Determining soft tissue related properties from x-ray imaging data

Abstract

An apparatus (100) for determining soft tissue related properties of a population of bones from X-ray imaging data is provided. The apparatus (100) comprises a data processor (110) configured to obtain X-ray imaging data of a region of interest comprising a population of bones, and determining a pose estimate of the group of bones and at least one geometric parameter (d, t, r) defined between the bones of the group of bones by using a 3D pose estimation computer model (CM) to which the X-ray imaging data is fed to determine soft tissue related properties of the group of bones.

Description

Technical Field

[0001] The invention relates to medical imaging and in particular to an apparatus and a computer-implemented method and a computer program element for determining soft tissue related properties of a bone constellation from X-ray imaging data. Background Art

[0002] In medical imaging, automatic (i.e. computer-assisted) fracture detection is widely addressed, since bones are clearly visible in, for example, musculoskeletal X-ray images. Typically, X-ray technology only allows for clear visualization of dense objects, such as bones. However, if less dense objects are to be examined by imaging, the data required to do so are not clearly visible in the X-ray imaging data, and therefore non-bone-related injuries or abnormalities or soft tissue injuries or abnormalities may not be directly observed or diagnosed therefrom. Instead, these may only be observed or diagnosed indirectly, and therefore only by experienced physicians, which does not allow for automation by computers.

[0003] Therefore, for the purpose of observing or diagnosing soft tissue related injuries or abnormalities, more time-consuming or cost-intensive image acquisition techniques, such as magnetic resonance tomography (MRT), are often indicated. However, this requires appropriate equipment, technicians, and the patient's patience, among other things. Summary of the invention

[0004] Therefore, means for automatically analyzing less dense objects in X-ray images may be needed.The objects of the invention are solved by the subject-matter of the independent claims, wherein further embodiments are incorporated in the dependent claims.

[0005] According to a first aspect, there is provided an apparatus for determining soft tissue related properties of a skeletal group based on X-ray imaging data. The apparatus comprises at least one data processor configured to: (i) obtain X-ray imaging data of a region of interest including the skeletal group; and (ii) determine a pose estimate of the skeletal group and at least one geometric parameter defined between bones of the skeletal group by utilizing a 3D pose estimation computer model to which the X-ray imaging data is fed to determine the soft tissue related properties of the skeletal group.

[0006] This allows an automatic analysis of less dense objects in the X-ray imaging data, i.e. based solely on the X-ray imaging data. In particular, this allows an analysis of the X-ray imaging data by a processor, i.e. a computer, to obtain properties related to a bone group that are not directly visible in the X-ray image, in particular at least one soft tissue related property. Such an automatic analysis saves significant costs, since no MRI is required. Furthermore, the automatic analysis requires less well-trained personnel, and the results of the analysis can be further processed computationally, automatically recorded, etc.

[0007] The inventors have discovered in a non-obvious manner that by considering or encoding at least one additional geometric parameter defined between the bones of a skeletal group, a significant source of information can be used to perform soft tissue-related analyses and / or diagnoses other than purely bone-related diagnoses (such as fracture detection) based solely on X-ray imaging data.

[0008] For example, the 3D pose estimation computer model may also be referred to as an extended and / or parameterized 3D joint model, wherein the 3D pose estimation computer model is configured to estimate pose parameters of a skeletal group, which may be a joint, such as an ankle joint, etc. The 3D pose estimation computer model is parameterized using the additional geometric parameters described herein. The term pose as used herein may refer to projection geometry (i.e., a narrow pose including viewing direction and detector position), joint posture (i.e., flexion parameters), etc. The basic 3D pose estimation computer model may include two modules or stages, wherein, in a first stage, pose discriminative features (bone contours plus some inner contours) are detected by a convolutional neural network (CNN) as a binary segmentation mask, and in a second stage, another CNN regresses all pose parameters based on these feature masks. An advantage of this approach is that the two components can be trained independently of each other, so that no pose ground truth labels are required for X-ray images. An exemplary 3D pose estimation model (not limiting the described aspects) that can be extended by the additional geometric parameters described herein is described in " et al., CNN-based pose estimation for assessing quality of ankle-joint X-ray images (SPIE Medical Imaging 2022), which is incorporated herein by reference in its entirety.

