Method for reconstructing a 2D image from a plurality of X-ray images

Abstract

The present invention relates to a method and a system for producing X-ray images from an object. According to this invention, a shift-and-add method is used for generating a stack of linear tomography planes each associated with a different area inside the object. A set of shift values is defined from the consideration of ensuring that said stack of linear tomography planes fills in the tomographic volume with a spatial density adequate to the application. If so required by the application, some focal planes can be selectively processed for sharpness reduction in some areas to control the depth-of-field. A focus stacking method is used to synthesize a single 2D X-ray image from UPI the tomographic stack of images. A depth map of in-focus areas from the linear tomography stack can be used for creating a 3D object model.

Claims

1. A method for processing a plurality of X-ray images of an object that have been obtained using an X-ray source and an X-ray detector, wherein each of the images corresponds to a different positioning of the X-ray source, the X-ray detector, and the object relative to each other, the method comprising: using a shift-and-add method for generating a plurality of focal planes from the plurality of X-ray images, each focal plane being associated with a different area inside the object; and using a focus stacking method for generating a 2D image from the plurality of focal planes.

2. The method according to claim 1, wherein the X-ray detector is an area mode X-ray detector operating in continuous frame capture mode.

3. The method according to claim 1, wherein the X-ray source and the X-ray detector are kept at a fixed positional relationship with respect to each other during the obtaining of the plurality of X-ray images.

4. The method according to claim 3, wherein, during the obtaining of the plurality of X-ray images, the X-ray source and the X-ray detector are moved, and the object is kept stationary.

5. The method according to claim 4, further comprising digitally correcting a generated focal plane to account for scale distortion caused by a movement of a center of rotation of the combination of the X-ray source and the X-ray detector during the movement thereof to produce a corrected focal plane.

6. The method according to claim 1, further comprising equaling a size of the focal planes at least in a direction where a shift of the underlying X-ray images has been applied during the shift-and-add method.

7. The method according to claim 1, further comprising storing the plurality of X-ray images from the X-ray detector in computer memory prior to performing the shift-and-add method or the focus stacking method.

8. The method according to claim 1, further comprising using a depth map of respective areas that are in sharpness in respective focal planes for constructing a 3D model of the object.

9. The method according to claim 1, wherein the shift-and-add method or the focus stacking method is performed in real-time using a Graphic Processing Unit.

10. The method according to claim 1, further comprising selective image processing of some areas of the focal planes before performing the focus stacking method, to control a depth-of-field (DOF) in the generated 2D image.

11. The method according to claim 10, wherein the selective image processing of some areas of the focal planes before performing the focus stacking method comprises blurring areas that do not correspond to a region of interest of the object.

12. The method according to claim 1, further comprising associating a set of shift values to X-ray images among the plurality of X-ray images, shifting the individual X-ray images by their associated shift value, and adding the plurality of X-ray images after performing said shifting.

13. The method according to claim 12, further comprising selecting the set of shift values for the shift-and-add method to achieve a non-uniform spatial density of the plurality of generated focal planes.

14. The method according to claim 1, wherein the plurality of X-ray images have been obtained using an X-ray source that emits X-rays having an X-ray energy distribution that alternates between two or more different distributions, wherein the plurality of X-ray images comprises a plurality of subsets of X-ray images, each subset of X-ray images corresponding to one particular X-ray energy distribution, the method further comprising performing the shift-and-add method and the focus stacking method for each subset of X-ray images separately to generate a respective 2D image.

15. The method according to claim 14, further comprising generating a single 2D image from the respective 2D images.

16. An X-ray system, comprising: an X-ray source; an X-ray detector; a moving unit for causing a relative movement between an object to be imaged and at least one of the X-ray detector and the X-ray source during a process of obtaining a plurality of X-ray images; and an image processing unit configured to generate a plurality of focal planes from the plurality of X-ray images using a shift-and-add method, each focal plane being associated with a different area inside the object, and generate a 2D image from the plurality of focal planes using a focus stacking method.

