Method and device for optimizing the positioning of a patient for x-ray

Abstract

The invention relates to a method and a device for optimizing the positioning of a patient undergoing X-ray projection imaging. The method comprises the steps of: providing a base image of a body part of a patient positioned for acquiring a projection image; identifying a plurality of central landmarks and a plurality of edge landmarks in the base image from a predefined set of landmarks, where each central landmark is located between at least two edge landmarks, and where equivalents of the landmarks in the patient are located on a plurality of triangles inclined with respect to a projection plane of the base image; determining a distance from at least one center landmark to at least two edge landmarks in the base image; calculating a ratio value of the calculated distance; determining the patient rotation angle from the ratio value and a given mapping function that maps the ratio value to the patient rotation angle; generating output data based on the determined patient rotation angle; and outputting the generated output data. The invention further relates to a medical X-ray system.

Description

Technical Field

[0001] This invention describes a method and apparatus for optimizing the positioning of a patient undergoing X-ray projection imaging, as well as a medical X-ray system. Background Technology

[0002] Automated patient positioning plays a crucial role during X-ray examinations. In fact, good patient positioning ensures diagnostic accuracy and supports image analysis. In cases of poor patient positioning, detailed information on how to achieve proper patient positioning is lacking.

[0003] There are known systems for patient localization examination that comprise two main components. The first component describes rule-based patient localization examination in a general manner. This component includes the steps of: segmenting structures from an image, and evaluating the segmented structures based on rules.

[0004] The second component is an end-to-end approach to quantifying internal patient rotation.

[0005] The first component is transparent (the segmented structure can be displayed to the user) but not quantized (the rotation angle is not quantized), and the second component is quantized (it is trained to regress the rotation angle) but not transparent (because it is an end-to-end method). This can be a problem because transparent quantization is not available. Summary of the Invention

[0006] The object of this invention is to improve known systems and methods, and to provide a method and apparatus for optimizing patient positioning during X-ray projection imaging, as well as a medical X-ray system, to overcome the aforementioned problems. In particular, the object of this invention is to provide a feedback loop for training technicians in X-ray examination setup, thereby improving the examination workflow and reducing the biomedical or clinical workload required to determine whether and how to repeat an X-ray examination.

[0007] This objective is achieved by the methods, apparatus, and medical X-ray systems described in accordance with this disclosure.

[0008] The object of this invention is to combine two components to provide a system that provides quantitative properties (its correction information) and transparency properties.

[0009] The method according to the present invention is used to optimize the positioning of a patient undergoing X-ray projection imaging. The method includes the following steps:

[0010] - Provides a base image of the patient's body parts that are located for obtaining projection images.

[0011] - Identify multiple central and multiple edge landmarks in the underlying image from a predefined set of landmarks, wherein each central landmark is located between at least two edge landmarks, and wherein the equivalents of the landmarks within the patient body are located on multiple triangles tilted relative to the projection plane of the underlying image.

[0012] - Determine the distances from at least one central landmark to at least two edge landmarks in the base image.

[0013] - Calculate the ratio of the calculated distances.

[0014] -The patient rotation angle is determined based on the ratio value and a given mapping function that maps the ratio value to the patient rotation angle.

[0015] - Output data is generated based on the determined patient rotation angle.

[0016] - Output the generated output data.

[0017] This method is used to suggest how to position the patient to obtain X-ray projection images, with the base image being a low-intensity guide image recorded before the main image. It can also be used to correct images or measurements acquired from them. In this case, the main image can also be used as the base image. All measurements taken on this image that would be affected by incorrect positioning can be corrected using the output data, as this is based on the (correct) patient rotation angle. If the base image (the acquired image) passes the positioning check, it can be used as the main image that can be examined. If not, additional images can be acquired until the patient is correctly positioned.