[0009] For example, feature detection for a 3D pose estimation model can be implemented as follows: As a preprocessing step for a feature detection network, a region of interest (ROI) can be automatically extracted from a given input X-ray image. The ROI cropping method can be based on a foveal fully convolutional network (FNet) that predicts the bone contours on a downsampled X-ray image and a probabilistic anatomical atlas of these bone contours. A similarity transformation with an additional flip degree of freedom for compensating for left / right laterality can be used to align the detected bone contours to the atlas. The resulting transformation can be used to map the anatomically defined ROIs of the joints from the atlas into the input image, and to resample the ROIs thus determined. A feature detection network can be trained on the ROI resampled at a high spatial resolution so as to sufficiently resolve the spatial group of contours for 3D pose prediction. For example, for a lateral (LAT) ankle X-ray image, an FNet can be trained to segment the external contour of the fibula and the selected prominent contours of the tibia, talus, and calcaneus. A contour-by-contour softmax layer can be attached to the FNet architecture. Thus, the network is configured to predict a heatmap and its complement for each target contour, compressing them into binary predictions by applying a pixel-wise argmax operation to each of these heatmap pairs, thereby allowing spatial coexistence of different contour categories. Training data can be generated by manually or automatically annotating structures in multiple X-ray images (e.g., multiple X-ray images, such as multiple, tens, hundreds, etc.), and discretized by contour annotations on pixels of the expanded ROI. Thus, the feature detection task can be formulated as a standard segmentation problem.

[0010] For example, pose estimation for a 3D pose estimation model can be implemented as follows: In order to train a pose estimation network without the need for X-ray images with pose truth, an articulated 3D model of a bone group (e.g., ankle joint) can be constructed from, for example, a sequence of MR images obtained with varying flexion angles. For example, a semi-manual segmentation of the tibia, fibula, talus, and calcaneus in one MR image enables the creation of a surface model represented using a triangular mesh. For each triangle of the mesh, features for model adjustment can be trained using image intensities, and the resulting model is used to depict the corresponding bone surface in other MR images of the sequence. A bone-by-bone rigid transformation can be applied to the model-based segmentation because all images are created with the same volunteer. By measuring the angle between the main axis of the tibia and the main axis of the calcaneus, a flexion angle δ can be assigned to each MR image. Smooth interpolation of the surface model with respect to δ enables the mesh group to be estimated at any flexion angle between approximately 80° and approximately 135°. In order to generate synthetic projection data based on this articulated 3D model, some pose parameters are defined. The central beam (or viewing direction) is determined by the angle θ relative to the head-to-foot axis and the angle θ relative to the medial-lateral axis in the transverse plane. To simplify the sampling of pose parameters for training the pose estimation network, it can be assumed that the detector and source rotate in a fixed C-arm-like configuration about the joint center, which means that and θ uniquely determine the positions of the source and detector. Here, the source-detector distance and the joint-detector distance can be selected to be within the realistic range of clinical ankle lateral examinations and to approximately match the ROI on which the feature detection network operates. The posture parameters may include the detector plane rotation angle γ, the translation 2D vector t, the scaling factor s, and the flexion angle δ. The basic 3D posture estimation computer model is configured to determine not only the acquisition parameters (such as the geometric relationship between the source, joint model and detector), but also some free parameters of the joint itself, such as scaling parameters and joint flexion angles. For example, this can be used to analyze and / or evaluate whether the projection image of the joint has been acquired in the desired posture, and therefore whether the image quality is sufficient to achieve the desired diagnosis. The basic 3D posture estimation computer model is not currently used for medical diagnosis of images.