17. The X-ray system according to claim 16, further comprising a mounting frame, wherein the X-ray source and the X-ray detector have a fixed positional relationship, wherein the X-ray system is configured to, using the moving unit, move the object relative to the X-ray source and X-ray detector, which are kept stationary relative to the mounting frame, or to move the X-ray source and the X-ray detector relative to the object, which is kept stationary relative to the mounting frame.

18. The X-ray system according to claim 16, further comprising a memory for storing the plurality of X-ray images from the X-ray detector prior to performing the shift-and-add method or the focus stacking method.

Description

(1) Next, the invention will be described in more detail by referring to the appended drawings, wherein:

(2) FIG. 1 presents two known configurations for X-ray detectors;

(3) FIG. 2 illustrates the concept of linear tomography;

(4) FIGS. 3A and 3B illustrate the focal trough;

(5) FIG. 4 illustrates the application of the shift-and-add method for generating a plurality of focal planes in correspondence with the present invention;

(6) FIG. 5 illustrates an example of a dental panoramic image;

(7) FIG. 6 illustrates the correction of a focal plane for spatial distortions;

(8) FIGS. 7A and 7B illustrate a comparison between dental panoramic images obtained using a prior art approach and an approach in accordance with the present invention;

(9) FIGS. 8A and 8B illustrate the dental focal trough in accordance with the present invention;

(10) FIG. 9 illustrates an application of the invention for security; and

(11) FIGS. 10A and 10B illustrate the principle of the shift-and-add method.

(12) X-ray imaging techniques predominantly use a cone-shaped X-ray beam. The reason is that practical sources of X-rays are X-ray tubes, which emit X-rays from a small area on the X-ray tube anode called the X-ray focal spot. Because the X-ray beam is conical there is a geometric magnification factor associated with a given X-ray setup. In most cases, it is advantageous to reduce the magnification factor as much as possible to avoid imaging artefacts like geometrical distortions and image blur caused by the finite X-ray tube focal spot size. To realize such, the SID in the imaging setup is usually selected to be as large as possible for a given power of the X-ray tube, expected exposure times, radiation leakage and other practical considerations. This approach is typical for both scanning and area mode X-ray imaging applications.

(13) Even if the magnification factor of the X-ray imaging setup is small, it still impacts the imaging results in linear tomography imaging with a TDI-mode detector. As mentioned above, only the object area, which has the same linear speed of moving over the detector image plane as the speed of TDI signal transfer, will be imaged in focus. This could result in only a part of the object being captured in focus in the TDI X-ray image.

(14) Certain X-ray detectors operate in TDI imaging mode by design, and the results of scanning X-ray examinations cannot be modified as the image of a single focal plane is outputted by the detector and it is being generated inside the detector in the analog domain. This situation is different for an area mode X-ray detector that is operating in continuous frame capture mode. The set of frames acquired from such detector during image acquisition can be digitally processed to generate multiple focal planes, using the shift-and-add method. The shift-and-add method constitutes in essence a ‘digital TDI’ image reconstruction technique. Therefore, the term ‘TDI’ will be applied to both traditional analog and digital TDI imaging modes without distinction, except when it is important to the subject matter. The shift value used for the shift-and-add method does not need to correspond exactly to the size of a single pixel or integer number times this size. It also could be a fraction of the native X-ray detector pixel size. The choice of shift values for the shift-and-add method is equivalent to adjustments made to the analog TDI signal transfer speed to control the relative speed of signal movement over the detector image plane, i.e. the smaller the shift value, the slower equivalent speed of object movement will be captured by the generated image and the closer the generated focal plane will be in space to the X-ray detector. This also implies that by varying the shift value it is possible to control the position of the corresponding focal plane over the SID. According to the invention, multiple focal planes can be generated from the same original set of X-ray images from the X-ray detector by varying the shift value for shift-and-add method. Generation or reconstruction of multiple focal planes from the same set of image data could substantially improve image quality and add valuable functionality in many X-ray imaging applications, examples of which will be provided below. The invention is however not limited to these applications.

