Method for positioning control of object under study prior to acquisition of projected x-ray image

Abstract

The invention relates to a computer-implemented method (10) for positioning control of an object under study prior to acquisition of a projected X-ray image of a region of interest, comprising:-receiving (11) object information of the object under study in an examination posture,-retrieving (12) an object model relating to the examination posture and the region of interest, -adapting (13) the object model to the subject under study on the basis of the object information,-evaluating (14) the examination posture on the basis of the adapted object model.

Description

Technical Field

[0001] The present invention relates to a computer-implemented method for positioning control of an object under investigation before acquiring a projection X-ray image of a region of interest, a computer-implemented method for creating a set of reference object models for at least one examination posture and at least one region of interest, a medical system, a computer program and a computer-readable medium, which provide an assessment of the examination posture before acquiring a projection X-ray image. Background Art

[0002] In X-ray imaging, projected X-ray images are common for many imaging tasks. In projected X-ray geometry, structures can overlap. X-rays are emitted by an X-ray source, partially absorbed by the tissue and bones of the object under investigation, and the remaining portion of the X-rays is finally registered by an X-ray detector. As output, an X-ray image can be displayed.

[0003] In musculoskeletal (MSK) X-ray examinations, length and angle measurements are often performed to determine orthopedic diseases. The accuracy of the measurements depends on the quality of the acquired images. There are various factors that reduce the image quality. For example, non-optimal patient positioning, especially relative to the X-ray source and the X-ray detector, may lead to overlapping of structures in the projected image and thus to incorrect measurements.

[0004] Radiographers are well trained to ensure the best possible patient positioning so that images can be used for the indications obtained from them. The patient's physical disability must be considered when finding the most appropriate positioning for that individual patient. The level of expertise of radiographers may vary. Therefore, the ability to find the best possible patient position may vary and depend on the radiographer.

[0005] To date, patient positioning is typically determined by a radiologist. There are few automated software solutions for specialized applications (e.g., chest x-rays) that provide feedback to the radiologist after the exam has been acquired. To date, there are no automated solutions that can determine patient positioning before an x-ray is issued. Summary of the invention

[0006] The invention is intended to overcome the problem that the image quality, in particular with regard to patient positioning, depends on the expertise of the radiographer who acquires the examination or, more precisely, the x-ray image.

[0007] It is an object of the present invention to provide a method for assessing patient positioning prior to acquiring an X-ray image by technical means.

[0008] According to the invention, this object is achieved by the features of the method and corresponding computer program product and system of the present application. Advantageous aspects, features and implementations and advantages are described in the embodiments and in the following description.

[0009] Independent of grammatical usage of the term, individuals with either male or female gender identity are included in the term.

[0010] The invention relates to a computer-implemented method for controlling the positioning of an object under investigation before acquiring a projected X-ray image of a region of interest, comprising:

[0011] - receiving object information of the object under investigation in an examination posture,

[0012] - retrieve the object model associated with the examination pose and the region of interest,

[0013] - adapting the object model to the object under investigation based on the object information, and

[0014] - Evaluating the examination posture based on the adapted object model.

[0015] The computer-implemented method according to the invention is used to control or check the positioning of an object under investigation before acquiring a projection X-ray image of a region of interest. Typically, the object under investigation is a patient.

[0016] The method comprises the steps of receiving, retrieving, adapting and evaluating, preferably in a given order. Object information of an object under investigation in an examination posture is received. An object model associated with the examination posture and a region of interest is retrieved, for example, from a database. The object model is adapted to the object under investigation based on the object information. The examination posture is evaluated based on the adapted object model.

[0017] In order to be able to use the method, at least one object model, preferably a plurality of object models is available. The inventors propose to create a plurality of reference object models based on an avatar-based simulation phase. This step may need to be performed only once in each measurement application. The results from this step can be used for the following steps: adapting the object model to the object under investigation, a decision or evaluation based on the matched avatar or the adapted object model, and / or a recommendation of an alternative positioning based on data of an individual patient or an individual object under investigation.

[0018] The purpose of the avatar-based simulation phase is to explore the relationship between the positioning of the patient and its influence on the measurement results when the patient is represented by an avatar or an object model. The results of the simulation phase provide the relationship between the positioning of the object under investigation that can be represented by the object model and the influence on the measurement results, in particular based on acquired X-ray images of the object under investigation positioned in that specific position and / or posture.

[0019] Preferably, different object models are provided. A specific object model can be selected from a plurality of object models. A plurality of object models can include object models of patients of young, old, normal, obese, slim, small, large or medium size. The object under study is preferably a human patient. An anatomical model of the relevant anatomical structure (e.g. based on a CT volume) can be used to refine the object model. In a preferred embodiment, in addition to the anatomical model, a motion model for the object model may also be required. The motion model can provide a model of the real motion of the skeleton. The object model includes at least one landmark in 3D for the corresponding measurement task or inspection, in particular a plurality of relevant landmarks. In a preferred embodiment, the object model must include relevant landmarks in 3D for the corresponding measurement task or inspection. The imaging process models the landmarks in the X-ray geometry by a forward projector. The landmarks of the object model are projected into the imaging plane in the X-ray geometry. The anatomical model and the motion model are used to modify the object model so as to provide a reference object model, wherein the patient position parameter p can be used to characterize the reference object model and / or the object model.

[0020] For a measurement task, which may be indicated by the index j but is omitted here in most formulas for better readability, the relevant patient position parameters p have to be defined. These include the patient pose, the patient orientation relative to the imaging device and the geometry of the imaging system.

[0021] Need to define the reference position p ref :

[0022] p ref =[p1 ref ,p2 ref …] T .

[0023] For this reference position, by applying the forward projection to the position p ref 3D landmarks of the object model at vol To obtain the projected 2D landmark l ref From p ref Let's start by using function m ref =M j (p ref ) to calculate the reference measurement m ref , where M j () describes the measurement process of a specific measurement or examination. The reference position can characterize the posture and / or position of a reference object model, in particular in a standard examination posture. The position can characterize an examination posture that may differ from the standard examination posture.

[0024] Similarly, we obtain the position p covering the parameter space i Measurement of m i . Measure m iTo the reference measurement m ref Deviation

[0025] d i =m ref -m i =M(p ref )-M(p i )=D(p i )

[0026] It can be given as a relative deviation or an absolute deviation. D() depends on p i The function of p i Calculate the predefined p ref Deviation of p i The corresponding value d i Stored in the database. Value d i Can be given as an absolute or relative value.