[0011] It should be noted that in the above-described base 3D pose estimation computer model, the bones of a skeletal group (e.g., joints) are constrained to have fixed groups relative to each other, which limits the ability of the 3D pose estimation computer model to adapt to a given X-ray image if those model constraints are not given, for example, due to pathological changes in the skeletal group.

[0012] In this paper, it is proposed to extend the above-mentioned (basic) 3D posture estimation computer model by additional free geometric parameters between bones and their matching, so as to also determine soft tissue related or correlation geometric parameters and determine them based on X-ray imaging data (i.e. X-ray images) alone. In addition to the angles between bones that can reflect typical joint movements (such as flexion, abduction, rotation, etc.) to be found in the acquired posture, the present disclosure focuses on at least one geometric parameter between bones in 3D described herein. The additional geometric parameters are a valuable source of information in various diagnostic tasks, not only for image quality assessment, but also in particular for medical diagnosis of images. In contrast, attempts to perform dislocation detection only in 2D images are limited to deviations (dislocations) only in the image plane and are affected by positioning errors whenever they exist. The device, system and method described herein can circumvent these two limitations because it is based on posture position estimation in 3D.

[0013] In addition to the at least one geometrical parameter, further relevant parameters may also be combined, for example in a linear manner or by logistic regression.

[0014] The apparatus may be implemented in software run by a data processor, in hardware or in a combination of hardware and software.The 3D pose estimation model may be stored as a program element in a computer memory and executed by a data processor, or may be implemented in a chip.

[0015] As used herein, a skeleton group may be any arrangement of bones of a human or animal body in a torso, skull, spine, arms, hands, legs, feet, etc. Specifically, the skeleton group may form joints or articulations of a body mapped in a 3D pose estimation model, such as knees, elbows, hands, shoulders, etc.

[0016] In addition, as used herein, soft tissue can be understood as being different from bone groups or bones, as an object of the body having a lower density than bones, due to which the soft tissue is not visible or poorly visible or almost invisible in X-ray imaging data. With respect to bone groups, soft tissue can include, for example, cartilage, ligaments, tendons, muscles, fascia, etc. Soft tissue can extend or be located between bones, adjacent bones, or at a (small) distance around bones.

[0017] It should be noted that the device may also include a suitable data interface configured to receive X-ray imaging data. The imaging data may be provided by a picture archiving and communication system (PACS), an X-ray detector, etc. In addition, the device may be configured to generate a report on soft tissue related properties after detection.

[0018] Optionally, the device may also be configured to trigger output of at least the soft tissue related property. Output may be understood as any type that can be used for further processing, such as data output for further processing by a data processor or computer, output of a property of output or information only for a user, etc. This may allow the determined soft tissue related property to be used for preferably automatic medical diagnosis.

[0019] Optionally, the apparatus may also be configured to interpret and / or visualize the soft tissue related properties in or based on the 3D pose estimation computer model. For example, the determined soft tissue related properties may be graphically illustrated and / or interpreted in a graphical representation of the 3D pose estimation computer model. They may be graphically highlighted therein, labeled in text form, etc.

[0020] Optionally, the device can also be configured to determine the pathological changes and / or abnormalities related to soft tissues, particularly automatically, according to the properties related to at least one soft tissue. This can include any type of dislocation in or of a bone group that may involve damaged soft tissues. For example, dislocated (too far, too narrow) bones may indicate, for example, ligament tearing due to trauma, ligament weakness (osteoarthritis risk), cartilage degeneration (osteoarthritis), other bone fractures (trauma, osteoporosis), other muscle or tendon ruptures (trauma, overload), congenital anomalies, posture effects (errors), etc. In addition, for example, there may also be injuries that are more difficult to detect from X-ray images than dislocated feet. In traumatic situations such as motorcycle or bicycle accidents, hand injuries are often overlooked when focusing on other life-threatening injuries to the abdomen and head. Once the patient is otherwise stable, the device, system, and method described herein can be used to mark suspicious cases for early disposal.

