Self-calibration process of digital breast tomosynthesis imaging device

Through data-driven motion estimation algorithm and iterative correction geometric transformation, the problem of patient movement and mechanical inaccuracy in finite angle digital breast tomography is solved, the image quality is improved and the radiation dose is reduced, and the possibility of three-dimensional tomography imaging is realized sex.

CN112804945BActive Publication Date: 2024-11-26PLANMED OY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201980066204.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2018-09-14
Filing Date
2019-09-16
Publication Date
2024-11-26
Estimated Expiration
2039-09-16

AI Technical Summary

Technical Problem

In finite angle digital breast tomography, patient movement leads to geometric inconsistency of X-ray projected images, reducing the accuracy of the reconstruction image, and the prior art is difficult to effectively correct mechanical inaccuracy and patient movement, affecting image quality and radiation Dosage control.

Method used

Using data-driven motion estimation algorithm, the relative positions of the X-ray source and detector are retrospectively estimated based on the content of the X-ray projected image, and the projected geometry shape is optimized through iterative correction geometric transformation to achieve geometric consistency of image reconstruction, and Correct the slight movement of the patient without relying on the landmark component.

Benefits of technology

Improve image clarity, detail level and contrast, reduce rescans due to degraded image quality, reduce radiation dose to patients, and enable existing digital mammogram devices to achieve three-dimensional tomography.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112804945B_ABST
    Figure CN112804945B_ABST
Patent Text Reader

Abstract

The present invention relates to a calibration process particularly suitable for limited angle digital breast tomosynthesis (DBT). The method comprises acquiring a set of X-ray projection exposure images, forming an initial estimate of projection geometry corresponding to each X-ray projection exposure image, computing intermediate DBT reconstructions, establishing a set of rigid transformation parameters corresponding to the calibration results applied to the initial projection geometry estimate of each X-ray projection exposure image, and computing a final DBT reconstruction using the set of X-ray projection exposure images and the final calibrated estimate of projection geometry.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a method for self-calibration of projection geometry and correction of patient motion in the field of limited-angle digital breast tomosynthesis. Background Art

[0002] The following disclosure relates to X-ray digital breast tomosynthesis, in which multiple X-ray projection images acquired from limited tomosynthesis angles are used to reconstruct cross-sectional images of a patient's anatomy. The duration of X-ray projection image acquisition in limited angle tomosynthesis is typically about 10 seconds while the X-ray tube (X-ray source) and sensor (X-ray detector) physically travel a spatial trajectory corresponding to the limited acquisition angle.

[0003] It should be known that the imaging trajectory has sufficient accuracy and the imaged object should remain sufficiently still during the X-ray projection image acquisition so that the reconstructed image is clear and consistent with the human anatomy, because the measurements of the projection images are assumed to represent a co-registered integrated view of the stationary object. This results in a set of geometrically consistent measurements that can be used to reconstruct an attenuation distribution representing the anatomy under investigation. Whenever these assumptions are violated, the accuracy of the reconstructed image decreases because the projection measurements become inconsistent with each other.

[0004] The accuracy of the final image generally depends on how accurately the assumptions made during the reconstruction process correspond to the actual physical image acquisition process. The estimated spatial positions of the X-ray source and detector corresponding to each acquired X-ray image affect the calculation of the ray paths during the reconstruction process. Due to inherent manufacturing and operating tolerances, potential deformations of the imaging device, and the patient leaning against the imaging device, the achieved rotation angles and positions often deviate from the ideal values ​​according to the assumed form of the imaging trajectory. However, systematic deviations can be accounted for by using different calibration methods, which are repeated after a certain period of time or operation cycle.

[0005] In general, the most important source of geometric inaccuracy is the potential movement of the patient during the acquisition of the X-ray projection images. In digital breast tomosynthesis, this potential movement is mitigated to some extent because the imaged anatomy (breast) is usually physically compressed during imaging. However, if the imaged object moves during the acquisition of the X-ray projection images, the effective spatial paths of the ray measurements become inconsistent with each other. Although it is well known that the patient should not move and is conventionally instructed not to do so, the patient usually cannot remain completely still during the X-ray projection acquisition. For example, the patient may be startled by sudden movements or sounds made by the imaging equipment.

[0006] In all practical measurements, the above-mentioned inaccuracies are usually present to some extent. In the worst case, the resulting geometric inconsistencies of the projection image measurements may even require a repeat scan after the radiologist has checked the quality of the image reconstruction. This is undesirable, since the radiation dose associated with X-ray image acquisition is intended to be kept as low as possible.

[0007] Mammography is an X-ray imaging method widely used for breast cancer screening. In screening studies, mammography has been reported to achieve a sensitivity of 90-93% and a specificity of 90-97%. Mammography is estimated to reduce breast cancer mortality by 35% in women over 50 years of age and by 25-35% in women aged 40 to 50 years.

[0008] Mammographic images are examined for various abnormalities in the breast, such as calcifications, which are small deposits of calcium in the tissue. Calcifications are usually not detectable by palpation but are visible in X-ray images. Clusters of small calcium deposits, so-called microcalcifications, indicate increased breast cell activity, which may be associated with breast cancer.

[0009] In conventional mammography, the imaged breast is usually compressed between two compression plates and exposed to radiation at least twice, from above and from an oblique direction. If necessary, an additional third image is taken from the side. However, in the resulting two-dimensional X-ray radiograph, strongly attenuating tissue may obscure other structures located on the same line of sight between the X-ray source and the detector. This has led to the application of three-dimensional tomography in mammographic examinations, which requires the acquisition of X-ray exposure images from several angles.

[0010] Mammography can be performed using a three-dimensional tomosynthesis reconstruction process, in which several two-dimensional X-ray projection images of the breast are acquired at different projection angles, e.g., within a range of ±15 degrees from the vertical axis, based on which well-known algorithms can be used to reconstruct tomographic images. Performing the reconstruction requires the availability of a sufficiently accurate representation of the three-dimensional projection geometry describing the relative positions of the X-ray source, the detector, and the image grid used for the reconstruction.

[0011] As discussed, based on multiple two-dimensional X-ray projection images, a three-dimensional image can be reconstructed using appropriate algorithms. This imaging modality is generally referred to as digital breast tomosynthesis (DBT), where a narrow, limited slice angle is used and a relatively small number of X-ray projection images are acquired, but where on the other hand the in-plane spatial resolution is also required for the detection of microcalcifications.