[0027] The object information is provided by receiving object information of the object under investigation in an examination posture, for example based on information acquired by an external source or sensor or a combination of an X-ray detector and an X-ray source, in particular a previously taken X-ray image acquisition. The object information may be a depth image of the object under investigation. The object information may be determined based on the image of the object under investigation. The object under investigation is preferably positioned according to an examination posture associated with the X-ray system. A specific examination posture is assigned to a specific X-ray examination. The examination posture may describe, for example, a standing, sitting or lying posture. The examination posture may include an indication of an angle or rotation of a body part.

[0028] According to one aspect of the invention, the object information is determined based on an image of the surface of the object under examination. The image of the surface may be acquired by a sensor or an external source. The external source or sensor that records the external appearance of the patient, in particular does not record the interior of the object under investigation, may include:

[0029] -RGB camera device,

[0030] - Depth camera,

[0031] -Structured light camera device,

[0032] - infrared camera, and / or

[0033] -Light Detection and Ranging (LiDAR) camera device.

[0034] The external source or sensor may be located at the medical system, at the imaging device, or at various locations in the examination room.

[0035] The object information may be determined, for example, based on a landmark detection method. The object information may be determined, for example, based on the detection of the outline of the object under investigation. The object information may include parameters related to the size and / or weight of the object under investigation. The size may describe the whole body length, or a measure of a body part or body region. The object information may include the outer shape of the object under examination, in particular the outer shape that can be observed from the position of the X-ray source. The outer shape may be provided with a grid representation including depth information of at least a portion of the object under examination.

[0036] According to one aspect of the invention, object information is determined based on an X-ray image. The X-ray image may be a low-dose X-ray pre-taken image. The low-dose X-ray pre-taken image may show some information about the interior of the object being studied, but has less details than a complete X-ray image. An image may be acquired using an X-ray pre-taken, in which case the image includes information about the internal appearance of the body of the object being examined. The X-ray image may show details in a two-dimensional plane, such as contours, as well as some information about the internal structure of the object being studied. The internal structure may include information about dense objects such as bones, about soft tissue and / or about air volume. The X-ray image provides two-dimensional information displayed by grayscale values.

[0037] By retrieving an object model related to an examination posture and a region of interest, an object model is selected according to the examination posture and the region of interest, and the object model is provided to the system.

[0038] In a next step, the avatar can be matched to the investigated object. This means that the object model can be adapted to the investigated object based on the object information. The object model can be referred to as an avatar. The object model can be adapted or modified based on the object information. The object information can include parameters such as the dimensions of the anatomical structure or the dimensions of the outer shape of the investigated object.

[0039] Data from an external source can be used to adapt the parameters of the avatar to the real patient. The object information can be acquired by an external source such as a camera. The object information may include parameters describing at least the external shape of the object under study (respectively the patient). The object information is used to adapt the object model to the object under study. Preferably, the object information includes a set of parameters describing the object under study and the object model. The parameters may be, for example, the width of the chest of the object under study and the object model, respectively. The object information is used to adapt the object model in such a way that at least the external shape of the object is represented by the adapted object model.

[0040] In order to match an object model to the object under investigation, one has to find the correct surface transformation between the surface of the object model and the object under investigation. Various methods for this task have been proposed in the literature and can be used for this task. The work by Shuai Lin, Yu-Kun Lai, Ralph R. Martin, Shiyao Jin, Zhi-Quan Cheng, Color-aware surface registration, Computers & Graphics, Volume 58, 2016, Pages 31-42, https: / / doi.org / 10.1016 / j.cag.2016.05.007, for example, introduces a new color-aware registration method for rigid and non-rigid registration, in which the input from a depth camera will provide an initial alignment and a deformable registration step will improve it to obtain a surface with high-quality texture. This step is known to have a certain accuracy range due to the individual characteristics of the objects under investigation. However, one can use this step or this method to obtain a statistical estimate with a certain confidence range.

[0041] According to one aspect of the invention, the object model is adapted based on the object information by using a surface transformation function. The surface or external shape of the object model can be adapted to the surface or external shape of the object under investigation. The object information can be used to make the surface of the object model consistent with the surface of the object under investigation at least to a certain extent.

[0042] Surface transformations of human body shapes based on depth images are well known. These transformations involve analysis and manipulation of depth images to extract meaningful information about the shape and structure of the human body. By understanding the surface properties of the human body shape, gesture recognition and pose estimation can be performed based on (depth) images.

[0043] Depth images capture distance information about objects in a scene, typically represented as a two-dimensional matrix of depth values. These depth values ​​provide a measure of the distance between the camera and the surface of the object being viewed. For the human body, depth images can provide valuable insight into the shape, pose, and movement of an individual because they capture the three-dimensional geometry of the human form.

[0044] Surface transformation in this context can involve various techniques for processing and analyzing depth images to extract useful features or parameters. One of the basic tasks is the extraction of human skeleton structure or pose estimation. This process involves detecting key joints such as head, shoulders, elbows, wrists, hips, knees, and ankles and estimating their positions in three-dimensional space.

[0045] Another important surface transformation is shape analysis, which aims to extract shape descriptors and features from depth images. Shape analysis techniques can provide insights into body proportions, body segment lengths, and even detect abnormalities or deviations from normal body structure. The information collected by surface transformation can include parameters that describe the object under study based on object information (e.g., depth images).

[0046] According to one aspect of the invention, the surface transformation function is a rigid registration function or a non-rigid registration function.

[0047] Rigid transformations are transformations that maintain the shape and size of an object. These transformations include translation, rotation, and reflection. Transformation involves moving an object or shape without changing its orientation or size. It is characterized by moving the position of an object in a specific direction. Rotation involves rotating an object or shape around a fixed point (often called the center of rotation). The object maintains its size and shape but is repositioned at a different angle. Reflection is a transformation that involves flipping or mirroring an object or shape on a line or plane. The object appears as a mirror image of itself but retains its shape and size.

