Device for assisting in determining functional subdivision of arterial blood supply region in organ of subject

Abstract

It is an object of the present invention to provide a device 110 that allows a user to be assisted in determining an improved subdivision of an arterial blood supply region in an organ of a patient 121. An image providing unit 111 provides an image of an organ of a subject. A geodesic distance map generation unit 113 generates a geodesic distance map for the region of interest on the basis of the anatomical image, in which voxels of the geodesic distance map are given by cumulative path costs from a predetermined seed to the corresponding voxels, and in which the path costs are determined on the basis of a cost function, the cost function takes into account vascularization filter responses and / or contrast agent densities determined based on the images, respectively. A geodesic distance map providing unit 114 provides the geodesic distance map to the user and / or further processes the geodesic distance map to determine an arterial blood supply region.

Description

Technical Field

[0001] The present invention relates to a device for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object, a system including such a device, a method, and a computer program product. Background Art

[0002] Determining the arterial blood supply regions in an organ of an object is very important for the diagnosis and treatment of many medical conditions, especially for the diagnosis and treatment of myocardial conditions. Generally, medical images acquired under the influence of a contrast agent are used to determine the perfusion in different tissue regions of the organ, and thus to determine the blood supply to these regions from the corresponding arteries. In order to determine the different blood supply regions of the arteries based on perfusion, it is known to use a schematic geometric pre-determined blood supply region atlas for the corresponding organ. In this way, the blood supply regions are roughly rigidly mapped onto the organ without using the information provided by the functional images (e.g., the contrast agent uptake and perfusion conditions obtained when determining the blood supply regions). Therefore, it would be advantageous if a practicing physician could be assisted in determining the arterial blood supply regions of an organ for each individual patient. Summary of the Invention

[0003] An object of the present invention is to provide a device, a system including such a device, a method, and a computer program product that allow a user to be assisted in determining an individual functional subdivision of an arterial blood supply region in an organ of a patient, in particular an improved one.

[0004] In a first aspect of the present invention, a device for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object, wherein the device includes: a) an image providing unit for providing an image of an organ of an object, wherein the image is an anatomical image acquired after administration of a contrast agent; b) a region of interest defining unit for defining a region of interest in the image for which the arterial blood supply region is to be determined; c) a geodesic distance map generating unit for generating a geodesic distance map for the region of interest based on the anatomical image, wherein the voxels of the geodesic distance map are given by the cumulative path cost from a predetermined seed to the corresponding voxel, and wherein the path cost is determined based on a cost function that takes into account a vascularity filter response and / or a contrast agent density determined respectively based on the image; d) a geodesic distance map providing unit for providing the geodesic distance map to a user and / or for further processing to determine the arterial blood supply region.

[0005] Since a geodesic distance map for a volume of interest is generated based on an image, wherein the voxels of the geodesic distance map are given by the cumulative path cost from a predetermined seed to the respective voxel, and wherein the path cost is determined based on a cost function that takes into account a vascularity filter response and / or a contrast agent density determined based on the image respectively, information about the blood supply to a certain region of tissue is clearly displayed on the geodesic distance map. Moreover, since the geodesic distance map is generated based on an image of the respective volume of interest of the patient, the determined result of the arterial blood supply region can be individually adapted to the respective patient. Thus, when determining a functional subdivision of the arterial blood supply region in an organ of an object, the device allows for improved assistance to the user, in particular, personalizing the determined arterial blood supply region.

[0006] The device is configured to assist in determining a functional subdivision of the arterial blood supply region in an organ of an object. The device can be implemented in the form of hardware and / or software provided by a general or special purpose computer system. In particular, the device can also be implemented in distributed computing. For example, the device can be implemented as part of a computer network, wherein the functions of the device are provided by different processors, servers, or computer systems that can be located at the same or different locations. The object can be a human or an animal, and can also be referred to as a patient hereinafter. The organ for which the arterial blood supply region is to be determined can be any organ of a human or an animal, but is preferably the heart of the object.

[0007] The image providing unit is configured to provide an image of the organ of the object. In particular, the image providing unit can refer to a storage unit or can be communicatively coupled to a storage unit, wherein the image has been stored on the storage unit, and the image providing unit is configured to provide the image stored on the storage unit. Additionally, the image providing unit can also be a receiving unit, which is used to receive an image, for example, from an input unit or directly from an image acquisition unit via an interface, and is used to provide the received image. Moreover, the image providing unit can also be an image acquisition unit, for example, a CT or MRI unit that is configured to acquire an image and is also configured to provide the image. The image is an anatomical image acquired after administration of a contrast agent, such that the image provides information about the transport and perfusion of fluid in the patient's vascular system (especially the arterial system). Preferably, the anatomical image is an anatomical volume image (i.e., a 3D anatomical image). This is particularly suitable if the organ is the patient's heart. However, in the case where the organ is naturally flat or is physically flattened, for example, for image acquisition, the anatomical image can also be a 2D image. In a preferred embodiment, the anatomical image is a volume image of the coronary arteries of the patient's heart.

[0008] The region of interest defining unit is configured to define a region of interest in the image for which the arterial blood supply region is to be determined. In particular, the region of interest can be any region of the anatomical image, even the entire image, and can be defined by the user or automatically during the user-machine interaction. For example, the region of interest defining unit can be configured to roughly segment the main anatomical structures of the anatomical image using known algorithms and then outline the anatomical structure of interest as the region of interest. However, the automatic definition of the region of interest can also be simpler, for example, by simply defining the region at a specific predetermined position of the anatomical image. In addition, more complex algorithms can be used, such as machine learning algorithms that have been trained to recognize and define regions of interest in the image. If the user is to define the region of interest, the region of interest defining unit can be configured to, for example, present the anatomical image to the user on a display, and then the user can use an input device (such as a mouse or keyboard) to indicate the outline of the region of interest for defining the region of interest. In addition, during the interaction, the region of interest defining unit can be configured to provide the user with the anatomical image on which the region of interest is automatically defined, where, in this case, the user can then accept or modify the corresponding region of interest. It should be noted here that a rather rough definition of the region of interest has been found to be suitable for this method. In particular, even if the region of interest refers to the entire image, the method can be performed with a suitable accuracy.