[0012] A typical modern digital mammography apparatus comprises a frame component and a C-arm or corresponding structure that can rotate relative to the frame. One end of the C-arm houses an X-ray source and the other end houses a radiation detector. The imaged breast is usually positioned close to the detector and is compressed between the plates during X-ray exposure acquisition. The X-ray source is usually a relatively heavy component, which poses a challenge to the mechanical design of the mammographic imaging device. The relatively large mass is conducive to an X-ray projection image acquisition strategy in which the C-arm moves continuously in its trajectory to avoid mechanical vibrations caused by acceleration and deceleration of the C-arm and excessive duration of X-ray exposure. However, this requires short single X-ray exposure times of approximately less than 50 milliseconds to avoid motion blur of the exposure and the use of a sufficiently powerful X-ray source that further increases its mass and the associated mechanical challenges.

[0013] In DBT imaging of clinically relevant structures of very small dimensions, it is crucial to obtain a sufficiently accurate representation of the three-dimensional imaging geometry that expresses the relative positions of the X-ray source and X-ray detector with respect to the image grid used for tomosynthesis reconstruction. That is, the spatial resolution of the reconstruction depends heavily on the consistency and accuracy of the imaging geometry used as input to the reconstruction algorithm. This requires either a high-accuracy mechanical design that can accurately reproduce the nominal imaging trajectory, or a calibration method that can routinely measure or infer the actual reproduced form of the X-ray source and detector trajectories with respect to the imaged field of view, preferably for each individual image acquisition. The average lower retail price of DBT imaging devices, compared to CBCT imaging devices, which typically rely on intermittently performed geometry calibration procedures, places additional constraints on the feasible mechanical accuracy of their moving parts, thus favoring solutions based on continuously measuring or inferring the reproduced imaging geometry rather than relying on it to match a previous calibration procedure. In particular, the prospect of upgrading existing digital mammography imaging devices to enable DBT imaging without changing the core mechanical design is of clinical and commercial interest. This further highlights the need to be able to observe the imaging geometry as it is reproduced by the device.

[0014] Computational methods have been developed to address the problem of geometric inaccuracies. In the methods reported in the literature, virtual motion of the X-ray source and X-ray detector by means of a rigid geometric transformation in a fixed coordinate system has been applied to model and compensate for rigid motion during projection image acquisition. In the latest methods related to medical tomographic imaging, such geometric transformations are optimized by maximizing the sharpness of the resulting reconstruction. Typically, this correction process is performed iteratively.

[0015] In DBT imaging, the use of a fixed coordinate system for modeling and compensation of geometric inconsistencies or patient motion is not ideal, since the inherent geometric degrees of freedom are not separated by the coordinate system. In DBT imaging devices, in particular the X-ray beam diverges and forms a pyramidal cone. Therefore, shifts along iso-rays adjacent to the center of the X-ray source and the X-ray detector will only affect the magnification, while in-plane shifts along the pixel array of the X-ray detector will result in a maximum shift of the imaged object within its projection image. Furthermore, preventing the net transformation due to the applied geometric correction by means of known methods of rigid registration of the resulting corrected DBT reconstruction and the uncorrected DBT reconstruction is computationally expensive, especially if it is repeatedly applied during the geometry correction process.

[0016] Due to the high geometric accuracy required in DBT imaging, dedicated methods have been designed to ensure this. An example of these is to embed sufficiently radiopaque or radiolucent landmark objects in the field of view of the DBT imaging device. Based on the landmark positions in the X-ray projection images, the achieved X-ray projection geometry can be inferred and this information incorporated in the DBT image reconstruction process. However, the manufacturing cost of such precise embedded landmark components is huge, their visibility requirements in the X-ray projection images limit the applicable field of view, and the resulting geometric representation cannot take into account any movement of the patient relative to the imaging device and the landmarks due to, for example, discomfort, breathing or heartbeat, although the patient's breast remains compressed and mostly remains stationary during the X-ray projection image acquisition. Summary of the invention

[0017] In one embodiment of the disclosed process, a data-driven algorithm utilizing an intrinsic rotating coordinate system attached to the spatial position and orientation of the X-ray source and detector can be used to establish an improved estimate of the actual imaging geometry corresponding to the measured X-ray projection images. The net transformation in the fixed reference coordinate system is prevented by calculating and subtracting the net transformation directly from the estimated correction geometric transformation parameters. Thus, the relative position of the X-ray source and detector system with respect to the patient and vice versa is retrospectively estimated and improved based on the X-ray projection image content. The purpose of the correction process is to improve the resulting image quality by improving the geometric consistency of the X-ray image measurements and thereby better meet the requirements of image reconstruction. The intrinsic coordinate system enables the applied geometric degrees of freedom to be defined in a manner corresponding to their relative importance to the correction process and the resulting image quality.

[0018] The process can take as input the data typically required for image reconstruction: a set of X-ray projection images and an estimate of the projection geometry corresponding to the spatial position of the X-ray source and the X-ray detector during the acquisition of the X-ray projection images. Similar to the corresponding method, an intermediate reconstruction is first calculated using the estimated projection geometry. Then, a correction geometric transformation is established that improves the geometric correspondence of each accessed X-ray projection image with the remaining X-ray projection images, wherein the transformation corresponds to a virtual movement of the X-ray source and the detector during image acquisition. In particular, a rotation coordinate system specific to the projection images is applied to determine the correction geometric transformation. The correction process can be iterated multiple times, including the calculation of intermediate reconstructions using the current estimate of the projection geometry and the subsequent optimization of the correction transformation. After the correction geometric transformation is established, the final DBT reconstruction is calculated using the X-ray projection images and the final estimate of the projection geometry corresponding to the correction transformation.

[0019] In another embodiment, the present invention and its preferred embodiments aim to enable three-dimensional tomosynthesis reconstruction from a single two-dimensional X-ray projection image taken by a digital mammography device by retrospectively determining a reconstructed form of the imaging geometry of each projection image in conjunction with an image reconstruction process. The present invention enables an accurate representation of the imaging geometry without resorting to a landmark-based calibration process and also enables the utilization of existing digital mammography equipment whose mechanical precision is limited to its originally intended use of two-dimensional mammographic acquisition and which is not capable of accurately reproducing a given X-ray exposure trajectory or inherently enabling accurate measurement of the positions of X-ray sources and detectors. Therefore, a preferred embodiment of the present invention comprises upgrading the software of existing digital mammography equipment to also be applicable to three-dimensional tomosynthesis imaging.

[0020] In the present invention, a self-calibration process is applied prior to the final DBT image reconstruction process. During the calibration process, a data-driven motion estimation algorithm is applied to establish mutual geometric correspondence between acquired X-ray projection images starting from an initial estimate of the projection geometry that reflects the approximate geometric form of the X-ray exposure trajectory. In essence, the position of the X-ray source and detector relative to the patient and the reconstructed images is retrospectively estimated based on the X-ray projection image data. Although in DBT imaging, it can be expected that the main source of geometric uncertainty and consistency between X-ray projection images is caused by the reproduction accuracy of the X-ray exposure acquisition trajectory, the method is also able to correct for any patient movement that may have occurred despite the fact that the imaged breast is typically compressed during imaging. Given the high accuracy requirements associated with the small physical dimensions of the structure of interest, this has a clear advantage over landmark assembly-based calibration methods, which can only capture the position of the X-ray source relative to the X-ray detector, but not the position of the patient.