(15) One of the X-ray imaging applications where the method of this invention could improve the quality of X-ray examination is dental panoramic X-ray imaging.

(16) FIG. 4 shows the principle of reconstructing multiple tomographic planes for an X-ray panoramic application using a circular geometry, which is chosen for simplicity. A gantry with an X-ray detector (11) and X-ray source (10) orbits the patient's head (12) at constant speed. The X-ray source (10) emits a narrow beam (13) of X-rays, which is captured by the X-ray detector (11). The X-ray detector (11) is operating in continuous frame capture mode at constant frame rate, producing a ‘movie’ M of X-ray images A, B, C, and so forth.

(17) According to the invention, multiple focal planes I_FP1-I_FP3 are produced using the shift-and-add method with different shift values. FIG. 4 shows the dependence of the position for generated focal planes FP1, FP2, and FP3, on the shift value (not to scale). A focal plane closer to X-ray source (10) requires a large shift value to be applied to the underlying X-ray images A, B, C, in order to generate the corresponding focal plane as features in this plane have higher relative speed of moving over the X-ray detector plane than the other shown planes.

(18) No single reconstructed focal plane captures all the required features of jaw anatomy in focus as the curvatures of the generated focal planes and the jaw profile are very different. FIG. 5 shows only a few focal planes reconstructed from digital data acquired by the X-ray detector. By changing the shift values for the shift-and-add method, a stack of focal planes can be reconstructed. Furthermore, the set of shift values for the shift-and-add method could be selected in the way the said stack of focal planes will fill in the volume occupied by the jaw with small spacing between the focal planes. An obvious inference to this is the fact that the image of interest is becoming ‘distributed’ over the totality of focal planes. According to the invention, in order to ‘extract’ a single panoramic X-ray image the totality of focal planes is processed using a focus stacking operation. An example of such an image is provided in FIG. 5.

(19) As stated above, the in-focus image contents in the focal planes at different distances from the X-ray source exhibit different magnification, and the position of the focal plane over the SID defines the magnification. It could be trivially shown that the single parameter that defines said position of the focal plane (FPn) over the SID, in a simplified geometry as shown in FIG. 4, is the shift value of the shift-and-add method used for reconstructing said focal plane. Therefore, the focal planes could be re-sized to the same magnification in digital domain, based solely on the respective shift values used in the shift-and-add method.

(20) Because the rotation of the gantry in actual commercial panoramic X-ray machines is not circular, the method above will not fully compensate for differences in magnification for generated focal planes in practical imaging conditions and the panoramic image will show some scale distortions. For example, the front (anterior) teeth will be shown larger in scale compared to the teeth in the molar (posterior) jaw area. FIG. 6 shows how the original focal plane (top) could be digitally corrected for scale distortions caused by movement of the center of rotation of the X-ray console, i.e. the gantry with the X-ray source and X-ray detector, to produce a corrected focal plane (bottom). This additional scaling depends on the specific design of the panoramic machine, which is known or can be derived through a straightforward calibration routine.

(21) It is important to recognize that this correction in the digital domain is not artificial and it achieves identical or similar results as other existing imaging techniques and algorithms for creating panoramic images. The reason is these other approaches are also compensating for magnification errors in the panoramic setup by other aspects of panoramic machine design, like the mechanical design of the rotating parts, X-ray detector frame rate, image reconstruction, etc. Any deviations of the actual anatomy of the patient from assumed curves of upper and lower jaws in these systems result in scale distortions in the traditional panoramic image too Small amounts of scale distortions in panoramic X-ray images are acceptable in the dental industry, as X-ray panoramic images are not used for biometrical purposes. Dental OEMs typically put a disclaimer in the user documentation of their systems, stating that images are produced based on assumptions of standard or average anatomy of the patient.

