input/output unit 20 are connected. In this device, the bus 50 may be connected to other devices that store various external reference data.

The program storage unit 60 includes an image input unit 11, an image discrimination unit 12, an image standardization unit 13, a shape construction unit 14, a quality evaluation unit 15, a shape measurement unit 16, and a blood vessel shape data output unit 17. I have it. The data storage unit 70 stores the medical image data input by the input/output unit 20. The data storage unit 71 stores blood vessel shape data 21, image appropriateness 22, image correction degree 23, lesion site/type 24, blood vessel shape measurement data 25, and quality score 26.

The constituent elements (the image input unit 11, the image discrimination unit 12, the image standardization unit 13, the shape construction unit 14, the quality evaluation unit 15, the shape measurement unit 16, and the blood vessel shape data output unit 17) are actually Is constituted by computer software stored in the storage area of the hard disk, and is called by the CPU 18, expanded on the RAM 19 and executed, so that it is configured and functions as each component of the present invention.

FIG. 5 shows a flow for constructing a blood vessel shape by this device. Hereinafter, each step shown in this figure will be described in order.

(Medical image acquisition (step S1))
In step S1, a medical image is input to the image input unit 11 of this system.

In this embodiment, the input medical image is a device capable of acquiring a tomographic image of a target blood vessel region such as MRA (magnetic resonance image), CTA (X-ray computed tomography image), and DSA (angiography image). In addition, it may be obtained by various devices such as US (ultrasonic image), IVUS (intravascular ultrasonic image), and OCT (near infrared image) capable of acquiring image data of a target blood vessel region. ..

The target blood vessel portion may be, for example, a cerebral artery, a coronary artery, a carotid artery, an aorta, or another target blood vessel portion of the subject.

As described above, the quality of the medical image, that is, the image characteristic changes depending on the device used, the imaging target, the imaging condition, and the like. Below, the factors that cause qualitative differences in medical images will be explained systematically.

There is a standard for medical images, which is called DICOM (Digital Imaging and Communication in Medicine), which indicates image attributes, and many medical images comply with this standard. However, little standardization has been done on the quality of medical images. Therefore, if appropriate correction processing is not performed on the medical images captured by the various devices described above, different blood vessel shapes may be constructed according to the image characteristics of the medical images. That is, when a blood vessel shape is constructed from a medical image obtained with different devices and imaging conditions by a method as shown in FIGS. 3A to 3D, even if the same blood vessel is imaged, the shape is the same. Not necessarily, it is expected that the results of blood flow analysis will change and affect the accuracy of the analysis results.

As shown in FIG. 6, the factors that cause the difference in the image characteristics of medical images are roughly classified into “device characteristics 31” and “imaging characteristics 32”, and the device characteristics 31 are “maker 33” and “modality”. 34”, the imaging characteristic 32 can be divided into an “imaging condition 35” and an “imaging target 36”.

Here, the "maker 33" is a company that develops an imaging device. The "modality 34" is an imaging device, and is, for example, the above-mentioned MRA, CTA, DSA, US, IVUS, or OCT device. The difference in the measurement principle of various imaging devices has a strong influence on image characteristics, as will be described later.

The "imaging condition 35" can be subdivided into (1) measurement conditions and (2) injection conditions. “Measurement conditions” are meant to include the intensity of irradiation waves such as magnetic fields and X-rays, and spatial and temporal resolution. An image construction method for acquiring a two-dimensional tomographic image group as an image set is included in the measurement conditions. This "image construction method" differs depending on the measurement principle of the imaging device. That is, it depends on modality. For example, in the case of MRA, since imaging/measurement is performed while scanning in the body axis direction, usually a tomographic image can be directly acquired. On the other hand, in the case of CTA or DSA, since the helical method or tomography method is used, a tomographic image is acquired by converting the measurement data using a special reconstruction function. Further, in the case of IVUS or OCT, since the inside of the blood vessel is scanned and imaged while rotating the sensor at the tip of the catheter, a tomographic image in the blood vessel traveling direction is acquired by correcting the difference in the local coordinate system at each imaging position. .. Further, the image construction method also includes a subtraction method that removes bones and calcification by subtracting the images before and after the contrast in imaging using a contrast agent. The subtraction method is extremely useful in constructing blood vessel shapes because it can remove the effects of bone and calcification. The “injection condition” is an injection condition of a contrast agent injected from a vein or an artery in order to make a blood vessel image clear. The type, concentration, injection rate, injection time, etc. of the contrast agent affect the image quality.

The "imaging target 36" can be subdivided into "patient status" and "imaging site". The “patient situation” is related to the degree of vascular lesion and the presence or absence of treatment. Examples of the "vascular lesion" include those caused by arteriosclerosis. When arteriosclerosis progresses, it often accompanies calcification, but in an image including such an arteriosclerosis site, local halation may occur in the image due to calcification (or advanced calcification), Affects the construction of blood vessel shape. The “presence or absence of treatment” is, for example, whether or not a treatment device such as a stent, a coil, or a clip is indwelling in the target blood vessel site. In a patient who has undergone indwelling treatment with such a treatment device, noise may occur around the treatment device and the image quality may deteriorate. The “imaging region” refers to a factor that affects the image quality in a region-dependent manner. For example, in the case of MRA imaging in the vicinity of the nasal cavity, there is a tendency for spatial variation of the magnetic field due to the air layer. In addition, when imaging around a hard tissue such as a bone by CTA, it tends to be difficult to distinguish the bone from the blood vessel.

As described above, the factors that cause the difference in the image characteristics are influenced by an extremely large number of parameters, and the qualitative image correction of the medical image according to the factors is indispensable for ensuring the accuracy of the blood flow analysis.

Therefore, in each of the following steps, image correction is performed according to various factors.

(Image discrimination step (step S2))
In this step S2, the image discriminating unit 12 encodes and digitizes two pieces of "image attribute information" and "image analysis information" as image proper information, and processes them as input images for blood flow analysis based on these. Whether or not input is possible is determined.