[0021] The self-calibration process involves measuring a set of digitized mammographic X-ray projection exposure images, transferring the images to a reconstruction workstation, establishing an initial estimate of the projection geometry corresponding to each X-ray projection exposure image, computing intermediate DBT reconstructions and iteratively establishing a set of rigid transformation parameters for each X-ray projection exposure image to refine the initial projection geometry estimate. Applying the rigid transformation corresponds to a virtual motion of the X-ray source and detector that makes them correspond to the actual observed X-ray projection image content. This is possible by adaptively determining the imaging geometry using the measurement data itself.

[0022] The self-calibration process is preferably performed iteratively so that an intermediate DBT image is reconstructed before each iteration. Then, for each measured X-ray projection image, a digitally reconstructed radiograph (DRR) image corresponding to each set of evaluated rigid transformation parameters is calculated from the DBT reconstruction. The validity of the transformation parameters, i.e. the validity of the virtual motion candidates of the X-ray source and detector for a particular X-ray projection image, is established by measuring the similarity of the calculated DRR images with the physically measured X-ray projection images. Optimization of the rigid transformation parameters is then based on finding the extreme values ​​of the similarity measure between the measured X-ray projection images and the corresponding calculated DRR projection images according to the rigid transformation parameters. The process is iterated multiple times, where the rigid transformation parameters specific to the X-ray projection image are used in the next iteration to calculate an updated intermediate DBT reconstruction and further serve as an initial guess for subsequent parameter optimization.

[0023] Benefits of the disclosed embodiments of the present invention include improvements in projection geometry consistency, which results in higher image quality in terms of clarity, detail level, and contrast. Another benefit is that radiation dose to the patient can be reduced by potentially preventing rescans due to image quality degradation associated with projection geometry.

[0024] With particular regard to mammography, the disclosed self-calibration process can be utilized when it is desired to upgrade existing digital mammographic imaging equipment to enable three-dimensional tomosynthesis imaging. Three-dimensional tomosynthesis increases the likelihood of early detection of breast cancer and can make tomosynthesis mammographic devices more affordable. Another benefit is that not only mechanical inaccuracies can be corrected, but also slight movements of the patient during projection image acquisition can be corrected, which is not possible when using, for example, fixed calibration landmark assemblies. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings provide a further understanding of the disclosed patient movement correction method. In the drawings:

[0026] Figure 1 An example of a typical mammography apparatus is shown.

[0027] Figure 2 The principle of typical tomosynthesis imaging is shown.

[0028] Figure 3 One embodiment of a patient movement correction method is shown as a flow chart.

[0029] Figure 4 A detailed description is shown Figure 3 Flow chart of step 202.

[0030] Figure 5 Illustration of fixed and rotating coordinate systems in the context of tomosynthesis imaging. DETAILED DESCRIPTION

[0031] A typical description of the imaging geometry includes the 3D position of the focus of the X-ray source and the center of the X-ray detector, as well as information sufficient to uniquely determine the orientation of the X-ray detector. Such information may include, for example, rotation angles in a reference coordinate system that will determine the orientation of the horizontal and vertical axes of the detector pixel array. The description of the projection geometry is usually based on physical measurements of the imaging device and knowledge of the ideal exposure trajectory corresponding to a given imaging procedure. In addition, a periodic calibration process is usually performed to ensure sufficient accuracy of the projection geometry.

[0032] The basic assumption used in the reconstruction process is that the imaged anatomy has remained sufficiently stationary during the acquisition of the X-ray projection images. The rationale is that the X-ray projections should represent co-registered measurements of the stationary object, which can then be consistently combined to reconstruct the structure of the object. Systematic errors in the projection geometry can be compensated by regular calibration measurements that remove the effects of any deviations from the assumed ideal X-ray projection image acquisition trajectory.

[0033] A more difficult problem is caused by patient movement during imaging, which is unpredictable and random in nature. Although this is prevented to some extent by supporting the patient during imaging, the relatively long duration of X-ray projection image acquisition in DBT imaging (approximately 10 seconds) makes it impractical to completely eliminate patient movement. Moreover, elderly patients who routinely attend mammography examinations find it more difficult to remain still during imaging.

[0034] In the event of significant patient movement during X-ray projection image acquisition, noticeable streaking or blurring artifacts will appear in the reconstructed image, which in the worst case may render the image useless for the intended medical purpose. Repeating the scan may provide better results, but at the expense of additional radiation dose.

[0035] In the disclosed retrospective method for compensating for the adverse effects of undesired movement, an intermediate reconstruction is first calculated using the 2D X-ray projection images and the estimated imaging geometry. It is sufficient to calculate the intermediate reconstruction at a coarser resolution than is normally used when reconstructing for diagnostic purposes. The intermediate reconstruction is used to aggregate information from all X-ray projection images, wherein the appearance of the reconstruction reflects the mutual geometric consistency of the measured 2D X-ray projection images.

[0036] The optimization of the projection image specific geometry is based on the similarity of the measured physical X-ray projection images and the corresponding reprojection data of the intermediate reconstruction forming the digitally reconstructed radiograph (DRR). The basic principle is that when the estimated projection geometry is consistent, the reconstructed image, in the sense of the forward projection operation trying to satisfy the measurements conveyed by the projection images, will, by definition, produce a reprojection that closely matches the measured data. In case of a geometric mismatch, the reprojection of the intermediate reconstruction will deviate from the measured projection. Since the intermediate reconstruction serves as an aggregation of all projection images, its reprojection reflects the sum of all projection images, and it can be expected that the maximum similarity of the measured X-ray projection image and the forward projection image is maximized when the corresponding reprojection geometry matches the average correct projection geometry of the intermediate reconstruction.

[0037] In a typical DBT imaging device, the emitted X-ray beam diverges and forms a pyramidal cone. The deviation along the iso-ray adjacent to the center of the X-ray source and the X-ray detector will only affect the magnification, while the deviation along the plane of the X-ray detector will result in the maximum deviation of the imaged object within its projection image. From this perspective, the disclosed method adopts a rotating coordinate system that is attached to the physical position of the X-ray source and the detector during image acquisition. That is, by setting one axis to coincide with the iso-ray pointing from the X-ray source to the center of the pixel array of the X-ray detector, and setting two other perpendicular axes to coincide with the orthogonal projection of the pixel array axis of the X-ray detector to the plane perpendicular to the iso-ray. By limiting the geometric transformation of the projection geometry of each projection image along these intrinsic axes, the geometric degrees of freedom whose importance is different in the sense of geometric accuracy can be separated.