[0018] The patient rotation angle is an angle in a predefined coordinate system (“reference frame”). This reference frame is preferably a Cartesian coordinate system and should be related to the coordinate system of the imaging system. Particularly preferably, the X and Y axes are in a horizontal plane, and the Z axis is arranged vertically. In this reference frame, the patient rotation angle reflects the arrangement of the patient’s coordinate system (“patient system”) relative to the reference frame. In a preferred patient system, the X-axis is the longitudinal axis, the Y-axis is the frontal axis, and the Z-axis is the sagittal axis. For many examinations, a single angle (e.g., according to the Z-axis) may be sufficient as the patient rotation angle. However, typically, rotation can occur around any of the three spatial axes. This means that the patient rotation angle can have one entry (an angle relative to one axis), two entries (two angles, each relative to a different axis), or three entries (an angle relative to the X-axis, an angle relative to the Y-axis, and an angle relative to the Z-axis). Regarding specific axes: if the patient is positioned such that the axis of the patient system is parallel to the corresponding axis of the reference frame, the corresponding entry for the patient rotation angle is 0°. Where the patient system should be parallel to the reference frame, the patient rotation angle is a measure of patient deviation. In cases where the correct position of the patient will include a given positional angle of the patient system's axis relative to the reference frame (e.g., the patient is lying on their side with the sagittal axis at 90° relative to the Z-axis (positional angle)), the difference between this positional angle (e.g., 89° in this case) and the patient's rotation angle will be a measure of the patient's deviation (1° in this example).

[0019] First, a baseline image of the patient's body part positioned for acquiring projection images must be provided. This may include acquiring the baseline image or retrieving it from a database (e.g., via PACS). The baseline image can be a (low-intensity) guiding image or the master image of the examination. The baseline image shows the patient positioned in the location to be examined (or being examined) so that the patient's rotation angle can be determined during the imaging process.

[0020] In this base image, at least three landmarks are identified. For example, for a chest examination, points on the clavicle, especially the inner point, and points on the spine can be three landmarks. Since the region of interest is given, it is easy to provide a predefined set of landmarks from which to select landmarks.

[0021] One of the three landmarks is the central landmark, and the other two are edge landmarks. In cases with more than three landmarks, additional central and / or edge landmarks may exist. However, regardless of the number of landmarks identified, each central landmark lies between at least two edge landmarks. This means that, in practice, a central landmark lies between at least two edge landmarks.

[0022] Consider three landmarks: one central landmark and two edge landmarks, located within a triangle (in 3D space) within the patient's body. This triangle is tilted relative to the 2D projection plane of the underlying image. This means that, when the underlying image was acquired, the central landmark within the patient's body lies either above or below the line between the two edge landmarks. Therefore, as the patient rotates, the central landmark on the projection plane (in a series of underlying images) appears to move relative to the edge landmarks. Thus, the position of the central landmark relative to the edge landmarks provides a measure of the patient's rotation angle.

[0023] To estimate the patient rotation angle around one axis, two edge landmarks are sufficient; for all three axes, one central landmark and three edge landmarks are sufficient. The three edge landmarks should form a triangle on the projection plane (i.e., in the base image). The patient rotation angle around the Z-axis can be derived from the rotation of the three edge landmarks, and the patient rotation angles around the X and Y axes can be derived from the position of the central landmark relative to the three edge landmarks. The central landmark and the three edge landmarks together form three triangles, thus forming a tetrahedron.

[0024] It should be noted that, theoretically, when the triangle is well-known, the patient rotation angle can be directly calculated from the positions of the central and edge landmarks, as the ratio of the distances from the central to the edge landmarks would be a direct measure of the patient rotation angle. However, in reality, the triangle is not always known. Furthermore, symmetry cannot be used in all cases (e.g., for imaging the chest) because the projection of the patient's body will be distorted with the patient's rotation angle.

[0025] Therefore, after calculating the distances from the center landmark to at least two edge landmarks in the base image and determining the ratio of the calculated distances, this ratio is not directly used as a measure of the patient's rotation angle. Instead, a given mapping function is used to determine the patient's rotation angle based on the ratio. This mapping function maps the ratio to the patient's rotation angle. It should be noted that the expression "ratio" indicates that the value is based on a ratio of distances. It can be calculated, in particular, by difference and / or division.