FIG. 7 is a configuration diagram of the image determination unit 12.

The image discriminating unit 12 receives a medical image, and has a configuration including (1) a device characteristic discriminating unit, (2) an image capturing characteristic discriminating unit, and (3) an image proper discriminating unit. Outputs the image adequacy.

First, when a medical image is input (step S2-0), the device characteristics are discriminated as image adequacy information by the manufacturer discrimination unit and the modality discrimination unit (steps S2-1-1 and S2-1). 2). Thereby, the manufacturer and modality of the device that gave the medical image are determined. Since the manufacturer and modality information is added to the medical image, it is based on that.

Next, the image pickup characteristics are discriminated by the image pickup condition discriminating unit and the image pickup target discriminating unit (steps S2-2-1 and S2-2-2). Since the information on the imaging characteristic is added to the medical image, it is performed based on it. The information regarding the imaging condition is added as attribute information to the medical image, and is therefore based on the attribute information (step S2-2-1).

In the image pickup target discriminating unit, image analysis is performed, various noises are discriminated, and quantitatively output as image adequacy information (step S2-2-2). Specifically, (1) device-derived ε ₁ , (2) contrast agent-derived ε ₂ , (3) blood vessel-derived ε ₃ , (4) hard tissue-derived ε ₄ , (5) site-derived ε ₅ , (6). Each noise of the treatment-derived ε ₆ is calculated. In this embodiment, noise is defined as an error factor in performing blood flow analysis.

The specific calculation method will be described below.
(1) Device-Derived Noise Device-derived noise is device-specific background noise included in the background of an image. There are random noise that is uniformly dispersed in the entire image and bias noise that appears in a striped pattern in the local area of the image. In this embodiment, the characteristics of these noises are digitized by a test image obtained by a phantom. More specifically, random noise is quantified and output as a high-frequency component by Fourier transform, and bias noise is quantified as a low-frequency component for output.
(2) Contrast agent-derived noise Contrast agent-derived noise is detected as uneven brightness values due to poor mixing of the contrast agent and blood. When the contrast agent and blood are mixed well, the change in the brightness value of the contrast agent shows a certain characteristic regardless of the location, and therefore it can be determined. For example, when there is no unevenness in the brightness value, it is the same in the direction orthogonal to the blood vessel, and decreases gradually in the traveling direction. The poor mixing of the contrast agent and blood is quantified by analyzing the signal intensity of the brightness value of the contrast agent.
(3) Blood vessel-derived noise Blood vessel-derived noise is noise associated with deviation or adhesion of blood vessel shape. In lesions such as aneurysms and stenosis, the shape of blood vessels deviates from the normal value. If a blood vessel shape that deviates excessively is used, the accuracy of blood flow analysis itself may not be guaranteed. For example, it is difficult to accurately quantify the stenosis itself in an excessively stenotic blood vessel. The same is true for adhesions. The blood vessel-derived noise is calculated based on the quantification of the shape of the blood vessel, and the region that is not suitable for blood flow analysis is defined as the blood vessel-derived noise and specified.
(4) Hard tissue-derived noise Hard tissue-derived noise is noise associated with bone and calcification. Bone and calcification are observed as high-brightness signals exceeding the brightness value of the contrast agent. Since it is a high-intensity signal, it is likely to be erroneously recognized as a part of the lumen of the blood vessel even though it is inside or outside the blood vessel wall. Therefore, here, a signal due to bone and calcification is detected as noise. Bone is an anatomically continuous high-intensity signal, and calcification is a discontinuous high-intensity signal isolated from the surroundings. Based on the difference in the characteristics of the signal strength, both noises are detected and quantified to make a determination.
(5) Site-originated noise Site-originated noise is detected by the degree of luminance change in the blood vessel surrounding area. For accurate blood vessel shape construction, it is desirable that there is sufficient contrast around the blood vessel as compared with the inside of the blood vessel. On the other hand, noise components such as hard tissues such as bone and calcification, soft tissues such as brain parenchyma tissue, and air layers around the nasal cavity exist around blood vessels. Here, the site-derived noise is quantified and discriminated by analyzing the change in the brightness value of the blood vessel surrounding region.
(6) Treatment-originated noise Treatment-originated noise is, for example, noise peculiar to the treatment device when the irradiation wave from the diagnostic device is reflected on the treatment device when various treatment devices such as stents, coils, and clips are already implanted in the body. It is a phenomenon that appears. There is a feature that it is observed as a high-luminance signal having a spatially constant pattern, and noise is detected by detecting the pattern and quantifying the signal intensity.

Next, the image adequacy is discriminated by the image adequacy calculation unit and the input propriety discriminating unit (step S2-3-1 and step S2-3-2). The image adequacy calculating unit defines the sum of the above noises as the noise amount and calculates (step S2-3-1). That is, the total noise amount is

Becomes Here, α and β are coefficients and multipliers, which are unique values for each noise. In this embodiment, these values are optimized by machine learning according to the increase in the number of cases. In this machine learning, if machine learning is performed in the hospital, the degree of optimization will differ depending on the facility due to the difference in the number of cases, so the image analysis results will be transferred to an external data server outside the hospital via a dedicated line, etc. Achieve homogenization of values between facilities by performing machine learning in multi-institution network through external data server. The coefficients/multipliers α and β obtained here are appropriately written and updated in the image analysis information template, and are used for the calculation of the sum total of the noise amount and the calculation of the image adequacy described below.

On the other hand, the input permission/inhibition determination unit determines the input permission/inhibition based on the image attribute information (input permission determination template 40 based on the image attribute information) and the image analysis information (input permission determination template 41 based on the image analysis information) (step S2-3-2).

An input permission/inhibition determination template (hereinafter, referred to as “image attribute information template”) 40 based on image attribute information is a template of inputtable manufacturers, modalities, and imaging conditions. This is a template having a function of outputting 1 if not possible.

