Apparatus, system and method for generating medical image of region of interest of subject indicative of contrast enhanced region

Abstract

An apparatus for generating a medical image of a region of interest (ROI) of a subject indicative of a contrast enhanced region is presented. The device obtains a spectral image dataset of the ROI, the spectral image dataset comprising two or more spectral images; and segmenting an anatomical sub-ROI of the ROI in a virtual single energy image and / or a full energy image of the two or more spectral images. The device determines distribution information of at least one of the anatomical sub-ROIs in at least one contrast-sensitive image of the two or more spectral images, and determines distribution information of at least one of the anatomical sub-ROIs in at least one contrast-sensitive image of the two or more spectral images. And determining a contrast enhancement region by determining a threshold value based on the distribution information of the at least one anatomical sub-ROI and determining pixels of the at least one anatomical sub-ROI having a value higher than the threshold value. The device outputs a medical image in which the contrast enhancement region is indicated.

Description

Technical Field

[0001] The present invention relates to apparatus, systems and methods for generating medical images of a region of interest (ROI) of an object (e.g., a patient) that indicate a region of contrast enhancement. Background Technology

[0002] Delayed myocardial enhancement is a valuable imaging technique for diagnosing a variety of diseases, including myocardial viability, cardiomyopathy, myocarditis, and other infiltrative myocardial processes. While late enhancement has primarily been evaluated on MR, its feasibility on CT has recently been demonstrated, offering significantly wider availability and much lower cost than MR.

[0003] More generally, contrast-enhanced imaging (preferably early enhancement in the early stage of contrast agent application and late enhancement in the later stage of contrast agent application) is of interest for different medical applications and the diagnosis of different organs (e.g., myocardial disease, heart disease, liver perfusion analysis and stroke analysis). Summary of the Invention

[0004] The purpose of this invention is to provide an apparatus, system, and method.

[0005] In a first aspect of the invention, an apparatus is provided for generating a medical image of a region of interest (ROI) of an object, indicating a region of contrast enhancement, the apparatus comprising: An input unit is configured to obtain a spectral image dataset of the ROI of the object, the spectral image dataset comprising two or more spectral images; The processing unit is configured as follows: Segmenting the ROI into multiple anatomical sub-ROIs in virtual single-energy images and / or full-energy images of the two or more spectral images; Determine the distribution information of at least one anatomical ROI in at least one contrast-sensitive image from the two or more spectral images, which has higher contrast sensitivity and / or quantitative accuracy than the virtual single-energy image used for segmentation, the distribution information indicating the tonal distribution of the pixels of the corresponding anatomical ROI; and The contrast enhancement region of the at least one anatomical ROI in the at least one contrast-sensitive image is determined by determining a threshold based on the distribution information of the at least one anatomical ROI and identifying pixels of the at least one anatomical ROI with pixel values ​​higher than the determined threshold; and An output unit is configured to output a medical image of the ROI generated based on the spectral image dataset, indicating the determined contrast-enhanced region in the medical image.

[0006] In another aspect of the invention, an imaging system is provided, comprising: An imaging device configured to acquire a spectral image dataset of a region of interest (ROI) of an object, the spectral image dataset comprising two or more spectral images; The apparatus according to any one of the preceding claims, the apparatus being configured to generate a medical image of a region of interest of an object, indicating a region of contrast enhancement, based on an acquired spectral image dataset; and The display unit is configured to display synthesized medical images.

[0007] In yet another aspect of the invention, a corresponding method, a computer program, and a non-transient computer-readable recording medium are provided, the computer program including program code units configured to cause the computer to perform the steps of the method disclosed herein when the computer program is executed on a computer, the non-transient computer-readable recording medium storing a computer program product therein, the computer program product causing the execution of the method disclosed herein when run by a processor.

[0008] Preferred embodiments of the invention are defined by the dependent claims. It should be understood that the claimed methods, systems, computer programs, and media have preferred embodiments similar to and / or identical to the claimed systems, particularly those as defined in the dependent claims and as disclosed herein.