[0009] The geodesic distance map generating unit is configured to generate a geodesic distance map for the region of interest based on the anatomical image. In particular, the voxels of the geodesic distance map are given by the cumulative path cost from a predetermined seed to the corresponding voxel. Note that the formula "voxels given by..." means determining the value of the voxel such that the expression can also be formulated as "the value of the voxel of the geodesic distance map is given by the cumulative path cost from a predetermined seed to the corresponding voxel". Preferably, the cumulative path cost of the voxel from the predetermined seed to the corresponding voxel is determined by calculating the path costs of all possible precursor voxels (i.e., all voxels in the neighborhood of the corresponding voxel that have been assigned to the path) and determining the precursor voxel with the lowest cost. Generally, the path costs of the possible precursor voxels are the calculated path cost values for the corresponding values, i.e., as a result of calculating the corresponding path costs for the corresponding possible precursor voxels. Preferably, the cumulative path cost of the corresponding voxel is then determined based on the path cost of the lowest-cost precursor voxel plus the step cost of the path from the precursor voxel to the corresponding voxel, where the step cost is calculated using a cost function. Thus, preferably, the cumulative path cost is determined such that each voxel has only one precursor voxel with respect to one path, but can include multiple consecutive voxels with respect to that path. Thus, starting from the predetermined seed, the path can be divided into multiple paths downward, but each voxel can only be connected to one seed.

[0010] Typically, the path cost refers to a value associated with a certain path or a part of a path from a predefined starting point (i.e., the seed) to the corresponding voxel, and this value is determined based on a cost function. In most cases, when determining the path cost, the absolute value of the path cost is not relevant, but only the relationship of the cost of one path to another path contains relevant information, such as information about the arterial blood supply area. However, if needed, the cost function can be adjusted so that the absolute value of the path cost can also be utilized, for example, by normalizing, weighting, or otherwise adjusting the corresponding cost function. Preferably, the geodesic distance map is generated starting from a predefined seed using a forward propagation method based on a priority queue. Typically, not only one seed can be utilized, but also multiple seeds (e.g., two or more seeds) can be used, and the corresponding geodesic distance maps can be calculated for each of the different seeds. In particular, this can have the advantage of providing further information about different arterial blood supply areas. For example, if the corresponding path cost strongly depends on the seed, the development of the corresponding path cost can indicate that a particular voxel belongs to different arterial blood supply areas. Preferably, the seed is set by the user (e.g., a practicing physician) at or near the expected starting point of the arterial blood supply system. However, the seed can also be automatically set based on the corresponding predefined characteristics of the corresponding region of interest. Preferably, the seed is automatically determined by using a heuristic or a machine learning model. This model preferably detects the location in the arterial root from which the blood flows into the region of interest. For example, for the heart, this can be the location of the orifice on the ascending aorta. However, if the model cannot detect the corresponding seed location (e.g., due to anatomical variations), then preferably a graphical display is used to manually set the seed.

[0011] A cost function refers to a function that considers the vesselness filter response determined based on an image respectively and / or the contrast agent density. Preferably, the cost function determines a path cost such that a voxel being part of a blood vessel has a high probability and a good alignment between the current blood vessel direction estimation result and the previous blood vessel direction estimation result results in a low path cost, and vice versa. Additionally, if the contrast agent density is considered additionally or alternatively, preferably, a high contrast agent estimation result in the region of a voxel causes a low path cost in the cost function, and vice versa. A corresponding vesselness filter (especially a local vesselness filter) that determines the local vesselness fraction for a voxel and optionally also determines the vesselness direction is known. For example, a vesselness filter using a Hessian matrix including the second derivative of a voxel can be used, or a vesselness filter using a gradient with the first derivative of a voxel or a vesselness filter using search rays such as a radial structure tensor can be used. Generally, the vesselness filter response determines a fraction indicating the probability that a voxel is part of a blood vessel in a medical image. The contrast agent density can be determined based on the image, also using corresponding known methods (e.g., determining the gray value of a voxel and / or one or more surrounding voxels to determine the amount of contrast agent indicated by the gray value). Since the contrast agent exists in the blood vessel at the highest concentration (especially in the artery as shown in the image), the contrast agent density also indicates the presence of a blood vessel at the position of the voxel. Therefore, the cost function determines the path cost based on the probability of the presence of a blood vessel (especially an artery) in the region of the corresponding voxel for which the path cost should be determined. This allows the algorithm to follow the arterial blood vessel until the blood is distributed into the tissue and thus clearly indicates which tissue region is supplied by which artery. In a preferred embodiment, both the vesselness filter response and the contrast agent density are used. Additionally, the cost function can also consider additional features. For example, if the blood vessel segmentation for the region of interest is available, the cost function can further utilize the blood vessel segmentation to determine the path cost. Furthermore, the cost function can also consider the alignment of the path direction with the general direction of the blood vessel (e.g., an artery). In particular, the alignment with the general direction of the blood vessel from the root to the blood vessel and then to its target tissue (e.g., from the root of the aorta to the apex of the left ventricle). Preferably, the direction refers to the blood flow direction through the blood vessel. For this additional feature, the lower the path cost, the more the path direction follows the general direction of the blood vessel. Additionally, the cost function can be configured to consider the local blood vessel radius. For example, if available, the local blood vessel radius can be determined based on the segmentation result. Preferably, the local blood vessel radius is considered by comparing it with the generally decreasing overall blood vessel thickness. This gradual decrease in blood vessel thickness is usually towards the lower branches of the blood vessel (e.g., towards the apex of the ventricle). For example, the cost function can determine that the cost of a voxel with a longer path is higher because it becomes less and less likely that the voxel is still part of the path of the blood vessel.

[0012] In a preferred embodiment, the cost function f has the following form:

[0013] f = d[1 + exp(-αD c + βV)]

[0014] where d refers to the voxel adjacent distance, D c refers to the contrast agent density, V refers to the vascularity filter response, and the coefficients α and β refer to optimizable variables.

[0015] The geodesic distance map providing unit is configured to provide the geodesic distance map to the user and / or for further processing to determine the arterial supply area. Since the geodesic distance map includes all the information required to determine the arterial supply area, the user can determine the arterial supply area by simply viewing the geodesic distance map very easily. However, the geodesic distance map can also be further processed, for example, for automatically determining the arterial supply area.