[0021] Optionally, the device may also be configured to compare geometric parameters and / or posture estimates with corresponding thresholds to determine soft tissue related lesions and / or abnormalities. For example, a geometric parameter may be a distance between two bones, including a relative distance, which may have a normal value between a lower threshold and an upper threshold, such that damage, abnormality, etc. may be determined by comparing the geometric parameter with this normal value range. In addition, by comparing with the corresponding threshold, it may be determined whether the posture was correct during image acquisition, thereby avoiding misdiagnosis. For example, deviations from normal parameters or thresholds may be used as additional diagnostic information in routine radiology services also for asymptomatic situations. They may indicate, for example, congenital anomalies or give prognostic parameters, such as a risk score for developing osteoarthritis. A database of typical or normal bone distances may be provided and used for diagnosis, so that each individual patient is evaluated with respect to statistical values.

[0022] Optionally, the device can also be configured to adjust the corresponding threshold when the implant is present in or associated with the bone group. For example, after trauma, the bones can be fixed surgically using implants and other devices, which can both strongly constrain possible parameter changes and introduce parameter settings that are not typical for normal anatomical structures (e.g., outside the learning range). This situation may require the introduction of additional constraints and / or conditions on a point-to-point / situation-specific basis. In this regard, the device can be configured to obtain information about implants or other bone fixation devices, for example, from data entered by a user, from patient data, through feature detection in X-ray imaging data, etc., and the impact of implants on soft tissue can be taken into account. For example, it is conceivable that relatively small distances between bones caused by implants or bone fusions are given correction values ​​to exclude misdiagnosis.

[0023] Optionally, the apparatus may also be configured to combine at least one geometric parameter, at least one geometric parameter and / or soft tissue related property with patient related data to determine soft tissue related lesions and / or abnormalities. The patient related data may include, for example: the age of the patient, as some findings are more or less likely depending on age; the body mass index (BMI) of the patient, as this may affect the distances between bones, etc. The patient's pre-existing medical conditions, etc. may also be taken into account as additional parameters. This may make the automated diagnosis more precise.

[0024] Optionally, the device may also be configured to trigger an alert to be output based on the determined soft tissue related lesions and / or abnormalities. For example, a reader of the images with limited expert knowledge, such as MSK applications, may be alerted to "incidental findings". Often, reader or diagnostic attention is drawn to major structural lesions, such as large fractures or other bony injuries, which may cause more subtle (e.g. soft tissue) lesions or abnormalities to be overlooked. Here, automatic alerts may be provided and may also be used for additional (secondary) findings to be automatically populated into the report scheme.

[0025] Optionally, the 3D pose estimation computer model can be parameterized by parameters including one or more of: flexion angle of the skeletal group, viewing direction of the skeletal group, X-ray detector position relative to the skeletal group, translation in the X-ray detector plane, rotation in the X-ray detector plane, scaling in the X-ray detector plane, and geometric parameters, wherein the 3D pose estimation computer model is trained to regress one or more parameters.

[0026] Optionally, the at least one geometric parameter may include one or more of the following: distance between bones, orthogonal translation and rotation. For example, distance between bones (such as talus-tibia distance) encoded as additional free parameters in the 3D pose estimation computer model is an important source of information for diagnosing patients. The distance between bones can reflect the size of the joint space (e.g., upper ankle joint space, etc.). Thus, low distance values ​​can indicate cartilage degradation, such as due to age or osteoarthritis, and very high numbers can indicate ligament tears and foot dislocations.

[0027] Optionally, the device can also be configured to determine an axis between two adjacent bones of the skeletal group and measure the distance between the two adjacent bones along the axis. For example, the distance between bones (such as the talus-tibia distance) encoded as an additional free parameter in the 3D pose estimation computer model is an important source of information for diagnosing patients.