[0009] This invention is based on the concept of identifying anatomical sub-regions of interest (sub-ROIs, i.e., portions of ROIs) on a virtual single-energy image (also referred to as a virtual single-energy image, abbreviated as VMI; optionally, on a full-energy image as an alternative or supplementary option), segmenting contrast-enhanced regions (especially late-enhanced regions) on spectral image reconstruction; and deriving one or more case-specific thresholds on the spectral map with the highest contrast sensitivity (i.e., on at least one contrast-sensitive image with higher contrast sensitivity and / or quantitative accuracy than the virtual single-energy image used for segmentation). For example, myocardial regions and several calibration regions (e.g., anatomical sub-ROIs including the myocardium, left ventricle, right ventricle, aorta, etc.) can be segmented on one or more virtual single-energy images.

[0010] Determine distribution information (sometimes referred to as dynamic distribution) in one or more contrast-sensitive images. This distribution information involves intensity distribution (rather than spatial distribution) and indicates the tonal distribution of pixels in the corresponding anatomical ROI.

[0011] Then, one or more determined case-specific thresholds (e.g., one threshold for each anatomical sub-ROI) are used to determine pixels in the corresponding anatomical sub-ROI that have pixel values ​​above the threshold, so that in the final medical image of the ROI generated from the spectral image dataset, the contrast-enhanced region can be indicated and visualized so that the user (e.g., physician) can easily identify the contrast-enhanced region, thereby supporting the user in making a diagnosis.

[0012] The spectral image dataset of the object's ROI (e.g., a single spectral CT image) is used as input. This dataset typically includes two or more spectral images, such as one or more base images, a single E40 keV image (or other single-energy image), an iodine map, a Zeff image, etc. Spectral images are usually acquired simultaneously (when using a dual-layer detector or photon-counting detector) or almost simultaneously (when using a dual-source detector, where there is a very small time shift). It is possible to copy the ROI from one spectral image to other spectral images. Anatomical sub-ROIs should be segmented in at least one of the spectral images, but the spectral images may differ for different anatomical sub-ROIs. Segmentation of one anatomical sub-ROI may be sufficient, as the region inside and outside the sub-ROI can be used as two separate sub-ROIs.

[0013] Virtual single-energy images, or in conventional cases (full-energy) images, are preferably used to segment anatomically relevant regions (referred to herein as anatomical sub-ROIs), thereby creating a trade-off between anatomical structures and contrast agents. Such images can also generally be used for contrast agent uptake measurements, i.e., for determining distribution information of the tonal distribution of pixels indicating the corresponding anatomical sub-ROI. However, preferably, different spectral images, such as iodine density maps or electron density maps (which are not suitable for segmentation), are used to determine distribution information.

[0014] For segmentation, different methods or known algorithms can typically be applied, and the processing unit can be configured to segment the one or more anatomical ROIs by using one or more of the following: model-based segmentation methods, trained computer systems, and trained algorithms.

[0015] According to an embodiment, the processing unit is configured to erode and / or allow the user to erode multiple segmented anatomical sub-ROIs. This allows for manual or automatic correction of the sub-ROI contours, which further improves segmentation accuracy and ensures that only desired anatomical structures (e.g., desired organs (e.g., myocardium)) are visible in the segmented sub-ROIs.

[0016] According to another embodiment, the processing unit is configured to determine a histogram and / or one or more histogram derived values, specifically one or more of the median, mean, and mode, as distribution information for the anatomical ROI. Preferably, for each anatomical ROI, a histogram and one or more histogram derived values ​​are determined. In this context, the mode should be understood as the data value that appears most frequently in the dataset. For a symmetric histogram, the values ​​of the mean, median, and mode are all the same and all located at the center of the distribution, but typically they are not the same, and the median is likely the most robust value.