[0026] The mapping function can be created by a machine learning model trained using a base image of the patient and a given patient rotation angle as baseline ground truths. Preferably, the base image is a synthetic image of a 3D model of the patient lying on a surface in many different poses at a given patient rotation angle. Therefore, a preferred method for training is to generate synthetic X-ray images (training images) of the patient rotation angle, then, knowing the true patient rotation angle in each training image, calculate the ratio values, and finally find the accurate mapping function by fitting a polynomial, for example, via value pairs, to a series of ratio values ​​for both parameters and the corresponding patient rotation angles.

[0027] The determined patient rotation angle can be used to estimate whether the patient is properly positioned. For example, the optimal positioning is when the patient's chest should be examined and the patient is accurately supine with a 0° patient rotation angle. If the patient rotation angle starts from 0°, a rearrangement can be performed.

[0028] Therefore, output data is generated based on the determined patient rotation angle. This output data can be a message to the user and can include only the patient rotation angle. The output can also be used for automated placement and can include commands designed to control the automated positioning mechanism.

[0029] The generated output data is then output. It can be displayed on a monitor or sent to the automatic positioning mechanism.

[0030] This method offers a transparent and simple approach, providing transparent and quantitative results using only one component. Even when using AI, the method's clear steps—and especially the use of the mapping function—always provide insight into determining the patient's rotation angle.

[0031] The device according to the invention is used to optimize the positioning of a patient undergoing X-ray projection imaging. The device includes the following components:

[0032] - A data interface designed to receive baseline images of body parts of a patient located for acquiring projection images.

[0033] - A landmark unit designed to identify multiple central and multiple edge landmarks in a base image from a predefined set of landmarks, wherein each central landmark is located between at least two edge landmarks, and wherein the equivalent of a landmark within the patient body lies on a triangle tilted relative to the projection plane of the base image.

[0034] - Distance cell, which is designed to determine the distance from the center landmark to at least two edge landmarks in the base image.

[0035] - Ratio unit, which is designed to calculate the ratio value of the calculated distance.

[0036] - A correction unit designed to determine the patient rotation angle based on a ratio value and a given mapping function that maps the ratio value to the patient rotation angle.

[0037] - Output unit, which is designed to generate output based on the determined patient rotation angle.

[0038] - Data interface, which is designed to output the generated output.

[0039] In practice, considering a certain type of image view, such as the frontal view, anatomical structures will be positioned in corresponding locations as defined by clinical guidelines. Biomedical technicians and clinicians typically adhere to these guidelines during examinations.

[0040] In the example use case of a chest X-ray, there are four criteria for a well-defined chest X-ray. One of these criteria is that the spinous processes of the thoracic vertebrae should form a vertical line equidistant from the medial end of the clavicle. This distance ratio (ratio value ρ) can be calculated based on the shortest distance A from the medial end of the left clavicle to the spine and the shortest distance B from the medial end of the right clavicle to the spine as: ρ = (BA) / (|B| + |A|).

[0041] At least for symmetrical positioning, the ratio value is preferably chosen such that ρ = 0 when A = B and the range is from -1 to 1. More preferably, the ratio value (ρ) is a dimensionless quantity without a definite physical meaning.

[0042] To calculate the patient rotation angle α, a mapping function is needed. To create this mapping function, it is preferable to calculate the rotation angle α relative to several different patients from a synthetic X-ray image derived from the CT volume. G The training images, where α G These were then used as the baseline ground truth. In these training images, relevant anatomical structures (e.g., clavicle, spine) were automatically segmented (segmentation in 3D and forward projection to 2D). Based on these structures, the ratio value ρ was calculated. For CT volume and α... G Repeat the process for many different combinations (which can be greater than 1000). From all samples, fit a mapping function (e.g., using a third-order polynomial). Furthermore, confidence intervals can be created.

[0043] This gives the function ρ = f(α), which can be inverted to give the mapping function α = f -1 (ρ). Based on the α value of the mapping (e.g., through thresholding), the category (e.g., good / bad positioning) can be determined.

[0044] This method can be applied to all body parts susceptible to localization problems. This could be the chest, hips, knees, femur, or head. In this case, synthetic data can be created based on the body part of interest, and later, a ratio value ρ can be defined as the ratio of distances between given landmarks based on clear measurements in the image, and ultimately, a mapping function can be found between ρ and α.

