RU2575495C2 - Method and device for analysing region of interest in object using x-rays - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: invention can be used to analyse a region of interest in an object using X-rays. The invention comprises (a) providing measurement data by using a differential phase contrast X-ray imaging system, and (b) analysing characteristics of the object in the region of interest. The measurement data comprise a two-dimensional or three-dimensional set of pixels, wherein for each pixel the measurement data comprise three types of image data spatially aligned with each other, including (i) absorption representing image data A, (ii) differential phase contrast representing image data D, and (iii) coherence representing image data C. The analysis step is based, for each pixel, on a combination of at least two of information comprised in the absorption representing image data A and information comprised in the differential phase contrast representing image data D and information comprised in the coherence representing image data C.

EFFECT: high reliability of segmentation or classification of internal structures in an object of interest.

12 cl

Description

FIELD OF TECHNOLOGY

The present invention relates to a method and apparatus for analyzing an area of interest in an object using x-rays. In addition, the invention relates to a computer program product configured to implement such a method, as well as a computer-readable medium containing a computer program product.

BACKGROUND OF THE INVENTION

X-ray imaging systems are standardly used to study objects of interest, in particular in medical applications. Recently, F. Pfeiffer et al., Phys. Rev. Lett. 98 (10), 108105, 2007, proposed an improved method for x-ray imaging, referred to as differential phase contrast imaging on gratings (DPCI). An example of a differential phase contrast imaging system is described in WO 2010/109368 A1.

SUMMARY OF THE INVENTION

There may be a need for improved capabilities for analysis, for example, such as segmentation or classification, of internal structures in an object of interest.

Such a requirement may be met by an object claimed in the independent claims. Embodiments of the invention are described in the dependent claims.

According to a first aspect of the present invention, a method for analyzing an area of interest in an object is provided. The method includes: (a) providing measurement data by means of a differential phase contrast x-ray imaging system, and (b) analyzing the characteristics of an object in a region of interest. Here, the measurement data comprises a two-dimensional or three-dimensional set of pixels, where for each pixel the measurement data contains three types of image data spatially aligned with each other, which include (i) image data A representing absorption, (ii) image data D representing differential phase contrast; and (iii) C image data representing coherence. The analysis step is based, for each pixel, on a combination of at least two of the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast and the information contained in the image data C representing coherence. For example, it may be advantageous to base the analysis step, for each pixel, on a combination of the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast.

The essence of the present invention is to identify that, taking into account, for example, both the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast, the analysis of the measurement data will make it possible to have a higher quality and accuracy than analysis based only on image data representing absorption, as is usually done. In particular, in segmentation methods, criteria for detecting boundaries are often based on gradients in the absorption image. Since information on such gradients can be obtained with high accuracy from image data D representing the differential phase contrast, an additional evaluation of these data to obtain an analysis of x-ray measurement data can improve the overall analysis result. The results of the analysis can be used, for example, in mammography, radiography and computed tomography systems.

Additionally, an indication of the reliability of the analysis result can be provided for each pixel based on the information contained in the image data C representing coherence. In other words, information about the loss of coherence, i.e. increasing the incoherence of the X-ray beam passing through the region of interest can be used to provide an indication of how reliable the information contained in the image data A representing the absorption and the image data D representing the differential phase contrast, and thus it can be assumed how reliable the results of the analysis of this data are. Such information can be very useful for the doctor in the planning and implementation of, for example, surgery. Alternatively, it is possible to use image data C representing coherence in order to provide other information. For example, incoherence may be due to the strong predominance of fibrous structures, such as fibrin filaments, in a direction parallel to the lattice structure of the interferometer.

The analysis may include the segmentation of various sub-regions within the region of interest based on a combination of the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast. Then, a segmented visualization of the anatomical features within the area of interest can be provided to an observer, such as a doctor.

Segmentation methods can include surface rendering, volume rendering, projection with maximum intensity, interactive refinement of the display, building up areas, front propagation, segmentation at a set level or segmentation based on a model.