[0017] According to another embodiment, the processing unit is configured to determine the distribution information in at least one of an iodine density map, an electron density map, and an effective atomic number (Zeff) map. These images typically show higher contrast sensitivity and / or quantitative accuracy than single-energy images (e.g., single E40 keV or single E60 keV images or single-energy images for another energy), indicating a sufficient accuracy to represent the trade-off between contrast agent and anatomical structure. Virtual single-energy images or, under normal circumstances (full-energy) images, may be more advantageous for segmenting anatomically relevant parts and are therefore used for segmentation according to the invention.

[0018] There are several embodiments of how to use distribution information to determine one or more contrast-enhancing regions (ROIs). In one embodiment, the processing unit may be configured to: determine multiple thresholds for each anatomical ROI, relative to the median, mean, or mode of the pixel values ​​of the anatomical ROI, and use the multiple thresholds to determine the contour regions of the anatomical ROI. The output unit may then be configured to output the contour regions as contrast-enhancing regions.

[0019] Therefore, the processing unit can be configured to determine the threshold based on the median, mean, or mode of the anatomical ROI, specifically through a linear combination or nonlinear regression of the median, mean, or mode.

[0020] In another embodiment, the processing unit is configured to determine the threshold(s) by using a metric, particularly a histogram similarity metric. Preferably, a metric capable of comparing two histograms and calculating a scalar value describing the mutual similarity between the two histograms is used. Thus, the processing unit can be configured to: minimize the metric between the contrast-enhanced region and the remaining region of the anatomical ROI, and / or maximize the metric between the contrast-enhanced region and the blood pool region. Therefore, two different objectives can be optimized independently or simultaneously. For the first objective (minimizing the metric), the threshold (for each trial generated to the assumed trial separation in the two regions (e.g., foreground / background)) can be varied until the similarity between the contrast region and the non-contrast region (measured by the metric, particularly the histogram similarity metric) becomes minimal (optimally distinguishable). For the second objective (maximizing the metric), the threshold can be varied until the similarity between the assumed contrast region and a region definitively known to be filled with (pure) contrast agent (a so-called blood pool region, such as a cardiac ventricle or aorta) becomes highest.

[0021] In another embodiment, the output unit is configured to output the medical image of the ROI in the form of a slice view and / or a cylindrical projection view and / or a map view, and / or to indicate the determined contrast enhancement area by using different colors or grayscale values.

[0022] The spectral image dataset can typically be a spectral CT image dataset or a photon-counting CT image dataset, depending on the type of detector used to acquire the spectral image dataset. Additionally, the spectral image dataset may include two or more of the following: a spectral base image, a single E40 keV image, an iodine map image, and a Zeff image.

[0023] In addition to the aforementioned devices, the proposed imaging system also includes imaging equipment (e.g., a spectral CT scanner) and a display unit (e.g., a computer or workstation screen). Furthermore, a user interface (e.g., a touchscreen, a pointing device, a computer mouse, etc.) may be provided, configured to allow users to provide user input to interactively label the anatomical ROI in at least one of two or more spectral images from the spectral image dataset. Attached Figure Description

[0024] These and other aspects of the invention will become apparent and elucidated from and with reference to one or more embodiments described below. In the following figures: Figure 1 A schematic diagram of an embodiment of the system according to the present invention is shown.

[0025] Figure 2 A schematic diagram of an embodiment of the device according to the present invention is shown.

[0026] Figure 3 A schematic diagram of an embodiment of the method according to the present invention is shown.

[0027] Figure 4 An exemplary cardiac CT scan with semantic segmentation of anatomical sub-ROIs is shown.

[0028] Figure 5 An exemplary visualization of a case-specific late-enhancing region relative to the myocardium is shown. Detailed Implementation

[0029] Figure 1 A schematic diagram of an embodiment of a system 1 for generating a medical image of a region of interest (ROI) of an object, indicating a region of contrast enhancement, according to the present invention, is shown. System 1 includes an imaging device 2 for acquiring a spectral image dataset of the ROI of the object, a device 3 for generating a medical image of the ROI of the object, indicating a region of contrast enhancement, based on the acquired spectral image dataset, and a display unit 4 for displaying the synthesized medical image.