[0038] When each X-ray projection image is subjected to a separate geometric transformation, the resulting average transformation may have a net effect on the reconstructed image. For example, this manifests itself as an overall net shift or rotation of the reconstructed anatomy relative to the uncorrected anatomy. The net transformation of the anatomy may have an adverse effect on the suitability of the reconstructed image. Therefore, a compensation method for the net transformation is also disclosed. The net transformation can be estimated by mapping the transformation corresponding to each transformed projection image from a rotating coordinate system to a fixed coordinate system. For example, by assuming that each projection image is displaced only along its horizontal axis, the corresponding displacement in the fixed coordinate system can be calculated based on the known horizontal axis of the projection-specific rotating coordinate system, and the average value can be taken to represent the net displacement in the fixed coordinate system. By the inverse of the above mapping, the inverse of the net transformation can be mapped back to the rotating coordinate system and subtracted from the transformation specific to the projection image. Therefore, the net transformation in the fixed coordinate system is eliminated.

[0039] In the optimization process, the goodness of the corrective geometric transformation of a given X-ray projection image is measured by the similarity of the forward projection image corresponding to the transformed projection geometry to the original X-ray projection image. The similarity of the re-projection and the X-ray projection can be measured, for example, by means of the mean square error, correlation coefficient or gradient correlation coefficient of the images. The best (in the sense of the similarity measure) geometric transformation for each projection image of a given intermediate DBT reconstruction can then be determined by finding the extreme value of the similarity measure between the forward projection image and the X-ray projection image according to the parameters of the geometric transformation.

[0040] The optimization process then comprises computing intermediate DBT reconstructions using an initial estimate of the X-ray projection geometry; accessing all or a subset of the measured X-ray projection images; establishing (for each accessed projection image) a projection image-specific correction transformation by finding the maximum similarity between the measured X-ray projection image and the corresponding forward projection of the intermediate reconstruction according to the parameters of the geometric transformation performed in the rotated coordinate system; estimating the net transformation in the fixed reference coordinate system and subtracting the corresponding transformation from the transformation parameters in the rotated coordinate system; and computing the final reconstruction when it is estimated that sufficient correction results have been obtained.

[0041] Figure 1 A typical configuration of a digital mammography apparatus is shown. Figure 1The mammographic X-ray device (1) comprises a frame portion (10) and a C-arm (11) attached thereto (more typically an arm structure (11)). An X-ray radiation source (12) is arranged inside the upper part of the C-arm (11) and its cover, which is arranged to generate an X-ray beam that passes through an upper compression plate (14) of the mammographic X-ray device (when such an upper compression plate is attached to the device) and is directed towards an X-ray detector (18) located in a detector housing (13). The detector housing (13) or a corresponding structure is typically arranged inside a lower tray structure (15), in which a grid structure (17) can be integrated to absorb radiation scattered by the imaging object. The lower tray structure (15) can be a fixed structure in the device, or it can be detachable. Since the top surface of the lower tray structure (15) is typically used as a platform on which the breast is positioned for imaging, the structure is also typically referred to as a lower compression plate. In accordance with Figure 1 In the solution of the present invention, the lower tray structure (15) is arranged with a connecting unit (16) so as to enable the lower tray structure to be detachably connected to the mammography device (1).

[0042] A typical modern digital mammography device is motorized so that a C-arm (11) can be moved in a vertical direction and rotated about an axis, which is usually a horizontal axis connecting the C-arm (11) to a frame (10) of the device. The C-arm (11) can be realized in two parts, whereby the basic construction of the device can include a frame part (10) standing essentially vertically or a frame part (10) fixed to, for example, a wall or a ceiling and an arm structure (11) connected thereto, which is arranged to be rotatable relative to a horizontal rotation axis, at its opposite ends, essentially at a first end a radiation source (12) including a focus is placed, and at a second end an image data receiving unit (18) is placed. The arm structure (11) can be realized so that the first end and / or the second end of the arm structure (11) can be independently rotated relative to the horizontal rotation axis. The arrangement can also include a control unit, which includes an information recording unit and a unit for processing information related to imaging, in particular for processing image information.

[0043] Figure 2 A typical tomosynthesis imaging principle is shown, where individual X-ray exposure images of the breast are taken at different projection angles (eg, about ±15 degrees from the vertical axis). Figure 2 An arrangement is shown in which only the radiation source (12) is deflected relative to the position of the breast being imaged, but arrangements are also known in which the C-arm (11) as a whole is rotated so that the breast remains positioned in its position, but the detector (18) follows the movement of the radiation source (12) on the opposite side of the breast being imaged. The proposed self-calibration method is applicable to either arrangement.

[0044] Figure 3 The steps of an embodiment of the present invention for patient motion correction are described, which can be used, for example, to process Figure 1 The method is based on finding a corrective geometric transformation for the initially estimated projection geometry in a rotating coordinate system attached to the positions of the X-ray source 105 and the X-ray detector 109 .

[0045] exist Figure 3 In step 200 of the method, input data is acquired, the input data comprising a measured X-ray projection image and an initial estimate of a projection geometry corresponding to the acquisition process. The projection geometry determines an estimated orientation of the X-ray detector 109. The form of the projection geometry is typically based on a geometric calibration process that utilizes, for example, a known reference phantom with radiopaque markers.

[0046] In step 201, an intermediate reconstruction for the projection optimization method is calculated using the input data acquired in step 200. A downsampled version of the input data may be used, since a high spatial resolution is usually not required as in images for diagnostic purposes. An intermediate reconstruction is understood to aggregate all available physical and geometric information acquired during the measurement process of step 200. In case of geometric inconsistencies, this is reflected by the intermediate reconstruction (e.g. by ambiguity of the reconstruction details).

[0047] In step 202, a corrective geometric transformation of the projected image is established in the rotating coordinate system. The purpose of the geometric transformation is to compensate for the geometric inconsistencies inherent in the initial estimate of the projection geometry. Figure 4 The details of step 202 are explained, but more generally, a corrective geometric transformation is sought by finding the best geometric transformation. The goodness of a transformation is defined by assigning a similarity value to it. The similarity value is calculated by comparing a re-projected digitally reconstructed radiograph (DRR) of an intermediate image reconstruction with a corresponding measured X-ray projection image, where the applied projection geometry corresponds to the evaluated geometric transformation. A higher similarity value is considered to be an indication of a better corrective geometric transformation. Applying a rotated coordinate system to the geometric transformation enables the separation of geometric degrees of freedom according to their importance for the problem in conical projection geometry.