[0045] The device is preferably designed to perform the method according to the invention. The functions of the components of the device have been described above.

[0046] The medical X-ray system according to the invention includes the device according to the invention and / or is designed to perform the method according to the invention.

[0047] Some of the units or modules of the present invention mentioned above can be implemented, wholly or partially, as software modules running on the processor of a computing system. The advantage of implementing them primarily as software modules is that applications already installed on existing computing systems can be updated with relatively little effort to install and run these units of this application. The object of the present invention is also achieved by a computer program product having a computer program that can be directly loaded into the memory of a computing system, and the computer program product including program units for performing the steps of the method when the program is executed by the computing system, at least those steps that can be executed by a computer. In addition to the computer program, such a computer program product may also include additional parts such as documentation and / or additional components, and hardware components such as hardware keys (dongles, etc.) to facilitate access to the software.

[0048] Computer-readable media, such as memory sticks, hard drives, or other removable or permanently mounted carriers, can be used to transmit and / or store executable portions of a computer program product, making these portions readable from the processor unit of a computing system. The processor unit may include one or more microprocessors or equivalents.

[0049] Particularly advantageous embodiments and features of the invention are provided by the technical solutions of this disclosure, as disclosed in the following description. Features of different claim categories may be suitably combined to provide other embodiments not described herein.

[0050] According to the preferred method, the patient's body parts, particularly organs and / or bones, are segmented in a base image using a machine learning model trained for segmentation. Techniques for image segmentation are well known in the art. Finding landmarks is highly advantageous for this invention.

[0051] According to the preferred method, the distance from the center landmark to the first edge landmark is A, and the distance from the center landmark to the second edge landmark is B. The ratio value R is calculated by normalizing the difference between A and B using the sum of A and B, so that the ratio value R lies between -1 and 1 (R = (AB) / (A+B) or R = (BA) / (A+B)). It should be noted that this calculation of the ratio value is advantageous for symmetrical cases. For asymmetrical cases, other formulas can be used to calculate the ratio value. Since a mapping function is used, the formula for calculating the ratio value is not important, but it is crucial to calculate the ratio value using the same formula from the underlying image that has already been used to determine the ratio value for the mapping function.

[0052] According to the preferred method, the patient rotation angle is the aberration angle between the patient and the desired rotation angle. This is simply a matter of choosing a suitable coordinate system for the patient rotation angle. For example, in a symmetrical case, when the patient is correctly positioned, the coordinate system should be chosen such that the patient rotation angle is 0°. Instead of choosing a coordinate system, a mapping function can be designed in an appropriate manner. For example, when 10° would be the correct angle, the mapping function can be designed such that subtracting 10° from the patient rotation angle results in 0° again indicating correct positioning.

[0053] According to a preferred method, a machine learning model is used to determine the patient rotation angle. This machine learning model is designed to output the patient rotation angle based on a ratio value as input. The machine learning model has been trained with a set of ratio values ​​as input, each with a given true patient rotation angle as a reference true value. As described above, preferably, for training, multiple synthetic X-ray images (especially CT images) of at least one body part of the patient projected at various given rotation angles are generated. For each synthetic X-ray image, landmarks are detected, a ratio value is calculated based on the distance between the landmarks, and the patient rotation angle is given as a reference true value based on the given rotation angle of the projection onto the corresponding synthetic X-ray image.

[0054] According to a preferred method, for multiple specific ratio values, a matching patient rotation angle is calculated, and a matching function is fitted to the specific values. Preferably, the matching function is a polynomial of order greater than two, and more preferably greater than three. In particular, preferably, the inverse function of the matching function is calculated as a mapping function to determine the patient rotation angle based on the ratio values.

[0055] According to the preferred method, the patient's body part is the chest, hips, knees, shoulders, feet, femur, or head.

[0056] According to a preferred method, the patient rotation angle is calculated using a given mapping function that provides the desired patient rotation angle value based on a certain ratio. Preferably, the mapping function monotonically increases or monotonically decreases and passes through the origin.

[0057] A preferred method includes the following steps:

[0058] - Provide a basic image of the patient's chest.

[0059] - Identify at least one landmark on each clavicle and one landmark on the spine in the baseline image.