The method may further comprise the steps of adjusting the properties of the visualized surfaces with respect to color, reflectivity, specularity and / or transparency, based on a combination of at least two of the information contained in the image data A representing the absorption and the information contained in the image data D, representing differential phase contrast, and information contained in image data C representing coherence. Here, various types of information sources contained in different image data sets A, D, C can be visualized in various ways so that a person looking can easily distinguish between the visualized information content of the displayed visualized surfaces.

In an alternative approach, the method may further comprise adjusting the opacity functions in volumetric visualizations based on a combination of at least two of the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast and information contained in the image data C representing coherence. Again, the person who is looking can get easy access to the visualized information content of the displayed volumes.

In order to provide additional informational content to the beholder, the method may further comprise transmitting signals about areas in which a predetermined quality requirement or accuracy requirement cannot be satisfied. Such requirements may be imposed, for example, due to clinical targeted use. For example, after or during the segmentation of parts of the region of interest, it can be determined based on the corresponding information contained in the image data D representing the differential phase contrast and / or the image data C representing the coherence, the part may or may not only respond minimally preliminary defined quality requirements or accuracy requirements and, if so, the segmented surfaces of this part can be visualized in a special way so that the beholder can easily to know the lack of quality / accuracy.

As an additional feature, the method may include adjusting the degree of freedom and / or sensitivity of user interaction for interactive refinement based on information contained in image data D representing differential phase contrast and / or image data C representing coherence.

In a particular embodiment, the method may further include classifying subdomains within the region of interest based on the results of the analysis, i.e. based on information included in image data D representing differential phase contrast and / or image data C representing coherence. Such a classification can be of high quality, since the additional information contained in the image data D and C can be used to refine the structural information contained in the image data A representing the absorption and to indicate the reliability of such information so that the updated structure can be analyzed for a more reliable classification of subdomains.

Finally, the method may further include selecting a minimum reliability based on clinical application requirements and pre-computing at least one of: segmentation results and classification results for various levels of reliability.

According to a second aspect of the present invention, there is provided a device for analyzing an area of interest in an object. The device is configured to implement the method described above. In particular, the device is configured to: (a) obtain measurement data from a differential phase-contrast x-ray imaging system and (b) analyze the characteristics of an object in the region of interest for each pixel. Here, the measurement data contains a two-dimensional or three-dimensional set of pixels, where for each pixel the measurement data contains three types of image data spatially aligned with each other, which include: (i) image data A representing absorption, (ii) image data D representing differential phase contrast, and (iii) image data C representing coherence. Here, the analysis step is carried out based on a combination of at least two of the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast and the information contained in the image data C representing the coherence.

According to a third aspect of the present invention, there is provided a computer program product that is configured to implement a method in accordance with the above first aspect when it is executed on a computer.

According to a fourth aspect of the present invention, there is provided a computer-readable medium on which a computer program product according to the preceding third aspect is stored.

These and other aspects of the invention will become more apparent from and explained with reference to the embodiments described hereinafter.

DETAILED DESCRIPTION OF EMBODIMENTS

The following explains a lot of embodiments of the analysis methods in accordance with the invention.

The diagnosis, as well as the planning and monitoring of the treatment of many diseases, are typically based on the analysis of x-ray images in order to determine the boundaries of anatomical objects, such as organs or lesions, and in order to classify anatomical anomalies according to their texture or shape of the border. Standard x-ray imaging is based on measuring the spatial distribution of the attenuation of incoherent x-rays in two or three dimensions using projection, reconstruction or tomographic methods, respectively.

The use of coherent X-rays for imaging measuring phase changes has recently been proposed by Pfeiffer et al. This method allows you to simultaneously register three images.

1. A standard attenuation image, which may be referred to herein as image data A representing absorption. This image depicts the spatial distribution of the linear attenuation coefficient, i.e. in mathematical terms, the imaginary part of the refractive index of a permeable medium. In two dimensions, this data contains linear linear attenuation integrals µ, while in three dimensions linear integrals can usually be reconstructed so that the data contains a map of the linear attenuation µ itself.