[0048] Non-rigid transformations, also called deformations, are transformations that can change the shape and size of an object. Unlike rigid transformations, non-rigid transformations can change the local geometry of an object. These transformations are used to model objects that can undergo elastic deformations, such as the human body. Non-rigid transformations can include scaling, shearing, bending, or warping.

[0049] Scaling is a non-rigid transformation that involves resizing an object or shape. It can increase or decrease the size of an object uniformly or along different dimensions. Shearing is a non-rigid transformation that skews the shape of an object along one or more axes. It distorts an object by displacing points in a specific direction proportional to the distance from a reference line. Bending or warping involves non-linear deformations that change the shape of an object in a localized manner. These transformations are used to model complex deformations in objects or surfaces.

[0050] In a next step, the examination posture is evaluated based on the adapted object model. The decision on positioning can be determined based on the matched avatar or the adapted object model. In this step, the results from the avatar-based simulation phase and the result of adapting the object model to the investigated object based on the object information can be combined. The object model parameters of the adapted object model are used as input for querying the database created by the avatar-based simulation phase and obtaining the statistically most likely measurement deviation. In case the measurement deviation exceeds a predefined threshold, the human operator can be notified about the deviation.

[0051] According to one aspect of the present invention, the object model also involves object information. The object information may include parameters such as age, height, weight or other descriptive parameters of the object under study. The corresponding object model may be described by the same parameters with at least similar values.

[0052] According to one aspect of the invention, the adapted object model is defined by object model parameters. The object model parameters may include a set of parameters describing the object model. The object model parameters may be used for determining the deviation. By querying a database based on the object model parameters, the deviation may be estimated or determined.

[0053] According to one aspect of the invention, a reference object model is selected from a plurality of reference object models based on object model parameters, the reference object model having a specified positioning deviation compared to a standard inspection posture. The reference object model may be stored in a database. Due to the finiteness of the database, the object model parameters may be used to select the reference object model from a plurality of reference object models. The reference object model may be selected based on a comparison of object model parameters describing the adapted object model with corresponding parameters of the reference object model.

[0054] According to one aspect of the invention, an output is generated if the positioning deviation of the selected reference object model exceeds a predefined threshold. If the positioning deviation exceeds the predefined threshold, the user can receive a message. Exceeding the predefined threshold can correspond to a misalignment or poor positioning of the object under investigation, which will result in a poor image for the imaging task or examination. The output can be displayed on a display. The output can be communicated to the user via optical, acoustic or tactile means. The amount of deviation can be provided to the user. Thus, the user can decide whether it is necessary or possible to change the position or posture of the object under investigation, especially with respect to individual circumstances or individual parameters.

[0055] According to one aspect of the invention, the output is a proposal for an improved examination posture, the proposal being determined based on individual parameters of the subject under investigation. According to one aspect of the invention, the individual parameters include individual limitations of mobility or individual limitations due to pain associated with the examination posture. The individual parameters may describe, for example, the amount of pain associated with the movement of a joint or the fixation of a joint, for example by a plaster cast.

[0056] In a preferred embodiment, a recommendation of an alternative positioning is made to the user. When the patient positioning is assessed as suboptimal, in particular when the statistical measurement deviation exceeds a certain threshold, the system can suggest a more suitable positioning of the object under investigation. The suggestion or proposal can also include individual limitations of the patient, for example due to physical disabilities.

[0057] Mathematically, the new proposed patient positioning can be described by the avatar parameters:

[0058] psugg =[p1 sugg ,p2 sugg ,…] T

[0059] For example, p1 sugg refers to the specific flexion of the knee, and p2 sugg Can refer to a specific internal rotation of the knee. The following relationship must be satisfied for the proposed new patient positioning:

[0060] D(p sugg )≤t D

[0061] where D() is the measured deviation depending on the avatar parameters (see the step of adapting the object model to the object under study based on the object information), and t D is the threshold value.

[0062] Some patient positions may not be easy to achieve or may be painful for individual patients, for example, due to physical disabilities. pat (p) is a function describing the cost associated with a particular patient positioning for a particular patient (how this is determined is described below). For positions that are not easy to do, the costs are very high because they are painful. The task is to find p sugg Make

[0063] D(p sugg )≤t D And C pat (p sugg )≤t C

[0064] Among them, t C is the maximum tolerable cost. The task can be reformulated to minimize the following function

[0065] f(p sugg )=D(p sugg )-b*C pat (p sugg )

[0066] Here, a weighting value b is selected between the measurement deviation and the cost.

[0067] There are standard methods for this optimization task. The function C must be determined individually for each patient. pat (p). Therefore, individual information about the patient is necessary to construct the function. For example, individual information can be obtained by using a room camera that records the patient's mobility. C can be determined based on it. pat Other sources of (p) are previous images or patient records. If individual patient information is not available, a statistical mean can be selected from the population.

[0068] In summary, the present invention solves this problem using the following steps in a preferred embodiment:

[0069] - Step 0: An avatar-based simulation phase, which can provide a set of reference object models as output;

[0070] - Step 1: Matching the avatar to the patient, which may include the steps of receiving object information of the investigated object in an examination posture, retrieving an object model associated with the examination posture and the region of interest, and adapting the object model to the investigated object based on the object information,

[0071] - step 2: decision based on the matched avatar, which may include evaluating the check gesture based on the adapted object model,

[0072] - Step 3 (optional): Recommendation of alternative positioning, which may include proposals for improved examination postures determined based on individual parameters of the subject under investigation.

[0073] The present invention provides an automated assessment that is less subjective and more time efficient than the visual determination by a radiographer or user, which is the current standard of care. As a unique feature, the pain of patient positioning for a particular patient is included in determining a more optimal patient positioning. Patient positioning can be determined before X-ray radiation is delivered to the patient, which is an advantage and helps to work according to the ALARA principle to minimize the patient's dose. In cases involving pre-captured images, only a small amount of X-ray radiation dose is required to determine the quality of positioning.

[0074] The invention also relates to a computer-implemented method for creating a set of reference object models for at least one examination posture and at least one region of interest, comprising:

[0075] - providing a plurality of different object models, the plurality of different object models comprising at least one landmark for a corresponding region of interest,

[0076] - adapting a plurality of different object models to at least one examination posture to generate a reference object model,

[0077] - for each reference object model, generating a forward projection of at least one landmark comprised by the reference object model in an X-ray geometry associated with the examination pose and the region of interest,

[0078] - provides multiple different object models in standard examination postures,

[0079] - generating a forward projection of at least one landmark comprised by the object model in the standard examination pose in an X-ray geometry associated with the examination pose and the region of interest, and

[0080] - Determining the positioning deviation by comparing the forward projection of the reference object model with the forward projection of the corresponding object model in a standard examination pose.