[0048] In step 203, the net geometric transformation in the fixed coordinate system is subtracted. The fixed coordinate system is usually defined with reference to the static components of the imaging device. The net geometric transformation is calculated by linearly transforming the geometric transformation established in step 202 from the rotating coordinate system to the fixed coordinate system. Based on the known coordinate axes of the rotating coordinate system and the fixed coordinate system, the linear transformation can be easily obtained. After the net transformation is established, its inverse is linearly transformed from the fixed coordinate system to its rotating coordinate system. The inverse of the net transformation then corresponds to a set of geometric transformation parameters for each projected image in the rotating coordinate system. Adding these values ​​to the geometric parameter values ​​established in step 202 will cause the net transformation to be cancelled in the fixed coordinate system.

[0049] In step 204, the transformation parameter values ​​obtained from step 203 are applied to the initial projection geometry estimate to obtain a corrected projection geometry estimate. In particular, before proceeding to step 204, steps 201-203 may be iteratively repeated multiple times.

[0050] The corrected projection geometry estimate is used to compute a final DBT reconstruction in step 205. The final reconstruction is computed in the usual manner, with the exception of the case where steps 201-204 are not applied, where the initial projection geometry estimate is replaced by the corrected projection geometry estimate obtained in steps 201-204.

[0051] Figure 4 As an example, the Figure 3 Details of step 202 are provided. In step 300, the evaluated geometric transformation is applied to the initial geometry of the projection image in the rotating coordinate system. Again, the coordinate system is defined to be consistent with the spatial position and orientation of the X-ray source and the X-ray detector during the physical acquisition of the projection image under consideration. Applying the transformation in the rotating reference system involves a linear mapping from the fixed coordinate system (where the imaging geometry is usually defined) to the rotating coordinate system where the transformation is performed, followed by an inverse linear mapping from the rotating coordinate system to the fixed coordinate system. As a simple example of the effect achieved, in order to apply a translation along an axis corresponding to the horizontal axis of the X-ray detector (i.e., its projection to a plane perpendicular to the iso-rays), the initial projection geometry of the projection image is mapped to the rotating coordinate system, translated along this axis of the rotating coordinate system by a given amount corresponding to the virtual movement of the X-ray source and the detector, and then mapped back to the fixed coordinate system. The transformed projection geometry is expressed in the fixed coordinate system, and the transformed projection geometry obtained is used as input to the subsequent step 301.

[0052] In step 301, a re-projected DRR image of the intermediate DBT reconstruction calculated in step 201 is calculated using the transformed projection geometry obtained in step 300. The calculation of the DRR image can be performed using a standard algorithm, such as the Siddon ray casting method. In a typical re-projection algorithm, the input to the algorithm includes the endpoints of the X-ray representation and a source image, the DRR image of which is calculated, including knowledge of its spatial position and orientation expressed in the same coordinate system. In the described setup, this coordinate system corresponds to a fixed coordinate system. The effect of the geometric transformation applied in step 300 is to change the endpoints of each virtual X-ray path through an intermediate reconstruction, which propagates the effect of the geometric transformation to the obtained DRR image.

[0053] In step 302, the similarity between the DRR image obtained in step 301 and the X-ray projection image acquired in step 200 is evaluated. The similarity is based on a point-by-point comparison of the images using established methods, such as the mean square error of the images or their cross-correlation. The specific measure used to evaluate the similarity is not important to the described method. The obtained similarity values ​​are assigned to the geometric parameters given as input to step 300. A higher similarity is considered to be an indication of more suitable geometric transformation parameters.

[0054] In step 303, the optimal geometric transformation parameters are established by finding the parameters corresponding to the highest similarity value obtained by applying steps 300-302. In a typical setting, steps 300-302 are repeatedly evaluated by a suitable minimization algorithm such as the well-known Nelder-Mead simplex algorithm to establish the optimal geometric transformation parameters. These parameters are stored and assigned to a specific projection image until they are potentially modified by repeating steps 201-203.

[0055] Figure 5 The rotating coordinate system applied in step 202 is shown. In a DBT imaging device, such as Figure 1The arm structure (11) shown in the figure generally supports an X-ray source (12) and an X-ray detector (18). In DBT imaging, in particular the X-ray beam diverges and forms a pyramidal cone. Deviations along iso-rays adjacent to the centers of the X-ray source (12) and the X-ray detector (18) will only affect the magnification, while deviations corresponding to the axes of the pixel array of the X-ray detector will result in maximum displacement of the imaged object within its projected image. From this point of view, the disclosed method employs a rotated uvw coordinate system that is attached to the physical position and orientation of the X-ray source during image acquisition. By restricting the geometric transformation of the projection geometry of each projected image along these intrinsic axes, it is possible to separate geometric degrees of freedom whose importance differs in the sense of geometric accuracy. The fixed coordinate system for calculating the net transformation is represented by the xyz coordinate system.

[0056] The u-axis of the coordinate system is set to coincide with an iso-ray pointing from the X-ray source (12) to the center of the X-ray detector (18), and the v-axis is set to coincide with the planar projection of the horizontal axis of the pixel array of the X-ray detector (18) onto a plane perpendicular to the iso-ray. For example, these degrees of freedom are used for the rigid transformation:

[0057] 1. Longitudinal offset along the isotropic ray (u-axis); 2. Lateral offset along the projected horizontal axis (v-axis) of the X-ray detector pixel array; 3. Vertical offset along the vertical axis (w-axis) of the projection of the X-ray detector pixel array; 4. Rotation around the u-axis (roll angle); 5. Rotation around the v-axis (pitch angle); 6. Rotation around the w-axis (yaw angle).

[0058] The result of the disclosed patient motion correction method is an improved estimate of the projection geometry corresponding to the physical acquisition of X-ray projection images and, correspondingly, DBT reconstructed images, wherein the effects of geometric inconsistencies are reduced.

[0059] A preferred embodiment of the invention comprises a mammographic imaging arrangement for tomosynthesis, comprising a mammographic imaging device (1) comprising a frame part (10) standing substantially vertically or a frame part (10) that can be fixed to a wall or ceiling, an arm structure (11) connected to the frame part (10) and rotatable about a horizontal rotation axis, wherein an X-radiation source (12) comprising a focal spot is arranged substantially at a first end at the opposite ends of the arm structure (11), and an image data receiving unit (18) is arranged substantially at a second end. A lower tray structure (15) is also arranged in connection with the second end of the arm structure (11), the lower tray structure (15) being located substantially on top of the image data receiving unit (18). The arrangement further comprises a control unit for moving at least one part of the arm structure (11) and controlling the operation of the image data receiving unit (18), and a processing unit comprising an information recording unit and a unit for processing information.

[0060] The tomosynthesis imaging treatment may be started by positioning the patient for imaging, in connection with the arrangement of the unit described above with the breast placed on top of the lower tray structure (15) and possibly made immovable by being compressed by the upper compression plate (14). If the arm structure (11) of the apparatus or the part of the arm structure containing the radiation source (12) has not yet been driven to the first projection angle of the imaging treatment, it will be driven thereto, after which the arrangement is ready for the first exposure. The image information associated with the exposure detected by the detector (18) is read out for recording, and the imaging is repeated at small projection angle intervals, for example at projection angle intervals of 2 degrees.

