Segmentation of a result image taking into account contour significance data

Abstract

Method for segmenting result image data (EBD) of an object under investigation (3) from measurement data (MD) acquired during a relative rotational movement between a radiation source (8) of an X-ray imaging system and the object under investigation (3), comprising the following steps: - Reconstructing (S1) initial image data (IBD) from the measurement data (MD), - Deriving (S5) contour data (KD) from the initial image data (IBD), - Calculating (S4) contour significance data (KSD) from the measurement data (MD) and / or the initial image data (IBD), - Calculating (S6) the result image data (EBD) using the contour data (KD) and the contour significance data (KSD), and - Segmenting the result image data (EBD).

Description

The invention relates to a method for segmenting result image data of an object under investigation from measurement data, a corresponding computer program, a corresponding data carrier, a corresponding control and computing unit and a corresponding X-ray image acquisition system. Medical imaging systems are characterized by the fact that internal structures of an object or patient can be examined without having to perform surgical procedures on them. Examples of such imaging systems or imaging devices are ultrasound systems, X-ray machines, X-ray computed tomography (CT) systems, positron emission tomography (PET) systems, single-photon emission tomography (SPECT) systems or magnetic resonance (MR) systems. In particular, X-ray imaging devices enable tomographic image generation, whereby a number of projections of the object under investigation are taken from different angles. From these projections, a two-dimensional cross-sectional image or a three-dimensional volume image of the object can be calculated. An example of such a tomographic imaging technique is the aforementioned X-ray CT scan. Methods for scanning an object with a CT system are well-known. These include, for example, circular scans, sequential circular scans with feed, or spiral scans. Other scanning techniques not based on circular motion are also possible, such as scans with linear segments. Using at least one X-ray source and at least one opposing detector, X-ray attenuation data of the object are acquired from different angles. These attenuation data or projections are then processed using appropriate reconstruction methods to create cross-sectional or volumetric images of the object. DE 10 2011 086 456 A1 discloses a method for reconstructing an image data set based on a projection data set obtained using an X-ray computed tomography device. Due to their non-invasive functionality, medical imaging devices already play a significant role in patient examinations. The images of a patient's internal organs and structures generated by these systems are used for a wide variety of applications, such as preventive examinations (screening), tissue sampling (biopsy), diagnosing the causes of diseases, planning and performing surgeries, and preparing for therapeutic procedures. In radiation therapy, for example, radiological data is needed to plan the radiation treatment with regard to dose distribution. The goal is for the dose in the area being treated to be above a certain threshold, while keeping it as low as possible in the surrounding tissue, especially in sensitive organs, to avoid secondary damage. For this and many other applications and tasks mentioned above, the segmentation of specific target structures is useful or even necessary. Such target structures can be, for example, defined bone structures, specific organs, vascular structures, or even defects or lesions such as tumors, which must first be identified and, if necessary, extracted from the image data. Segmentation generally refers to the creation of thematically related regions by grouping adjacent pixels according to a specific criterion. This criterion could, for example, be membership in a particular structure. The image data belonging to the structure can then be marked and / or virtually separated from the remaining image data and viewed separately, or made available for further analysis. A reliable and sufficiently accurate spatial delimitation of the segmentation of structures is essential for many applications. Organs or lesions are identified through manual contouring or automatic segmentation. Manual contouring, in which an operator uses a graphical user interface to draw boundary lines or points on a screen while viewing the image data, and segmentation is then based on these lines, is more reliable in terms of the accuracy of assigning pixels to structures and is therefore still considered the standard. However, it is extremely time-consuming, as the markings must be placed layer by layer by the operator. Automatic segmentation algorithms generally allow for a reduction in time and personnel costs while simultaneously increasing the objectivity of the segmentation. Besides primitive algorithms with linear edge detectors (Sobel / Scharr operator), more advanced algorithms consider the statistical significance of contours; that is, the segmentation weights the linear edge response with the background noise. However, the resolution and statistical properties in tomographic image datasets are non-trivial; the noise is non-stationary and anisotropic, and the resolution