(22) Equipment for panoramic X-ray imaging exhibits a variety of system architecture decisions to enable panoramic imaging. Actual machines could be substantially different in the way the image acquisition is implemented. The type of orbital movement, detector frame rate, pixel size, detector orientation, SID, distance of X-ray detector from the patient, scanning time, could differ. These and other system level decisions will determine how many tomographic slices at which spacing will be required to generate a satisfactory panoramic X-ray image. FIG. 7 shows panoramic images of a dental skull phantom produced from the same set of frames captured by an area mode X-ray detector in a commercial panoramic imaging system. The image on the left is produced by an existing technique of reconstructing the panoramic image using the assumption of average patient anatomy, while the image on the right is produced using the method of this invention. Both images show very good sharpness and a high degree of similarity otherwise.

(23) FIG. 7B provides a zoomed view on the anterior region of the images in FIG. 7A. It can be seen that reconstruction using the method of this invention overcomes a slight misplacement of the phantom, and captures details of the front teeth with better resolution than the standard reconstruction technique. It is important to note that the image on the right in FIG. 7B, corresponding to the figure on the right in FIG. 7A, is produced without using any actual information regarding SID, X-ray detector frame rate, acquisition time, etc. from the dental X-ray system. The only assumption used in reconstruction is that both orbital mechanical movement and detector frame rate were kept constant during image acquisition.

(24) An additional benefit of the method of this invention in panoramic imaging applications, potentially simplifying the design of future dental panoramic systems, is that the large DOF provided by a synthetic 2D image produced by focus stacking of the generated focal planes removes the need for movement of the center of rotation of the gantry with the X-ray source and X-ray detector, enabling significant reductions in cost and complexity of the dental X-ray system. Simple circular rotation of the gantry will produce geometrical distortion caused by differences in magnification for anterior and posterior teeth, which can still be effectively corrected in the digital domain during post-processing of the image.

(25) FIGS. 8A and 8B show the approximate position of the focus trough for a panoramic imaging modality, or the DOF in more generic terms, using the method of this invention for producing a 2D panoramic image for a simple circular geometry (FIG. 8A) and an industry standard gantry rotation (FIG. 8B). The area of the focus trough expands over the minimum required area as shown in FIG. 3A. This expansion does not add any new features to the panoramic image, as the added areas contain only soft tissue not visible in the panoramic X-ray image because of the relatively hard spectrum of the panoramic X-ray beam that reaches the X-ray detector. Still, if the reduction or any other change to the shape or a position of the DOF is required, it can be achieved by intentional image blurring in the respective areas of relevant focal planes in the generated stack of focal planes. The focus stacking procedure will omit information from these areas and the resulting DOF will exclude the locations in the volume occupied by the processed areas. Said blurring could be achieved by, for example, applying a Gaussian blur filter or any other low pass image filtering method.

(26) The method of the invention produces a tomography image stack, in which image coordinates can be calculated to correspond precisely to spatial coordinates. This means that another dental X-ray imaging application, cephalometry, which is the analysis of the dental and skeletal relationships in a human head, is also suitable for the application of the method of this invention. The current approach for cephalometry is to use a large SID up to 2 meters in order to reduce the difference in magnification ratios at the front and the back of a human head. This allows for a cephalometric X-ray image to be obtained that can be used for measurement of angles, distances, etc., for the purpose of planning the targeted treatment. Modern dental equipment enables cephalometric imaging in dedicated versions of dental X-ray systems only, and these require a substantially larger working space in the X-ray room compared to panoramic-only X-ray systems. The method of this invention will allow for generating a cephalometric image using a similarly compact system geometry as used for panoramic imaging.

(27) Another medical X-ray application that could benefit from the method of this invention is mammography. A current trend in mammography is the growing proliferation of Digital Breast Tomosynthesis (DBT), which promises a noticeable increase in sensitivity, i.e. the ability to detect objects inside the breast, and specificity, i.e. the ability to classify detected objects into malignant or benign classes, over traditional 2D X-ray mammography. Primary objects of interest for X-ray mammography are either large areas exhibiting a small contrast difference to normal glandular tissues, i.e. tumors, or objects with high contrast and dimensions smaller than or comparable to the X-ray detector pixel size, i.e. calcifications. In both cases, the mammography diagnostic system should exhibit excellent contrast sensitivity to enable positive identification and further classification of these objects of interest by a trained radiologist. In patients with dense breast tissue, the X-ray beam attenuation is higher than typical and objects of interest show reduced visibility when using traditional mammography techniques. This increases the false negatives diagnostics rate, and results in an increase in missed malignant cancers.