[0016] In a preferred embodiment, the device further includes a geodesic distance map processing unit and a mapping unit. The geodesic distance map processing unit is configured to process the geodesic distance map such that the arterial supply area of the organ in the region of interest is determined based on the geodesic distance map. The mapping unit is configured to map the determined arterial supply area to the representation of the organ. In particular, the geodesic distance map providing unit can provide the geodesic distance map to the geodesic distance map processing unit. The geodesic distance map processing unit can be configured to process the geodesic distance map to present it to the user, where the user can then use a corresponding input device to determine the arterial supply area of the organ in the region of interest, for example, by providing a contour, a sketch, or an overlay indicating the corresponding supply area in the geodesic distance map. However, the geodesic distance map processing unit can also be configured to automatically process the geodesic distance map. For example, the corresponding rules and characteristics of the arterial supply area can be stored and used to determine the arterial supply area based on the geodesic distance map. For example, based on the determined path to the corresponding tissue area provided by the geodesic distance map, which part of the artery supplies blood to the corresponding area can be determined. Preferably, to determine the supply area for a voxel of the geodesic distance map, the steepest descent of the path cost from the voxel to a corresponding known and optionally labeled blood vessel (e.g., one of the renal arteries) is followed. The boundary between two supply areas can also be determined by following the steepest descent of the path cost of two adjacent voxels of the geodesic distance map that are not connected via a path to a common precursor voxel (i.e., the voxels belonging to the paths of the two voxels to a common seed). If the length difference of the paths following the steepest descent to a common precursor voxel is greater than a predetermined threshold, the adjacent voxels belong to different supply areas, and the boundary between the supply areas is determined between the adjacent voxels.

[0017] Then, the mapping unit is configured to map the determined arterial blood supply region to the representation of the organ, for example, by providing overlays, visible contours, color indications, etc. to the image or atlas representation of the organ. Then, the map thus determined can be presented to the user as an arterial blood supply map, and the user can be assisted in diagnosing and treating the corresponding patient.

[0018] In one embodiment, the device further includes an optimization unit configured to optimize the cost function by comparing the generated geodesic distance map with the contrast agent distribution in the anatomical image, wherein the cost function is optimized based on the comparison, and the geodesic distance map is generated based on the optimized cost function. Generally, the cost function can be optimized by optimizing one or more variables of the cost function. The variables of the cost function can refer to, for example, weights used to weight the different effects of the cost function on the path cost (e.g., the effects of the vascularity filter response and the contrast agent density on the result of the path cost calculation). However, the variables can also refer to other quantities that affect the result of the cost function and are not determined, for example, from the anatomical image. Generally, such variables can be predetermined, for example, based on the user's experience, but preferably these variables will be adjusted independently for each case to improve the accuracy of the corresponding geodesic distance map obtained. However, the cost function can also be optimized by changing the influence characteristics of the cost function (i.e., by changing the cost function itself). For example, instead of using only the vascularity filter response as an influence characteristic, the contrast agent density can be additionally or alternatively used in the changed cost function. Also, for example, the type of vascularity filter used to determine the vascularity filter response can be changed. This allows, for example, the cost function to be further optimized such that the cost function takes into account the most influential influence characteristics.

[0019] Since the contrast agent distribution is a directly measurable variable (i.e., part of the provided anatomical image) and indicates the blood flow through the blood vessels in the region of interest that should be mapped by the geodesic distance map, optimizing the cost function based on the comparison between the geodesic distance map and the contrast agent distribution in the anatomical image allows the cost function to be adjusted independently for each patient and thus provides a more accurate geodesic distance map. Any known optimization method (e.g., an iterative optimization method, where one or more iterative steps are performed until the difference between the generated geodesic distance map and the contrast agent distribution is minimized with respect to the corresponding cost function used) can be used to perform the optimization.

[0020] Preferably, the comparison is based on determining a correlation measure between the geodesic distance map and the contrast agent density in the anatomical image. Using a correlation measure indicating the correlation between the geodesic distance map and the contrast agent density has the following advantages: the absolute values of the geodesic distance map and the contrast agent density are less relevant when comparing than the relative values, and thus are less relevant than the structures of the corresponding geodesic distance map and contrast agent density. This allows for a fast and accurate comparison between the geodesic distance map and the contrast agent density in the anatomical image. Preferably, the correlation measure uses the absolute magnitude of the cross-correlation and / or cross-entropy between the geodesic distance map and the contrast agent density in the anatomical image.

[0021] In one embodiment, the cost function is optimized with respect to one or more variables, the one or more variables being linear weights of the quantities of the cost function, and wherein the optimization of the one or more variables is based on: solving a system of linear equations based on the cost function for each voxel of the geodesic distance map. For example, the cost function can include a predetermined number of influencing factors, such as, for example, a vascularity filter response and / or a contrast agent density, wherein each of these influencing factors includes at least one weighted quantity, and in particular, the influencing factors themselves can be linearly weighted quantities. The path cost for a voxel is typically the sum of the path costs of all previous voxels on the corresponding path of the voxel, and these previous voxel paths also respectively depend on a corresponding predetermined number of weights M. For N voxels, this thus results in a system of linear equations with N equations and M unknown weights. Such a system is overdetermined and can be solved using known methods (e.g., using methods of linear algebra).

[0022] In one embodiment, the cost function is derived as regression values from the voxel neighborhood for each voxel. In particular, in this case, the cost function can be regarded as a regression function implemented to determine regression values based on the voxel neighborhood of each voxel, where the regression function can be determined using corresponding regression analysis. Preferably, in this case, the cost function as a regression function linearly depends on the neighborhood of the corresponding voxel according to which the regression value is calculated. For example, the cost function can include a matrix multiplication of K weights and the neighborhood of the voxel. Then, a predetermined number N of voxels in the geodesic distance map can be used to form a system of linear equations of N equations for the K unknown weights of the cost function, where the system of linear equations can then also be solved using known methods (in particular, using linear algebra). However, alternatively, a machine learning regression algorithm (such as a neural network) can be used to determine the regression values. In this case, the machine learning regression algorithm can be trained to determine the path cost of a voxel based on the anatomical image values of adjacent voxels. The training dataset can include the corresponding anatomical images, thus including the anatomical image values of adjacent voxels, and include the corresponding desired path cost (i.e., the geodesic distance map) for the corresponding voxels. Then, the machine learning regression algorithm can be parameterized based on the training dataset. In both cases, the vesselness filter results and / or the contrast agent density are considered through the corresponding weights and parameters used by the regression function.