[0030] Imaging device 2 is a CT device, particularly a spectral CT or photon-counting CT device, such as a conventionally used CT device, used to acquire image data of an object, such as image data of a specific body region of a patient. The imaging device acquires a spectral image dataset comprising two or more spectral images. The spectral image dataset may include one or more base images, single E40 keV images (or other single-energy images or other energies (e.g., 60 keV, 80 keV), etc.), full-energy images, iodine maps, Zeff images, etc., which may be acquired simultaneously or subsequently (almost simultaneously), depending on the implementation of the imaging device (particularly its detectors and radiation sources).

[0031] Device 3 may be implemented as a processor or computer, or implemented within a processor or computer, for example, in software and / or hardware. For example, a programmed processor may be included in device 3, which can run computer programs that can be stored in memory accessible by the processor.

[0032] The display unit 4 may be, for example, a monitor or screen of a PC, workstation, tablet computer, etc. (e.g., a computer that displays device 3 and performs a method according to the invention for generating a medical image of an object's ROI indicating a region of contrast enhancement).

[0033] Device 3 can obtain (i.e. retrieve or receive) image datasets (e.g., patient records stored in a hospital’s file management system) directly from imaging device 2 or from storage device 5 (e.g., cache or memory or image library) via, for example, a wired or wireless network.

[0034] A user interface 6 (e.g., touchscreen, pointing device, computer mouse, etc.) may be provided, enabling the user to provide user input to, for example, interactively label or correct the contour of the sub-ROI in at least one of two or more spectral images in a spectral image dataset.

[0035] Figure 2 A schematic diagram of an embodiment of a device 3 for generating a medical image of a region of interest (ROI) indicating a contrast-enhanced area according to the present invention is shown. The device may include circuitry performing the functions of the device, such as a processor, processing circuitry, a computer, dedicated hardware, etc. In another embodiment, such as... Figure 2 As shown, individual units or elements that together represent the circuit of device 3 can be used.

[0036] In this embodiment, device 3 includes an input unit 31 configured to input a spectral image dataset of the ROI of the input object, the spectral image dataset comprising two or more spectral images. The spectral image dataset can be obtained, for example, from imaging device 2 or storage device 5. For this purpose, input unit 31 can be directly or indirectly (e.g., via a network or bus) coupled to or connected to imaging device 2 or storage device 5. Therefore, the input unit can be, for example, a (wired or wireless) communication interface or data interface, such as a Bluetooth interface, Wi-Fi interface, LAN interface, HDMI interface, direct cable connection, or any other suitable interface that allows data transmission to device 2.

[0037] Device 3 further includes a processing unit 32, which is configured to perform the steps of the method according to the invention, as shown below. Figure 3 As explained, the processing unit 32 can be any kind of device configured to process the acquired spectral image dataset and generate a medical image indicating a contrast-enhanced region based on the spectral image dataset. The processing unit 32 can be implemented in software and / or hardware as, for example, a programmed processor, computer, laptop computer, PC, workstation, etc.

[0038] Device 3 also includes an output unit 33 configured to output a medical image of the ROI generated based on a spectral image dataset, indicating the identified contrast-enhanced region in the medical image. The output unit 33 can typically be any interface that provides the generated visualization (e.g., transmits the generated visualization to another device or provides the generated visualization for retrieval by another device, such as transmitting the generated visualization to another computer or directly transmitting the generated visualization to display 4 for display). Therefore, the output unit 33 can typically be any (wired or wireless) communication or data interface.

[0039] Figure 3 A flowchart illustrating a first embodiment of a method 100 for generating a medical image of a region of interest (ROI) indicating a contrast-enhancing area according to the present invention is shown. The steps of the method can be performed by device 3, wherein the main steps of the method are performed by processing unit 32. The method can be implemented, for example, as a computer program running on a computer or processor.