2. A gradient image, which in two dimensions can be represented in the form of gradients of linear integrals of decrease in the refractive index of X-rays ∇δ (associated with the phase gradient of the front of the X-ray waves), and which in three dimensions can be represented in the form of a map of decrease in the refractive index of X-rays δ, and which may be referred to herein as image data D representing differential phase contrast. This image represents an additional new image depicting a phase gradient in the coherence direction, i.e. in mathematical terms, the projection of the gradient of linear integrals of the real part of the refractive index onto the coherence direction (perpendicular to the lattice structures of the interferometer) in two dimensions, due to the spatially varying refractive index of x-rays in a permeable medium.

3. An image of incoherence, which can be a quantitative measure of ε the loss of visibility when an object is transmitted by X-rays, and which can be called coherence image data C in this document. This image is an additional new image depicting the loss of coherence of the original x-rays that they were exposed to when passing through through Wednesday.

The problem this disclosure is directed at is the poor contrast between the different types of soft tissues and the corresponding complex differentiation in the absorption images. This can be overcome at least partially by including information from δ and ε, i.e. image sets D and C, in the improvement of low-contrast edges or boundaries between structures.

In the analysis of x-ray images, such as image segmentation, visualization and refinement of anatomical structures, based on computerized image segmentation and classification of x-ray images, the reliability of the segmentation algorithms is highly dependent on changes in the signal value when the boundary between two image areas is crossed. The less pronounced this difference, the more difficult it becomes to draw a segmentation line between the two areas. For example, this may be true for the area of the tumor surrounded by healthy tissue in a standard attenuation-based image. The main idea underlying this disclosure is that the criteria for detecting the boundary can be based not only on gradients in the absorption image, but also on directly measured gradients of decrease in the refractive index of x-rays δ.

The following are details of various embodiments of border detection, image segmentation, and other x-ray image analysis procedures.

If the segmentation algorithm detects boundaries based on the condition that the absolute value of the image gradient in the attenuation image | ∇µ | greater than the predefined threshold T, i.e.:

| ∇µ |> T,

, т.е. условиеbut the image noise σ is not far enough from the difference in the average signal levels in two regions

, i.e. condition

not sufficiently implemented, the algorithm is unlikely to be able to identify the boundary between the regions to be segmented, and noise will significantly affect the segmentation results. In DPCI imaging, the X-ray refractive index decreasing gradient ∇δ is measured directly and therefore a better measure can be provided for detecting the boundaries between the two regions. In addition, the gradient operation necessary for the image attenuation enhances the noise present in the image and is not necessary for the phase gradient.

An alternative criterion for detecting the boundary between two different regions can be based on a combination of all three measured values of µ, ∇δ, ε, in a differential phase-contrast installation, for example, in the following form

,

,

where the subscript M of the phase gradient is used to indicate that it is a directly measured quantity, as opposed to the case of attenuation, where the gradient operator is explicitly applied to the measured value of µ. In the above equation, the two-dimensional spatial dependence of all measured values is clearly expressed. Since the loss of coherence in the x-ray phase, as reflected by an increase in ε, is a measure for the reliability of the phase gradient information, the function f must be a monotonically decreasing function of its argument in order to serve as a suitable weight function for the importance of the phase gradient to find borders. The value of α allows you to adjust the appropriate absolute balance between the two terms, and it will mainly be determined by noise and contrast levels in the initial measurements µ, ∇δ.

In another implementation, the three measured values can also be used to improve the edges by combining the information contained in the values of µ and ∇δ with ε as a control parameter that controls the extent to which the gradient information from and ∇δ are used to improve the edges in µ.

In yet another implementation, similar methods can be used to improve the edges on reconstructed tomographic images, where the noise power spectra of the reconstructed attenuation images µ and energy densities δ will be very different due to the different nature of the measured source data.