depends on the position in the measurement field and the direction. Therefore, this consideration based solely on the image data is only partially successful without additional information about the data acquisition. Consequently, the results of automatic segmentation are regularly corrected manually. In contrast, the object of the present invention is to provide an improved segmentation of result image data based on image data reconstructed from tomographic measurement data. This problem is solved according to the invention by a method for segmenting result image data, a corresponding computer program, a corresponding machine-readable data carrier, a control and computing unit for segmenting result image data and a corresponding X-ray image acquisition system according to the dependent claims. The inventive solution to the problem is described below with respect to both the claimed method and the claimed device. Features, advantages, or alternative embodiments mentioned herein are also applicable to the other claimed items and vice versa. In other words, claims relating to a device, for example, can also be further developed with features described or claimed in connection with a method. The corresponding functional features of the method are thereby realized by corresponding modules or units. The invention is based on the consideration that by appropriately manipulating image data during image reconstruction, subsequent automatic segmentation can be significantly simplified and thereby made more reliable. This is achieved by incorporating statistical properties of the image data not only as additional information during segmentation, but also directly into the reconstruction of the image data from the measurement or raw data. In other words, the inventors have recognized that the information content of reconstructed image data with regard to depicted contours can be improved for a subsequent segmentation step by emphasizing contours contained in the image data according to their statistical significance in a resulting image.The subsequent segmentation can therefore do without additional information or further knowledge about, for example, data acquisition or image reconstruction. Accordingly, the invention relates to a method for segmenting result image data of an object under investigation from measurement data acquired during a relative rotational movement between a radiation source of an X-ray imaging system and the object under investigation, comprising the following steps: - Reconstructing initial image data from the measurement data, - Deriving contour data from the initial image data, - Calculating contour significance data from the measurement data and / or the initial image data, - Calculating the result image data using the contour data and the contour significance data, and - Segmenting the result image data. The measurement data corresponds to the raw data or X-ray projections acquired by an X-ray detector. Initial image data is generated from this raw data using a so-called "neutral" reconstruction. The initial image data can be a two-dimensional or three-dimensional image set. The "neutral" reconstruction corresponds to a conventional, well-known reconstruction method that can be used for diagnostic purposes. The reconstruction method can be selected so that the initial image data is neutral with regard to the contour information it contains. This means that the contours are not enhanced or, more generally, manipulated by the reconstruction. Contour data can then be derived from the initial image data. This contour data represents all edges or contours depicted in the initial image data according to their position and direction.Derivation, in this context, refers to any analysis, manipulation, or evaluation of the initial image data suitable for determining the contour data. In a further step of the process, contour significance data is calculated. The inventor recognized that this data can be derived either from the initial image data or directly from the measurement data. Both the contour data and the contour significance data are then incorporated into the calculation of the resulting image data. Unless otherwise specified by the method according to the invention, the sequence of the steps included is arbitrary and variable. Calculating contour significance data as a step in image reconstruction provides complete knowledge about the resolution and statistical properties of the image data. In particular, anisotropic or directional noise effects can be taken into account during the reconstruction process, which significantly increases the reliability of the significance of detected structures. According to one aspect of the invention, calculating the contour significance data includes calculating local contour information from the initial image data. The term 'local' indicates that the contour information is considered individually for each image area or part. An image area can be, for example, an image element, i.e., a pixel or voxel, but also an area in the image formed by several, for example, adjacent or at least contiguous image elements. According to another aspect, the local contour information includes the contour amplitude and / or the contour direction. This contour information can be obtained, for example, by applying a simple edge detector to the initial image data. The detection filter