[0081] A set of reference object models is created for at least one examination pose and at least one region of interest. The region of interest may be specified by an imaging task. The region of interest typically relates to an anatomical region of the object under investigation, such as a chest, a shoulder, or any other part of the human body.

[0082] A plurality of different object models including at least one landmark for a corresponding region of interest are provided. The plurality of different object models may be stored in a database. The plurality of different object models may be retrieved from the database.

[0083] A plurality of different object models are adapted to at least one inspection posture to generate a reference object model. At least one inspection posture may deviate from a standard inspection posture. The plurality of inspection postures may provide various deviations from the standard inspection posture, wherein the deviations may range from small deviations to strong deviations.

[0084] For each reference object model, a forward projection of at least one landmark included by the reference object model is generated in an X-ray geometry associated with the examination pose and the region of interest.

[0085] Furthermore, a plurality of different object models in standard examination postures are provided. A forward projection of at least one landmark comprised by the object model in the standard examination posture is generated in an X-ray geometry associated with the examination posture and the region of interest.

[0086] The positioning deviation is determined by comparing the forward projection of the reference object model with the forward projection of the corresponding object model in a standard examination pose.

[0087] In order to be able to use the method, at least one object model, preferably a plurality of object models, is provided. The inventors propose to create a plurality of reference object models based on an avatar-based simulation phase. This step may need to be performed only once in each measurement application. The results from this step can be used for the following steps: adapting the object model to the object under investigation, a decision or evaluation based on the matched avatar or the adapted object model, and / or a recommendation of an alternative positioning based on data of an individual patient or an individual object under investigation.

[0088] The purpose of the avatar-based simulation phase is to explore the relationship between the positioning of the patient and its influence on the measurement results when the patient is represented by an avatar or an object model. The results of the simulation phase provide the relationship between the positioning of the object under investigation that can be represented by the object model and the influence on the measurement results, in particular based on acquired X-ray images of the object under investigation positioned in that specific position and / or pose.

[0089] Preferably, different object models are provided. A specific object model can be selected from a plurality of object models. A plurality of object models can include object models of patients of young, old, normal, obese, slim, small, large or medium size. The object under study is preferably a human patient. An anatomical model of the relevant anatomical structure (e.g. based on a CT volume) can be used to refine the object model. In a preferred embodiment, in addition to the anatomical model, a motion model for the object model may also be required. The motion model can provide a model of the real motion of the skeleton. The object model includes at least one landmark in 3D for the corresponding measurement task or inspection, in particular a plurality of relevant landmarks. In a preferred embodiment, the object model must include relevant landmarks in 3D for the corresponding measurement task or inspection. The imaging process models the landmarks in the X-ray geometry by a forward projector. The landmarks of the object model are projected into the imaging plane in the X-ray geometry. The anatomical model and the motion model are used to modify the object model so as to provide a reference object model, wherein the patient position parameter p can be used to characterize the reference object model and / or the object model.

[0090] For a measurement task, which may be indicated by the index j but is omitted here in most formulas for better readability, the relevant patient position parameters p have to be defined. These include the patient pose, the patient orientation relative to the imaging device and the geometry of the imaging system.

[0091] Need to define the reference position p ref :

[0092] p ref =[p1 ref ,p2 ref …] T .

[0093] For this reference position, by applying the forward projection to the position p ref 3D landmarks of the object model at vol To obtain the projected 2D landmark l ref From p ref Let's start by using function m ref =M j (p ref ) to calculate the reference measurement m ref , where M j () describes the measurement process of a specific measurement or examination. The reference position can characterize the posture and / or position of a reference object model, in particular in a standard examination posture. The position can characterize an examination posture that may differ from the standard examination posture.

[0094] Similarly, we obtain the position p covering the parameter space i Measurement of m i. Measure m i To the reference measurement m ref Deviation

[0095] d i =m ref -m i =M(p ref )-M(p i )=D(p i )

[0096] It can be given as a relative deviation or an absolute deviation. D() depends on p i The function of p i Calculate the predefined p ref Deviation of p i The corresponding value d i Stored in the database. Value d i Can be given as an absolute or relative value.

[0097] The present invention also relates to a medical X-ray system, comprising:

[0098] - a first interface configured for receiving object information of an object under investigation in an examination posture,

[0099] - a second interface configured for retrieving an object model associated with the examination posture and the region of interest,

[0100] - a third interface configured for adapting the object model to the object under investigation based on the object information,

[0101] - a fourth interface configured for evaluating the examination gesture based on the adapted object model.

[0102] The invention also relates to a computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method according to the invention.

[0103] The invention also relates to a computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] Examples of embodiments of the invention are explained in more detail below with the aid of the drawings.

[0105] Figure 1 is a schematic representation of a method according to the invention for positioning control of an object under investigation prior to acquiring a projected X-ray image of a region of interest;

[0106] Figure 2is a schematic representation of a method according to the present invention for creating a set of reference object models for at least one examination posture and at least one region of interest; and

[0107] Figure 3 is a schematic representation of a medical system according to the present invention. DETAILED DESCRIPTION

[0108] Figure 1 An exemplary embodiment of a computer-implemented method 10 according to the present invention is shown. This is a method 10 for positioning control of an object under investigation before acquiring a projection X-ray image of a region of interest, comprising:

[0109] - receiving 11 object information of the object under investigation in an examination posture,

[0110] - retrieving 12 an object model associated with the examination posture and the region of interest,

[0111] - adapting 13 the object model to the object under investigation based on the object information,

[0112] - Evaluating 14 the examination posture based on the adapted object model.

[0113] The computer implemented method 10 according to the invention is used to control or check the positioning of an object under investigation before acquiring a projection X-ray image of a region of interest. Typically, the object under investigation is a human patient.

[0114] The method 10 comprises the steps of receiving 11, retrieving 12, adapting 13 and evaluating 14, preferably in the order given. Object information of an object under investigation in an examination posture is received. An object model associated with the examination posture and a region of interest is retrieved, for example, from a database. The object model is adapted to the object under investigation based on the object information. The examination posture is evaluated based on the adapted object model.