The image analysis information template (hereinafter referred to as “image analysis information template”) 41 is a template that has a function of outputting a value that is standardized by the total noise amount, that is, less than 1. Specifically, the coefficients α and β for each noise type with respect to the sum of the noise amounts shown in the above equation, and the threshold value for the total noise amount are stored. Then, as described above, α and β are constantly updated by machine learning. By comparing the threshold value with the noise amount, the image adequacy is determined by the ranking of A, B, and C in this embodiment.

The input availability determination unit determines whether the medical image can be input based on the image attribute and the image adequacy. For example, regarding the above-mentioned image attribute, it is determined that input is rejected for a specific manufacturer or modality. Regarding the image adequacy, those of rank C are judged to be input rejected.

After that, the image discrimination unit 12 outputs the medical image and the image adequacy (S2-4). In this embodiment, the medical image determined to be inputable and the image adequacy thereof are output. However, the present invention is not limited to this, and the image adequacy may be output even when it is determined that the input is rejected. .. Further, each calculated noise amount may be output.

(Image standardization process (step S3))
FIG. 8 is a configuration diagram of the image standardization unit 13.

The image standardization unit 13 receives a medical image and image adequacy information (including image adequacy), and has a configuration including a noise removal unit, an isotropic unit, a resolution increasing unit, and an image correcting unit. The standard medical image, the image adequacy, and the image correction degree are output.

First, when the medical image and the image adequacy are input to the image standardization unit 13 (S3-0), the noise removal unit removes noise included in the original medical image (step S3-1). The noise here is derived from (1) device, (2) contrast agent, (3) blood vessel, (4) hard tissue, (5) site, and (6) treatment. Since the identification of the noise portion has already been completed by the image discrimination unit 12 as the discrimination of the image adequacy information, this noise removal unit removes various noise components from the specific portion of the original image.

The noise is specified in either the spatial domain or the frequency domain. In this embodiment, the spatial domain indicates that the image is directly analyzed, and the frequency domain indicates that the image is frequency-analyzed by using Fourier transform or wavelet analysis. The noise removing unit removes a noise component in either the spatial domain or the frequency domain according to the difference in the quantification method of various noises.

Next, the isotropic section and the high resolution section respectively perform the isotropic (step S3-2) and the high resolution (step S3-3) by image interpolation. In the isotropic process, anisotropy of spatial resolution is corrected and an image composed of cubic voxels is acquired. With higher resolution, the apparent spatial resolution is increased.

Next, the image correction unit performs the image correction in two stages (step S3-4).

First, as the primary image correction, the device characteristics and the imaging characteristics are corrected (step S3-4-1). In this correction, the device characteristic, that is, the image distortion due to the device, and the image pickup characteristic, that is, the image distortion due to the image pickup are corrected.

The image distortion caused by the device is the distortion given to the entire image by the difference in manufacturer or modality. For example, MRA acquires a slice image while scanning an imaging cross section in the body axis direction, CTA or DSA scans a tomographic image while rotating the device, and IVUS or OCT scans a catheter-type image sensor placed in a blood vessel. While acquiring the slice image. The difference in these modalities causes image distortion unique to the device. In this embodiment, the characteristics of such image distortion are quantified in advance from an image test using a phantom, and the distortion correction is performed using a correction line example that is optimized to minimize image distortion from the test image using the phantom.

The image distortion derived from the image pickup is the image pickup condition or the image pickup target gives to the entire image. The image distortion due to the imaging condition includes that due to the image reconstruction method. Similar to the above, this distortion is also corrected by using the correction matrix obtained by the phantom experiment.

After the above primary image correction is completed, the process proceeds to the secondary image correction described later. The primary image correction targets distortion of the entire image, whereas the secondary image correction targets local blood vessel shape distortion. In the blood vessel shape construction in this embodiment, the area inside and outside the blood vessel is divided as described later. This area division is to judge the inside and outside by applying some criterion to the change in the brightness value from inside the blood vessel to outside the blood vessel, but the change in the brightness value in the boundary area inside and outside the blood vessel is It is desirable to be constant regardless of the position inside. This is because when the change in the brightness value near the boundary between the inside and outside of the blood vessel (boundary area) differs depending on the position in the target blood vessel, local distortion may occur in the constructed blood vessel shape.

For example, there may be a case where the blood vessel portion having a linear shape and the blood vessel portion having a curved shape have different brightness value changes in the boundary region inside and outside the blood vessel, and based on such a medical image without performing any correction processing. When the blood vessel shape is constructed, it is expected that the constructed blood vessel shape will be distorted differently depending on the position. More specifically, in the case of non-contrast MRA, the brightness value itself becomes flow rate-dependent, so that the brightness value changes between the outer circumference and the inner circumference of the blood vessel, which may cause distortion in the constructed blood vessel shape. .. Also, in CTA and DSA using a contrast agent, the brightness value is concentration-dependent. That is, in CTA or DSA using a contrast agent, the change in the brightness value due to the flow velocity does not occur as in the case of non-contrast MRA, but the contrast agent has a higher specific gravity than blood and is uniformly dispersed in the blood vessel cross section. Since it is not always the case, the poor mixing of the contrast medium and blood may bring place-dependence on the change of the brightness value in the boundary region inside and outside the blood vessel.

Therefore, the image correction unit calculates the brightness gradient from the brightness value around the blood vessel wall of the medical image subjected to the first image correction (step S3-4-2), and corrects the brightness based on the change amount of the brightness gradient. The location to be set is detected (step S3-4-3). The image correction unit calculates the brightness gradient value with respect to the two axes of the long axis and the short axis. Here, the long axis is a coordinate axis taken in the blood vessel running direction, and the short axis is a coordinate axis taken in the blood vessel orthogonal direction. A method of calculating the blood vessel traveling/orthogonal direction will be described in detail in the shape construction (step 4). In this embodiment, a brightness gradient value inside and outside the blood vessel is calculated, and a location where the brightness gradient value is equal to or less than a threshold value is detected as a correction required location. This is because the location where the brightness gradient value is low is the location where the flow velocity is reduced, which may distort the shape of the blood vessel wall.