observes and identifies intensity changes in the initial image data and generates the contour information from this information. Image elements or areas with significant intensity changes are recognized as contours in the initial image data and, for example, reassigned an intensity value corresponding to the intensity change. For instance, the edge detector can be a convolution of the initial image data with a filter matrix. The edge detector could be an operator from the class "identity minus low-pass filter," specifically a Scharr or Sobel operator. The contour amplitude does not directly correspond to the filtering but represents a scalar measure of the magnitude of the direction-dependent contributions.The contour direction results, for example, from the directional components of the edge detector, which each describe the change in a specific spatial direction, for example Kx(x,y,z), Ky(x,y,z), Kz(x,y,z) as a change along the Cartesian axes, or more compactly written as a vector. The contour amplitude results as the magnitude of this vector, i.e., where the spatial dependence has been omitted for the sake of clarity. According to another aspect, calculating contour significance data includes calculating local statistical information within the initial image data. The term "local" refers to individual image areas or parts, or more generally, their location in space, with respect to this statistical information. Here, an image area can be, for example, an image element such as a pixel or voxel, but also an area within the image comprised of several, for example, adjacent or at least connected, image elements. Additionally, in this context, "local" also denotes the directional dependence of the statistical information. This allows for the anisotropy of the noise in the initial image data to be taken into account. By determining this statistical information, statistical properties are already considered during the reconstruction process. According to another aspect, the local statistical information is the local standard deviation of the noise. To determine this, image-based approaches are available, such as those described in patent applications DE 10 2004 008 979 A1 or DE 10 2005 038 940 A1, the disclosures of which are expressly incorporated in full into the present application. Alternatively, the variance of the measurement data can be analyzed and the local standard deviation derived from it. For this purpose, the method described, for example, in Proc. SPIE 6510, Medical Imaging 2007: Physics of Medical Imaging, 651023 (March 14, 2007; doi: 10.1117 / 12.713692) is available, which is also expressly and fully incorporated into the present application. In particular, the evaluation of the measurement data to determine the local standard deviation of the noise makes it possible to take the directional dependence of the noise into account.As an alternative to the local standard deviation, the variance of the noise can also be used as statistical information or quantities derived from it, such as percentiles, within the scope of the invention. According to another aspect, calculating contour significance data includes determining a contour-to-noise ratio. This is derived, for example, as the quotient of the local contour information and the local statistical information. The contour-to-noise ratio corresponds to a signal-to-noise ratio and relates the local contour information to the local statistical information. In other words, the contour-to-noise ratio normalizes the edge signal in the image data to the noise. According to a further aspect of the invention, calculating the contour significance data comprises mapping the contour-to-noise ratio to a value between 0 and 1 using a restricted significance function. By means of this step, the method according to the invention discriminates between contours in the image data that are significant or insignificant compared to the noise by mapping the contour-to-noise ratio to a significance level between 0 and 1. Significant contours are visible in the background noise, whereas insignificant contours do not stand out from the noise signal. The significance function f(t) is restricted in the sense that According to another aspect, the significance function f(t) is either a continuous function or a step function. The significance function f(t) in the form of a step function definitively determines the relevance of a contour by means of its threshold. According to another aspect of the invention, the significance function f(t) takes the following form: where the parameter c determines the transition between significant and insignificant. Any other functions are also conceivable. Another aspect of deriving contour data involves high-pass filtering of the initial image data. The high-pass filtered image data represents a differential signal corresponding to the direction or orientation of an edge or contour. In other words, wherever an edge is present in the initial image data, the high-pass filtering highlights this edge in the contour data. A characteristic feature of the contour data is the so-called "light-dark fringe" along the contour's path. According to a further aspect of the invention, the high-pass filter has a frequency response whose transfer characteristic vanishes at spatial frequency 0 and assumes values ​​greater than 1 as the spatial frequency increases. In