Apparatus, method and computer program for separating film spaces

Abstract

The present invention relates to an apparatus and a method for dividing a membranous spacing. The method comprises the following steps: segmenting a left ventricular outflow tract LVOT of a heart in a 3D image data set; determining wall thickness information indicating spaced wall thicknesses at different locations of a portion of the segmented LVOT oriented toward the right ventricle and the right atrium; mapping the determined wall thickness information onto a surface of the portion of the segmented LVOT; and segmenting the membranous spacing in the 3D image dataset based on the mapped wall thickness information.

Description

Technical Field

[0001] The present invention relates to a device, a method and a computer program for dividing a membranous septum. Background Art

[0002] The so-called membranous septum (i.e., the membranous fibrous component of the entire cardiac septum) is a small structure in the heart that lies between the left ventricular outflow tract (LVOT) and the right ventricle / atrium and is characterized by a thin wall between the left and right ventricles at the level of the tricuspid annulus. It indicates the location of the conduction pathway in the heart. Certain geometric properties of the membranous septum are associated with the risk of requiring a pacemaker after transcatheter aortic valve replacement (TAVR).

[0003] The membranous septum is a small and non-invasive structure on a non-planar surface that is difficult to find and annotate in 3D image datasets (such as cardiac computed tomography angiography (CTA) image datasets), especially in standard planar slice views. In addition, anatomical knowledge is available that can help identify the membranous septum when used appropriately. In this way, large datasets as well as cumbersome and thus expensive manual annotation (such as that required by neural network-based methods) should be avoided.

[0004] The paper titled "Virtual septal myelectomy for preoparative planning in hyperhropic cardiomyopathy"

[0005] (published by Takayama Hiroo et al. in The Journal of Thoracic and Cardiovascular Surgery, Vol. 158, No. 2, August 1, 2019, pp. 455-463) discloses a virtual myectomy (VM) technique that uses three-dimensional reconstruction of gated cardiac computed tomography (CT) to assist in the intraoperative objective assessment of resection adequacy. Summary of the Invention

[0006] An object of the present invention is to provide a device, a method and a computer program for effectively and reliably dividing a membranous septum in an automated manner.

[0007] In a first aspect of the present invention, a device for dividing a membranous septum is proposed, the device comprising a circuit configured to:

[0008] Divide the left ventricular outflow tract LVOT of the heart in a 3D image dataset;

[0009] Determine wall thickness information that indicates the wall thickness of the septum at different positions of the portion of the segmented LVOT that is oriented towards the right ventricle and right atrium;

[0010] Map the determined wall thickness information onto the surface of the portion of the segmented LVOT; and

[0011] Segment the membranous septum in the 3D image dataset based on the mapped wall thickness information.

[0012] In a further aspect of the invention, there is provided a corresponding method, computer program, which includes program code modules and a non-transitory computer-readable recording medium. When the computer program is executed on a computer, the program code modules are used to cause the computer to perform the steps of the methods disclosed herein. The non-transitory computer-readable recording medium stores a computer program product therein, and the computer program product, when run by a processor, causes the methods disclosed herein to be executed.

[0013] Preferred embodiments of the invention are defined in the dependent claims. It should be understood that the claimed methods, computer programs, and media have similar and / or identical preferred embodiments as the claimed system, particularly as defined in the dependent claims and as disclosed herein.

[0014] The present invention is based on the concept of automatically segmenting the membranous septum by segmenting the LVOT and using the segmented LVOT to determine the thickness information of the septum adjacent to the LVOT on the sides of the right ventricle and right atrium. Based on this thickness information, the membranous septum can be found, particularly as the part of the entire cardiac septum with the minimum thickness. Mapping the region of interest onto a two-dimensional reformat can simplify the subsequent image processing steps for manually and / or automatically delineating and quantifying the membranous septum.

[0015] According to an embodiment, the circuit is further configured to segment the membranous septum by applying any one of the following: thresholding of the surface of the portion of the segmented LVOT, independent component analysis, and k-nearest neighbor clustering. Generally, for this purpose, any standard method for segmentation can be applied on the surface of the LVOT.

[0016] The circuit can also be configured to segment the LVOT by adapting a cardiac model to the 3D image dataset using model-based segmentation. This provides a good starting point for subsequent processing steps.