[0023] In one embodiment, the optimization is based on determining a mismatch error as the difference between the compared geodesic distance map and the contrast agent density, and backpropagating the mismatch error along the dendrogram structure of the geodesic distance map to the corresponding cost function in order to optimize the cost function. In this case, the relative mismatch error (i.e., the difference between the voxel in the geodesic distance map and the contrast agent density) can be directly determined. Since it is known that the mismatch error is necessarily caused by, for example, the path cost of the calculated path of the corresponding voxel, the variables used to determine the corresponding path cost in the cost function can be modified until the mismatch error is minimized. This direct optimization of the cost function is possible because each voxel belongs to only one path, i.e., is part of a dendrogram structure (i.e., a tree structure), such that it is clear that the corresponding mismatch in the geodesic distance map and the contrast agent density of the voxel can only be caused by the corresponding cost function of this one path. Generally, the mismatch error is a relative quantity used to determine which voxels have a particularly high mismatch compared to other voxels. In particular, the difference between the geodesic distance map and the contrast agent density is substantially the same for all regions (i.e., voxels) where the geodesic distance map generally follows the contrast agent density, and the difference is higher in cases where the geodesic distance map strongly deviates from the contrast agent density. This functional relationship is independent of the specific values of the geodesic distance map and the contrast agent density, and thus independent of the fact that the geodesic distance map and the contrast agent density represent different quantities. Additionally, as part of the optimization process in which the mismatch error is to be optimized, a local or absolute minimum of the mismatch error is determined, and this local or absolute minimum can have any value as long as the value refers to a minimum. This optimization is used to optimize the cost function such that when the cost function is calculated, it applies a higher weight to the influence on the contrast agent density.

[0024] In one embodiment, the image providing unit is configured to provide additional images, wherein the additional images are anatomical images acquired after administering a contrast agent at different vital parameters of the patient, wherein the geodesic distance map generating unit is configured to generate an additional geodesic distance map based on the additional images, and wherein the device further includes an optimization unit configured to compare the geodesic distance map with the additional geodesic distance map for optimizing the cost function. Generally, different vital parameters of the patient can refer to any vital parameters suitable for causing differences in blood vessels (especially in the arteries of the patient). For example, different vital parameters of the patient can be caused by different pressure states of the patient (such as rest state vs. exertion state). Additionally, different pharmacological means (e.g., by administering vasoactive drugs) can be used to initiate such different vital parameters of the patient.

[0025] In one embodiment, the image providing unit is configured to provide a dynamic time series of images acquired during the propagation of the contrast agent through the tissue of the organ of the patient as an image, wherein the geodesic distance map generating unit is configured to: a) generate geodesic distance maps respectively based on at least two images in the dynamic time series of images, thereby generating at least two geodesic distance maps, and b) i) generate a geodesic time curve for at least one point of interest based on the at least two geodesic distance maps, and ii) generate a contrast agent time curve for the at least one point of interest based on the dynamic time series of images, wherein the device further includes an optimization unit configured to correlate the geodesic time curve with the contrast agent time curve for optimizing the cost function. Preferably, a geodesic distance map is generated for each image, and each image is a part of the dynamic time series of images acquired during the propagation of the contrast agent to the tissue of the organ of the patient. This allows for more accurate time curves. Generally, the time curve indicates such a flow of the contrast agent through the corresponding part of the patient's body represented by the voxels in the corresponding image. Thus, the geodesic time curve provides this information based on the geodesic distance map, and the contrast agent time curve directly provides this information based on the measured contrast agent density provided by the corresponding images of the dynamic time series. Therefore, the relative courses of the time curves (i.e., the geodesic time curve and the contrast agent time curve for the voxel) should be the same. Then, the correlation between the geodesic time curve and the contrast agent time curve can be determined using, for example, the corresponding correlation method described above to determine the difference, and an optimization algorithm (e.g., an iterative algorithm) can be used to minimize the difference between the contrast agent time curve and the geodesic time curve with respect to the cost function. In particular, the overall correlation of multiple geodesic time curves and contrast agent time curves for multiple selected voxels (i.e., points of interest) can be used as the objective function for optimizing the cost function.

[0026] In a further aspect of the present invention, there is provided a method for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object, wherein the method comprises: a) providing an image of the organ of the object, wherein the image is an anatomical image acquired after administration of a contrast agent; b) defining a region of interest in the image for which the arterial blood supply region is to be determined; c) generating a geodesic distance map for the region of interest based on the image, wherein the voxels of the geodesic distance map are given by the cumulative path cost from a predetermined seed to the respective voxel, and wherein the path cost is determined based on a cost function that takes into account a vasculature filter response and / or contrast agent density determined respectively based on the image; d) processing the geodesic distance map such that the arterial blood supply region of the organ in the region of interest is determined based on the geodesic distance map; and e) mapping the determined arterial blood supply region to a representation of the organ. The method is a computer-implemented method that can be implemented on any computer hardware using a corresponding software product.

[0027] In a further aspect of the present invention, there is provided a system for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object, wherein the system comprises: a) an image acquisition unit for acquiring an image of the organ of the object after administration of a contrast agent; and b) the device as described above.

[0028] In a further aspect of the present invention, there is provided a computer program product for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object, wherein the computer program product causes the device as described above to perform the method as described above.

[0029] It should be understood that the device, system, method, and computer program product as described above have similar and / or identical preferred embodiments, in particular the embodiments defined in the dependent claims.

[0030] It should be understood that the preferred embodiments of the present invention can also be any combination of the dependent claims or the above embodiments and the corresponding independent claims.

[0031] These and other aspects of the present invention will be apparent with reference to the embodiments described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In the following drawings:

[0033] Figure 1 There is schematically and exemplarily shown a system for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object,

[0034] Figure 2A flowchart of a method for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object is schematically and exemplarily shown.

[0035] Figure 3 An exemplary illustration of determining an anatomical region of interest is shown.

[0036] Figure 4 An exemplary illustration of the development of a geodesic distance map based on one seed is shown, and

[0037] Figure 5 An exemplary illustration of the determination of the arterial blood supply region and the development of the geodesic distance map is shown. Detailed Description

[0038] Figure 1 A system for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object (e.g., a patient) is schematically and exemplarily shown. The system includes an imaging acquisition unit 120 configured to acquire images of a patient 121 lying on a patient table 122. Generally, the acquired images are anatomical images acquired after administration of a contrast agent. The image acquisition unit 120 can be, for example, a CT system that allows acquisition of three-dimensional CT images of the anatomical structure of the patient 121. In a preferred embodiment, the anatomical structure of the patient 121 is the patient's heart. Generally, for the anatomical structure of an organ that is a certain organ of a patient, the arterial blood supply region is to be determined, for example, for diagnostic or treatment purposes.