other words, low spatial frequencies are eliminated or suppressed by the high-pass filter, while higher spatial frequencies, and thus edge or contour information in the initial image data, are amplified. The high-pass filter therefore ensures that contours in the contour data are emphasized. According to a further aspect of the invention, the resulting image data is generated based on the product of the contour data and the contour significance data. This procedure corresponds to the calculation of a contour image weighted by the contour significance data. The multiplication is performed, for example, for each image element or image area. According to an alternative approach, the resulting image data is generated based on the sum of the initial image data and the product of the contour data and the contour significance data. According to this approach, the resulting image data corresponds to a contour image that exhibits contours emphasized depending on their local significance. In contrast to the contour image presented here, the previous contour image weighted according to significance is only of limited use for diagnostic purposes due to its significant deviation of image properties from the initial image data. However, with regard to the subsequent segmentation algorithm, the resulting image data from both alternatives provide comparable improvements, as both deliver the same information when analyzing differential signals, for example, from simple edge detectors. Segmentation during post-processing can now be simplified.This is possible because information regarding the significance of structures is already contained in the resulting image data in the form of selective edge enhancement. Edge detection can then be easily performed using a linear edge detector and a corresponding threshold based on the resulting image data. According to a further aspect, the method according to the invention is carried out iteratively. In each iteration step, the contour data weighted with significance, or a fraction thereof corresponding to the number of iterations, is added to the result image data of the previous iteration step as a correction term. The iteration can be terminated, for example, after a predetermined number of iteration steps, or alternatively, when a predetermined termination criterion is met, for example, when the deviation between the result image data of the last and penultimate iterations falls below a predetermined value. The initial image data can be incorporated into each iteration step, for example, to determine deviations of the current result from it. The invention further relates to a computer program with program code for carrying out all process steps according to the method according to the invention when the program is executed on a computer. This makes the method reproducible and less prone to errors on different computers. The invention also relates to a machine-readable data carrier on which the computer program described above is stored. The invention further relates to a control and computing unit for segmenting result image data of an object under investigation from measurement data acquired during a relative rotational movement between a radiation source of an X-ray imaging system and the object under investigation, comprising a program memory for storing program code, wherein the program memory contains program code that carries out a method according to the method according to the invention. The invention also relates to a control and computing unit for segmenting result image data of an object under investigation from measurement data acquired during a relative rotational movement between a radiation source of an X-ray imaging system and the object under investigation, comprising: - a reconstruction unit configured to reconstruct initial image data from the measurement data, - a derivation unit configured to derive contour data from the initial image data, - a first computing unit configured to calculate contour significance data from the measurement data and / or the initial image data, - a second computing unit configured to calculate the result image data using the contour data and the contour significance data, and - a segmentation unit configured to segment the result image data. Ultimately, the invention relates to an X-ray imaging system with a control and computing unit according to the invention. The invention will now be described and explained in more detail with reference to the embodiments shown in the figures. Figure 1 shows an X-ray imaging device according to an embodiment of the invention, Figure 2 shows the method according to the invention as a block diagram according to an embodiment of the invention, Figure 3 shows a flowchart of the method according to the invention according to an embodiment of the invention, Figure 4 shows an exemplary frequency response of a high-pass filter used in an embodiment of the invention, and Figure 5 shows an exemplary comparison of image data reconstructed according to the prior art and within the framework of the method according to the invention. Fig. 1 shows an X-ray imaging device using the example of an X-ray computed tomography (CT) scanner. The CT scanner shown here has an imaging unit 17 comprising a radiation source 8 in the form of an X-ray source and a radiation detector 9 in the form of an X-ray detector. The imaging