(28) The DBT technique requires the acquisition of multiple images at different angles of incidence of the X-ray beam while the breast and X-ray detector are held static. Because the internal structure of the breast is observed at different angles, the detectability of objects inside the breast is improved. A DBT stack of slices is generated from the totality of acquired individual images at different angles and presented to the radiologist as a stack of slices for individual review. This is often accompanied by a single synthetic 2D image containing a ‘summary’ of objects of interest from all slices.

(29) This invention significantly improves the way the DBT procedure is executed. From a single scan of a breast with an area mode X-ray detector operating in continuous frame capture mode, the full tomosynthesis stack of slices can be generated. System image acquisition and image generation for a DBT system utilizing the method of this invention will not be different from the description of the panoramic X-ray modality.

(30) As previously explained, the narrow DOF of X-ray detectors operating in TDI mode limits their use to niche applications. An example of narrow DOF applications like this is X-ray PCB solder inspection. In this case, the area of interest is a shallow plane and the TDI DOF is adequate for the imaging requirements in this application.

(31) Because of their large DOF and ability to image continuously while providing output imaging data in real time, the majority of scanning X-ray imaging is still performed using line-scan X-ray detectors. This invention eliminates the weaknesses of TDI X-ray imaging, and allows for upgrading from line-scan X-ray detectors to better performing X-ray imaging devices that offer a number of very substantial benefits in various application areas.

(32) Industrial and security X-ray inspection applications of continuously moving objects are currently almost exclusively using line-scan X-ray detectors. An example of such application is airport security baggage inspection. A large DOF is a critical performance requirement for this application and, in combination with the low cost of line-scan X-ray detectors, results in the dominant position of line-scan X-ray detectors in these types of applications. This invention overcomes the inherent DOF limitations of the otherwise advantageous TDI imaging principle, and makes it possible to upgrade the application with better performing imaging technology.

(33) FIGS. 9A-9C illustrate the image quality improvements when applying the method according to the invention for the case of conveyer belt applications. The object for examination was a cardboard box approximately 20 cm high and filled with different objects. The transportation system allowed for linear movement of the object in relation to a stationary X-ray source and X-ray detector, while the X-ray detector was operating in continuous frame capture mode.

(34) As explained earlier, the speed control of TDI operation of the X-ray detector allows for adjustment of the position of TDI focal plane over the SID. FIG. 9 (left) and FIG. 9 (middle) show the imaging results when the TDI focal plane is at the bottom and at the top of the box, respectively. It is obvious that the narrow DOF of a standard TDI operation makes it impossible to reveal the contents of the box with adequate overall sharpness. The image in FIG. 9 (right) shows the synthetic 2D image from a tomographic stack of 19 images, generated in accordance with the present invention using the same data as used to generate the images in FIG. 9 (left) and (right). All of the objects in the image in FIG. 9 (right) are in focus and are easily recognizable as the DOF for this experiment has been effectively increased to 20 cm.

(35) All embodiments above are provided for imaging modes using a poly-energetic X-ray beam, and X-ray detectors generating an output signal proportional to the total amount of energy captured by the pixel. Substantially higher information content can generally be produced by X-ray images where the X-ray setup is capable of energy discrimination. Additional information from spectrum changes for the object under study provides means to differentiate between various materials or differences in composition of the object, as well as increases the dynamic range of the X-ray images. The proposed method of this invention is not dependent on the technology for X-ray capture in the X-ray detector, and therefore is in general compatible with dual energy and energy discrimination principles of X-ray imaging using energy discrimination at detector level.