[0115] In order to be able to use the method 10, at least one object model is provided, preferably a plurality of object models. The inventors propose to create a plurality of reference object models based on an avatar-based simulation phase. This step may need to be performed only once in each measurement application. The results from this step can be used for the following steps: adapting the object model to the object under investigation, a decision or evaluation based on the matched avatar or the adapted object model, and / or a recommendation of an alternative positioning based on data of an individual patient or an individual object under investigation.

[0116] The purpose of the avatar-based simulation phase is to explore the relationship between the positioning of the patient and its influence on the measurement results when the patient is represented by an avatar or an object model. The results of the simulation phase provide the relationship between the positioning of the object under investigation that can be represented by the object model and the influence on the measurement results, in particular based on acquired X-ray images of the object under investigation positioned in that specific position and / or posture.

[0117] Preferably, different object models are provided. A specific object model can be selected from a plurality of object models. A plurality of object models can include object models of young, old, normal, obese, slim, small, large or medium-sized patients. The object under study is preferably a human patient. An anatomical model of the relevant anatomical structure (e.g. based on a CT volume) can be used to refine the object model. In a preferred embodiment, in addition to the anatomical model, a motion model for the object model may also be required. The motion model can provide a model of the real motion of the skeleton. The object model includes at least one landmark in 3D for the corresponding measurement task or inspection, in particular a plurality of relevant landmarks. In a preferred embodiment, the object model must include relevant landmarks in 3D for the corresponding measurement task or inspection. The imaging process models the landmarks in the X-ray geometry by a forward projector. The landmarks of the object model are projected into the imaging plane in the X-ray geometry. The anatomical model and the motion model are used to modify the object model so as to provide a reference object model, wherein the patient position parameter p can be used to characterize the reference object model and / or the object model.

[0118] For a measurement task, which may be indicated by the index j but is omitted here in most formulas for better readability, the relevant patient position parameters p have to be defined. These include the patient pose, the patient orientation relative to the imaging device and the geometry of the imaging system.

[0119] Need to define the reference position p ref :

[0120] p ref =[p1 ref ,p2 ref …] T .

[0121] For this reference position, by applying the forward projection to the position p ref 3D landmarks of the object model at vol To obtain the projected 2D landmark l ref From p ref Let's start by using function m ref =M j (p ref ) to calculate the reference measurement m ref , where M j() describes the measurement process of a specific measurement or examination. A reference position may characterize a posture and / or a position of a reference object model, in particular in a standard examination posture. The position may characterize an examination posture which may differ from the standard examination posture.

[0122] Similarly, we obtain the position p covering the parameter space i Measurement of m i . Measure m i To the reference measurement m ref Deviation

[0123] d i =m ref -m i =M(p ref )-M(p i )=D(p i )

[0124] It can be given as a relative deviation or an absolute deviation. D() depends on p i The function of p i Calculate the predefined p ref Deviation of p i The corresponding value d i Stored in the database. Value d i Can be given as an absolute or relative value.

[0125] The object information is provided by receiving 11 object information of the object under investigation in an examination posture, for example based on information acquired by an external source or sensor, or an X-ray detector in combination with an X-ray source, in particular a previously taken X-ray image acquisition. The object information may be a depth image of the object under investigation. The object information may be determined based on the image of the object under investigation. The object under investigation is preferably positioned according to an examination posture associated with the X-ray system. A specific examination posture is assigned to a specific X-ray examination. The examination posture may describe, for example, a standing, sitting or lying posture. The examination posture may include an indication of an angle or rotation of a body part.

[0126] In a preferred embodiment, the object information is determined based on an image of the surface of the object under examination. The image of the surface may be acquired by a sensor or an external source. External sources or sensors that record the external appearance of the patient, in particular without recording the interior of the object under investigation, may include:

[0127] -RGB camera device,

[0128] - Depth camera,

[0129] -Structured light camera device,

[0130] - infrared camera, and / or

[0131] -Light Detection and Ranging (LiDAR) camera device.

[0132] The external source or sensor may be located at the medical system, at the imaging device, or at various locations in the examination room.

[0133] The object information may be determined, for example, based on a landmark detection method. The object information may be determined, for example, based on the detection of the outline of the object under investigation. The object information may include parameters related to the size and / or weight of the object under investigation. The size may describe the whole body length, or a measure of a body part or body region. The object information may include the outer shape of the object under examination, in particular the outer shape that can be observed from the position of the X-ray source. The outer shape may be provided with a grid representation including depth information of at least a portion of the object under examination.

[0134] In another embodiment, object information is determined based on an X-ray image. The X-ray image may be a low-dose X-ray pre-taken image. A low-dose X-ray pre-taken image may show some information about the interior of the object being studied, but with less detail than a complete X-ray image. An image may be acquired using an X-ray pre-taken, in which case the image includes information about the internal appearance of the body of the object being examined. The X-ray image may show details in a two-dimensional plane, such as contours, as well as some information about the internal structure of the object being studied. The internal structure may include information about dense objects such as bones, about soft tissue, and / or about air volume. The X-ray image provides two-dimensional information displayed by grayscale values.

[0135] By retrieving 12 an object model associated with the examination posture and the region of interest, the object model is selected according to the examination posture and the region of interest, and the object model is provided to the system.

[0136] In a next step, the avatar can be matched to the investigated object. This means that the object model can be adapted to the investigated object based on the object information. The object model can be referred to as an avatar. The object model can be adapted or modified based on the object information. The object information can include parameters such as the dimensions of the anatomical structure or the dimensions of the outer shape of the investigated object.

[0137] Data from an external source may be used to adapt the parameters of the avatar to the real patient. The object information may be acquired by an external source such as a camera. The object information may include parameters describing at least the external shape of the investigated object (respectively the patient). The object information is used to adapt the object model to the investigated object. Preferably, the object information comprises a set of parameters describing the investigated object and the object model. The parameters may be, for example, the width of the chest of the investigated object and the object model, respectively. The object information is used to adapt the object model in such a way that at least the external shape of the object is represented by the adapted object model.