[0061] The measured raw data corresponding to the X-ray exposure typically represents an intensity or proportional measure of the incident X-ray radiation. A well-known logarithmic mapping is applied to the measured raw image to derive an X-ray projection image that is approximately proportional to the spatial pencil integral of the total X-ray attenuation distribution of the imaged breast. Pre-processing of the measured X-ray exposure image may be arranged to be performed on the imaging device or on an image processing workstation.

[0062] After the pre-processed X-ray projection images are acquired and transferred to an image processing workstation or similar device, an iterative process for determining the imaging geometry for each X-ray exposure is performed. If the X-ray detector is fixed, then the imaging geometry is determined by the 3D position of the X-ray source relative to the detector. If the detector moves slightly during the X-ray exposure acquisition, then the position of the detector is also required to fix the imaging geometry. This is also necessary to correct for any patient movement and is assumed in the following. Therefore, the projection geometry description includes the X-ray source (focal point) and the detector position and the height of the detector. The X-ray source position is expressed as a point in a 3D reference coordinate system.

[0063] In an iterative process, the three-dimensional positions of the X-ray radiation source (12) and the detector (18) corresponding to the exposure of each X-ray projection image are estimated by using predetermined estimates of the positions of the X-ray radiation source (12) and the detector (18) as initial guesses and then optimizing a corrective rigid transformation to obtain a refined estimate that maximizes the geometric correspondence between each measured X-ray projection image and all other measured X-ray projection images. It should be understood that due to the aforementioned aspects related to the limited mechanical and positional accuracy of digital mammography equipment, the initial estimate of the X-ray source position is actually always not ideal.

[0064] The geometric correspondence with all other X-ray projection images is measured by comparing each X-ray projection image with a corresponding digitally reconstructed radiograph (DRR) calculated from an intermediate DBT image reconstruction using a suitable forward projection algorithm. The DBT reconstruction can be calculated using a suitable algorithm (such as the well-known iterative SIRT algorithm). It is understood that in the ideal case where the projection geometry is known exactly and the ideal image reconstruction can be calculated with infinite accuracy from the ideal projection image measurements, for example based on the image element size and the physical model for X-ray propagation and detection, the calculated DRR image will be completely equivalent to the measured X-ray projection image. In addition, the intermediate DBT reconstruction is understood to contain substantially all available structural and geometric information from each X-ray projection image. Based on this reasoning, an iterative method is used to incrementally estimate the imaging geometry that maximizes the correspondence between the measured X-ray projection image and the calculated DRR projection image of the intermediate tomosynthesis reconstruction. The intermediate DBT reconstruction is usually calculated on a coarser sampling than the typical DBT image reconstruction used for medical diagnosis. This is done to improve the computational efficiency of the motion correction algorithm used in the self-calibration process.

[0065] The rigid transformation parameters of the projection geometry description of each X-ray projection image are initialized to an identity transformation that leaves the initial projection geometry description unchanged. For each iteration and each X-ray projection image, updated rigid transformation parameters are established that maximize the geometric correspondence between the measured projection image reconstructed by the intermediate DBT and the corresponding calculated DRR projection image. The degree of correspondence of each rigid transformation candidate, and thus their suitability, is then measured by a similarity measure calculated from the measured X-ray projection image and the calculated DRR projection image.

[0066] In summary, steps 300-304 describe an iterative optimization of projection-specific rigid transformation parameters in a self-calibration process, wherein each individual iteration comprises these steps. Each iteration begins by calculating an intermediate DBT reconstruction using measured X-ray projection images and a current estimate of the imaging geometry as given by an initial estimate of the variation of the imaging geometry uniquely determined by the projection-specific rigid transformation parameters, step 300. In the first iteration, the transformation parameters are set to an identity transformation, i.e., zero translation or rotation.

[0067] The validity of new candidate values ​​of the rigid transformation parameters of the projection geometry of each X-ray image in each iteration is evaluated by first calculating the corresponding DRR projection image of the current intermediate DBT reconstruction. The calculation of the DRR projection image begins by applying the candidate rigid transformation parameters to the initial estimate of the projection geometry of the current projection image, step 301. The rigid transformation is applied to the X-ray source and detector positions, which are then used to calculate the DRR projection images.

[0068] With the candidate rigid transformation applied to the projection geometry of the current X-ray projection image, a suitable forward projection algorithm is used to compute a DRR image of the current intermediate DBT reconstruction, step 302. The resulting DRR image is uniquely determined by the DBT reconstruction, and the projection geometry is derived by applying the rigid transformation parameters to the original initial projection geometry estimate.

[0069] The validity of candidate values ​​of the rigid transformation parameters is estimated by evaluating the similarity of the measured X-ray projection image and the DRR projection image calculated using the projection geometry obtained by applying the transformation parameters, step 303. The similarity is evaluated by calculating a suitable image similarity measure (such as image cross-correlation, gradient correlation, mutual information, or entropy of image differences). In view of the above considerations, a higher similarity between the measured X-ray projection image and the calculated DRR image is taken as an indication or more suitable rigid transformation parameters.

[0070] Optimization of the rigid transformation parameters is performed by finding the extreme values ​​of the similarity measure applied between the measured X-ray projection image and the DRR projection image calculated based on the rigid transformation parameters, step 304. The optimization is performed by iterating steps 301-303 until a predetermined condition is met. The predetermined condition can be, for example, a number of iterative cycles or an overall change in the projection-specific transformation parameters. After finding the optimal rigid transformation parameters corresponding to the extreme values ​​of the found projection similarity measure, the transformation parameters are stored for reuse in subsequent iterations starting from step 300. The sufficient number of iterations generally depends on the desired level of accuracy of the correction process and the accuracy of the initial estimate. It may be desirable for the iterative process to converge to a point where the projection geometry estimate no longer changes significantly between subsequent iterations and the calculation can be stopped.

[0071] The only result of the motion correction based self-calibration process is a set of rigid transformation parameters specific to the projection image, which are applied to the initial projection geometry estimate, producing an optimized estimate of the X-ray projection geometry. The optimized projection geometry estimate replaces the initial projection geometry estimate, and it is used as input for computing a final DBT reconstructed image representing the patient for medical purposes.

[0072] In the context of tomosynthesis in mammography, the present invention can be used with a limited number of projection images and using a limited tomosynthesis angle. The number of projection images can be only about a few dozen images, such as less than 20 images, or even less than 10 images. The tomosynthesis angle can be less than about 60 degrees, or even less than 40 degrees, such as 15-30 degrees.

[0073] The above method can be implemented as computer software executed in a computing device. When the software is executed in the computing device, it is configured to perform the above inventive method. The software is implemented on a computer readable medium, and the obtained software can be provided to the computing device.