[0060] - Calculate the distance A from the landmark on the spine to the landmark on the right clavicle and the distance B from the landmark on the spine to the landmark on the left clavicle in the base image.

[0061] - Preferably, the ratio value R of the calculated distance is calculated using the formula R = (AB) / (A+B) or R = (BA) / (A+B).

[0062] - Preferably, the patient's rotation angle is determined based on a ratio value using a given mapping function.

[0063] - Generate output based on the determined patient rotation angle.

[0064] - Output the generated output.

[0065] A preferred device includes a function unit designed to generate a mapping function from multiple calculations of the patient's rotation angle based on an input ratio value. Preferably, the device, and particularly the function unit, includes a machine learning model trained to calculate the patient's rotation angle based on the input ratio value.

[0066] The use of AI-based methods (AI: "artificial intelligence") is preferred for the methods according to the invention, particularly for image segmentation, landmark finding, or mapping function creation. Artificial intelligence is based on machine-based learning principles and typically uses adaptive algorithms that have been trained accordingly. The term "machine learning" is generally used for machine-based learning, but it also includes the principles of "deep learning."

[0067] The method may also include elements of "cloud computing." In the field of "cloud computing," IT infrastructure is provided through data networks, storage space or processing power and / or application software. Communication between the user and the "cloud" is achieved through data interfaces and / or data transmission protocols. In the context of "cloud computing," in a preferred embodiment of the method according to the invention, data is provided to the "cloud" via a data channel (e.g., a data network). The "cloud" includes (remote) computing systems, such as computer clusters that typically do not include the user's local machine. Particularly preferred is that the cloud service provides computing power comparable to that of the application software. Attached Figure Description

[0068] Other objects and features of the invention will become apparent from the following detailed description taken in conjunction with the accompanying drawings. However, it should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0069] Figure 1 An X-ray system having an apparatus according to the invention is shown;

[0070] Figure 2 A block diagram of the method according to the present invention is shown;

[0071] Figure 3 The segmentation process and landmark identification are shown;

[0072] Figure 4 The segmentation using a trained segmentation model is shown; and

[0073] Figure 5 The training of the machine learning model that generates the mapping function is shown. Detailed Implementation

[0074] Figure 1 A schematic diagram of an X-ray system 1 with a control device 2 is shown. It is equipped with a device 6 designed to perform the method according to the invention. The X-ray system 1 typically has an X-ray source 3, and while recording a projected image, it irradiates the patient P with a beam collimated by a collimator 5, such that the radiation falls on a detector unit 4 opposite to the X-ray source 3.

[0075] In the control device 2, only the components necessary for explaining the invention are shown. In principle, the X-ray system 1 and the associated control device 2 are known to those skilled in the art and therefore do not require detailed description.

[0076] The control device 2 includes a device 6 for optimizing the positioning of the patient during X-ray projection imaging. It includes a data interface 7, a landmark unit 8, a distance unit 9, a ratio unit 10, a correction unit 11 (with a function unit 13), and an output unit 12.

[0077] Data interface 7 is designed to receive a base image BI of a body part of patient P that is positioned to acquire a projected image, and data interface 7 is designed to output the generated output at the end.

[0078] The landmark unit 8 is designed to identify multiple central landmarks LC and multiple edge landmarks LE in the underlying image BI from a predefined set of landmarks, wherein each central landmark LC is located between at least two edge landmarks LE, and wherein the equivalent of the landmark in the patient P is located on a triangle tilted relative to the projection plane of the underlying image BI.

[0079] Distance cell 9 is designed to calculate the distances D2, D1 from the center marker LC to at least two edge markers LE in the underlying image BI.

[0080] The ratio unit 10 is designed to calculate the ratio value ρ of the calculated distances D2 and D1.

[0081] The correction unit 11 is designed to determine the patient rotation angle α based on the ratio value and a given mapping function F that maps the ratio value ρ to the patient rotation angle α.

[0082] The output unit 12 is designed to generate an output based on the determined patient rotation angle α.

[0083] Figure 2 A block diagram is shown for a method to optimize the positioning of a patient undergoing X-ray projection imaging.