[0028] Optionally, the at least one geometric parameter may comprise distances between bones, and a tree or graph is used to encode a plurality of distances between bones in the 3D pose estimation computer model, wherein the bones are vertices and the distances are modelled by edges.

[0029] Optionally, the at least one geometric parameter may be a variable parameter of the 3D pose estimation computer model. As described above, in the base 3D pose estimation computer model, the bones of the skeleton group are constrained to have fixed groups relative to each other. According to this example, the geometric parameter is a freely configurable parameter that can be varied. This may indicate soft tissue related properties.

[0030] According to a second aspect, there is provided a computer-implemented method for determining soft tissue-related properties of a skeletal group from X-ray imaging data. The method comprises: obtaining X-ray imaging data of a region of interest including the skeletal group; and determining a pose estimate of the skeletal group and relative distances between bones of the skeletal group by utilizing a 3D pose estimation computer model to which the X-ray imaging data is fed, the relative distances being assigned to at least one soft tissue-related property of the skeletal group.

[0031] The method may preferably be performed by using the apparatus according to the first aspect.

[0032] According to a third aspect, a computer program element is provided, which, when run by a processor, is configured to perform the method of the second aspect and / or control the apparatus according to the first aspect.

[0033] According to a fourth aspect, a computer-readable storage or transmission medium is provided, storing or carrying a computer program element according to the fourth aspect.

[0034] It should be noted that the above examples and embodiments can be combined with each other, regardless of the aspects involved. Thus, the method can be combined with structural features of the apparatus of other aspects, and likewise, the apparatus can be combined with features of each other and can also be combined with features described above with respect to the method.

[0035] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Exemplary embodiments of the present invention will be described in the following drawings.

[0037] Figure 1 An exemplary apparatus for determining soft tissue-related properties of a skeletal group based on X-ray imaging data is shown in block diagram form.

[0038] Figure 2 Figures 2A to 2FModel parameters of an exemplary 3D pose estimation computer model are illustrated in FIG.

[0039] Figure 3 An example graph is shown encoding adjacent bones in an exemplary bone group.

[0040] Figure 4. The corresponding X-ray images Figure 4A and 4B , geometric parameters for determining soft tissue-related properties of a bone group based on X-ray imaging data are shown.

[0041] Figure 5 A method for determining soft tissue-related properties of a skeletal group based on X-ray imaging data is illustrated in a flow chart.

[0042] List of reference numerals:

[0043] 100 devices

[0044] 110 Data Processor

[0045] CM3D Pose Estimation Computer Model

[0046] 200 imaging data sources

[0047] d, t, r geometric parameters DETAILED DESCRIPTION

[0048] Figure 1 An exemplary apparatus 100 for determining soft tissue related properties of a bone group based on X-ray imaging data is shown in a schematic block diagram. The apparatus 100 can be any suitable computer and includes a data processor 110, which is operatively connected to a suitable data interface and can be operatively connected to further computer peripherals. The X-ray imaging data can be provided as input data by a data source 200 (such as an X-ray detector, a picture archiving and communication system (PACS) or the like) to the apparatus 100, in particular the data processor 110. The output data OUT of the data processor 110 can be provided for further processing to another computer device, a software module, an output device (such as a display, a reporting system, etc.), a PACS, etc.

[0049] The data processor 110 is configured to obtain X-ray imaging data of a region of interest (ROI) including the skeletal group. These X-ray imaging data may be obtained from a data source 200. In addition, the data processor 110 is configured to determine a pose estimate of the skeletal group and at least one geometric parameter defined between bones of the skeletal group using the obtained X-ray imaging data to which a 3D pose estimation computer model CM is fed or input, so as to determine soft tissue related properties of the skeletal group.