[0138] In order to match an object model to the object under investigation, one has to find the correct surface transformation between the surface of the object model and the object under investigation. Various methods for this task have been proposed in the literature and can be used for this task. The work by Shuai Lin, Yu-Kun Lai, Ralph R. Martin, Shiyao Jin, Zhi-Quan Cheng, Color-aware surface registration, Computers & Graphics, Volume 58, 2016, Pages 31-42, https: / / doi.org / 10.1016 / j.cag.2016.05.007, for example, introduces a new color-aware registration method for rigid and non-rigid registration, in which the input from a depth camera will provide an initial alignment and a deformable registration step will improve it to obtain a surface with high-quality texture. This step is known to have a certain accuracy range due to the individual characteristics of the objects under investigation. However, one can use this step or this method to obtain a statistical estimate with a certain confidence range.

[0139] In a preferred embodiment, the object model is adapted based on the object information by using a surface transformation function. The surface or outer shape of the object model can be adapted to the surface or outer shape of the object under investigation. The object information can be used to make the surface of the object model consistent with the surface of the object under investigation at least to a certain extent.

[0140] Surface transformations of human body shapes based on depth images are well known. These transformations involve analysis and manipulation of depth images to extract meaningful information about the shape and structure of the human body. By understanding the surface properties of the human body shape, gesture recognition and pose estimation can be performed based on (depth) images.

[0141] Depth images capture distance information about objects in a scene, typically represented as a two-dimensional matrix of depth values. These depth values ​​provide a measure of the distance between the camera and the surface of the object being viewed. For the human body, depth images can provide valuable insight into the shape, posture, and movement of an individual because they capture the three-dimensional geometry of the human body.

[0142] Surface transformation in this context can involve various techniques for processing and analyzing depth images to extract useful features or parameters. One of the basic tasks is the extraction of human skeleton structure or pose estimation. This process involves detecting key joints such as head, shoulders, elbows, wrists, hips, knees, and ankles and estimating their positions in three-dimensional space.

[0143] Another important surface transformation is shape analysis, which aims to extract shape descriptors and features from depth images. Shape analysis techniques can provide insights into body proportions, body segment lengths, and even detect abnormalities or deviations from normal body structure. The information collected by surface transformation can include parameters that describe the object under study based on object information (e.g., depth images).

[0144] In a preferred embodiment, the surface transformation function is a rigid registration function or a non-rigid registration function.

[0145] Rigid transformations are transformations that maintain the shape and size of an object. These transformations include translation, rotation, and reflection. Transformation involves moving an object or shape without changing its orientation or size. It is characterized by moving the position of an object in a specific direction. Rotation involves rotating an object or shape around a fixed point, often called the center of rotation. The object maintains its size and shape but is repositioned at a different angle. Reflection is a transformation that involves flipping or mirroring an object or shape on a line or plane. The object appears as a mirror image of itself but retains its shape and size.

[0146] Non-rigid transformations, also called deformations, are transformations that can change the shape and size of an object. Unlike rigid transformations, non-rigid transformations can change the local geometry of an object. These transformations are used to model objects that can undergo elastic deformations, such as the human body. Non-rigid transformations can include scaling, shearing, bending, or warping.

[0147] Scaling is a non-rigid transformation that involves resizing an object or shape. It can increase or decrease the size of an object uniformly or along different dimensions. Shearing is a non-rigid transformation that skews the shape of an object along one or more axes. It distorts an object by displacing points in a specific direction proportional to the distance from a reference line. Bending or warping involves non-linear deformations that change the shape of an object in a localized manner. These transformations are used to model complex deformations in objects or surfaces.

[0148] In a next step, the examination posture is evaluated based on the adapted object model. The evaluation of the positioning can be determined based on the matched avatar or the adapted object model. In this step, the results from the avatar-based simulation phase and the result of adapting the object model to the investigated object based on the object information can be combined. The object model parameters of the adapted object model are used as input for querying the database created by the avatar-based simulation phase and obtaining the statistically most likely measurement deviation. In case the measurement deviation exceeds a predefined threshold, the human operator can be notified about the deviation.

[0149] In a preferred embodiment, the object model also involves object information. The object information may include parameters such as age, height, weight or other descriptive parameters of the object under study. The corresponding object model may be described by the same parameters with at least similar values.

[0150] In a preferred embodiment, the adapted object model is defined by object model parameters. The object model parameters may include a set of parameters describing the object model. The object model parameters may be used for the determination of the deviation. By querying a database based on the object model parameters, the deviation may be estimated or determined.

[0151] In a preferred embodiment, a reference object model is selected from a plurality of reference object models based on object model parameters, the reference object model having a specified positioning deviation compared to a standard examination posture. The reference object model may be stored in a database. Due to the finiteness of the database, the reference object model may be selected from a plurality of reference object models using object model parameters. The reference object model may be selected based on a comparison of object model parameters describing the adapted object model with corresponding parameters of the reference object model.

[0152] In a preferred embodiment, an output is generated if the positioning deviation of the selected reference object model exceeds a predefined threshold. If the positioning deviation exceeds the predefined threshold, the user can receive a message. Exceeding the predefined threshold can correspond to a misalignment or poor positioning of the object under investigation, which will result in a poor image for the imaging task or examination. The output can be displayed on a display. The output can be communicated to the user via optical, acoustic or tactile means. The amount of deviation can be provided to the user. Thus, the user can decide whether it is necessary or possible to change the position or posture of the object under investigation, especially with respect to individual circumstances or individual parameters.

[0153] According to one aspect of the invention, the output is a proposal 15 for an improved examination posture, the proposal being determined based on individual parameters of the subject under investigation. According to one aspect of the invention, the individual parameters include individual limitations of mobility or individual limitations due to pain associated with the examination posture. The individual parameters may describe, for example, the amount of pain associated with the movement of a joint or the fixation of a joint, for example by a plaster cast.

[0154] In a preferred embodiment, a recommendation of an alternative positioning is made to the user. When the patient positioning is assessed as suboptimal, in particular when the statistical measurement deviation exceeds a certain threshold, the system can suggest a more suitable positioning of the object under investigation. The suggestion or proposal can also include individual limitations of the patient, for example due to physical disabilities.