[0040] One of the goals of the concept achieved by the proposed method is to automatically segment and quantify contrast-enhanced regions using scan-specific thresholding, leveraging different spectral maps (scan reconstruction) from various building blocks (steps). In exemplary, non-limiting applications of the invention, which will be explained below, late-enhancing regions in the myocardium should be segmented and quantified to visualize regions with enhanced contrast agents or to reduce potentially deficient regions indicated by contrast agents (e.g., regions with circulatory disorders).

[0041] In the first step 101, a spectral image dataset of the ROI of the object is obtained, which includes two or more spectral images. The spectral image dataset can be obtained directly from an imaging device (e.g., a spectral or photon-counting CT scanner) or from a storage device (e.g., an image library).

[0042] In the second step 102, multiple sub-ROIs of the ROI are segmented in a virtual single-energy image (VMI) of two or more spectral images. Alternatively or additionally, a full-energy image can be used for segmentation.

[0043] In an exemplary application, cardiac segmentation of the spectral VMI is performed; that is, the cardiac region is automatically detected and segmented (mapped) into several anatomical ROIs, such as the myocardium, left ventricle, right ventricle, aorta, etc. This is in Figure 4 The middle image shows, Figure 4A cardiac CT scan with semantic segmentation of the following anatomical sub-ROIs is shown 10: myocardium 11, left ventricle 12, right ventricle 13, ascending aorta 14, descending aorta 15. Sub-ROIs can be defined to improve robustness if desired (e.g., by (manual or automatic) erosion of the segmented region).

[0044] Automatic segmentation can be based on, for example, model-based segmentation (MBS) or other machine learning-based (ML-based) semantic segmentation (e.g., deep CNNs). The segmentation (e.g., the segmentation algorithm) should (solely or at least primarily) rely on a contrast-sensitive image quality (VMI) with high contrast sensitivity (e.g., VMI 40 keV) because most of the contrast agent has been washed away in post-enhancement images. However, the VMI should still have a sufficiently conventional appearance to make standard segmentation models applicable.

[0045] In the third step 103, distribution information of at least one sub-ROI among two or more contrast-sensitive images is determined from at least one of the two or more spectral images, which has higher contrast sensitivity and / or quantitative accuracy than the VMI used for segmentation. The distribution information indicates the tonal distribution of the pixels of the corresponding sub-ROI. Typically, the VMI (or full-energy image) used for segmentation can be used in this step, but for this contrast agent uptake measurement, a different spectral image is preferably used.

[0046] In an exemplary application, region histogram analysis can be performed on contrast agent-sensitive (CA-sensitive) images. Histograms and histogram-derived values ​​(e.g., mean, median, or mode) are calculated for all anatomical sub-ROIs (particularly the myocardium, left and right ventricular blood pools, aorta, etc.). This should be performed on one or more spectral maps (e.g., iodine density maps, electron density maps, effective atomic number (Zeff) maps) that offer the highest contrast sensitivity and / or quantitative accuracy.

[0047] In step 104, a contrast-enhanced region of at least one anatomical sub-ROI is determined by determining a threshold based on the distribution information of the determined at least one anatomical sub-ROI and identifying pixels of the at least one anatomical sub-ROI with pixel values ​​higher than the determined threshold. Different options are available for implementing this step to identify contrast-enhanced regions.

[0048] Depending on the first option, a pseudo-color map or contour region is used, which can be scaled using the median, mean, or mode. For example, in an exemplary application, within the myocardial region, the contour region is calculated on the spectral volume using a series of thresholds relative to the median (or mean or mode) value of the anatomical sub-ROI (e.g., using the median (or mean or mode) value of the myocardium as the zero level and the median (or mean or mode) value of the blood pool region as the maximum level). The resulting contour region is provided to the user (e.g., medical practitioners, physicians, etc.) as a contrast enhancement candidate region (e.g., as a post-contrast enhancement candidate region) for visual evaluation or interactive selection.