[0074] The disclosed method can be implemented as computer software executed in a computing device. The software is implemented on a computer readable medium, and the result can be provided to the computing device.

[0075] As described above, the components of the exemplary embodiments may include computer-readable media or memory for maintaining instructions programmed according to the teachings of the present embodiment and for maintaining data structures, tables, records and / or other data described herein. Computer-readable media may include any suitable medium for participating in providing instructions for execution to the processor. Common forms of computer-readable media may include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, any other suitable magnetic media, CD-ROMs, CD±Rs, CD±RWs, DVDs, DVD-RAMs, DVD±RWs, DVD±Rs, HD DVDs, HD DVD-Rs, HD DVD-RWs, HDDVD-RAMs, Blu-ray discs, any other suitable optical media, RAMs, PROMs, EPROMs, flash-EPROMs, any other suitable memory chips or cassettes, carriers or any other suitable media from which a computer can read.

[0076] In summary, embodiments of the present invention include a self-calibration process for digital breast tomosynthesis reconstruction, including:

[0077] Acquiring a set of X-ray projection images including a limited number of two-dimensional images of an imaged patient over a limited tomographic angle using an X-ray imaging unit including an X-ray source and an X-ray detector;

[0078] defining an initial projection geometry estimate describing the spatial position and orientation of an X-ray source and an X-ray detector during acquisition of an X-ray projection image;

[0079] Computing intermediate DBT reconstructions using the X-ray projection images and the initial projection geometry estimate;

[0080] determining a projection image-specific correction geometry transformation for the initial projection geometry estimate based on the X-ray projection image, the initial projection geometry estimate, and the intermediate DBT reconstruction;

[0081] and computing a final DBT reconstruction using the X-ray projection images and a corrected estimate of the projection geometry corresponding to a determined corrective geometric transformation, wherein the corrective geometric transformation used for the initial projection geometry estimate is determined in a projection image-specific rotated coordinate system corresponding to the spatial position and orientation of the X-ray source and the X-ray detector during acquisition of the X-ray projection images.

[0082] The projection image-specific rotational coordinate system can be defined by setting one axis to coincide with an isotropic ray pointing from the X-ray source to the center of the pixel array of the X-ray detector, and setting the other two perpendicular axes to coincide with the orthogonal projections of the pixel array axes of the X-ray detector to planes perpendicular to the isotropic ray.

[0083] Another possibility for defining a projection image-specific rotational coordinate system is to arrange two perpendicular axes to coincide with the pixel array of the X-ray detector and to arrange a third perpendicular axis to coincide with the normal to the pixel array of the X-ray detector.

[0084] According to one embodiment, a rotational coordinate system specific to the projection image may be defined by setting two perpendicular axes to coincide with orthogonal geometric projections of the edges of the X-ray detector onto planes perpendicular to the iso-rays, and setting a third axis to coincide with the iso-rays.

[0085] According to one embodiment, a rotational coordinate system specific to the projection image may be defined by setting two perpendicular axes to coincide with the edges of the X-ray detector and setting a third axis to coincide with the normal of the X-ray detector.

[0086] The projection geometry of the re-projected digitally reconstructed radiograph is obtained by applying a geometric transformation in a rotated coordinate system to the initial projection geometry estimate.

[0087] The corrective geometric transformation is determined by finding the maximum similarity between the X-ray projection image and the corresponding re-projected digitally reconstructed radiograph of the intermediate DBT reconstruction.

[0088] The net transformation in the fixed coordinate system is calculated and subtracted from the corrected geometric transformation determined in the rotating coordinate system.

[0089] The net transformation in the fixed coordinate system is calculated by mapping the geometric transformation specific to the projection image from the rotated coordinate system to the fixed coordinate system, taking the average thereof, and subtracting the average from the geometric transformation after mapping from the fixed coordinate system to the rotated coordinate system.

[0090] The computation of the intermediate DBT reconstruction is iterated multiple times and a corrected geometric transformation is determined, wherein the corrected projection geometry estimate after each iteration serves as the new initial projection geometry estimate for the next iteration and the final corrected projection geometry estimate corresponds to the corrected projection geometry estimate of the last applied iteration.

[0091] Embodiments include: a computer program comprising a computer program code configured to perform a method according to the above disclosure when executed in a computing device; and an apparatus comprising:

[0092] at least one processor configured to execute a computer program; and

[0093] at least one memory configured to store a computer program and associated data;

[0094] The device is connectable to a medical imaging device and is configured to perform a method according to the above disclosure.

[0095] Another embodiment is a method for a digital breast tomosynthesis (DBT) self-calibration process, comprising:

[0096] Acquiring a set of X-ray projection exposure images;

[0097] forming an initial estimate of the projection geometry corresponding to each X-ray projection exposure image; and further

[0098] calculating intermediate DBT reconstructions; and

[0099] Establishing a set of rigid transformation parameters corresponding to the calibration results for each X-ray projection exposure image and applied to the initial projection geometry estimate;

[0100] A final DBT reconstruction is calculated using the set of X-ray projection exposure images and a final calibrated estimate of the projection geometry.

[0101] The final DBT reconstructed image is calculated using the original X-ray exposure and a refined projection geometry estimate based on an established set of rigid transformation parameters.

[0102] The method also includes performing a projection image-specific rigid transformation of the initial projection geometry estimate by applying an established set of rigid transformation parameters.

[0103] The transformation includes translation and rotation.

[0104] The set of rigid transformation parameters applied to the current projection geometry estimate for a given projection image results in a modified or refined form of the projection geometry corresponding to the virtual motion of the X-ray source and detector.

[0105] Before computing the intermediate DBT-reconstructed DRR images, a rigid transformation is performed on the corresponding initial projection geometry estimate.

[0106] The method also includes performing a rigid transformation in a rotated coordinate system specific to the projected image.

[0107] The method also includes measuring the similarity of the calculated DRR image and the corresponding measured X-ray projection image.

[0108] The method also includes optimizing projection-specific rigid transformation parameters by finding an extremum of a similarity measure between the measured X-ray projection image and the calculated DRR projection image based on the rigid transformation parameters.

[0109] The method performs multiple iterations and uses the transformed parameters in the next iteration, which are then used as initial guesses for subsequent parameter optimization.

[0110] In one embodiment, a computer program comprises computer program code configured to, when executed in a computing device, perform the method according to the above discussion.

[0111] One embodiment is an apparatus comprising:

[0112] at least one processor (112) configured to execute a computer program; and at least one memory (113) configured to store the computer program and related data;

[0113] Wherein the device is connectable to a medical imaging arrangement and is configured to perform the method as discussed above.

[0114] Another embodiment is a medical digital breast tomosynthesis imaging arrangement, comprising:

[0115] Medical imaging equipment; and

[0116] The device as disclosed above, wherein the device is connected to the medical imaging device.