[0155] Mathematically, the new proposed patient positioning can be described by the avatar parameters:

[0156] p sugg =[p1sugg ,p2 sugg ,…] T

[0157] For example, p1 sugg refers to the specific flexion of the knee, and p2 sugg Can refer to a specific internal rotation of the knee. The following relationship must be satisfied for the proposed new patient positioning:

[0158] D(p sugg )≤t D

[0159] where D() is the measured deviation depending on the avatar parameters (see the step of adapting the object model to the object under study based on the object information), and t D is the threshold value.

[0160] Some patient positions may not be easy to achieve or may be painful for individual patients, for example, due to physical disabilities. pat (p) is a function describing the cost associated with a particular patient positioning for a particular patient (how this is determined is described below). For positions that are not easy to do, the costs are very high because they are painful. The task is to find p sugg Make

[0161] D(p sugg )≤t D And C pat (p sugg )≤t C

[0162] Among them, t C is the maximum tolerable cost. The task can be reformulated to minimize the following function

[0163] f(p sugg )=D(p sugg )-b*C pat (p sugg )

[0164] Here, a weighting value b is selected between the measurement deviation and the cost.

[0165] There are standard methods for this optimization task. The function C must be determined individually for each patient. pat (p). Therefore, individual information about the patient is necessary to construct the function. For example, individual information can be obtained by using a room camera that records the patient's mobility. C can be determined based on it. pat Other sources of (p) are previous images or patient records. If individual patient information is not available, a statistical mean can be selected from the population.

[0166] In step 9, object information is determined based on the image of the surface of the examined object. The object information may be represented by a depth or an X-ray image of the investigated object. The object information may be determined based on the X-ray image.

[0167] In step 13, the object model is adapted based on the object information by using a surface transformation function. The surface transformation function is a rigid registration function or a non-rigid registration function.

[0168] In step 12, the object model also relates to object information, such as age or height. In step 13, the adapted object model is defined by object model parameters.

[0169] In step 14, a reference object model is selected from a plurality of reference object models based on the object model parameters, the reference object model having a specified positioning deviation compared to the standard inspection pose. The object model may be stored in a database.

[0170] If the positioning deviation of the selected reference object model exceeds a predefined threshold, an output is generated.

[0171] In step 15, an output is generated and is a proposal for an improved examination posture, which is determined based on individual parameters of the subject under study. The individual parameters may include parameters regarding pain. The individual parameters include individual limitations of mobility or individual limitations due to pain associated with the examination posture.

[0172] Figure 2 An exemplary embodiment of a computer-implemented method 20 for creating a set of reference object models for at least one examination posture and at least one region of interest is shown, comprising:

[0173] - providing 21 a plurality of different object models, the plurality of different object models comprising at least one landmark for a corresponding region of interest,

[0174] - adapting a plurality of different object models to at least one examination posture to generate a reference object model,

[0175] - for each reference object model, generating 23 a forward projection of at least one landmark comprised by the reference object model in an X-ray geometry associated with the examination pose and the region of interest,

[0176] - Provides 24 different object models in standard examination postures,

[0177] - generating 25 a forward projection of at least one landmark comprised by the object model in the standard examination pose in an X-ray geometry associated with the examination pose and the region of interest, and

[0178] - Determining 26 the positioning deviation by comparing the forward projection of the reference object model with the forward projection of the corresponding object model in a standard examination pose.

[0179] A set of reference object models is created for at least one examination pose and at least one region of interest. The region of interest may be specified by an imaging task. The region of interest typically relates to an anatomical region of the object under investigation, such as a chest, a shoulder, or any other part of the human body.

[0180] In step 21, a plurality of different object models including at least one landmark for a corresponding region of interest are provided. The plurality of different object models may be stored in a database. The plurality of different object models may be retrieved from the database.

[0181] In step 22, a plurality of different object models are adapted to at least one inspection posture to generate a reference object model. At least one inspection posture may deviate from a standard inspection posture. The plurality of inspection postures may provide various deviations from the standard inspection posture, wherein the deviations may range from small deviations to strong deviations.

[0182] In step 23 , for each reference object model, a forward projection of at least one landmark comprised by the reference object model is generated in an X-ray geometry associated with the examination pose and the region of interest.

[0183] Furthermore, a plurality of different object models in standard examination poses are provided in step 24. In step 25, a forward projection of at least one landmark comprised by the object model in the standard examination pose is generated in an X-ray geometry associated with the examination pose and the region of interest.

[0184] In step 26 , the positioning deviation is determined by comparing the forward projection of the reference object model with the forward projection of the corresponding object model in the standard examination pose.

[0185] In order to be able to use the method for positioning control of the object under investigation before acquiring a projection X-ray image of a region of interest, at least one object model, preferably a plurality of object models, is provided. The inventors propose to create a plurality of reference object models based on an avatar-based simulation phase. This step may need to be performed only once in each measurement application. The results from this step can be used for the following steps: adapting the object model to the object under investigation, a decision or evaluation based on the matched avatar or the adapted object model, and / or a recommendation for an alternative positioning based on data of an individual patient or an individual object under investigation.

[0186] The purpose of the avatar-based simulation phase is to explore the relationship between the positioning of the patient and its influence on the measurement results when the patient is represented by an avatar or an object model. The results of the simulation phase provide the relationship between the positioning of the object under investigation that can be represented by the object model and the influence on the measurement results, in particular based on acquired X-ray images of the object under investigation positioned in that specific position and / or pose.

[0187] Preferably, different object models are provided. A specific object model can be selected from a plurality of object models. A plurality of object models can include object models of young, old, normal, obese, slim, small, large or medium-sized patients. The object under study is preferably a human patient. An anatomical model of the relevant anatomical structure (e.g. based on a CT volume) can be used to refine the object model. In a preferred embodiment, in addition to the anatomical model, a motion model for the object model may also be required. The motion model can provide a model of the real motion of the skeleton. The object model includes at least one landmark in 3D for the corresponding measurement task or inspection, in particular a plurality of relevant landmarks. In a preferred embodiment, the object model must include relevant landmarks in 3D for the corresponding measurement task or inspection. The imaging process models the landmarks in the X-ray geometry by a forward projector. The landmarks of the object model are projected into the imaging plane in the X-ray geometry. The anatomical model and the motion model are used to modify the object model so as to provide a reference object model, wherein the patient position parameter p can be used to characterize the reference object model and / or the object model.