[0050] Soft tissue related properties may include any soft tissue in, at and / or around the bone group included in the ROI (i.e., the X-ray imaging data (input data)). With respect to the bone group, the soft tissue may include, for example, cartilage, ligaments, tendons, muscles, fascia, etc. The soft tissue may extend or be located at a (small) distance between bones, adjacent bones, or around bones.

[0051] The at least one geometric parameter may include one or more of the following: distances between bones, orthogonal translations and rotations. The at least one geometric parameter is a variable parameter of the 3D pose estimation computer model CM.

[0052] The 3D pose estimation computer model CM may be a computer program element (implemented in the apparatus 100 , stored in a computer memory, etc.) and may be executed by the data processor 110 .

[0053] In addition, the device 100 (e.g., the data processor 110) can be configured to determine soft tissue-related lesions and / or abnormalities based on at least one soft tissue-related property. In order to determine the soft tissue-related property and / or determine the soft tissue-related injury, lesion and / or abnormality, the device 100 can also be configured to compare the geometric parameters and / or posture estimates with corresponding thresholds. The device 100 (e.g., the data processor 110) can also be configured to output at least one of the following items determined: the soft tissue-related property and / or the determined injury, lesion and / or abnormality. The device 100 can also be configured to trigger an alarm to be output based on the determined soft tissue-related lesions and / or abnormalities.

[0054] Optionally, the device 100 may also be configured to interpret and / or visualize soft tissue-related properties in or based on the 3D posture estimation computer model CM. Furthermore, in order to determine the at least one soft tissue-related property, the device 100 may also be configured to determine the axis between two adjacent bones of the bone group and measure the distance along the axis between the two adjacent bones as an example of at least one geometric parameter. Furthermore, in order to determine the at least one soft tissue-related property, the device 100 may also be configured to adjust the corresponding threshold when the implant is present in or associated with the bone group. Optionally, the device 100 may also be configured to combine at least one geometric parameter, at least one geometric parameter and / or soft tissue-related property with patient-related data to determine soft tissue-related lesions and / or abnormalities.

[0055] Figure 2 Figures 2A to 2F2 illustrates exemplary model parameters of an exemplary 3D pose estimation computer model CM utilized by the data processor 110. This is merely an exemplary selection of possible model parameters, wherein not all model parameters must be considered or encoded. It should be noted that at least one geometric parameter as explained above is specific to Figures 2A to 2F The additional free parameters of the model parameters shown in . FIG. 2A to FIG. 2F Each of is an example of a ROI comprising a bone group, which here is a joint, in particular an exemplary ankle joint. The 3D pose estimation computer model CM may also be referred to as a 3D joint model.

[0056] Figure 2A The model parameter is the flexion angle. Figure 2B Model parameters for viewing direction are illustrated, where (1) indicates azimuthal rotation and (2) indicates polar rotation. Figure 2C The model parameters shown are the detector positions. Figure 2D Model parameters of the translator in the detector plane are illustrated. Figure 2E The model parameter is the rotation in the detector plane. Figure 2F Scaled model parameters in the detector plane are illustrated, where scaling may include, for example, indicating changes in joint size or image sampling.

[0057] The 3D pose estimation model CM is trained to regress the model parameters considered or encoded therein. For example, such training of the 3D pose estimation model CM can be performed using artificially generated images using 3D joint shape models and their possible poses. This is not limited in this article, as the training data can also be generated in different ways.

[0058] Figure 3 An example graph is shown that encodes adjacent bones in an example bone group. Figure 3 In the example shown (which should only be used for better illustration), the group of bones included in the ROI is again the ankle joint, which includes the fibula (a), tibia (b), talus (c) and calcaneus (d), which can be shown as a graphic encoding these bones. For example, the at least one geometric parameter includes one or more of the following: the distance d between the bones, the orthogonal translation t and the rotation r. Note again that except FIG. 2A to FIG. 2F In addition to the model parameters shown, each of the geometric parameters of distance d, orthogonal translation and rotation r are model parameters that can be encoded in the 3D pose estimation model.