unit 17 rotates about a system axis 5 during the acquisition of X-ray projections, and the X-ray source emits radiation 2 in the form of X-rays during the acquisition. The X-ray source is an X-ray tube. The X-ray detector is a line detector with multiple lines. The subject 3, here a patient, lies on a patient table 6 during the acquisition of projections. The patient table 6 is connected to a table base 4 in such a way that it supports the patient table 6 with the patient 3. The patient table 6 is designed to move the patient 3 along an acquisition direction through the opening 10 of the acquisition unit 17. The acquisition direction is generally defined by the system axis 5, around which the acquisition unit 17 rotates during the acquisition of X-ray projections. In a spiral acquisition, the patient table 6 is continuously moved through the opening 10 while the acquisition unit 17 rotates around the patient 3 and acquires projection data. Thus, the X-rays describe a spiral on the surface of the patient 3. The X-ray imaging device includes a computer system 12, which is connected to a display unit 11, for example, for the graphical display of reconstructed X-ray images, such as a reconstructed image data set, and an input unit 7. The display unit 11 can be, for example, an LCD, plasma, or OLED screen. It can also be a touch-sensitive screen, which also serves as the input unit 7. Such a touch-sensitive screen can be integrated into the imaging device or be part of a mobile device. The input unit 7 can be, for example, a keyboard, a mouse, a touchscreen, or a microphone for voice input. The input unit 7 can also be configured to detect user movements and translate them into corresponding commands. The computer system 12 communicates with the rotatable imaging unit 17 for data exchange. Control signals for X-ray image acquisition are transmitted from the computer system 12 to the imaging unit 17 via an interface unit 21 and the connection 14. Various scan protocols, each tailored to a specific examination type, can be stored in a memory 24 and selected by the user before the projection data acquisition. The imaging unit 17 is controlled according to the selected scan protocol. Simultaneously, the interface unit 21 captures acquired measurement data (MD) in the form of projection data for further processing in a control and computing unit 16, described in more detail below, or in the corresponding individual components. The connection 14 is implemented in a known manner, either wired or wirelessly. The computer system 12 includes a reconstruction unit 23, which is configured to reconstruct initial image data IBD in the form of an image data set from the measurement data MD using known reconstruction methods. For this purpose, one or more reconstruction algorithms can be stored in a memory 24 of the computer system 12 for retrieval or selection by the reconstruction unit 23. The computer system 12 includes a derivation unit 22, which processes the initial image data IBD in such a way as to generate contour data KD from it. For this purpose, the derivation unit 22 applies, for example, a suitable high-pass filter H to the initial image data IBD, as will be explained in more detail below. The derivation unit 22 is in data communication with the reconstruction unit 23 to provide the initial image data IBD. The computer system 12 also includes a first processing unit 20, which is configured to calculate contour significance data (CSD) from the initial image data (IBD) and / or the measurement data (MD). For this purpose, the first processing unit 20 is connected to the interface unit 21 for receiving the measurement data (MD) and to the reconstruction unit 23 for receiving the initial image data (IBD). The first processing unit 20 is configured to determine, firstly, local contour information (LKI), for example, in the form of contour amplitude and / or contour direction, and secondly, local statistical information (LSI), for example, in the form of the local standard deviation of the noise. To do this, the first processing unit 20 applies the necessary data or image processing steps to either the initial image data (IBD) or the measurement data (MD), as described in more detail below.Furthermore, the first processing unit 20 is also configured to calculate a contour-to-noise ratio KSNR from the calculated local contour information LKI and the calculated local statistical information LSI, and to map this ratio to a significance value between 0 and 1 using a significance function f(t). For this purpose, the first processing unit 20 is connected to memory 24, in which various possible significance functions f(t) are stored for recall by the first processing unit 20. Depending on the application, it is thus possible to adapt the significance function f(t). Furthermore, computer system 12 also includes a second processing unit 18, which is configured to calculate result image data EBD from the contour data KD and the contour significance data KSD. These result image data are characterized by contours that are emphasized according to their local contour significance. The second and first processing units 18 and 20 are also connected via data links. In this case, the reconstruction unit 23, the derivation unit 22, and the first and second computation units 20 and 18 are designed as separate modules within the computer system 