Then, the secondary image correction is performed on the portion requiring correction based on the morphological information or the blood flow information (step S3-4-4). In this embodiment, the morphological information and the blood flow information are a blood vessel statistical model and a blood flow statistical model calculated as statistical average values. The blood vessel statistical model is a vessel in which the blood vessel is a tapered tube in the flow direction and its shape change is replaced by a mathematical model. Moreover, the blood flow statistical model is a model in which the change pattern of the brightness value that appears in a blood flow-dependent manner is replaced with a mathematical model. For example, the blood flow statistical model is a mathematical model of blood flow information obtained by analyzing the actual blood flow information measured by the phase contrast MRI method or the spatial distribution of flow velocity dependence. These statistical models are acquired by machine learning from a standard medical image database. In this embodiment, the standard medical image database is composed of a multi-institution shared network arranged outside the hospital in order to eliminate the difference between the facilities.

When the secondary image correction is completed, the standard medical image, the image adequacy, and the image correction degree are output (step S3-4-5). In this embodiment, the image correction degree is the size of the area of the correction portion.

(Shape construction (step 4))
FIG. 9 is a configuration diagram showing the blood vessel shape construction unit 14.

Inputs are medical images standardized in the above process, image adequacy, and image correction, and outputs are blood vessel shape data, lesion site/type, image adequacy, and image correction.

The blood vessel shape construction unit 14 includes a region division (rough extraction) unit, a structure analysis unit, a region division (precision extraction) unit, an adhesion detection/separation unit, a region setting unit, and a shape construction unit.

The area dividing unit executes adaptive blood vessel area division that performs area division based on an adaptive criterion provided for each blood vessel, and recursive blood vessel area division that repeatedly performs fine extraction or fine extraction.

In the region setting, labeling each blood vessel and lesion site, and referring to the anatomical template, the computer sets the blood vessel region necessary for blood flow analysis is called automatic blood vessel region setting.

The blood vessel shape construction unit 14 first performs a rough extraction by temporarily calculating a threshold value (step S4-1-1-1) and binarizing the entire image (step S4-1-1-2). As the threshold value, one calculated from a histogram of brightness values of the entire image by setting a predetermined reference is used. However, when it is determined that the degree of adhesion detection, which will be described later, is high, the threshold value is made stricter. Although the blood vessel shape can be obtained to some extent when binarization is performed on the entire image, the blood vessel extraction accuracy is rough because a unique threshold is given to the entire image.

Next, the blood vessel shape construction unit performs thinning (step S4-2-1). Thinning is a technique for acquiring the center line of a blood vessel. Although several techniques are well known, in this embodiment, for example, the core of the blood vessel is extracted by facing the skin from the outer peripheral side of the blood vessel to the center one skin at a time, and a spline curve in the blood vessel traveling direction, etc. Is fitted to form the center line.

After that, the blood vessel shape construction unit 14 graphs the blood vessel structure using the acquired center line of the blood vessel (step S4-2-2). That is, that is, the end portion/branch point of the blood vessel center line is specified, and a graph showing the connection relationship between the node (node) and the side (edge) as shown in FIG. 11B is acquired. Hereinafter, each blood vessel portion corresponding to the center line between the end points/branching points is referred to as a blood vessel element.

Next, image analysis is performed based on the traveling direction of each side graphed (step S4-2-3). More specifically, a tomographic image perpendicular to the blood vessel traveling direction is specified at each point in the blood vessel traveling direction, and the change characteristic of the brightness value in the boundary region inside and outside the blood vessel is acquired for each blood vessel element from the tomographic image group between each node. To do. That is, an adaptive inside/outside determination criterion is obtained for each blood vessel element. In this embodiment, the half value of the difference between the brightness value on the blood vessel center line and the background brightness value is used as the inside/outside determination criterion. By using the inside/outside determination criteria for each blood vessel element, binarization can be adaptively performed for each blood vessel in the subsequent extraction processing. In this structural analysis, the change characteristic of the brightness value may be acquired for each blood vessel element, or the change characteristic of the brightness value may be acquired for each minute region in which one blood vessel element is subdivided.

Next, the blood vessel shape construction unit 14 realizes accuracy improvement by performing area division for each blood vessel element and then performing area division again from thinning again. By repeating the region division, the centerline can be specified in a place-dependent manner, and thereby the accuracy of the region division of the blood vessel can be improved. In rough extraction, one criterion was applied to the whole as described above, but in precise extraction, it is optimized for each site even in one blood vessel by repeating precision extraction that adaptively applies the criterion for each blood vessel element. Apply the specified criteria. In this embodiment, the blood vessel region specified by the rough extraction is used as a reference, and the difference with the blood vessel region obtained each time the precision extraction is performed is quantified. Then, by determining the convergence of this difference, a condition for ending the precise extraction is given. That is, when the difference exceeds the threshold value, the precision extraction is repeated, and when the difference becomes equal to or less than the threshold value, the precision extraction ends.

Next, the blood vessel shape construction unit 14 detects and separates adhesions (step S4-3). As shown in FIG. 10, the adhesion here means a phenomenon in which blood vessels originally do not adhere to each other, but they seem to adhere to each other due to insufficient image accuracy. Hereinafter, this phenomenon is referred to as "adhesion" or "adhesion artifact".

Normally, a blood vessel is a loop circuit of a closed loop at the whole body level, but no loop is formed in a relatively microscopic region handled by blood flow analysis. Therefore, in this embodiment, the presence/absence of adhesion is detected depending on the presence/absence of a loop (step S4-3-1). Specifically, as shown in FIG. 11, the adhesion site is detected by graphing the blood vessel shape and depth-first search of the graph. That is, the end point and the branch point are detected, and the connection relationship between them is examined.