[0039] The system 100 includes means 110 for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object (e.g., a patient 121). Thus, the means 110 includes an image providing unit 111 configured to provide an image of an organ of an object (e.g., a patient 121). For example, the image providing unit 111 can be communicatively coupled to the image acquisition unit 120 for directly receiving the anatomical image of the patient 121 and providing the anatomical image. However, the image acquisition unit 120 can also first store the acquired images in a storage device (e.g., a cloud storage device), and then the image providing unit 111 can be configured to access the storage device and receive the image from the storage device for providing the image.

[0040] The region of interest defining unit 112 is configured to determine a region of interest in the image for which the arterial supply region is to be determined. For example, the region of interest defining unit 121 can present the image to the user by means of an output unit (such as the display 117), whereupon the user can then use an input unit (such as a mouse or keyboard 118) to define the region of interest. For example, the user can draw the contour of the region of interest on the image including the region of interest. However, the region of interest defining unit 112 can also be configured to automatically define the region of interest, for example, based on predetermined criteria. For example, the region of interest defining unit 112 can be configured to segment the anatomical structure within the image and can be provided with a predetermined rule that determines whether the segmented part of the heart is the region of interest or any other part of the segmented anatomical structure. Generally, the region of interest can also be the entire image, in which case, the region of interest defining unit can simply determine that the entire image is the region of interest.

[0041] Then, the geodesic distance map generating unit 113 of the apparatus 110 is configured to generate a geodesic distance map for the region of interest based on the anatomical image. Generally, the value of a voxel of the geodesic distance map is given by the cumulative path cost from a predetermined seed to the corresponding voxel. For example, the geodesic distance map generating unit 113 can present the image to the user via the display 117, for example, such that the user can indicate one or more seeds from which the geodesic distance map is to be determined. This allows the user to influence the generation of the geodesic distance map via seed determination. However, one or more predetermined seeds can also be automatically set by the geodesic distance map generating unit 113, for example, based on predetermined rules, predetermined positions, predetermined characteristics of the image, segmentation results of the image, etc. Starting from one or more predetermined seeds, for example, a forward propagation method based on a priority queue can then be used to calculate the cumulative path cost for the corresponding voxels in front of the forward propagation. To calculate the cumulative path cost, preferably, for the corresponding voxel, it is determined which adjacent voxel that is part of the path includes the lowest and highest path costs. Then, that voxel can be regarded as the precursor voxel of the corresponding voxel along the path to which the corresponding voxel belongs. Then, the cumulative path cost is calculated by using the path cost of the path to the precursor voxel plus the path cost from the precursor voxel to the corresponding voxel. For example, since the path cost for the corresponding precursor voxel is known (i.e., is provided by the value of the voxel), the corresponding value of the precursor voxel can simply be added to the additional path cost determined for the path between the precursor voxel and the corresponding voxel.

[0042] If a geodesic distance map is calculated using a priority queue-based forward propagation method, the calculation of voxels follows the following algorithm. First, one or more seeds are submitted as the first entry to a waiting queue for the initialization of the algorithm. Then, the previous propagation is iterated by extracting from the waiting queue the waiting entry with the current lowest cost (i.e., the previous voxel with the lowest path cost among all voxels in the previous waiting queue). Then, the entry (i.e., the voxel) is processed by searching for adjacent voxels that have not been queued and processed so far (i.e., voxels for which the path cost has not been determined so far), and the entry is input into the waiting queue. For example, as described above, for each voxel for which the path cost has not been determined so far, its cost is determined as the sum of the current voxel (i.e., the parent node) and the corresponding currently added additional step for the potential precursor voxel. Thus, in this embodiment, the previous and thus the determination of the geodesic distance map propagates fastest along the path with the lowest cost (i.e., along the expected path of blood from an artery into tissue). The path propagation algorithm can be stopped when the queue is empty (i.e., when the path cost has been determined for all voxels), or when the next calculated path cost exceeds a certain threshold (e.g., when all possible blood supply paths have been followed into tissue), or when the calculation exceeds a certain predetermined calculation time or computer memory constraint.

[0043] The path cost between voxels along a path is determined based on a cost function that takes into account the vasculature filter response and / or the contrast agent density. Preferably, the cost function takes into account these two features that can be determined based on the image. Using the vasculature filter response and / or the contrast agent density to calculate the cost function allows increasing the path cost of voxels in parts that are not blood vessels (in particular, parts that are not arteries), but for example parts of tissue, and decreasing the path cost of voxels that have a high likelihood of being part of a blood vessel (e.g., an artery). This allows generating a geodesic distance map during the previous propagation to identify the path along which a fluid (in particular blood) travels and dissipates into tissue.

[0044] The geodesic distance map providing unit 114 can then provide the user with a geodesic distance map, for example, by using the display 117. The user can then easily distinguish different arterial supply regions in the organs of the object based on the geodesic distance map. However, the user can then also input one or more additional seeds, for example, when the arterial supply regions are not yet clear, and restart the generation of the geodesic distance map based on the one or more modified seeds. However, the geodesic distance map providing unit 114 can also provide the determined geodesic distance map to the optional geodesic distance map processing unit 115. The geodesic distance map processing unit 115 is then configured to process the geodesic distance map, in particular to determine the arterial supply regions of the organs in the region of interest based on the geodesic distance map. For example, in this case, the user can directly indicate the arterial supply regions in the geodesic distance map, for example, by using a contour. However, the geodesic distance map processing unit 115 can also be configured to automatically determine the arterial supply regions of the organs in the region of interest based on one or more features of the corresponding geodesic distance map. Then, the optional mapping unit 116 can be configured to map the determined arterial supply regions to a representation of the organ (for example, an anatomical image of the organ and / or any other representation, such as a contour representation, a segmented part of the organ, a atlas representation of the organ, etc.). Based on this map, the user can directly view the arterial supply regions in the organ and base further diagnostic and treatment options on this knowledge. In particular, with the device 110, the arterial supply regions in the organ can be determined very accurately for each individual patient.