(36) Another method of dual-energy imaging can be realized by imaging with alternating X-ray tube high voltage. An example of such application is Contrast Enhanced Digital Mammography (CEDM), where the difference between two breast images taken each having different high voltage settings applied to the X-ray tube is evaluated. Images are taken within quick succession with contrast agent injected into the breast before the examination. Using the method of this invention would allow performing contrast enhanced DBT, which is beyond the capabilities of any of commercial DBT system currently on the market.

(37) In general, the method of the invention allows the use of the generated 2D and 3D images for performing measurements of distances, angles, etcetera between the visualized features of the object with high precision, e.g. for biometrical purposes.

(38) FIGS. 10A and 10B illustrate the known principle of the shift-and-add method that is used in the present invention. In FIG. 10A, the X-ray source and X-ray detector have a fixed mutual positioning and are moved between two positions relative to an object having internal features A and B. More in particular, the X-ray source is at a position S.sub.1 when recording a first image, depicted on the left in FIG. 10B, and at a position S.sub.2 when recording a second image depicted on the right in FIG. 10B. Similarly, FIG. 10A illustrates two center lines C.sub.1, C.sub.2 that represent the center line of the X-ray detector at the time of recording image 1 and image 2, respectively. In FIG. 10A, the movement of the X-ray detector and X-ray source is such that a distance between feature A of the object and the X-ray detector remains at value h.sub.1, whereas a distance between feature A of the object and the X-ray source, or at least a line on which the X-ray source moves, is denoted by h-h.sub.1. Here, h denotes the source to image distance (SID).

(39) Line 101 indicates an X-ray that is emitted by the X-ray source at position S.sub.1 and which passes through features A and B inside the object to be imaged. Consequently, in image 1 these features will appear at the same position in the image.

(40) Line 102 indicates an X-ray that is emitted by the X-ray source at position S.sub.2 and which passes through feature A inside the object to be imaged. As can be seen, line 102 does not pass through feature B. Consequently, in image 2 these features will not appear at the same position in the image.

(41) FIG. 10A illustrates that feature A is at a distance h.sub.1 relative to the X-ray detector. Using conventional geometry, one can deduce that d.sub.1/h.sub.1=d/(h−h.sub.1). The offset between C.sub.1 and feature A in image 1, which equals the offset between C.sub.1 and feature B in image 1, is denoted by O.sub.A,B,1 and can be calculated using O.sub.A,B,1=d+d.sub.1=(h/(h−h.sub.1))×d.

(42) Using conventional geometry, one can also deduce that d.sub.2/h.sub.1=(d+dy)/(h−h.sub.1). The offset between C.sub.2 and feature A in image 2 is denoted by O.sub.A,2 and can be calculated using O.sub.A,2=dy+d+d.sub.2=(h/(h−h.sub.1))×d+(h/(h−h.sub.1))×dy.

(43) Comparing images 1 and 2, it can be concluded that feature A has been shifted in these images by an amount of D=O.sub.A,B,1−O.sub.A,2=(h/(h−h.sub.1))×dy.

(44) According to the invention, a new image can be constructed by using images 1 and 2, wherein the positional shift is compensated for. For instance, to compute a pixel value in the combined image at position (x,y), the pixel value of image 1 at position (x,y) could be added to the pixel value of image 2 at position (x, y+D). In this manner, an image will be obtained wherein features of the object in the plane defined by z=h.sub.1 will be in focus and wherein other features are blurred. Such image is referred to as a focal plane.

(45) A plurality of focal planes can be calculated for various values of z. This allows focused information to be obtained for different positions z inside the object. These focal planes can be used for generating a final image using a focus stacking method.

(46) From the equation D=O.sub.A,B,1−O.sub.A,2=(h/(h−h.sub.1))×dy stated above, it can be concluded that if the focal plane corresponds to an area inside the object close to the X-ray source, i.e. h.sub.1 approaches h, D will become very large. On the other hand, if the focal plane corresponds to an area inside the object close to the X-ray detector, D will approach dy.

(47) Although the invention is explained using detailed embodiments thereof, the skilled person readily understands that various modifications are possible without deviating from the scope that is defined by the appended claims and their equivalents.