[0059] With respect to the exemplary ankle joint, as an example of a bone group, the introduction of the talus-tibia distance d as an additional free parameter allows the 3D pose estimation computer model to adapt more accurately to situations with different or unusual distances. Once the model is parameterized and adapted to the image, the talus-tibia distance d encoded in the (one or more) model parameters provides diagnostically valuable information. For example, the distance d can reflect the size of the upper ankle joint gap, where low values ​​can indicate cartilage degradation due to age or osteoarthritis, and high values ​​can indicate a tear of the ligament and / or a foot dislocation.

[0060] To implement a 3D pose estimation computer model CM with additional geometric parameters, by way of example, a tree or graph may be used to encode multiple distances between bones in the 3D pose estimation computer model, where bones are vertices and distances are modeled by edges.

[0061] Furthermore, by way of example, to determine the distance d, the apparatus 100 (eg, the data processor 110) may be configured to construct an axis between two adjacent bones and measure the distance d along the axis.

[0062] Figure 4. The corresponding X-ray images Figure 4A and 4B The geometric parameters used to determine the soft tissue related properties of a bone group from X-ray imaging data are shown in FIG. Figure 4A and Figure 4B , distance d, translation t, and / or rotation r are associated with scaphoid-lunate widening, which can be considered an indicator of degenerative arthritis.

[0063] As described above, the apparatus 100 is configured to utilize the 3D pose estimation computer model CM, which is augmented by including at least one additional free parameter (i.e., at least one geometric parameter, such as distance d, translation t, and / or rotation r) to determine scaphoid-lunate widening as an example of a soft tissue related property. The apparatus 100 can utilize comparisons of the scaphoid-lunate widening to normal values, statistics, etc. to determine injuries, lesions, abnormalities, etc. in the soft tissue, even though the soft tissue is not visible in the X-ray imaging data.

[0064] Figure 5A computer-implemented method for determining soft tissue-related properties of a skeletal group based on X-ray imaging data is illustrated in a flow chart. The method can be performed by the above-mentioned device 100. The method comprises: as a step S1, obtaining X-ray imaging data of a region of interest including a skeletal group; and as a step S2, determining a pose estimate of the skeletal group and relative distances between bones of the skeletal group by utilizing a 3D pose estimation computer model CM to which the X-ray imaging data is fed, the relative distances being assigned to at least one soft tissue-related property of the skeletal group. As described above, at least one soft tissue-related property can be used for further diagnostic purposes, etc.

[0065] According to a further aspect, a computer program element for controlling the X-ray imaging system of the third aspect is provided, the computer program element being adapted to perform the method steps of the fourth aspect when run by a processing unit.

[0066] According to a further aspect, a computer readable medium is provided, having stored therein the computer program element of the fifth aspect.

[0067] Therefore, the computer program element can be stored on a computer unit, which can also be an embodiment of the present invention. The computing unit can be adapted to perform or induce the steps of the above method. In addition, it can be adapted to operate the components of the above device. The computing unit can be adapted to automatically operate and / or execute the user's commands. The computer program can be loaded into the working memory of a data processor. Therefore, the data processor can be equipped to perform the method of the present invention.

[0068] This exemplary embodiment of the invention covers both the computer program with interventions installed from the beginning and the computer program which by an update turns an existing program into a program using the invention.

[0069] The computer program may be stored and / or distributed on a suitable medium, such as an optical storage medium or solid-state medium provided 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. However, the computer program may also exist on a network such as the World Wide Web and can also be downloaded from such a network into the working memory of a data processor.

[0070] According to a further exemplary embodiment of the present invention, a medium for making a computer program element available for downloading is provided, the computer program element being arranged to perform a method according to one of the previously described embodiments of the present invention.