For example, the branch structure of a blood vessel is represented in the graph shown in FIG. 11B processed based on the blood vessel shape shown in FIG. 11A. In this figure, the circles indicate nodes, and end points or branch points. When adhesion occurs, it appears as a closed loop (loop) in this graph, and this closed loop is detected by performing a depth tip search of the graph. This is an operation that follows an edge from the initial node to pass through all nodes. The rules of how to follow are as follows.
(1) When a branch point is reached, select an edge that has not passed and proceed to the next node.
(2) When reaching the end point, return to the fork point in front.
(3) If there is no side that has not passed when returning to the branch point, return to the preceding branch point.

In FIG. 11C, the number of each node represents the visit order of the branch depth priority search. In addition, the solid arrow indicates the outward path and the dotted arrow indicates the return path. In the example shown in FIG. 11(c), a closed circuit occurs between No. 5 and No. 6. When the depth-first search is performed, the node No. 5 advances to Nos. 6, 7, 8 and returns to No. 6. Here, since the outward route has passed through the lower side of the figure, the upper side has not yet passed. However, if you choose the upper side, you will arrive at No. 5, which has already been visited. If there is no closed circuit, only the returning route returns to the visited node, but if there is a closed route, the visited node is reached on the outward route. Therefore, it is possible to detect a closed circuit by recording the passing conditions of the node and the side, and regard the node (No. 6 in the example shown in FIG. 11(c)) at the time when the closed circuit is determined as the adhesion site. You can

When the blood vessel shape constructing unit 14 detects the adhesion site, it then separates the adhesion site (step 4-3-2). This adhesion separation is performed by setting a certain threshold and evaluating the degree of adhesion. If the degree of adhesion is equal to or greater than the threshold value, the degree of adhesion is positively relaxed by returning to the starting point of the region division and performing binarization based on a stricter criterion. If the degree of adhesion is less than or equal to the threshold value, the adhesion site is separated based on the running direction and shape of the blood vessel connected to the adhesion site. The separation process is performed based on the blood vessel region, the center line, the running direction of the blood vessel, and the cross-sectional area. The blood vessel region, the center line, and the running direction of the blood vessel are acquired by region segmentation and structural analysis, and the cross section of the blood vessel is obtained by measuring the cross section perpendicular to the blood vessel running direction of the blood vessel region.

The flow of the separation process is as follows.
(1) Obtain the adhesion section from the change in the cross-sectional area of the blood vessel.
(2) The centerline of the adhesion section is estimated using the centerlines of the blood vessels before and after the adhesion section.
(3) The shape of each blood vessel is estimated by fitting two contacting ellipses to the contour of the blood vessel surface in the blood vessel cross section of the adhesion section.
(4) Divide the adhesion section cross section into two.

The respective separation processes (1) to (4) will be described in more detail below with reference to FIG. First, the cross-sectional area of the blood vessel before and after the adhesion section is measured. Since two blood vessels are integrated in the part where adhesion is made, when the cross-sectional area is plotted along the running of the blood vessel, only the adhesion section is shown as shown in the lower graph in Fig. 12(a). The cross-sectional area increases. By detecting this change, the adhesion section is determined.

Next, when focusing on the centerline of the blood vessel, the originally two centerlines shift inward and finally come into contact with each other to form a branch point. What is indicated by the alternate long and short dash line in the upper diagram of FIG. However, since it is originally two blood vessels, it should have two curves that do not intersect, as shown by the dotted line in the upper diagram of FIG. Therefore, it is estimated by interpolating the original center line of the adhesion section from the center lines before and after it. Then, using the estimated center line, the contour of each blood vessel is estimated on the cross section of the adhesion section. FIG. 12B shows a cross section perpendicular to the blood vessel traveling direction in the adhesion section. The two black dots 51 indicate the estimated centerline. Assuming that the cross section of the blood vessel is an ellipse, the two ellipses 52 are fitted to the outline 53 of the blood vessel in the adhesion section under the condition that the centers of the two ellipses 52 are known and are in contact with each other. The two ellipses to which the dotted line of FIG.12(b) fits are shown. Finally, the inside of the adhesion zone is divided into two regions according to the ratio of the diameters of the two blood vessels, and the separation process is completed.

Next, the blood vessel shape construction unit 14 performs labeling for each blood vessel (step S4-4-1). Labeling of blood vessels is performed by identifying main blood vessels and tracking each blood vessel based on graph data. The computer automatically labels each blood vessel by matching it with the anatomical template showing the anatomical position and orientation. At this time, the position suspected to be a vascular lesion is specified. A vascular lesion is a lump or narrowing of a blood vessel, and the shape of the blood vessel locally changes. The type and site of the lesion are identified by analyzing the shape change in the blood vessel running direction for each blood vessel. For example, a lump has a locally increased cross-sectional area, whereas a stenosis has a locally reduced cross-sectional area. The lesion site/type is specified according to the degree of change in the cross-sectional area.

Next, the blood vessel shape construction unit 14 deletes unnecessary blood vessels for constructing a blood vessel shape for blood flow analysis by filtering (step S4-4-2). The unnecessary blood vessels are small blood vessels and short blood vessels that are not suitable for blood flow analysis. Since the diameter and length of the blood vessel are acquired by thinning and graphing, filtering is performed by matching with the specified threshold. After that, a region for blood flow analysis is automatically extracted according to the type and site of the lesion (step S4-4-3). Here, it is decided with reference to the template prepared in advance.

Next, the blood vessel shape construction unit 14 forms a blood vessel phase by polygons by the marching cube method or the like, and acquires blood vessel shape data for blood flow analysis with appropriate smoothing (step S4-5). Note that, here, the method is not limited to the marching cum, and a vessel wall surface configured by a mathematical model may be used and the method is not limited.

Then, the blood vessel shape construction unit 14 outputs the blood vessel shape data, the lesion site/liquor identification data, the image adequacy, and the image correction degree (step S4-6).