12, which exchange data with each other where necessary. Alternatively, all the aforementioned units can also be integrated into the control and computation unit 16, either in the form of a physical or functional integrity. The computer system 12 can interact with a computer-readable data carrier 13, in particular to carry out a method according to the invention by means of a computer program with program code. Furthermore, the computer program can be stored on the machine-readable carrier in a retrievable manner. In particular, the machine-readable carrier can be a CD, DVD, Blu-ray Disc, a memory stick, or a hard drive. The reconstruction unit 23, the derivation unit 22, the first and second processing units 20 and 18, and the control and arithmetic unit 16 can be designed in the form of hardware or in the form of software. For example, the control and arithmetic unit 16 is designed as a so-called FPGA (acronym for "Field Programmable Gate Array") or comprises an arithmetic logic unit. In the example shown here, at least one computer program is stored in memory 24 of computer system 12, which performs all the process steps of the method according to the invention when the computer program is executed on the computer. The computer program for executing the process steps of the method according to the invention comprises program code. Furthermore, the computer program can be designed as an executable file and / or be stored on a different computing system than computer system 12. For example, the X-ray imaging device can be designed such that computer system 12 loads the computer program for executing the method according to the invention into its internal working memory via an intranet or the Internet. Fig. 2 shows the method according to the invention in a possible embodiment in a block diagram. Measurement data MD is available in the form of a multitude of projections acquired from different directions. These are converted into initial image data IBD in the form of an image dataset using a known reconstruction method, which in itself does not result in edge enhancement, for example, filtered back projection (FBP). The measurement data MD and the initial image data IBD contain information regarding the resolution and statistics of the dataset, which is not taken into account in a conventional reconstruction and is therefore no longer available for image post-processing. The method according to the invention takes this information into account. In particular, the method considers direction-dependent effects, i.e., directional noise.This information is extracted using the following steps of the procedure and made available for post-processing. A linear edge detector is applied to the initial image data set (IBD) to obtain local contour information (LKI). Local statistical information (LSI) is additionally obtained from the initial image data set (IBD) or from the measurement data (MD). In this embodiment, this is the local standard deviation of the noise in the data set. The local contour information (LKI) is then normalized to the local statistical information (LSI), resulting in a contour-to-noise ratio (KSNR), which is used to determine the local contour significance data (KSD). The contour significance data (KSD) is a measure of whether a contour or edge in the initial image data (IBD) stands out significantly above the noise in the image data or disappears within it.The initial image data (IBD) is also used to determine contour data (KD) in the form of a contour dataset. This is done by applying a high-pass filter (H) to the initial image data (IBD). The high-pass filtering generates a differential edge signal. In contrast, the local contour information (LKI) represents a scalar quantity. Finally, the contour data (KD) is weighted with the contour significance data (KSD). This generates result image data (EBD) that represent contour profiles amplified according to their significance. The result image data (EBD) is segmented, making it particularly suitable for simple, linear edge detection. Fig. 3 shows the inventive method in an exemplary embodiment as a flowchart. The first step S1 reconstructs a neutral three-dimensional image dataset I(x, y, z) from a plurality of projections of an image data acquisition, where x, y, z denote the coordinates of each voxel in the image data. The image dataset is neutral in the sense that it has not undergone any edge steepening as a result of the reconstruction itself. As already mentioned, a filtered backprojection can be applied; other reconstruction algorithms are also possible. In step S2, the local edge amplitude K(x,y,z) and, additionally, the local direction of the edge are calculated using a simple, linear edge detector, for example, a Sobel operator. The direction of the edge results, for example, from the directional selectivity of the filter used.The Sobel operator would employ a filter mask for horizontal, vertical, and / or diagonal edges. The edge direction is defined for each voxel entirely by its normal vector. The edge detector acts as a filter applied to a voxel and its neighboring voxels in spatial space via a convolution operation. The size of the area surrounding the voxel included in the filtering is determined by the filter itself and can therefore change. Alternatively, edge detection can be performed in frequency space using a Fourier transform