[0084] First (left), a base image BI of the body part of patient P is provided for positioning to obtain a projection image, for example, a base image BI that is recorded or downloaded from a database.

[0085] In step I, the base image BI is segmented, and landmarks are identified from a predefined set of landmarks. Here, a central landmark LC on the spine and two edge landmarks LE on the clavicle are identified in the base image BI.

[0086] In step II, the distances D2 and D1 from the center landmark LC to the two edge landmarks LE are determined in the base image BI.

[0087] In step III, the ratio ρ of the calculated distances D2 and D1 is calculated.

[0088] In step IV, the patient rotation angle α is determined based on the ratio value ρ and a given mapping function F that maps the ratio value ρ to the patient rotation angle α.

[0089] In step V, output data O is generated based on the determined patient rotation angle α. This can be a message for the user or a command designed to control the automatic positioning mechanism. The generated output data O is then output.

[0090] Figure 3 The segmentation process and landmark identification are illustrated. Typically, this could be... Figure 2 Step 1. First, a base image BI showing body part B is provided; in this example, body part B is the chest. Then, the image information is segmented into three segments S: the left and right clavicles and the spine. After this (right), the landmarks LC and LE are identified, which are points on the spine (center landmark LC) and the medial ends of the left and right clavicles (edge ​​landmarks LE).

[0091] Figure 4 The segmentation using the trained segmentation model SM is shown. Wherein Figure 3 The base image is shown. Figure 4 The process is illustrated. First, the base image BI is input into the segmentation model SM, which has been trained using training data T including real or synthetic images and segmented skeletons. The segmentation model SM segments the base image BI. Then, the shape of the segment S is examined and markers LE and LC are added. This can be accomplished using another trained machine learning model or through conventional algorithms.

[0092] Figure 5The training of a machine learning model M that generates the mapping function F is illustrated. Training data T of segmented images (preferably with landmarks LC, LE) and a given patient rotation angle α are fed as ground truth values ​​into the machine learning model, which calculates a ratio value ρ and creates pairs of ratio values ​​ρ and their corresponding patient rotation angles α. These pairs are then collected in a graph, and a polynomial function is fitted to these values ​​(right). This function, or its inverse (depending on the fit), is the mapping function F. In the example shown, the fitted function is the mapping function F.

[0093] Although the present invention has been disclosed in the form of preferred embodiments and variations thereof, it should be understood that many additional modifications and variations can be made thereto without departing from the scope of the invention. For clarity, it should be understood that the use of "a" or "an" throughout this application does not exclude a plurality, and "comprising" does not exclude other steps or elements. The expression "a plurality" means "at least one". References to "unit" or "device" do not exclude the use of more than one unit or device.

Claims

1. A method for optimizing positioning of a patient (P) for X-ray projection imaging, comprising the steps of: - providing a base image (BI) of a body part of a patient (P) positioned for acquiring a projection image, - identifying a plurality of center landmarks (LC) and a plurality of edge landmarks (LE) in the base image (BI) from a predefined set of landmarks, wherein each center landmark (LC) is located between at least two edge landmarks (LE), and wherein equivalents of the landmarks in the patient (P) are located on a plurality of triangles that are tilted with respect to a projection plane of the base image (BI), - determining distances (D1, D2) of at least one center landmark (LC) in the base image (BI) to the at least two edge landmarks (LE), - calculating a ratio value (p) of the calculated distances (D1, D2), - determining a patient rotation angle (a) from the ratio value (p) and a given mapping function (F) that maps ratio values (p) to patient rotation angles (a), - generating output data (O) based on the determined patient rotation angle (a), - outputting the generated output data (O).

2. The method of claim 1, wherein, The body part of the patient (P), in particular organs and / or bones, is segmented (S) in the base image (BI), preferably by using a machine learning model (M) trained for segmentation (S).

3. The method according to one of the preceding claims, wherein, The distance (D1, D2) of the center landmark (LC) to a first edge landmark (LE) is A, the distance (D1, D2) of the center landmark (LC) to a second edge landmark (LE) is B, and the ratio value (p) R is calculated by normalizing the difference of A and B to the sum of A and B.