A combined gradient of µ, ∇δ, ε can be used not only in segmentation algorithms based on the gradient threshold, but also in all kinds of segmentation algorithms that use the edge quality function for each pixel (increases with the gradient value), or the path cost function for each pixel (decreases with the magnitude of the gradient). Then, in all these segmentation algorithms (for example, shortest path algorithms, graph cut algorithms, domain growth algorithms, etc.), the normal gradient is replaced by a combined gradient, i.e. from µ, ∇δ, ε.

Any method of segmentation or pattern recognition, the work of which is based on criteria based on the intensity that controls the internal operation of the algorithm, is a candidate for the benefit of embedding a combined analysis of at least two of the three sets of input image data A, D, and C.

The following briefly describes four options for the implementation of a computerized analysis of phase-contrast x-ray images for commonly used methods of segmentation:

- Area enhancement: Area extension methods segment the image into areas exhibiting similar intensity characteristics. Starting from one or several starting points, the areas are increased by merging with the image elements, provided that their intensity characteristics are quite similar to those of this area. Thus, the area is expanded until either more sufficiently similar image elements are found or a certain size or shape limit is reached. The merger, as well as the criteria for the limits, can be refined with respect to the resolution of the intensity using the gradient image information, i.e. through the use of higher resolution grayscale, when adjacent image elements are very similar and located in areas of well-preserved x-ray coherence. This information is key when achieving reliable segmentation is necessary in areas where clinically significant anatomical structures are located in close proximity, but only slightly differ in intensity values, for example, CT images of the abdominal cavity. Using information about incoherence, the reliability of the obtained segmentation results can be noted, and the areas that do not meet the quality requirements set by the target application can be transmitted, for example, by using color overlay on graphic symbols to attract the attention of clinicians to these areas as sections requiring further study from a clinical point of view.

- Front propagation: Front propagation methods segment an image by modeling wavefront propagation. The wave front advances iteratively in fixed time intervals with the propagation velocity determined by the intensity characteristics of the image areas that the wave overcomes. By decreasing speeds along significant intensity gradients, wave propagation leads to a halt indicating the end result of the segmentation procedure. The speed function can be fine-tuned according to the information on the image of the intensity gradient with high resolution in order to accurately direct the wave front through the regions of good x-ray coherence, i.e. high quality intensity information. Sites of low-quality gradient information can either be avoided or quickly passed according to the quality requirements of the intended application. As indicated above, areas where segmentation accuracy is limited can be noted and reported to the clinical user.

- Segmentation at a set level: methods with a set level are closely related to front propagation. The same wave equations are used, however, wave propagation can also occur back in time. Thus, the results of segmentation can be iteratively refined when errors occur, for example, violating certain limits of shape or size, penetrating into adjacent segments, etc. The detection of such segmentation errors, as well as the control of back propagation, can be fine-tuned using a gradient image and an incoherence image, controlling, for example, the inclusion or exclusion of areas of strong coherence deterioration, increasing the sensitivity of wave propagation to high-resolution attenuation gradients with iteration, etc. .

- Model-based segmentation: here is an anatomical model that describes location, size, shape, contrast, etc. the anatomical object is characterized and corrected by the anatomy of the patient of interest. This adjustment is controlled by two functions: the force member, which tries to align the anatomy as accurately as possible with the significant intensity measurements found in the image, and the control member, which prevents the model from swaying too much around structural noise. Gradient image and quality image are useful in fine-tuning these key terms, for example, adjusting the model stiffness using the degree of coherence, to continue the model smoothly through areas of unreliable intensity information, where the model might otherwise fall into the structural noise trap. Similarly, and depending on the clinical issue to be resolved, the force term can be finely tuned to make the model sensitive to small but clinically significant changes in intensity, in order to separate the boundaries of the tumor from neighboring critical organs, performing this only in areas where such minor changes can be detected reliably.

Four embodiments may be sufficient to illustrate the effect that phase contrast imaging can have on intensity-based segmentation of medical images.