of the entire image volume, multiplication by the frequency representation of the filter, and an inverse Fourier transform. The procedure generally depends on the size of the image volume and the filter design. In step S3 of the procedure, the local expected standard deviation of the noise is determined.This can be achieved either by using the image data I(x, y, z), whereby, for example, as in DE 10 2004 008 979 A1, one-dimensional variances for each voxel in a specific spatial region or radius R are determined for numerous, preferably isotropically distributed, spatial directions, starting from the voxel under consideration, from which the respective standard deviation can be derived in a known manner by taking the square root. Alternatively, the local, direction-dependent standard deviation is determined using the projections. Alternatively, only coarse directions are determined, where (x, y, z) describe a coarser grid than that of the image data. The determination of the values ​​necessary for calculating the significance would then be carried out by interpolation, for example multi-linearly, of the pre-calculated coarse support points. This approach proves to be considerably faster and more memory-efficient.In a further step S4, the local contour significance S(x,y,z) is calculated from the local contour amplitude K(x,y,z) and the direction-dependent local standard deviation, where the significance function f(t) takes the form f(t) = 1 - exp (-(t / c)²). Alternative significance functions with a range of values ​​between 0 and 1 are also possible. The contour significance S(x,y,z) considers the quotient of the local contour amplitude K(x,y,z) and the direction-dependent local standard deviation. This corresponds to a normalization step. Using a high-pass filter H, such as the one shown in Fig. 4, the image data set I(x, y, z) is converted into a contour data set Δ(x, y, z) in a further step S5 according to the formula. The high-pass filtering amplifies higher-frequency signals, which is represented by the slope of the curve in Fig. 4 above the value 1. For low-frequency signals, the filter's transfer function approaches 0, meaning that low-frequency image components are either not represented or are only suppressed in the contour data set. Other high-pass filters with similar properties can be used in the same way. In step S6, the contour data set is multiplied by the local contour significance. This product corresponds to the resulting image data set I'(x, y, z) according to Step S6 corresponds to significance weighting of the contour data. This step ensures that only the high-frequency signal components in the image data that are significant and therefore identified as contours are amplified, while insignificant high-frequency signal components, such as those caused by noise, are not. This distinguishes the method from simple linear edge amplification, which amplifies all high-frequency components, including noise, regardless of their origin. According to an alternative variant, the product of the contour data set and local contour significance described above is added to the neutral image data set I'(x, y, z) in the form of a correction term. The resulting image is then obtained according to According to another variant, the product of contour data set and local contour significance described above is used as an addition to the update term within an iterative reconstruction algorithm. The resulting image is then obtained according to where Ij+1 and Ij denote the image data sets of the j+1-th and j-th iterations, respectively. α describes a constant factor that indicates the strength of the correction by the update term [Λ(Ij - I0) + βR(Ij) + γSΔ]. In this variant, this factor is composed as follows: Λ is a suitable factor that takes into account the statistics of the image data. R corresponds to a regularization applied to the image Ij of the j-th iteration to suppress the iteration-induced increase in noise. β represents a constant factor that determines the strength of the regularization. γ represents a constant factor that determines the concentration of the additive according to the invention. The resulting image is then subjected to segmentation, using a simple, linear edge detector. The sequence of steps in the method according to the invention is arbitrary, unless dictated by the invention itself, and can be changed without departing from the scope of the invention. Individual steps or aspects of the embodiments of the invention are, of course, interchangeable where appropriate and in line with the invention. Figure 5 shows two reconstructed image datasets, A and B, as examples. For simplicity, the representation is two-dimensional. Image A corresponds, for example, to an input image dataset I(x, y, z) which was reconstructed using conventional methods. Image B corresponds to a result image dataset I'(x, y, z) calculated using the method according to the invention. Image B exhibits significantly enhanced contrast at the interfaces of different tissue types compared to Image A. A characteristic feature is the light-dark fringe resulting from the high-pass filtered contour data Δ(x, y, z). It is also evident that the image quality in Image B is maintained compared to Image A, since the method according to the invention discriminates between the contour signal and the noise signal during amplification.