[0017] In a preferred embodiment, the heart model includes an indication of the region where the membranous septum should be found and / or includes information that allows generating two or more planes perpendicular to the LVOT and / or allows determining the aortic valve annulus plane. This improves the segmentation of the membranous septum. Thereby, the information can include, for example, geometric information in the form of landmarks with coordinates, such as the centroid of the aortic valve, the aortic valve normal vector, the annulus plane normal, the ascending aorta normal vector, and / or the septum centroid.

[0018] In a practical implementation, the circuit can also be configured to determine wall thickness information by:

[0019] Determining an axis approximating the centerline of the LVOT;

[0020] Defining a set of rays that are perpendicular to the axis and / or the surface and pass through the wall of the LVOT in the segmented portion of the LVOT;

[0021] For the rays of the set, determining corresponding wall thickness information; and

[0022] Mapping the wall thickness information to the position on the surface where the corresponding ray passes through the wall of the LVOT.

[0023] Thereby, the circuit can also be configured to determine the axis by:

[0024] Determining the aortic valve annulus plane before determining the wall thickness information; and

[0025] Defining an axis perpendicular to the aortic valve annulus plane.

[0026] In another embodiment, the circuit can also be configured to determine the axis by fitting a line through the centroids of two or more surface loops of the ascending aorta, particularly as provided by an adapted geometric model (model-based segmentation).

[0027] Even further, in another embodiment, the circuit can also be configured to determine a wall thickness value or a proxy value of the wall thickness value of the thickness of the septum along the corresponding ray as wall thickness information.

[0028] Thereby, as a proxy value of the wall thickness value, the average image value (or pixel value or grayscale value) of a ray segment of a corresponding ray that starts at the surface of the LVOT and has a predetermined or adaptable length. The image value is related to X-ray opacity (radiation opacity, typically in Hounsfield units HU), and the X-ray opacity is in turn related (approximately linearly) to mass density (relative to water and air). Thus, for a hypothetical constant material, the cumulative density yields an approximate mass, which translates to length (thickness).

[0029] In addition, the wall thickness value can be determined based on the length of a ray segment of a corresponding ray that originates at the surface of the LVOT and ends at a change in an image value (or pixel value or grayscale value) indicative of tissue change.

[0030] According to another embodiment, the wall thickness information can be determined by using a trained algorithm or computer system, particularly a trained machine learning, a trained classifier, or a trained neural network, which has been trained to estimate the wall thickness of an organ structure in a 3D image dataset of the heart. This avoids the need to apply a dedicated algorithm, but allows the use of artificial intelligence for this task.

[0031] The circuit can also be configured to determine the axis by calculating a distance map inside the LVOT and maximizing the distance from a straight line in the LVOT to the surface of the LVOT.

[0032] According to another embodiment, the circuit is further configured to determine the region in which the membranous septum should be found by:

[0033] Segmenting the right ventricle and the right atrium;

[0034] Finding the point closest to the LVOT;

[0035] Identifying the ray passing through that point; and

[0036] Defining an angular range for rotating the ray about the axis and a translation range for moving the ray up or down along the axis.

[0037] According to yet another embodiment, the circuit is further configured to:

[0038] Utilize a geometric model represented by a triangular mesh to identify 3D coordinates for a corresponding ray outward from the LVOT manifold surface, at which the ray intersects one of the mesh triangles in the mesh, which intersects the triangle vertices; and

[0039] Use the 3D coordinates as the starting point of a ray for estimating the local thickness of the septum.

[0040] The circuit can also be configured to visualize the determined wall thickness information and / or the segmented membranous septum on the surface of the segmented portion of the LVOT or on a flattened reformatting of the segmented portion of the LVOT. This helps the user understand the segmentation results and make a diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] These and other aspects of the invention will be apparent and elucidated with reference to the (one or more) embodiments described below. In the following drawings

[0042] Figure 1 shows a schematic diagram of a part of the anatomical structure of the human heart;

[0043] Figure 2 shows a schematic diagram of an embodiment of a device according to the present invention;

[0044] Figure 3 shows a flowchart of an embodiment of a method according to the present invention;

[0045] Figure 4 shows a flowchart of another embodiment of a method according to the present invention;

[0046] Figure 5 shows a schematic diagram of a heart model and a left ventricle into which additional information is encoded;

[0047] Figure 6 shows two cross-sectional views of the heart taken along different planes.

[0048] Figure 7 shows a cross-sectional view indicating the search space in the z direction.

[0049] Figure 8 shows cross-sectional views from different perspectives.