[0049] According to the second option, thresholding is applied using an optimized linear combination of the median (or mean or mode). A single scan-specific threshold is calculated (on the spectral map) using the median (or mean or mode) value of the anatomical sub-ROI, according to a formula (e.g., linear combination or nonlinear regression). The parameters have preferably been optimized (relatively few) on a limited number of early cases with manually set ground truth values. In an exemplary application, regions in the myocardium exceeding this scan-specific threshold are labeled and quantified as contrast-enhanced regions, particularly late-enhanced regions.

[0050] According to the third option, an automatic scan-specific threshold (e.g., for myocardium) is applied. This automatic scan-specific threshold can be automatically derived by an algorithm that identifies a threshold that minimizes a histogram metric (e.g., mutual information, entropy, Kullback-Leibler divergence, covariance, distance, etc.) between contrast-enhanced regions (e.g., late-enhanced regions of myocardium) and the remaining regions of the anatomical sub-ROI (e.g., the remaining regions of myocardium), and simultaneously or alternatively maximizes a histogram metric between contrast-enhanced regions and blood pool regions. Regions in the anatomical sub-ROI (e.g., myocardium) exceeding this scan-specific threshold are labeled and quantified as contrast-enhanced regions, e.g., late-enhanced regions. In other words, different thresholds can be "tested," and a histogram is prepared for all pixels with pixel values ​​below the corresponding test threshold. The more diverse the histograms, the better the corresponding threshold. The optimal threshold is then finally used.

[0051] In step 5, 105, a medical image of the ROI generated from the spectral image dataset is output, indicating the identified contrast-enhanced regions. Various options exist for this visualization of the results. For example, the results of each of the above options in step 4, 104, can be represented in a standard slice-by-slice view (or a reformulated view, e.g., as a short-axis or long-axis view) and a standardized cylindrical projection view (“bull’s eye”). Furthermore, all results can be summarized (e.g., mean, median, or mode) based on myocardial fragment atlases (as defined by the American Heart Association (AHA)).

[0052] Figure 5 An exemplary visualization of a case-specific late-enhancing region relative to the myocardium (scanned 7 minutes after CA injection) is shown. This example visualization uses color scales to represent spectral iodine densities above and below the median of the myocardium, which are automatically segmented by MBS based on a virtual single-energy 40keV image, as shown. Figure 5 The outline of the center is shown. Figure 5 Specifically, the following representations are shown: standard HU values ​​in a slice-by-slice long axis view 20 (center), cylindrical projection 21 (bull's eye with AHA fragments; top left), 3D mesh view 22 (top right), and flattened projection 23 (bottom). Each of these views 20 to 23, or a combination of two or more of these views, can represent a medical image of the object's ROI, indicating the contrast-enhanced region generated according to the invention.

[0053] According to an embodiment, for all the above options, a previous three-dimensional smoothing (e.g., Gaussian or edge-preserving) of the selected spectral map (not the sub-ROI) is performed, such that the resulting contrast-enhanced region exhibits a finer or coarser granularity depending on the smoothing filter width.

[0054] This invention provides one or more of the following advantages.

[0055] Conventional (non-spectral) CT imaging is the established standard, but it is less sensitive to contrast agents compared to dual-energy, dual-slice, or photon-count CT. For example, the anatomical contours of the myocardium (especially the left ventricle) are not well visible on conventional CT because most of the contrast agent is washed away during the later enhancement phase. This anatomical contour can be visualized using the present invention.

[0056] Contrast enhancement is optimally achieved on spectral maps (e.g., iodine density, electron density, effective atomic number, etc.). However, standard semantic segmentation models (e.g., MBS, deep CNNs) may be unsuitable for these image types (or require costly retraining). This invention addresses this issue by using single-energy images for initial segmentation. Furthermore, as the analysis algorithm, no ML training data (along with its annotation, sampling, imaging protocol coverage, and regulatory efforts) is required. Moreover, standard segmentation models developed on conventional CT data can be used without retraining the semantic segmentation model on spectral CT images.