(From shape measurement to output (steps S5 to S7))
Next, the operations of the shape measuring unit 15 and the quality evaluation unit 16 will be described with reference to FIG.

In this case, inputs are (1) blood vessel shape data, (2) lesion site/type, (3) image adequacy, (4) image correction degree, and outputs are (1) blood vessel shape data, (2) blood vessel. It is the shape measurement data, (3) lesion site/type, (4) lesion shape measurement data, and (5) quality score.

The shape measuring unit 15 calculates polygonalized blood vessel shape data (((1) blood vessel shape data and (2) lesion site/kind input in step S5-0) in each of the labeled blood vessel/lesion areas. )) is used to measure the shape (step S5-1). For a normal blood vessel site, an equivalent diameter is calculated by constructing an orthogonal cross section based on the center line of the blood vessel, which is the basis of shape measurement. In this embodiment, in addition to basic values such as length, surface area, and volume, blood vessel strain and the like calculated from the degree of deviation from a perfect circle are measured. The shape of the lesion is measured according to the type. For example, in the case of a bump, the neck length, bump depth, aspect ratio, surface area, volume, etc. are measured.

After the shape measurement, the quality evaluation unit 16 performs quality evaluation on the obtained blood vessel shape data (step S6). In this embodiment, the quality evaluation is a comprehensive evaluation value of the blood vessel shape data that is scored for three of (1) image adequacy, (2) image correction degree, and (3) shape construction accuracy. .. (1) and (2) are as described above. (3) The shape construction accuracy is the blood vessel wall identification accuracy in this embodiment. Although there is no single method for quantifying the blood vessel wall identification accuracy, in this embodiment, the blood vessel shape data that has been polygonized is quantified by superimposing it on a medical image. That is, the brightness gradient value near the blood vessel wall is used.

More specifically, as shown in FIG. 14A, a line segment Xi is formed in the direction perpendicular to the blood vessel surface with respect to the center of gravity of the minute surface (minute triangular element) of the polygonalized blood vessel shape data, and the line concerned is formed. Calculate the brightness gradient along the minute. FIG. 14B shows a graph in which the horizontal axis represents Xi and the vertical axis represents luminance value. As shown in the figure, the brightness value decreases from inside the blood vessel to outside the blood vessel. The steeper the degree of decrease, the clearer the contrast between the inside and outside of the blood vessel, which means that the blood vessel shape construction accuracy is higher. Therefore, the quality evaluation unit 16 calculates the brightness gradient for each triangular element of the blood vessel shape data. FIG. 14C is a histogram of the brightness gradient in each triangular element of one blood vessel element. In this embodiment, a histogram is created for each blood vessel element, but the present invention is not limited to this. After that, a threshold is set for the histogram to quantify the shape construction accuracy. For example, as shown by the shaded area in FIG. 14C, when the brightness gradient is less than or equal to the threshold value, it is determined that the quality is degraded, and the ratio below the threshold value is calculated.

After that, the quality evaluation unit 16 calculates the quality evaluation value X from the following formula using (1) image adequacy (α), (2) image correction (β), and (3) shape construction accuracy (γ). ..

Here, the respective proportional coefficients k _α , k _β and k _γ are values standardized so that the maximum value of the quality evaluation value X is 1, and are individually assigned to each value.

Finally, the device 10 outputs blood vessel shape data, blood vessel shape measurement data, lesion site/type, lesion shape measurement data, quality score and the like (step S7). The quality score may be the quality evaluation value X itself, or the quality evaluation value X may be classified into grades and expressed as characters. Alternatively, the blood vessel shape data may be superimposed and displayed.

In addition, the configuration of each part of the device in the present invention is not limited to the illustrated configuration example, and various modifications can be made as long as substantially the same action is exhibited.

For example, in the above embodiment, the image determination unit 12 determines the image adequacy based on the amount of noise, but the present invention is not limited to this. For example, based on image analysis information acquired by other image analysis. The image suitability may be determined. Furthermore, various noise amounts are not limited to those shown in the above embodiment, and other noise amounts may be quantified by image analysis.

Further, in the above embodiment, the manufacturer, modality, etc. are discriminated based on the data added to the medical image. However, the present invention is not limited to this. In this case, the image attribute information including the manufacturer and modality may be acquired from the imaging device that has captured the medical image via a network, a cable, or the like. In the above embodiment, a database for machine learning is installed outside the hospital, but the invention is not limited to this.For example, a database is installed in the hospital and communicates with other facilities via a network. The database may be updated or machine learning may be performed.

Further, in the above-described embodiment, the image standardization unit 13 acquires the image correction degree based on the size of the area of the correction location, but the present invention is not limited to this. Other information indicating the degree of correction before and after the correction may be acquired as the image correction degree.

Further, for example, in the above-described embodiment, the blood vessel shape constructing unit 14 uses the blood vessel region specified by the rough extraction as a reference to quantify the difference from the blood vessel region obtained each time the precision extraction is performed, and the convergence of the difference is determined. Although the end condition of the precise extraction is given by the judgment, the way of giving the end condition is not limited to this.As long as the quality of the blood vessel shape for blood flow analysis can be guaranteed, other end conditions can be used. It may be. For example, the termination condition may include the number of precision extractions.

Further, the device in the above embodiment is provided with an output unit, and outputs blood vessel shape data, blood vessel shape measurement data, lesion site/type, lesion shape measurement data, and quality score, but these data are part or all of them. May be transmitted by wire or wirelessly and displayed on other devices, such as other personal computers or laptops, or smartphones, tablets and the like.

Further, in the present invention, in addition to the processing in each part of the apparatus described in the above embodiment, the apparatus according to the present invention may be added with other processing suitable for constructing a precise blood vessel shape without user dependency. .. Further, it should be understood that the blood vessel shape constructing apparatus, the method thereof, and the computer program of the present invention can be applied to various uses as long as they have substantially the same operation.