[0071] It should be noted that embodiments of the present invention are described with reference to different subject matters. Specifically, some embodiments are described with reference to method type claims, while other embodiments are described with reference to apparatus type claims. However, it will be apparent to those skilled in the art from the above and following descriptions that, unless otherwise indicated, in addition to any combination of features belonging to one type of subject matter, other combinations between features relating to different subject matters are also considered to be disclosed by the present application.

[0072] All features may be combined to provide synergistic effects that are more than the simple sum of the features.

[0073] While 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.

[0074] Other variations to 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.

[0075] In the claims, the word "comprising" does not exclude other elements or steps, and the word "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. Although certain measures are recited in mutually different dependent claims, this does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. An apparatus (100) for determining soft tissue-related properties of a bone group based on X-ray imaging data, the apparatus comprising a data processor (110), the data processor being configured to: obtaining the X-ray imaging data of the region of interest including the bone group; and The soft tissue-related properties of the skeletal group are determined by determining a pose estimate of the skeletal group using a 3D pose estimation computer model (CM) to which the X-ray imaging data is fed and at least one geometric parameter (d, t, r) defined between the bones of the skeletal group.

2. The device according to claim 1, wherein: The device (100) is also configured to trigger output of at least the soft tissue-related property.

3. The device according to claim 1 or 2, wherein: The apparatus (100) is further configured to interpret and / or visualize the soft tissue related properties in or based on the 3D pose estimation computer model.

4. The device according to any one of the preceding claims, wherein The apparatus (100) is further configured to determine soft tissue related lesions and / or abnormalities based on the at least one soft tissue related property.

5. The device according to claim 4, wherein: The apparatus (100) is further configured to compare the geometric parameters and / or the posture estimate with corresponding thresholds to determine the soft tissue related lesions and / or abnormalities.

6. The device according to claim 5, wherein: The device (100) is also configured to adjust the corresponding threshold when an implant is present in or associated with the bone group.

7. The device according to claim 5 or 6, wherein: The apparatus (100) is further configured to combine the at least one geometric parameter and / or the soft tissue-related property with patient-related data to determine the soft tissue-related lesions and / or abnormalities.

8. The device according to any one of claims 3 to 7, wherein: The device (100) is also configured to trigger an alarm to be output based on the determined soft tissue related lesions and / or abnormalities.

9. The device according to any one of the preceding claims, wherein The 3D pose estimation computer model is parameterized by parameters, wherein the parameters include one or more of the following: the flexion angle of the skeletal group, the viewing direction of the skeletal group, the X-ray detector position relative to the skeletal group, the translation in the X-ray detector plane, the rotation in the X-ray detector plane, the scaling in the X-ray detector plane, and the geometric parameters, wherein the 3D pose estimation computer model is trained to regress the one or more parameters.

10. The device according to any one of the preceding claims, wherein The at least one geometric parameter comprises one or more of: a distance, an orthogonal translation and a rotation between the bones.

11. The device according to any one of the preceding claims, wherein The device (100) is also configured to determine an axis between two adjacent bones of the bone group and measure a distance between the two adjacent bones along the axis.

12. The device according to any one of the preceding claims, wherein The at least one geometric parameter comprises distances between the bones, and a tree or graph is used to encode a plurality of distances between bones in the 3D pose estimation computer model, wherein the bones are vertices and the distances are modeled by edges.

13. The device according to any one of the preceding claims, wherein The at least one geometric parameter is a variable parameter of the 3D pose estimation computer model.

14. A computer-implemented method for determining soft tissue-related properties of a skeletal group from X-ray imaging data, comprising: Obtaining (S1) X-ray imaging data of a region of interest including the bone group; and A pose estimate of the skeletal group and relative distances between bones of the skeletal group are determined (S2) by utilizing a 3D pose estimation computer model to which the X-ray imaging data is fed, the relative distances being assigned to at least one soft tissue-related property of the skeletal group.

15. A computer program element which, when run by a processor, is configured to perform the method according to claim 14.

Images