[0045] Figure 2 A flowchart schematically and exemplarily shows a method for assisting in the functional breakdown for determining the arterial supply regions in an organ of an object. Generally, the method 200 can be carried out by and in accordance with the principles described with respect to Figure 1 and the device 110. In a first step 210, the method includes providing an image of the organ of the object, wherein the image is an anatomical image acquired after administration of a contrast agent. In the next step 220, a region of interest is defined in the image for which the arterial supply regions are to be determined. Furthermore, the method 200 includes a step 230 of generating a geodesic distance map of the region of interest based on the anatomical image according to the principles described with respect to Figure 1 and the device 110. In step 240, the geodesic distance map can then be provided to the user or alternatively used for further processing in step 250, in which the geodesic distance map is processed such that the arterial supply regions of the organs in the region of interest are determined based on the geodesic distance map. In an additional optional step 260, the determined arterial supply regions can then be mapped to a representation of the organ.

[0046] In the following, preferred embodiments and examples are described in more detail. Generally, spectral CT assessment of the cardiac region has advantages over conventional (especially non-spectral) CT due to its excellent contrast agent sensitivity. Thus, spectral CT assessment of the cardiac region is beneficial for the functional assessment of the myocardium and its blood-supplying coronary arteries. The contrast agent uptake in blood vessels (especially coronary arteries) and muscle regions (especially the myocardium) can be used as a proxy for perfusion, even if the caliber of small tissue-embedded blood vessels is below the spatial resolution limit of the scanner. The perfusion in the myocardial region is functionally related to the blood-supplying coronary arteries, such that the regional relationship between coronary artery lesions and myocardial perfusion defects has diagnostic value. In order to determine the arterial blood supply regions in this context, schematic atlases of the myocardium are widely used, for example, the AHA segments, which are schematic, geometrically defined, and not patient-specific. Generally, the atlas is a rigid mapping to 17 coarse regions and does not utilize functional image information to indicate the blood supply range.

[0047] As described above, for example with respect to Figure 1 and Figure 2 has been described, the insight underlying the present invention is that the observed distribution of a contrast agent (e.g., iodine) in arteries and tissues (e.g., in coronary arteries and the myocardium) can be considered as a proxy for blood vessel and tissue perfusion and can be "explained" by a tree-like diagram structure rooted in the arterial blood supply and the corresponding blood vessel network (e.g., the coronary artery network). Utilizing this insight allows for the functional subdivision of an organ (e.g., the myocardium) into individual arterial blood supplies (i.e., perfusion ranges) based on a single, optionally static, given image after the administration of a contrast agent. Additionally, it can be considered that the computationally inexpensive and thus fast determination along the propagation path of high-contrast and blood vessel-like lines / regions in an anatomical image is related to the dissipation of arterial blood supply to highly perfused regions, and the dissipation path can thus be considered related to the functional blood supply range.

[0048] In the following, details of a preferred example of the present invention applied to determine the blood supply regions in the myocardium of a patient's heart are described. In a first step that can be performed, for example, by a region of interest defining unit in a preferred method, a region of interest is defined in an organ (e.g., the heart) shown by an anatomical image. In particular, a cardiac region of interest model can be provided in order to define the region of interest in the image. For example, in a grid model representation or a labeled volume image that produces labels (i.e., semantic segmentation) of the cardiac region of interest anatomical structures, the corresponding region of interest can be defined and transferred to the anatomical image. In this example, the myocardium surrounding the left or right ventricle and the orifice that is the root of the coronary artery tree can be defined as the region of interest, for example, as Figure 3 shown.

[0049] In the next step, based on the anatomical image, a geodesic distance map is calculated for the defined region of interest. For example, from one or more seeds (i.e., the tree roots), a local cost function is used to calculate the geodesic distance map. Each voxel of the geodesic distance map is given by the cumulative path cost from the seed to the point represented by the voxel. Then, the identity of the seed to which the path to the corresponding voxel belongs (i.e., the branch label) can be associated with the voxel and can be inherited from the path-predecessor voxel of the voxel, which is the adjacent voxel that has had the lowest cumulative path cost (i.e., distance) so far. Since each voxel of the geodesic distance map has a defined path-predecessor voxel, an implicit tree-like graph structure is constructed. Optionally, the variables used in the cost function can be further optimized. For example, the geodesic distance map is adjusted for the observed contrast appearance by optimizing the local cost function (i.e., the variables of the local cost function). Examples and possibilities of such optimization are described further below. Generally, the resulting geodesic distance map preferably depends on the local cost function such that a reduced cost is determined to increase the contrast agent density and the vasculature filter response. Then, the goodness of fit of the geodesic distance map can be evaluated by its correlation with the contrast agent density map. For example, where the anatomical image is a spectral CT image, as a perfusion surrogate. Then, the local cost function (which can be a global function with local independent variables) can be varied until an optimal correlation between the geodesic distance map and the observed contrast agent pattern is achieved. Such variation can be a choice of function or a variation of the variables (e.g., function coefficients), while the function independent variables are given by the observed volume of the image of interest. In a further step (e.g., performed by a geodesic distance map processing unit), the graph provided by the geodesic distance map can be partitioned into subtrees. For example, in the case where only a single seed is provided (e.g., in the aortic bulb) or a finer partition than the number of seeds provided is desired, the tree-like graph on which the geodesic distance map is based can be partitioned into subtrees (e.g., using dendrogram segmentation techniques). Then, the new path labels, and the new path costs (if applicable), are re-propagated to the voxels of the geodesic distance map. Additionally, the geodesic distance map processing unit can also be configured to map the perfusion extent (i.e., the blood supply region) determined based on the processing of the geodesic distance map to a graphical representation of the organ. For example, the resulting perfusion extent can be visualized graphically as a boundary or a pseudo-color scheme can be used to map the resulting perfusion extent. The mapping can be applied to graphical representations such as, for example, 3D rendering, a schematic 2D bull's-eye view, a flattened manifold (i.e., a "world map"), a scrollable slice-by-slice CT viewport, etc.

[0050] Generally, the main advantages of the present invention as described above by way of example are: the ability to present to the user a patient-specific functional perfusion range rather than a schematic atlas range, as they are derived from specific patient images (e.g., CT scans performed during a rest phase and a stress phase). Additionally, the presentation of the blood vessel flow in the coronary arteries can be an inherent part of the muscle region representation. Further, the described invention is particularly suitable for taking advantage of the additional benefits of spectral CT information (e.g., reducing the cost function along the path as the contrast agent density increases). Additionally, the method is algorithmically specific and tractable and can be computed within a typical interaction time span. Moreover, the method does not require a prior, explicit segmentation of the coronary arteries, which can be error-prone).