[0050] Figure 9 shows a cross-sectional view indicating the search space in the rotational direction.

[0051] Figure 10 Illustrates the results of an MBS-based implementation that gives a texture-covered manifold / graph.

[0052] Figure 11 Illustrates a manifold / graph with a covered area of the right ventricle and right atrium.

[0053] Figure 12 shows a diagram illustrating another exemplary implementation of the present invention. Detailed Description

[0054] Figure 1 shows a schematic diagram of the main parts of the anatomical structure of the human heart. The left ventricular outflow (LVO) is an anatomically complex region of the left ventricle (LV) that gives rise to the ascending aorta (AA), provides support for the aortic valve leaflets, and houses important components of the conduction system. The LV is located between the anterior leaflet of the mitral valve (MV) and the smooth left ventricular side of the muscular and membranous interventricular septum (IVS). The anterior leaflet of the MV (posterolateral wall) is displaced from the IVS by the LVO. The area of fibrous continuity is called the mitral-aortic valve interval fibrosa (MAIF).

[0055] The membranous septum (membranous IVS) is a small, compact fibrous part of the IVS that lies between the non-coronary and right coronary cusps and the septal leaflet of the tricuspid valve (TV), and is thus uninterrupted with the right wall of the aortic root at its upper end. Moreover, the membranous septum is partially atrioventricular and anatomically separates a small part of the right atrium closest to the septal leaflet of the TV from the LV. The atrioventricular (AV) bundle is present at the junction of the muscular and membranous IVS.

[0056] Figure 2 A schematic view of an embodiment of a device 1 for segmenting the membranous septum according to the present invention is shown. The device 1 includes a circuit 2, such as a processor or a computer, which is configured to perform the steps of a method 10 for segmenting the membranous septum, which is schematically illustrated in Figure 3 an embodiment of the method.

[0057] The first step 11 of the method 10 segments the left ventricular outflow tract (LVOT) of the heart in a 3D image dataset, which can be obtained directly from an imaging system 3 (e.g., a CT imaging system, an X-ray system, etc.) or from an image storage device or repository 4 (e.g., a hospital's picture archive or database). The 3D image dataset can be, for example, a cardiac computed tomography angiography (CTA) image dataset acquired from a patient, which can be processed immediately to provide important information to a physician (e.g., survival during surgery) or for making a diagnosis.

[0058] The second step 12 of the method determines wall thickness information that indicates the wall thickness of the septum at different positions of the part of the segmented LVOT that is oriented towards the right ventricle and the right atrium. As described above, the thickness of the septum varies along its longitudinal extent, such that the wall thickness information can be used to identify the membranous septum.

[0059] The third step 13 maps the determined wall thickness information onto the surface of the part of the segmented LVOT that is oriented towards the right ventricle and the right atrium, as used in step 12.

[0060] The fourth step 14 segments the membranous septum in the 3D image dataset based on the mapped wall thickness information. Thus, the region of the membranous septum can be segmented on the surface of the LVOT by thresholding, independent component analysis, k-nearest neighbor (k-NN) clustering, or another method.

[0061] The fifth (optional) step 15 can visualize (e.g., display on a screen 5) the wall thickness information and / or the segmentation result (the segmented membranous septum) on the surface of the LVOT or perform appropriate flattening reformatting. In addition, tools for interactive correction of the segmentation by a user can be provided, e.g., as a user interface.

[0062] The sixth (optional) step 16 may store wall thickness information and / or segmentation results, for example, together with the 3D image dataset or with information indicating that it belongs to the 3D image dataset. For example, it may be stored as an annotation of the 3D image dataset.

[0063] In an embodiment, the device may be implemented using dedicated units or modules, such as an input unit for obtaining the 3D image dataset, a processing unit for performing the steps of method 10, and an output unit for outputting the results.

[0064] The input unit may be directly coupled or connected to the imaging system, or may obtain (i.e., retrieve or receive) these signals from a storage device, buffer, network, or bus, etc. Thus, the input unit may 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 signal transmission to the device.

[0065] The processing unit may be configured to perform the steps of method 10 and may be any kind of module configured to process the 3D image dataset. It may be implemented in software and / or hardware, for example, as a programmed processor, computer, or app on a user device.

[0066] The output unit may generally be any interface that provides the determined results, for example, sending it to another device or making it available for retrieval by another device (such as a computer, tablet, hospital network, etc.). Thus, it may generally be any (wired or wireless) communication or data interface.