Claims (32)

An input availability determination unit that receives a medical image, determines the image attribute of the medical image and the adequacy as the medical image, and determines whether or not it can be processed as an input image based on them.
A medical image determined to be processable as the input image is corrected/corrected based on the determined image attribute and adequacy, the image quality of the medical image is standardized, and the degree of correction/correction is indicated. An image standardization unit that outputs the degree of correction,
A blood vessel shape construction unit that constructs a three-dimensional blood vessel shape based on the standardized medical image,
The blood vessel shape construction accuracy of the constructed three-dimensional blood vessel shape is determined, and a comprehensive quality evaluation of the constructed three-dimensional blood vessel shape is calculated and output based on the blood vessel shape construction accuracy, the adequacy, and the correction degree. A blood vessel shape quality evaluation unit,
A blood vessel shape data output unit that outputs the blood vessel shape data indicating the three-dimensional blood vessel shape together with the shape evaluation. The device of claim 1, wherein
The image attribute of the medical image is information including information on the manufacturer and modality of the imaging device that captured the medical image, and imaging conditions,
The adequacy as the medical image is information including the noise level of the medical image,
The input availability determination unit,
An apparatus for performing the above-mentioned determination of whether or not input is possible by referring to an image attribute and adequacy in a predetermined template. The device according to claim 2,
The apparatus according to claim 1, wherein the input availability determination unit determines a noise level by performing image analysis on a medical image. The device according to claim 3,
The apparatus according to claim 1, wherein the input availability determination unit calculates different types of noise components according to the image attributes by image analysis, and determines the noise level of each component based on the calculated noise components. The device of claim 1, wherein
The image standardizing unit standardizes image quality by performing correction/correction for removing a specific noise component from the medical image based on the image attribute and the adequacy. The device of claim 1, wherein
The image standardization unit corrects/corrects the image quality by performing isotropic as image quality standardization by interpolating spatial resolutions in the X-direction, Y-direction, and Z-direction in an XYZ coordinate system orthogonal to each other included in a medical image by interpolation. A device for performing the following. The device of claim 1, wherein
The apparatus is characterized in that the image standardization unit performs correction/correction for increasing the resolution as the image quality standardization by increasing the spatial resolution of the medical image by image interpolation. The device of claim 1, wherein
The image standardization unit executes, as the standardization of the image quality, a correction of image distortion derived from imaging and a correction of local shape distortion of a blood vessel shape. The device according to claim 8,
The image standardization unit performs correction of the local shape distortion of the blood vessel shape on the medical image that is determined to be the non-contrast MRA by the determination of the image attribute. The distortion is corrected based on the luminance information of the region corresponding to the inside and the outside of the blood vessel on the medical image so as to eliminate the flow velocity dependence between the inside and the outside of the blood vessel, which is the characteristic of non-contrast MRA. A device characterized by being performed. The device according to claim 9,
The image standardization unit detects a portion where the flow velocity dependency of the medical image is stronger than a predetermined value based on the luminance information of the portion corresponding to the vicinity of the blood vessel wall on the medical image, and based on the blood vessel morphology information or blood flow information. An apparatus for correcting the flow velocity dependence of the detected location by the above. The device according to claim 10,
The image standardization unit is configured to detect a portion where a brightness gradient value in each cross section orthogonal to a blood vessel center line in the vicinity of the blood vessel wall is equal to or less than a threshold value as a portion having a strong flow velocity dependency. .. The device according to claim 10,
When correcting the flow velocity dependency based on the morphological information of the blood vessel, the image standardization unit corrects the flow velocity dependency of the image by applying the statistical model of the blood vessel at the detected location. A device characterized by being. The device according to claim 12,
The apparatus is characterized in that the statistical model of the blood vessel is a mathematical model of changes in the blood vessel cross section in the blood vessel running direction obtained from the blood vessel shape database. The device according to claim 10,
When correcting the flow velocity dependence based on the blood flow information of the blood vessel, the image standardization unit corrects the flow velocity dependence of the image by applying a blood flow model of the blood vessel at the detected location. A device characterized by being. 15. The apparatus according to claim 14, wherein the blood flow model of the blood vessel is a mathematical model of blood flow information obtained by analyzing actual blood flow information measured by a phase contrast MRI method or spatial distribution dependent on the flow velocity. A device characterized by being a product. The device of claim 1, wherein
The image standardization unit removes bones and/or calcifications that are apt to be erroneously recognized as blood vessels as high-brightness signals from the medical image determined to be CTA (X-ray computed tomography image) by the determination of the image attribute. An apparatus characterized by performing correction/correction. The device according to claim 16, wherein
The image standardization unit removes bone using anatomical features of the bone, or removes bone using a composite statistical model of blood vessels, and the statistical model of blood vessels is obtained from a blood vessel shape database. An apparatus characterized by mathematically modeling a change in a blood vessel cross section in the obtained blood vessel traveling direction. The device according to claim 16, wherein
The apparatus according to claim 1, wherein the image standardization unit removes calcification based on a brightness characteristic specific to calcification that is not present in bones or blood vessels. The device of claim 1, wherein
The blood vessel shape construction unit,
A rough dividing unit that roughly divides the blood vessel region by image analysis of the medical image,
A structural division of the roughly divided blood vessel region, and a precision division unit that performs a fine division of the blood vessel region based on the result of the structural analysis,
And a detection unit for detecting the presence or absence of adhesion of a specific blood vessel based on the result of the structural analysis,
The blood vessel shape constructing unit re-executes the precision division of the blood vessel based on the determination of the presence or absence of the adhesion. The device of claim 1, wherein
The blood vessel shape construction unit,
A rough dividing unit that roughly divides the blood vessel region by image analysis of the medical image,
And a precision division section that performs a structural analysis of the roughly divided blood vessel region and performs a fine division of the blood vessel region based on the result of the structural analysis.