[0117] It is obvious to a person skilled in the art that, as technology advances, the basic idea of ​​a self-calibrating medical imaging device may be implemented in various ways.Therefore, the self-calibrating medical imaging device and its embodiments are not limited to the examples described above; rather they may vary within the scope of the claims.

Claims

1. A method for self-calibration processing for digital breast tomosynthesis (DBT) reconstruction, comprising: Using an X-ray imaging unit including an X-ray source and an X-ray detector, a plurality of two-dimensional X-ray projection images are acquired at tomographic angles; defining an initial projection geometry estimate describing the spatial position and orientation of an X-ray source and an X-ray detector during acquisition of the X-ray projection images; Computing intermediate DBT reconstructions using the X-ray projection images and the initial projection geometry estimate; determining a projection image-specific correction geometric transformation for the initial projection geometry estimate based on the X-ray projection image, the initial projection geometry estimate, and the intermediate DBT reconstruction; and computing a final DBT reconstruction using the X-ray projection images and a corrected estimate of the projection geometry corresponding to the determined corrective geometric transformation, wherein the X-ray projection images comprise a finite number of two-dimensional images acquired over a finite tomographic angle, and a corrective geometric transformation for initial projection geometry estimation is determined in a projection image-specific rotated coordinate system corresponding to the spatial position and orientation of an X-ray source and an X-ray detector during acquisition of the X-ray projection images, wherein the corrective geometric transformation is determined by finding a maximum similarity between the X-ray projection images and corresponding re-projected digitally reconstructed radiographs of an intermediate DBT reconstruction.

2. The method of claim 1 , wherein the projection image-specific rotational coordinate system is defined by setting one axis to coincide with an iso-ray adjacent to the center of a pixel array of an X-ray source and an X-ray detector, and setting two other perpendicular axes to coincide with orthogonal projections of the pixel array axes of the X-ray detector onto planes perpendicular to the iso-ray.

3. The method of claim 1 , wherein the projection image-specific rotated coordinate system is defined by setting two perpendicular axes to coincide with a pixel array of an X-ray detector and setting a third perpendicular axis to coincide with a normal to the pixel array of the X-ray detector.

4. A method according to claim 1 or 2, wherein the projection geometry of the re-projected digitally reconstructed radiograph is obtained by applying a geometric transformation in a rotated coordinate system to the initial projection geometry estimate.

5. The method according to any one of claims 1 to 3, wherein a net transformation in a fixed coordinate system is calculated and subtracted from a corrected geometric transformation determined in a rotating coordinate system.

6. A method according to any one of claims 1-3, wherein the net transformation in the fixed coordinate system is calculated by mapping a geometric transformation specific to the projection image from the rotating coordinate system to the fixed coordinate system and taking the average thereof, and subtracting the average from the geometric transformation after mapping from the fixed coordinate system to the rotating coordinate system.

7. The method according to any one of claims 1-3, wherein the calculation of the intermediate DBT reconstruction and the determination of the corrected geometric transformation are iterated multiple times, wherein the corrected projection geometry estimate after each iteration serves as the new initial projection geometry estimate for the next iteration, and the final corrected projection geometry estimate corresponds to the corrected projection geometry estimate of the last applied iteration.

8. The method according to any one of claims 1-3, wherein the limited number of projected images is about tens of images.

9. The method according to any one of claims 1-3, wherein the limited number of projected images is less than 20 images.

10. The method according to any one of claims 1-3, wherein the limited number of projected images is less than 10 images.

11. The method of any one of claims 1-3, wherein the limited tomographic angle is less than about 60 degrees.

12. The method of any one of claims 1-3, wherein the limited tomography angle is less than about 40 degrees, such as 15-30 degrees.

13. A computer program comprising computer program code configured to perform the method according to any of the preceding claims 1-12 when executed in a computing device.

14. An apparatus comprising: at least one processor configured to execute a computer program; as well as at least one memory configured to store a computer program and associated data; Characterized in that the device is connectable to a medical imaging device and is configured to perform the method according to any one of claims 1-12.

15. A medical digital breast tomosynthesis imaging arrangement comprising: Medical imaging equipment; as well as The apparatus according to claim 14, wherein the apparatus is connected to the medical imaging device.

16. A method for self-calibration processing of digital breast tomosynthesis (DBT), comprising: Acquiring a set of X-ray projection exposure images; forming an initial estimate of the projection geometry corresponding to each x-ray projection exposure image; Characterized in that the method further comprises: calculating intermediate DBT reconstructions; and establishing a set of rigid transformation parameters applied to the initial estimated projected geometry; and computing a final DBT reconstruction using the set of X-ray projection exposure images and a final calibrated estimate of projection geometry, wherein the final DBT reconstructed images are computed using the original X-ray exposures and a refined projection geometry estimate based on an established set of rigid transformation parameters, and Therein a set of established rigid transformation parameters is determined by finding the maximum similarity between the X-ray projection exposure images and the corresponding re-projected digitally reconstructed radiographs of the intermediate DBT reconstruction.

17. The method of claim 16, wherein the method further comprises performing a projection image-specific rigid transformation of the initial projection geometry estimate by applying an established set of rigid transformation parameters.

18. The method of claim 16 or 17, wherein the transformation comprises a translation and a rotation.

19. A method according to claim 16 or 17, wherein the set of rigid transformation parameters applied to the current projection geometry estimate for a given projection image results in a modified or refined form of the projection geometry corresponding to a virtual motion of the X-ray source and detector.

20. The method of claim 19, wherein prior to computing the intermediate DBT reconstructed DRR image, a rigid transformation is performed on the corresponding initial projection geometry estimate.

21. The method of claim 19, further comprising performing a rigid transformation in a rotated coordinate system specific to the projected image.

22. The method of claim 19, wherein the method further comprises measuring a similarity between the calculated DRR image and the corresponding measured X-ray projection image.

23. The method of claim 19, wherein the method further comprises optimizing projection-specific rigid transformation parameters by finding an extremum of a similarity measure between the measured X-ray projection image and the calculated DRR projection image based on the rigid transformation parameters.

24. The method of claim 19, wherein the method performs multiple iterations and uses the transformed parameters in the next iteration and then uses the transformed parameters as an initial guess for subsequent parameter optimization.

25. A computer program comprising computer program code configured to perform the method according to any of the preceding claims 16-24 when executed in a computing device.

26. An apparatus comprising: at least one processor configured to execute a computer program; as well as at least one memory configured to store a computer program and associated data; Characterized in that the device is connectable to a medical imaging arrangement and is configured to perform a method according to any of the preceding claims 16-24.

27. A medical digital breast tomosynthesis imaging arrangement comprising: Medical imaging equipment; as well as The apparatus according to claim 26, wherein the apparatus is connected to the medical imaging device.