[0067] Below, a more detailed embodiment of method 20 using model-based segmentation (MBS) according to the present invention will be explained, as Figure 4 illustrated by the flowchart shown.

[0068] As a first step 21 of the segmentation method, for example, a model-based method described in the following article is used to perform heart and aortic valve segmentation (which generally includes the above step 11 of segmenting the LVOT): "Patient Specific Models for Planning and Guidance of Minimally Invasive Aortic Valve Implantation" by Waechter I et al. in "Medical Image Computing and Computer-Assisted Intervention - MICCAI 2010" (Lecture Notes in Computer Science 2010, Volume 6361, 526 - 533). This method uses model-based segmentation to fit a heart model including a detailed model of the aortic valve (as shown in Figure 5A as shown (taken from this disclosure)) to the image. Additionally, information is encoded in the model, as shown in Figure 5B as shown (also taken from this disclosure), which allows, for example, constructing a plane perpendicular to the outflow tract or determining the aortic annulus plane. The information encoded in the model is supplemented with information about the region where the membranous septum can be found (for example, using information from publications or by using the region where the membranous septum has been found in a previous case, which is enlarged by a safety margin).

[0069] In a second step 22 (which is essentially the same as step 13), wall thickness information is mapped onto the outflow tract where the expected membranous septum is located. This can be done through the following sub-steps.

[0070] In a first sub-step 221, an axis close to the centerline of the outflow tract is determined. This can be done by determining the aortic annulus plane 30 (see Figure 6, which shows a cross-sectional view through the AA and LVOT) and defining an axis 31 perpendicular to this plane in the middle of the region where the annulus plane 30 intersects the outflow tract LVOT. Alternatively, the axis can be defined by fitting a line through the centroids of two or more surface loops of the ascending aorta, especially as provided by an adapted geometric pattern. Figure 6A shows a cross-sectional view through Figure 6B the annulus plane 30 as shown. Figure 6B shows a cross-sectional view 32 through another plane (relative to the long axis plane of the aortic valve (opposite to the long axis plane of the left ventricle)), which is perpendicular to the annulus plane 30 and passes through the axis 31. The line 33 indicates the AA ( Figure 6A and 6B ) and the walls of the LV and LVOT ( Figure 6B ).

[0071] In a second sub-step 222, the axis 31 is used to define a set of dense rays (in Figure 6A one ray is arranged in another plane 32), these rays are perpendicular to the axis 31, and pass through the wall 33 of the outflow tract in the region where the membranous septum can be found according to the markers encoded on the heart model. For each of the rays, a quantitative characteristic (proxy metric) of the septal wall thickness is calculated and mapped onto the mesh surface.

[0072] In a third sub-step 223, an example of such a quantity is, for example, the average gray value (or pixel value or image value) of a ray segment that starts at the surface of the LVOT and has a given length. The more uncontrasted tissue between the contrasted chambers, the smaller this average gray value will be. Alternatively, machine learning can be used to train a classifier that classifies a contour as belonging or not belonging to the septum, or to train a regression network that estimates the wall thickness.

[0073] In a third step 23, the region of the membranous septum is segmented, which can be done by thresholding, independent component analysis, k-NN clustering, or another method that uses the quantity that is characteristic of the wall thickness as a feature. Visualization and interactive correction of the segmentation result (e.g., by editing the contour on the curved surface after an appropriate transformation of the segmentation result) can be done using standard methods, such as by an electronic pen, a brush, etc., to draw or depict and modify the 2D contour on a 2D region.

[0074] Figures 7 to 11 The method and results are illustrated. Figure 7 Shows a cross-sectional view similar to the Figure 6B shown cross-sectional view. It indicates the search space in the z-direction (indicated by the arrow 40) along the axis 31 from the annulus plane 30 to the LV plane 34. The arrow 41 indicates the approximate position of the membranous septum. Fig. 8 shows a cross-sectional view similar to the one shown in Fig. 6, but in which a cross-sectional view of a different plane 35 rotated around the axis is shown compared to the Figure 6A shown plane 32 Figure 8B .

[0075] Fig. 9 shows a cross-sectional view similar to the cross-sectional view shown in Fig. 6a to indicate how the rays (indicated as ray 36 in Figure 9A and as ray 37 in Figure 9B ) rotate around the axis 31. This is the search space in the angular direction (indicated by the arrow 42).