The structural analysis acquires the center line of the blood vessel,
Once the center line of the blood vessel is obtained, the orthogonal image is acquired at each point on the center line, and the brightness value analysis of the orthogonal image is performed on each blood vessel.
The fine division unit divides a blood vessel region based on a result of the inside/outside determination of the blood vessel. The device of claim 20, wherein
The inside/outside determination by the luminance value analysis is based on a half value of the maximum luminance value in the centerline orthogonal image. The device of claim 20, wherein
The apparatus according to claim 1, wherein the blood vessel shape constructing unit increases the accuracy of blood vessel region division/extraction by recursively performing blood vessel inside/outside determination in the blood vessel region division/extraction. Computer software program executed by a computer, each instruction stored in the following storage medium:
A computer receives the medical image, determines the image attribute of the medical image and the adequacy as the medical image, and based on them, an input availability determination unit that determines whether the image can be processed as an input image,
The computer corrects/corrects the medical image determined to be processable as the input image based on the determined image attribute and adequacy, standardizes the image quality of the medical image, and corrects/corrects the medical image. An image standardization unit that outputs a correction degree indicating the degree,
A computer, a blood vessel shape construction unit that constructs a three-dimensional blood vessel shape based on the standardized medical image,
A computer determines the blood vessel shape construction accuracy of the constructed three-dimensional blood vessel shape, and calculates a comprehensive quality evaluation of the constructed three-dimensional blood vessel shape based on the blood vessel shape construction accuracy, the adequacy, and the correction degree. And output the blood vessel shape quality evaluation unit,
And a blood vessel shape data output section for outputting the blood vessel shape data indicating the three-dimensional blood vessel shape together with the shape evaluation. The computer software program of claim 23,
The image attribute of the medical image is information including information on the manufacturer and modality of the imaging device that captured the medical image, and imaging conditions,
The adequacy as the medical image is information including the noise level of the medical image,
The computer software program, wherein the input permission/inhibition determination unit causes the computer to determine the input permission/inhibition by referring to image attributes and adequacy in a predetermined template. The computer software program of claim 23,
The image standardization unit is a computer software characterized in that a computer corrects image distortion derived from imaging and locally corrects shape distortion of a blood vessel shape to standardize the image quality. program. The computer software program of claim 23,
The blood vessel shape construction unit,
A rough dividing unit that roughly divides the blood vessel region by image analysis of the medical image,
A structural division of the roughly divided blood vessel region, and a precision division unit that performs a fine division of the blood vessel region based on the result of the structural analysis,
And a detection unit for detecting the presence or absence of adhesion of a specific blood vessel based on the result of the structural analysis,
The computer software program, wherein the blood vessel shape constructing unit causes the computer to re-execute the precise division of the blood vessel based on the determination of the presence or absence of the adhesion. The computer software program of claim 23,
The blood vessel shape construction unit,
A rough dividing unit that roughly divides the blood vessel region by image analysis of the medical image,
And a precision division section that performs a structural analysis of the roughly divided blood vessel region and performs a fine division of the blood vessel region based on the result of the structural analysis.
The structural analysis is
Get the centerline of the blood vessel,
Once the center line of the blood vessel is obtained, the orthogonal image is acquired at each point on the center line, and the brightness value analysis of the orthogonal image is performed on each blood vessel.
The computer software program, wherein the precision dividing unit divides a blood vessel region based on a result of the inside/outside determination of the blood vessel. A computer implemented method comprising:
A computer receives the medical image, determines the image attribute of the medical image and the adequacy as the medical image, and based on them, an input availability determination step of determining whether the image can be processed as an input image,
The computer corrects/corrects the medical image determined to be processable as the input image based on the determined image attribute and adequacy, standardizes the image quality of the medical image, and corrects/corrects the medical image. An image standardization process that outputs a correction degree indicating the degree,
A computer, a blood vessel shape construction step of constructing a three-dimensional blood vessel shape based on the standardized medical image;
A computer determines the blood vessel shape construction accuracy of the constructed three-dimensional blood vessel shape, and calculates a comprehensive quality evaluation of the constructed three-dimensional blood vessel shape based on the blood vessel shape construction accuracy, the adequacy, and the correction degree. And output the blood vessel shape quality evaluation step,
And a blood vessel shape data output step of outputting blood vessel shape data indicating the three-dimensional blood vessel shape together with the shape evaluation. 29. The method of claim 28,
The image attribute of the medical image is information including information on the manufacturer and modality of the imaging device that captured the medical image, and imaging conditions,
The adequacy as the medical image is information including the noise level of the medical image,
In the input permission determination step, the computer executes the determination of the input permission by referring to the image attribute and the adequacy in a predetermined template. 29. The method of claim 28,
In the image standardization step, a computer performs a correction of image distortion caused by imaging and a correction of local shape distortion of a blood vessel shape to standardize the image quality. 29. The method of claim 28,
The blood vessel shape construction step,
A rough dividing step of roughly dividing the blood vessel region by image analysis of the medical image;
A structural division of the roughly divided blood vessel region, and a fine division step of performing a fine division of the blood vessel region based on the result of the structural analysis,
And a detection step of detecting the presence or absence of adhesion of specific blood vessels based on the result of the structural analysis,
In the blood vessel shape construction step, the computer re-executes the precision division of the blood vessel based on the determination of the presence or absence of the adhesion. 29. The method of claim 28,
The blood vessel shape construction step,
A rough dividing step of roughly dividing the blood vessel region by image analysis of the medical image;
The structure of the roughly divided blood vessel region is analyzed, and the fine division of the blood vessel region is performed based on the result of the structure analysis.
The structural analysis is
Get the centerline of the blood vessel,
Once the center line of the blood vessel is obtained, the orthogonal image is acquired at each point on the center line, and the brightness value analysis of the orthogonal image is performed on each blood vessel.
The precision dividing step divides a blood vessel region based on a result of the inside/outside determination of the blood vessel.

Images