[0051] In the following, some additional possible optional embodiments are described. In one embodiment, the geodesic distance map and the corresponding dendrogram of the geodesic distance map can be computed based on, for example, one or more seeds positioned at the orifices provided by the anatomical model. Additionally, a forward propagation method based on a priority queue can be utilized.

[0052] In one embodiment, the dendrogram for computing the geodesic distance map can be partitioned into blood supply regions. For example, based on the computed geodesic distance map and its implicit dendrogram, the segmentation operation for partitioning into the main branches (indicating the blood supply regions) can be accomplished, for example, by interactive user selection or automatically (e.g., by using the dendrogram in combination with a rule-based or atlas-based scheme) or by unsupervised clustering techniques.

[0053] Further, in the following, some embodiments of the optimization of the cost function that can be performed by an optional optimization unit are described in more detail. In one embodiment, a goodness-of-fit metric can be utilized for the optimization. For example, in order to optionally adjust (i.e., optimize) the cost function such that the geodesic distance map fits the observed perfusion appearance (i.e., the contrast agent image), a metric can be used to compare the computed geodesic distance map with the contrast agent image observed from spectral CT. For example, a low geodesic path cost is expected to be consistent with high perfusion. A suitable comparison metric is the absolute magnitude of the cross-correlation, or the cross-entropy between the computed geodesic distance map and the measured (i.e., the contrast agent density values observed in the anatomical region of interest, e.g., the coronary arteries and the myocardium). After optimization, the best achieved goodness-of-fit can be used as a confidence metric, i.e., if the correlation between the geodesic distance map and the observed iodine pattern is insufficient, the underlying assumptions are violated and a warning is issued to the user.

[0054] Typically, the local cost function can be considered as the reciprocal of the local velocity function and is a globally effective function, but its local cost function value depends on local properties (e.g., contrast agent density and magnitude, orientation, and width estimation results from a suitable vascularity filter, etc.). A very simple parametric cost function can be, for example, f = d[1 + exp(-αD c +βV)], where D refers to the voxel adjacent distance, D c refers to the contrast agent density, V refers to the vascularity filter response, and the coefficients α and β refer to optimizable variables. Then the function selection or function coefficients can be changed to optimize the goodness of fit.

[0055] In one embodiment, the variables of the cost function (e.g., diffusion parameters) can be fitted by general optimization. For example, if the local cost function contains only very few variables, a general numerical optimization scheme can be applied to optimize the correlation between the contrast agent density and the geodesic distance map in the volume of interest.

[0056] In one embodiment, the variables of the cost function (e.g., diffusion parameters) can be optimized by feature weight regression. For example, in the case of linear weights of the local cost function, a system of linear equations with M unknowns can be provided based on N voxels in the geodesic distance map in the volume of interest, which can be solved in closed form. Using the updated function weights (which are variables in this case), the geodesic distance map can be recalculated for multiple iterations until convergence.

[0057] In one embodiment, the variables of the cost function (e.g., diffusion parameters) can be optimized by feature weight backpropagation. Since the geodesic distance map implicitly provides a dendrogram structure, the mismatch error of the final graph nodes (i.e., voxel values) can be propagated back to each local cost function, and then each local cost function can be temporarily changed until convergence.

[0058] In one embodiment, the variables of the cost function (e.g., diffusion parameters) can be optimized by general non - linear regression. A non - parametric local cost function can also be used, which is derived from the adjacent voxel values of the local (e.g., 9×9×9) voxel neighborhood of a voxel as the regression value. The regression can be trained based on the corresponding training samples. Typically, for each voxel of the distance map and its path precursor voxel, there is a training sample, where the training sample is a set of two items: i) the adjacent image voxel values of the voxel determined based on, for example, an anatomical image, and ii) the desired path cost (i.e., the desired difference between the path cost of the voxel and the path cost of its precursor voxel). As a vector input regressor, for example, a neural network (NN), a support vector machine (SVM), or a random forest (RF) can be utilized.

[0059] In one embodiment, the optimization can be based on two phase images. In this case, two phase image volumes are provided that are acquired with different patient in - vivo parameters (e.g., when the patient is exerting force and when relaxed). Then, the optimization of the local cost function can be performed to satisfy the regions of interest in both images simultaneously. Then, the spatial consistency of the two computed geodesic distance maps can be used as another optimization and confidence criterion.

[0060] In one embodiment, the optimization can be based on a dynamic time - series scan. In a case where not only a single static spectral CT scan is available, but also a dynamic time - series of the scan volume is available, for each position in the volume of interest, a bolus curve of the contrast agent can be established and correlated with the "propagation" time (i.e., the cost value) of the geodesic distance map. This can be used (e.g., by modeling the relationship between vessel diameter and blood flow and the relationship between Hounsfield density and blood flow, optionally considering the incompressibility of fluids and calcifications, for example) for further optimization of the local cost function.

[0061] Figure 4 An exemplary illustration of the development of a geodesic distance map based on one seed is shown. In particular, Figure 4 The forward propagation of the evolution of the geodesic distance map starting from a seed (i.e., the root point in the upper - left corner of the image) using the local cost function is shown, where the cost decreases under high contrast agent density and high vascularity response.

[0062] Figure 5 An exemplary illustration of the determination of the arterial supply region and the development of the geodesic distance map is shown. In particular, Figure 5 Two sub - tree labels are shown (i.e., geodesic distance maps generated based on two seeds), which propagate with the evolving geodesic distance Figure 1 and define the affected area as an indication of the anatomical perfusion supply range. It should be noted here that this illustration is only a conceptual illustration and contains, for example, incorrect tracking at vessel crossings.

[0063] Although the above embodiments have been mainly described with respect to cardiac spectral CT, the present invention can also be applied in a similar manner to conventional CT, cardiac MR, or other cardiac modalities (such as ultrasound), where perfusion is manifested due to contrast agent uptake. In addition, the present invention can also be applied to organs other than the heart that have blood - supplying vessels (especially the lungs or the liver).

[0064] By studying the drawings, the disclosure, and the appended claims, those skilled in the art can understand and implement other variations of the disclosed embodiments when practicing the claimed invention.

[0065] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude a plurality.