[0076] Figure 10 Illustrates the results of the MBS-based implementation, which gives a manifold / graph that covers the texture on the model mesh surface with semi-transparent RV and RA. Figure 11 Illustrates the manifold / graph with the covered regions of RV and RA.

[0077] Figure 12 shows a diagram illustrating another exemplary embodiment of the present invention. As Figure 12A shown, starting from the centroid of the mesh triangles of the mesh modeling part of the cardiac anatomy (including the right ventricle), rays are sampled in a direction orthogonal to the LVOT axis 31. The rays can have a length in the range of 5 to 10 mm, for example, a length of 6 or 10 mm, where a predetermined value can be used and set to a value for the thin part covering the thickness of the septum. The starting point of the rays for the rays is typically on the LVOT / aortic surface. The average density (HU) of the rays is mapped to the drawing grayscale value color (image value or pixel value) of the triangular mesh. Thus, for example, light gray can indicate the thin low-density part of the septum, and dark gray can indicate the thick low-density part of the septum.

[0078] Figure 12B shows a flat (i.e., curve re-formatted) two-dimensional manifold, which is parameterized by cylindrical mapping of the LVOT and aortic mesh surfaces using height and angle relative to the LVOT axis. The sampling rays do not originate from the centroid of the sparse triangles, but are sampled in a dense manner from the flat two-dimensional manifold.

[0079] Figure 12C shows a diagram in which the right ventricle is hidden to improve visual inspection so as not to occlude the septum. Figure 12B The dense-sampled two-dimensional manifold shown is used for texture mapping and is folded around the triangular mesh representation of the LVOT and aortic surfaces, resulting in a finer and thus clearer representation.

[0080] Instead of using model-based segmentation in step 21, a neural network can also be used to segment the outflow tract into a voxel mask. Then, the axis for projecting the rays in sub-step 221 can be determined by calculating the distance map inside the outflow tract and maximizing the distance from a straight line in the outflow tract to the surface of the outflow tract. The rough region where the membranous septum can be found can be defined by: segmenting the right ventricle and right atrium, finding the point closest to the LVOT, identifying the ray passing through that point, and defining the angular range for rotating the ray around the axis and the translation range for moving the ray up or down along the axis. Identifying the membranous septum by projecting the rays and calculating the quantity indicating the wall thickness can be done as described for the example of using MBS in sub-steps 222 and 223.

[0081] The construction of the two-dimensional manifold from the triangular mesh and the LVOT axis can be performed as follows. For each cylindrical (z, f) coordinate on the LVOT axis fit, an axis-orthogonal ray can be projected from the axis, and the triangle vertex closest to the intersection (e.g., implemented as an efficient KD (K-dimensional) tree search) can be identified. The 3D coordinates of the corresponding manifold surface point can then be determined as the mean of, for example, the k nearest mesh vertex neighbors. Thus, all landmarks that can help define the membranous septum can be cross-correlated in all three representations (image volume, surface mesh, reformatted manifold) for both manual annotation and for automated algorithm purposes.

[0082] In summary, the present invention allows for the efficient and reliable segmentation of the membranous septum in an automated manner. The disclosed method can be used with other types of images such as MR or spectral CT images. Spectral CT images can be used to better resolve the material properties of the underlying tissue.

[0083] The segmentation results can be used to visualize the membranous septum and to extract measurements in transcatheter aortic valve replacement (TAVR) applications such as the distance between the aortic valve plane and the lower and / or upper edges of the septum, e.g., in a medical diagnostic view system such as Philips Intellispace Portal.

[0084] Although the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. Other variations of the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

[0085] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude a plurality. A single element or other unit may fulfill the functions of several items recited in the claims. Although specific measures are recited in mutually different dependent claims, this does not indicate that a combination of these measures cannot be used to advantage.

[0086] A computer program may be stored / distributed on a suitable non-transitory medium, such as an optical storage medium or a solid-state medium provided together with or as part of other hardware, but the computer program may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

[0087] Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. An apparatus (1) for segmenting membranous septa in a 3D image dataset, the apparatus comprising circuitry (2) configured to: Segment the left ventricular outflow tract (LVOT) of the heart in the 3D image dataset; Determine wall thickness information indicative of the wall thickness of the septum at different locations of the portion of the segmented LVOT that is oriented towards the right ventricle and right atrium; Map the determined wall thickness information onto the surface of the segmented portion of the LVOT; and Segment the membranous septum in the 3D image dataset based on the mapped wall thickness information.