[0057] Experimental evidence suggests that using a fixed threshold for significant enhancement is unsatisfactory because contrast agent concentrations generally depend on a number of imaging protocol details (contrast agent injection, volume, and timing) and patient-specific parameters (blood volume, circulation, weight, etc.). According to the present invention, an optimal case-specific threshold is determined for each scan.

[0058] For medical users, aggregating quantitative information from multiple three-dimensional spectral image volumes through unassisted, purely visual assessment is cumbersome, time-consuming, and error-prone. This invention overcomes this problem by automatically processing spectral images from a spectral image dataset.

[0059] This invention makes full use of multiple spectra from a single scan. In purely visual evaluation, understanding these multiple spectra side by side is tedious. Furthermore, some spectra are more useful for segmentation, while others are more useful for thresholding, zooming, and viewing.

[0060] The “enhancement” of contrast agents is a familiar radiological concept from conventional (non-spectral) imaging, which makes the thresholding and linear scaling of cardiac compartments relative to functionally meaningful ones (compared to “black box” AI decision-making algorithms) trustworthy and understood by medical users.

[0061] Cardiac spectral CT differs from other cardiac applications that use multiple images (such as MRI or PET-CT) because all images are acquired at exactly the same time and no additional registration step is required to compensate for patient motion. While the invention has been described above in relation to cardiac diseases, it can also be applied to other organs, such as liver perfusion analysis or stroke analysis.

[0062] Although optimal spectral images are used for processing, users may also have the option to switch between different spectral images / representations. Additionally, users can switch between different granularities / smoothnesses for contrast agent-sensitive spectra before thresholding / scaling (selection of spatial filter width).

[0063] In summary, embodiments of the present invention provide a combination of various spectral reconstructed image volumes (VMIs for semantic segmentation and quantitative spectra with values ​​thresholded relative to median concentration values ​​in the myocardial and blood pool regions) that result in a visual presentation of selectable conventional or spectral images.

[0064] This invention can be applied to realize or facilitate the diagnosis of various medical applications and different organs, such as myocardial diseases, heart disease, liver perfusion analysis, and stroke analysis.

[0065] While the invention has been described and illustrated in detail in the accompanying drawings and the foregoing description, such description is to be considered illustrative or exemplary rather than restrictive; the invention is not limited to the disclosed embodiments. Other variations of the disclosed embodiments will be understood and implemented by those skilled in the art in practicing the claimed invention upon studying the drawings, the disclosure, and the appended claims.

[0066] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude multiple. A single element or other unit can perform the function of several items recited in the claims. The mere fact that certain measures are recited in different dependent claims does not indicate that combinations of these measures cannot be advantageously used.

[0067] Computer programs can be stored / distributed on suitable non-transient media (e.g., optical storage media or solid-state media supplied together with or as part of other hardware), but can also be distributed in other forms (e.g., via the Internet or other wired or wireless telecommunications systems).

[0068] No reference numerals in the claims should be construed as limiting the scope.

Claims

1. An apparatus (3) for generating a medical image of a region of interest (ROI) of an object, indicating a region of contrast enhancement, said apparatus comprising: The input unit (31) is configured to obtain a spectral image dataset of the ROI of the object, the spectral image dataset comprising two or more spectral images; Processing unit (32), which is configured as follows: Segmenting the ROI into multiple anatomical sub-ROIs in virtual single-energy images and / or full-energy images of the two or more spectral images; Determine the distribution information of at least one anatomical ROI in at least one contrast-sensitive image from the two or more spectral images, which has higher contrast sensitivity and / or quantitative accuracy than the virtual single-energy image used for segmentation, the distribution information indicating the tonal distribution of the pixels of the corresponding anatomical ROI; and The contrast enhancement region of the at least one anatomical ROI in the at least one contrast-sensitive image is determined by determining a threshold based on the distribution information of the at least one anatomical ROI and determining the pixels of the at least one anatomical ROI with pixel values ​​higher than the determined threshold. as well as The output unit (33) is configured to output a medical image of the ROI generated based on the spectral image dataset, indicating the determined contrast enhancement region in the medical image.