[0066] A single unit or device can implement the functions recited in several of the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used advantageously.

[0067] A process performed by one or more units or devices (e.g., providing an anatomical image, defining a region of interest, generating a geodesic distance map, providing a geodesic distance map, etc.) can be performed by any other number of units or devices. These processes can be implemented as program code units of a computer program and / or dedicated hardware.

[0068] A computer program can be stored / distributed on a suitable medium (e.g., an optical storage medium or a solid-state medium) (provided together with other hardware or as part of other hardware), but can also be distributed in other forms (e.g., distributed via the Internet or other wired or wireless telecommunication systems).

[0069] Any reference signs in the claims shall not be construed as limiting the scope.

[0070] The object of the present invention is to provide a device that allows assisting a user in determining a subdivision of an arterial blood supply region in an organ of an improved patient. An image providing unit provides an image of an organ of an object. A geodesic distance map generating unit generates a geodesic distance map for a region of interest based on the anatomical image, wherein the voxel of the geodesic distance map is given by the cumulative path cost from a predetermined seed to the corresponding voxel, and wherein the path cost is determined based on a cost function that takes into account a vesselness filter response and / or a contrast agent density determined based on the image respectively. A geodesic distance map providing unit provides the geodesic distance map to the user and / or performs further processing to determine the arterial blood supply region.

Claims

1. A device for assisting in determining a functional subdivision of an arterial blood supply area in an organ of an object (121), wherein, The device (110) includes: - an image providing unit (111) configured to provide an image of an organ of an object (121), wherein the image is an anatomical image acquired after administration of a contrast agent, - a region of interest defining unit (112) configured to define a region of interest in the image for which the arterial blood supply region is to be determined, - a geodesic distance map generating unit (113) configured to generate a geodesic distance map for the region of interest based on the anatomical image, wherein a voxel of the geodesic distance map is given by a cumulative path cost from a predetermined seed to the corresponding voxel, and wherein the path cost is determined based on a cost function that takes into account a vasculature filter response and / or contrast agent density determined respectively based on the image, - a geodesic distance map providing unit (114) configured to provide the geodesic distance map to a user and / or for further processing to determine the arterial blood supply region.

2. The device according to claim 1, wherein The device (110) further includes a geodesic distance map processing unit (115) and a mapping unit (116), the geodesic distance map processing unit being configured to process the geodesic distance map such that the arterial blood supply region of the organ in the region of interest is determined based on the geodesic distance map, the mapping unit being configured to map the determined arterial blood supply region to a representation of the organ.

3. The device according to any one of claims 1 and 2, wherein The device (110) further includes an optimization unit configured to optimize the cost function by comparing the generated geodesic distance map with the contrast agent distribution in the anatomical image, wherein the cost function is optimized based on the comparison, and a geodesic distance map is generated based on the optimized cost function.

4. The device according to claim 3, wherein, The comparison is based on determining a correlation measure between the geodesic distance map and the contrast agent density in the anatomical image.

5. The apparatus according to claim 4, wherein, The correlation measure utilizes an absolute magnitude of a cross-correlation and / or cross-entropy between the geodesic distance map and the contrast agent density in the anatomical image.

6. The apparatus according to any one of claims 3 to 5, wherein The cost function is optimized with respect to one or more variables that are linear weights of quantities of the cost function, and wherein the optimization of the one or more variables is based on: solving a system of linear equations based on the cost function for each voxel of the geodesic distance map.

7. The apparatus according to any one of claims 2 to 5, wherein, The cost function is derived as a regression value from a voxel neighborhood for each voxel.

8. The device according to any one of claims 3 to 5, wherein, The optimization is based on determining a mismatch error as a difference between the compared geodesic distance map and the contrast agent density, and backpropagating the mismatch error along a trellis structure of the geodesic distance map to the corresponding cost function in order to optimize the cost function.

9. The device according to any one of the preceding claims, wherein, The image providing unit (111) is configured to provide additional images, wherein the additional images are anatomical images acquired after administering a contrast agent under different vital parameters of the patient, wherein the geodesic distance map generating unit (113) is configured to generate an additional geodesic distance map based on the additional images, and wherein the device (110) further includes an optimization unit configured to compare the geodesic distance map with the additional geodesic distance map for optimizing the cost function.

10. The device according to any one of the preceding claims, wherein, The image providing unit (111) is configured to provide a dynamic time series of images acquired during the contrast agent propagating through the tissue of the organ of the patient as images, wherein the geodesic distance map generating unit (113) is configured to: a) generate geodesic distance maps respectively based on at least two of the images of the dynamic time series, thereby generating at least two geodesic distance maps, and b) i) generate a geodesic time curve for at least one point of interest based on the at least two geodesic distance maps, and ii) generate a contrast agent time curve for the at least one point of interest based on the dynamic time series of the images, wherein the device (110) further includes an optimization unit configured to correlate the geodesic time curve with the contrast agent time curve for optimizing the cost function.

11. The device according to any one of the preceding claims, wherein, The cost function f has the following form: f = d[1 + exp(-αD c + βV)] where d refers to the voxel adjacent distance, D c refers to the contrast agent density, V refers to the vascularity filter response, and the coefficients α and β refer to optimizable variables.

12. The device according to any one of the preceding claims, wherein, The geodesic distance map is generated using a priority queue-based forward propagation method starting from the predetermined seed.

13. A method for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object (121), wherein, The method (200) includes: - providing (210) an image of an organ of an object (121), wherein the image is an anatomical image acquired after administering a contrast agent, - defining (220) in the image a region of interest for which the arterial blood supply region is to be determined, - generating (230) a geodesic distance map for the region of interest based on the anatomical image, wherein the voxels of the geodesic distance map are given by the cumulative path cost from a predetermined seed to the respective voxels, and wherein the path cost is determined based on a cost function that takes into account a vesselness filter response and / or a contrast agent density determined respectively based on the image, - providing (240) the geodesic distance map to a user and / or performing further processing to determine the arterial blood supply region.

14. A system for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object, wherein, The system (100) includes: - an image acquisition unit (120) for acquiring an image of an organ of an object after administering a contrast agent, - a device (110) according to any one of claims 1 to 12.

15. A computer program product for assisting in determining a functional subdivision of an arterial blood supply region in an organ of an object (121), wherein, The computer program product causes the device (110) according to any one of claims 1 to 12 to execute the method (200) according to claim 13.

Images