The following are briefly described some embodiments of the methods for rendering phase-contrast corrected surfaces or volumes:

- Surface rendering: surface rendering methods are based on computer representations of the surfaces of segmented objects, in the simplest form, for example, isosurfaces with equal intensities. The mapping of these surfaces is generated in accordance with some shading model that describes the local color, reflectivity, specularity and transparency of the surface, and some lighting model that describes the distribution of light, ambient light and the scattering properties of the environment. Shading and lighting models can be adjusted in accordance with the information obtained during phase-contrast x-ray imaging. Small-scale surface structures derived from gradient data can be extended to areas of well-preserved x-ray coherence by increasing reflectivity and specularity to appropriate surface areas to make such small-scale structures more visible. Similarly, surface imaging may have a more diffuse or fuzzy appearance in areas where segmented surface elements have less reliability, as indicated, for example, by poor X-ray coherence. Similarly, the color of the surface can be adjusted in order to report on areas for which the accuracy requirements imposed by the clinic cannot be met, as determined from the gradient and coherence information. If these areas coincide with areas of proximity of critical anatomical objects, for example, to separate the tumor from an adjacent vasculature or organ at risk, the transparency of the corresponding surfaces can be adjusted to provide a simultaneous view of neighboring objects.

- Volume rendering: volume rendering methods generate visual representations of objects that do not fit the definition of unique surfaces. The three-dimensional distribution of the initial intensity values or (often fuzzy) membership values is not subjected to segmentation procedures, but they are directly rendered into displays using opacity functions. These opacity functions determine in which direction the intensity values encountered along the rays passing through a set of three-dimensional data along the direction of view are embedded in the two-dimensional plane of view. Integration of gradient and incoherence information into the opacity function complements the resulting display. In bulk regions of well-preserved phase coherence, i.e. reliable information about the phase gradient, the opacity functions can be set up sensitive to small-scale variations in the attenuation data to make minor details more noticeable in the final display. Similarly, transparency (α-channel) can be adjusted to simultaneously view adjacent structures through areas where small anatomical details can be reliably resolved in order to clarify complex proximity relationships, for example, to separate a tumor from an adjacent vasculature, or in order to display organ infiltration at risk of a tumor object. RGB channels can be adjusted according to incoherence information to report areas with poorly reproduced anatomy using staining or fuzziness to provide clinically useful immediate feedback for areas requiring further study and refined, interactively controlled analysis.

- Projections with maximum intensity: volume renderings of particular importance in medical applications are projections with maximum intensity, generated by displaying the maximum intensity values encountered along the rays passing through a set of three-dimensional data along the direction of view. This method is often used to display vascular structures in order to examine vascular malformations, as well as the vascular network inside and outside organs or tumors. Small-scale vascular structures can be resolved in areas of well-preserved phase coherence by integrating small-scale gradient information into a refined analysis of the intensity distribution near the maximum intensity values. In combination with rotation around an axis orthogonal to the direction of coherence, the display can be updated accordingly to maintain improved visual analysis of complex vascular structures.

- Interactive refinement of the display: the generation of clinically acceptable visual representations of three-dimensional anatomical structures requires the simultaneous manipulation of a very large number of parameters that control the appropriate methods for rendering the surface or volume. The number of degrees of freedom involved often makes this task time-consuming and annoying for the user, in particular when it is difficult to predict the outcome of these manipulations. Using information on the phase gradient and incoherence, it is possible to determine intelligent presets of these parameters, for example, to complement the display in areas of well-preserved x-ray coherence and related gradient information. Similarly, the interaction sensitivity can be adjusted in accordance with the available gradient and coherence data in order to prevent the user from moving the system away from clearly defined displays too easily or with errors.

These four options for implementation may be sufficient to illustrate the effect that phase contrast imaging can have on improved visual representation of anatomical objects based on x-ray data.