2. The apparatus (1) according to claim 1, wherein, The circuit (2) is further configured to segment the membranous septum by applying any one of the following: thresholding of the surface of the segmented portion of the LVOT, independent component analysis, and k-nearest neighbor clustering.

3. The apparatus (1) according to any one of the preceding claims, wherein, The circuit (2) is further configured to segment the LVOT by adapting a heart model to the 3D image dataset using model-based segmentation, in particular using a heart model that includes an indication of the region where the membranous septum should be found, and / or includes information that allows the generation of two or more planes perpendicular to the LVOT and / or allows the determination of the aortic valve annulus plane.

4. The apparatus (1) according to any one of the preceding claims, wherein, The circuit (2) is further configured to determine the wall thickness information by: Determine an axis that approximates the centerline of the LVOT; Define a set of rays that are perpendicular to the axis and / or the surface and pass through the wall of the LVOT in the segmented portion of the LVOT; For the rays of the set, determine the corresponding wall thickness information; and Map the wall thickness information at the location on the surface where the corresponding ray passes through the wall of the LVOT.

5. The apparatus (1) according to claim 4, wherein, The circuit (2) is further configured to determine the axis by: Determine the aortic valve annulus plane before determining the wall thickness information; and Define an axis perpendicular to the aortic valve annulus plane.

6. The apparatus (1) according to claim 4, wherein, The circuit (2) is further configured to determine the axis by fitting a line through the centroids of two or more surface loops of the ascending aorta, in particular by fitting a line through the centroids of two or more surface loops of the ascending aorta provided by an adapted geometric model.

7. The apparatus (1) according to claim 4, wherein, The circuit (2) is further configured to determine a wall thickness value or a proxy value of the wall thickness value of the thickness of the septum along the corresponding ray as the wall thickness information.

8. The apparatus (1) according to claim 7, wherein, The circuit (2) is further configured to: determine the average image value of a ray segment that starts at the surface of the LVOT and has a predetermined or adaptable length as the proxy value of the wall thickness value for the corresponding ray; and / or determine the wall thickness value based on the length of a ray segment that starts at the surface of the LVOT and ends at a change in the image value indicating a tissue change for the corresponding ray.

9. The apparatus (1) according to claim 7, wherein, The circuit system (2) is further configured to determine the wall thickness information by using a trained algorithm or computer system, in particular a trained machine learning, a trained classifier, or a trained neural network, that has been trained to estimate the wall thickness of organ structures in a 3D image dataset of the heart.

10. The apparatus (1) according to claim 4, wherein, The circuit (2) is further configured to determine the axis by calculating a distance map inside the LVOT and maximizing the distance from a straight line in the LVOT to the surface of the LVOT.

11. The apparatus (1) according to claim 4, Among them, The circuit (2) is also configured to determine the region in which the membranous septum should be found by: Dividing the right ventricle and right atrium; Finding the point closest to the LVOT; Identifying the ray passing through that point; and Defining an angular range for rotating the ray about the axis and a translation range for moving the ray up or down along the axis.

12. The apparatus (1) according to claim 4, Among them, The circuit (2) is also configured to: Use a geometric model represented by a triangular mesh to identify 3D coordinates for corresponding rays outward from the LVOT manifold surface, at which the ray intersects one of the mesh triangles in the mesh and the ray intersects a triangle vertex; And Use the 3D coordinates as the starting point of the ray for estimating the local thickness of the septum.

13. The apparatus (1) according to any one of the preceding claims, Among them, The circuit (2) is also configured to visualize the determined wall thickness information and / or the segmented membranous septum on the surface of the segmented portion of the LVOT or on a flattened reformatting of the segmented portion of the LVOT.

14. A computer-implemented method (10) for segmenting membranous septa in a 3D image dataset, the method comprising: Segment (11) the left ventricular outflow tract LVOT of the heart in the 3D image dataset; Determine (12) wall thickness information indicative of the wall thickness of the septum at different positions in the portion of the segmented LVOT that is oriented towards the right ventricle and right atrium; Map (13) the determined wall thickness information onto the surface of the segmented portion of the LVOT; And Segment (14) the membranous septum in the 3D image dataset based on the mapped wall thickness information.

15. A computer program comprising program code modules which, when the computer program is executed on a computer, cause the computer to perform the steps of the method according to claim 14.

Images