2. The device according to claim 1, in, The processing unit (32) is configured to segment the one or more anatomical ROIs by using one or more of the following: a model-based segmentation method, a trained computer system, or a trained algorithm.

3. The device according to any one of the preceding claims, in, The processing unit (32) is configured to erode and / or allow users to erode multiple segmented sub-ROIs.

4. The device according to any one of the preceding claims, in, The processing unit (32) is configured to determine histograms and / or one or more histogram derived values, particularly one or more of the median, mean and mode, as the distribution information of the sub-ROI.

5. The device according to any one of the preceding claims, in, The processing unit (32) is configured to determine the distribution information in at least one of the iodine density map, the electron density map, and the effective atomic number map.

6. The device according to any one of the preceding claims, in, The processing unit (32) is configured to: determine multiple thresholds for the median, mean, or mode of the pixel values ​​relative to the anatomical ROI for each sub-ROI, and use the multiple thresholds to determine the contour regions of the anatomical ROI, and The output unit (33) is configured to output the contour area as a contrast enhancement area.

7. The device according to any one of claims 1-5, in, The processing unit (32) is configured to determine the threshold based on the median, mean, or mode of the anatomical ROI, specifically through a linear combination or nonlinear regression of the median, mean, or mode.

8. The device according to any one of claims 1-5, in, The processing unit (32) is configured to determine the threshold by using a metric, particularly a histogram similarity metric.

9. The device according to claim 8, in, The processing unit (32) is configured to minimize the metric between the contrast-enhanced region and the remaining region of the anatomical ROI, and / or maximize the metric between the contrast-enhanced region and the blood pool region.

10. The device according to any one of the preceding claims, in, The output unit (33) is configured to output the medical image of the ROI in the form of a slice view and / or a cylindrical projection view and / or a map view, and / or to indicate the determined contrast enhancement area by using different colors or grayscale values.

11. The device according to any one of the preceding claims, in, The spectral image dataset is a spectral CT image dataset, a photon-counted CT image dataset, an MRI image dataset, or a PET-CT image dataset, and / or includes two or more of the following: spectral base images, single E40KeV images, iodine maps, and Zeff images.

12. An imaging system (1), comprising: An imaging device (2) is configured to acquire a spectral image dataset of the region of interest (ROI) of an object, the spectral image dataset comprising two or more spectral images; The device (3) according to any one of the preceding claims is used to generate a medical image of the region of interest (ROI) of an object, indicating the contrast-enhanced region, based on the acquired spectral image dataset; as well as Display unit (4) is configured to display synthesized medical images.

13. The imaging system according to claim 12, It also includes a user interface (6) configured to allow users to provide user input to interactively label the sub-ROI in at least one of two or more spectral images in the spectral image dataset.

14. A method for generating a medical image of a region of interest (ROI) of an object, indicating a region of contrast enhancement, the method comprising: Obtain the spectral image dataset of the ROI of the object, the spectral image dataset comprising two or more spectral images; Segmenting the ROI into multiple anatomical sub-ROIs in virtual single-energy images and / or full-energy images of the two or more spectral images; Determine the distribution information of at least one sub-ROI in at least one contrast-sensitive image from the two or more spectral images that has higher contrast sensitivity and / or quantitative accuracy than the virtual single-energy image used for the segmentation, the distribution information indicating the tonal distribution of the pixels of the corresponding sub-ROI; The contrast enhancement region of the at least one anatomical ROI in the at least one contrast-sensitive image is determined by determining a threshold based on the distribution information of the at least one anatomical ROI and determining the pixels of the at least one anatomical ROI with pixel values ​​higher than the determined threshold. as well as Output a medical image of the ROI generated from the spectral image dataset, indicating the identified contrast-enhanced region in the medical image.

15. A computer program comprising program code units, the program code units being configured to cause the computer to perform the steps of the method according to claim 14 when the computer program is executed on the computer.

Images