Claims

Method for segmenting result image data (EBD) of an object under investigation (3) from measurement data (MD) acquired during a relative rotational movement between a radiation source (8) of an X-ray imaging system and the object under investigation (3), comprising the following steps: - Reconstructing (S1) initial image data (IBD) from the measurement data (MD), - Deriving (S5) contour data (KD) from the initial image data (IBD), - Calculating (S4) contour significance data (KSD) from the measurement data (MD) and / or the initial image data (IBD), - Calculating (S6) the result image data (EBD) using the contour data (KD) and the contour significance data (KSD), and - Segmenting the result image data (EBD). Method according to claim 1, wherein the calculation of the contour significance data (CSD) comprises a calculation (S2) of local contour information (LCI) from the initial image data (IBD). Method according to claim 2, wherein the local contour information (LKI) comprises the contour amplitude K(x ,y, z) and / or the contour direction φk(x, y, z). Method according to one of the preceding claims, wherein the calculation of the contour significance data (CSD) comprises a calculation (S3) of local statistical information (LSI) in the initial image data (IBD). Method according to claim 4, wherein the local statistical information (LSI) is the local standard deviation σ(φ; x, y, z) of the noise. Method according to one of the preceding claims, wherein the calculation of the contour significance data (CSD) comprises determining a contour-to-noise ratio (CNR). The method according to claim 6, wherein the calculation of the contour significance data (CSD) comprises mapping the contour-to-noise ratio (CNR) to a value between 0 and 1 using a restricted significance function f(t). Method according to claim 7, wherein the significance function f(t) is designed as a continuous function or as a step function. Method according to claim 7 or 8, wherein the significance function f(t) takes the following form: f(t) = 1 - exp(-(t / c)2). Method according to one of the preceding claims, wherein the derivation of the contour data (KD) comprises a high-pass filtering H of the initial image data (IBD). Method according to claim 10, wherein the high-pass filter H has a frequency response whose transmission vanishes at the spatial frequency 0 and assumes values ​​greater than 1 as the spatial frequency increases. Method according to one of the preceding claims, wherein the result image data (EBD) are formed as the product of the contour data (KD) and the contour significance data (KSD). Method according to any one of claims 1 to 11, wherein the result image data (EBD) is formed as the sum of the initial image data (IBD) and the product of the contour data (KD) and the contour significance data (KSD). Method according to one of the preceding claims, wherein the method is carried out iteratively. Computer program with program code for carrying out all process steps according to any one of claims 1 to 14, when the program is executed in a computer. Machine-readable data carrier (13) on which the computer program according to claim 15 is stored. Control and computing unit (16) for segmenting result image data (EBD) of an object of investigation (3) from measurement data (MD) acquired during a relative rotational movement between a radiation source (8) of an X-ray imaging system and the object of investigation (3), comprising a program memory for storing program code, wherein the program memory contains program code that performs a method according to one of claims 1 to 14. Control and computing unit (16) for segmenting result image data (EBD) of an object under investigation (3) from measurement data (MD) acquired during a relative rotational movement between a radiation source (8) of an X-ray imaging system and the object under investigation (3), comprising: - a reconstruction unit (23) configured to reconstruct initial image data (IBD) from the measurement data (MD) (S1), - a derivation unit (22) configured to derive contour data (KD) from the initial image data (IBD) (S5), - a first computing unit (20) configured to calculate contour significance data (KSD) from the measurement data (MD) and / or the initial image data (IBD) (S4), - a second computing unit (18) configured to calculate the result image data (EBD) using the contour data (KD) and the contour significance data (KSD) (S6), and - a Segmentation unit that is set up to segment the result image data (EBD). X-ray imaging system with a control and computing unit (16) according to claim 17 or 18 .

Images