[0188] For a measurement task, which may be indicated by the index j but is omitted here in most formulas for better readability, the relevant patient position parameters p have to be defined. These include the patient posture, the patient orientation relative to the imaging device and the geometry of the imaging system.

[0189] Need to limit the reference position pref:

[0190] p ref =[p1 ref ,p2 ref …] T .

[0191] For this reference position, by applying the forward projection to the position p ref 3D landmarks of the object model at vol To obtain the projected 2D landmark l ref From p ref Let's start by using function m ref =M j (p ref ) to calculate the reference measurement m ref , where M j() describes the measurement process of a specific measurement or examination. The reference position can characterize the posture and / or position of a reference object model, in particular in a standard examination posture. The position can characterize an examination posture that may differ from the standard examination posture.

[0192] Similarly, we obtain the position p covering the parameter space i Measurement of m i . Measure m i To the reference measurement m ref Deviation

[0193] d i =m ref -m i =M(p ref )-M(p i )=D(p i )

[0194] It can be given as a relative deviation or an absolute deviation. D() depends on p i The function of p i Calculate the predefined p ref Deviation of p i The corresponding value d i Stored in the database. Value d i Can be given as an absolute or relative value.

[0195] Figure 3 An exemplary embodiment of a medical X-ray system is shown, the medical X-ray system comprising:

[0196] a first interface 311 configured to receive object information of an object under investigation in an examination posture,

[0197] a second interface 312 configured for retrieving an object model associated with the examination gesture and the region of interest,

[0198] - a third interface 313 configured for adapting the object model to the object under investigation based on the object information,

[0199] A fourth interface 314 configured for evaluating the examination gesture based on the adapted object model.

[0200] The interfaces 311 to 314 may be included by the computing unit 35. The computing unit 35 may also include means for a computer-implemented method for creating a set of reference object models for at least one examination posture and at least one region of interest, comprising:

[0201] a providing unit 321 configured to provide a plurality of different object models, the plurality of different object models comprising at least one landmark for a corresponding region of interest,

[0202] an adaptation unit 322 configured for adapting a plurality of different object models to at least one examination gesture to generate a reference object model,

[0203] a generating unit 323 configured to generate, for each reference object model, a forward projection of at least one landmark comprised by the reference object model in an X-ray geometry associated with the examination pose and the region of interest,

[0204] a providing unit 324 configured for providing a plurality of different object models in standard examination postures,

[0205] a generating unit configured to generate a forward projection of at least one landmark comprised by the object model in the standard examination posture in an X-ray geometry associated with the examination posture and the region of interest, and

[0206] A determination unit 326 configured for determining a positioning deviation by comparing a forward projection of the reference object model with a forward projection of a corresponding object model in a standard examination pose.

[0207] In an alternative embodiment, units 321 to 326 are comprised by a single computing unit.

[0208] In a preferred embodiment, the X-ray system 30 is an X-ray imaging system. In alternative embodiments, the X-ray system 30 can be an angiography system, a fluoroscopy system, a mammography system or a computed tomography system. The X-ray system 30 comprises an X-ray source 31 which emits X-rays towards an object 33 under investigation. At least some of the X-rays can be registered by an X-ray detector 32. In a preferred embodiment, the external sensor is a (depth) camera 34. The camera 34, the X-ray source 31 and the X-ray detector 32 are connected to a computing unit 35.

[0209] Although the present invention has been further described in detail through the preferred embodiments, the present invention is not limited to the disclosed examples, and those skilled in the art may derive other variations therefrom without departing from the scope of protection of the present invention.

Claims

1. A computer-implemented method (10) for positioning an object under investigation prior to acquiring a projection X-ray image of a region of interest, comprising: - receiving (11) object information of the object under investigation in an examination posture, - retrieving (12) an object model associated with the examination posture and the region of interest, - adapting (13) the object model to the object under investigation based on the object information, - evaluating (14) the examination posture based on the adapted object model.

2. The method according to claim 1, wherein: The object information is determined based on an image of the surface of the inspected object.

3. The method according to one of the preceding claims, wherein: The object information is determined based on the X-ray image.

4. The method according to one of the preceding claims, wherein: The object model is adapted by using a surface transformation function based on the object information.

5. The method according to claim 4, wherein: The surface transformation function is a rigid registration function or a non-rigid registration function.

6. The method according to one of the preceding claims, wherein: The object model also relates to the object information.

7. The method according to one of the preceding claims, wherein: The adapted object model is defined by the object model parameters.

8. The method according to claim 7, wherein: A reference object model having a specified positioning deviation compared to a standard examination posture is selected from a plurality of reference object models based on the object model parameters.

9. The method according to claim 8, wherein: If the positioning deviation of the selected reference object model exceeds a predefined threshold, an output is generated.

10. The method according to claim 9, wherein: The output is a proposal for an improved examination posture, the proposal being determined based on individual parameters of the subject under investigation.

11. The method according to claim 10, wherein: The individual parameters include individual limitations of mobility or individual limitations due to pain associated with the examination posture.

12. A computer-implemented method (20) for creating a set of reference object models for at least one examination posture and at least one region of interest, comprising: - providing (21) a plurality of different object models, the plurality of different object models comprising at least one landmark for a corresponding region of interest, - adapting (22) the plurality of different object models to the at least one examination gesture to generate a reference object model, - for each reference object model, generating (23) a forward projection of the at least one landmark comprised by the reference object model in an X-ray geometry associated with the examination pose and the region of interest, - providing (24) said plurality of different object models in standard examination postures, - generating (25) a forward projection of the at least one landmark comprised by the object model in a standard examination posture in the X-ray geometry associated with the examination posture and the region of interest, and - determining (26) a positioning deviation by comparing a forward projection of the reference object model with a forward projection of a corresponding object model in the standard examination pose.

13. A medical X-ray system (30), comprising: - a first interface (311) configured to receive object information of an object under investigation in an examination posture, - a second interface (312) configured to retrieve an object model associated with the examination posture and the region of interest, - a third interface (313) configured to adapt the object model to the object under investigation based on the object information, A fourth interface (314) is configured to evaluate the examination gesture based on the adapted object model.

14. A computer program product comprising instructions which, when executed by a computer, cause the computer to perform the method according to one of claims 1 to 11.

15. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to perform the method according to one of claims 1 to 11.

Images