4. The method according to one of the preceding claims, wherein, The patient rotation angle (a) is an aberration angle of the patient (P) with respect to a desired rotation angle.

5. The method according to one of the preceding claims, wherein, The patient rotation angle (a) is determined with a machine learning model (M) designed to output a patient rotation angle (a) based on an input of a ratio value (p), the machine learning model having been trained by inputting a set of ratio values (p) with given true patient rotation angles (a) as ground truth values, Preferably, wherein for training, a plurality of synthetic X-ray images of at least one body part (B) of a patient (P) projected at various given rotation angles is generated, wherein for each synthetic X-ray image, landmarks are detected, a ratio value (p) is calculated from distances (D1, D2) between landmarks (LC, LE), and a patient rotation angle (a) is given as ground truth value based on the given rotation angle of the projection on the respective synthetic X-ray image.

6. The method according to one of the preceding claims, wherein, For a plurality of specific ratio values (p), a matching patient rotation angle (a) is calculated, and a matching function is fitted to the specific values, Preferably, wherein the matching function is a polynomial of more than second order, preferably of more than third order, Particularly preferably, wherein an inverse of the matching function is calculated as the mapping function (F) for determining a patient rotation angle (a) from a ratio value (p).

7. The method according to one of the preceding claims, wherein, The body part (B) of the patient (P) is the chest, the hip, the knee, the shoulder, the foot, the femur or the head.

8. The method according to one of the preceding claims, wherein, The patient rotation angle (a) is calculated using a given mapping function (F) which provides a value of the desired patient rotation angle (a) depending on a ratio value (p), Preferably, wherein the mapping function (F) is monotonically increasing or monotonically decreasing and passes through the origin.

9. The method according to one of the preceding claims, comprising the steps of: - providing a base image (BI) of the chest of the patient (P), - identifying in the base image (BI) at least a landmark on each clavicle and one landmark on the spine, - calculating a distance (D1, D2)A of the landmark on the spine to the landmark on the right clavicle and a distance (D1, D2)B of the landmark on the spine to the landmark on the left clavicle in the base image (BI), - calculating a ratio value (p)R of the calculated distances (D1, D2) preferably using the formula R = (A - B) / (A + B) or R = (B - A) / (A + B), - determining a patient rotation angle (a) from the ratio value preferably by using a given mapping function (F), - generating an output based on the determined patient rotation angle (a), - outputting the generated output.

10. An apparatus (6) for optimizing positioning of a patient (P) for X-ray projection imaging, comprising: - a data interface (7) designed for receiving a base image (BI) of a body part of a patient (P) positioned for acquiring a projection image, - a landmark unit (8) designed for identifying a plurality of central landmarks (LC) and a plurality of edge landmarks (LE) in the base image (BI) from a predefined set of landmarks, wherein each central landmark (LC) is located between at least two edge landmarks (LE), and wherein equivalents of the landmarks within the patient (P) are located on a plurality of triangles which are tilted with respect to a projection plane of the base image (BI), - a distance unit (D1, D2) (9) designed for determining distances (D1, D2) of the central landmarks (LC) to the at least two edge landmarks (LE) in the base image (BI), - a ratio unit (10) designed for calculating a ratio value (p) of the calculated distances (D1, D2), - a correction unit (11) designed for determining a patient rotation angle (a) from the ratio value (p) and a given mapping function (F) which maps ratio values (p) to patient rotation angles (a), - an output unit (12) designed for generating an output based on the determined patient rotation angle (a), - a data interface (7) designed for outputting the generated output.

11. The apparatus (6) of claim 10, comprising a function unit (13) designed for generating a mapping function (F) from a plurality of calculations of the patient rotation angle (a) depending on an input ratio value (p), Preferably, wherein, the apparatus (6), in particular the function unit, comprises a machine learning model (M) trained for calculating the patient rotation angle (a) depending on an input ratio value (p).

12. A medical X-ray system (1) comprising the apparatus (6) of claim 10 or 11 and / or designed for performing the method of one of claims 1 to 9.

13. A computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method according to any one of claims 1 to 9.

14. A computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the steps of the method according to any one of claims 1 to 9.

Images