It may be noted that the described methods can be carried out by computer. Accordingly, an embodiment of the present invention relates to a computer program product or a computer program element, which is characterized in that it is configured to carry out or control the corresponding steps of the method in the described method according to embodiments of the invention when it is executed on a computer. The computer program product, therefore, can be stored in a computer unit, which can be configured to implement or cause the implementation of the steps of the method. In addition, it can be configured to actuate the components of the device described above. The computing device may be configured to automatically power and / or execute user orders. The computer software product will be loaded into the working memory of the data processor. The data processor can thus be equipped to implement a method in accordance with embodiments of the invention. This embodiment closes as a computer program product that uses the invention directly from the beginning, as well as a computer program product that, by updating, turns an existing program into a program that uses the invention. A computer-readable medium on which such a computer program product is stored may be provided, for example, in the form of a CD-ROM. Alternatively, the computer program may also be provided over a network, such as the Worldwide Web, and may be loaded into the working memory of a data processor. Accordingly, the computer-readable medium according to one embodiment may also be an environment that makes a computer program product available for download.

It should be noted that embodiments of the present invention are described herein with reference to various objects. In particular, some embodiments are described with reference to the claims on methods, while other embodiments are described with reference to the claims on devices. However, a person skilled in the art and the above and the following description will understand, unless otherwise indicated, in addition to any combination of features related to objects of the same type, also any combination of features related to different objects are considered disclosed in this application . However, all features can be combined to provide synergistic effects that are more than just the sum of the features.

In the claims and description, the word “comprises” does not exclude other elements or steps, and the singular forms do not exclude plurals. It should also be noted that the reference position in the claims should not be construed as limiting the scope of the claims.

Claims (12)

1. A method for analyzing an area of interest in an object, comprising the steps of:
- provide measurement data by means of a differential phase-contrast x-ray imaging system, the measurement data comprising at least a two-dimensional set of pixels, where for each pixel the measurement data contains three types of image data spatially aligned with each other, including:
- image data A representing absorption,
- image data D representing the differential phase contrast, and
- image data C representing coherence,
- analyze the characteristics of the object in the area of interest, based on, for each pixel, a combination of at least two of the formation contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast and information contained in image data C representing coherence. 2. The method according to claim 1, further comprising the step of providing for each pixel an indication of the reliability of the analysis result based on the information contained in the image data C representing coherence. 3. The method according to claim 1 or 2, in which, at the stage of analysis, various sub-regions are segmented within the region of interest based on a combination of the information contained in the image data A representing the absorption and the information contained in the image data D representing the differential phase contrast. 4. The method according to claim 3, in which the segmentation methods comprise at least one of surface rendering, volume rendering, projection with maximum intensity, interactive refinement of the display, extension of areas, spreading of the segmentation front at a specified level and segmentation based on the model. 5. The method of claim 1, further comprising adjusting the properties of the surfaces to be rendered with respect to at least one of color, reflectivity, specularity and transparency, based on a combination of at least two of the information contained in the image data A, representing absorption, and information contained in image data D representing differential phase contrast, and information contained in image data C representing coherence. 6. The method of claim 1, further comprising adjusting the opacity functions in the visualization volume based on a combination of at least two of the information contained in the image data A representing the absorption and the information contained in the image data D representing differential phase contrast, and the information contained in the image data C representing coherence. 7. The method of claim 1, further comprising signaling to areas where a predetermined quality requirement cannot be met. 8. The method of claim 1, further comprising adjusting at least one of a degree of freedom and sensitivity of user interaction for interactive refinement. 9. The method according to claim 1, further comprising the step of classifying subdomains within the region of interest based on the results of the analysis. 10. The method according to claim 1, further comprising the step of selecting a minimum reliability based on clinical application requirements and at least one of the segmentation results and the classification results for various levels of reliability is pre-calculated. 11. A device for analyzing an area of interest in an object, the device being configured to
- receiving measurement data from a differential phase-contrast x-ray visualization system, wherein the measurement data contains at least a two-dimensional set of pixels, where for each pixel the measurement data contains three types of image data spatially aligned with each other, including:
- image data A representing absorption,
- image data D representing the differential phase contrast, and
- image data C representing coherence,
- analysis of the characteristics of the object in the region of interest for each pixel based on a combination of at least two of: information contained in the image data A representing the absorption and information contained in the image data D representing the differential phase contrast and information, contained in the image data C representing coherence. 12. A computer-readable medium containing computer-executable instructions for performing steps of a method for analyzing an area of interest in an object according to claim 1.