Estimation of lung volume from radiographic images.

Abstract

A computer-implemented method for estimating lung volume from radiographic images is provided, comprising the steps of: registering a two-dimensional radiographic image 12 of a patient's chest with a three-dimensional radiographic image 14 of the patient's chest to estimate 102 data describing a projection geometry 16 of an imaging device setting used to capture the two-dimensional radiographic image, estimating 104 at least one radiographic magnification factor 18 related to the imaging device setting, and calculating 106 an estimated lung volume 20 using the two-dimensional radiographic image and the at least one radiographic magnification factor.

Description

[Technical field]

[0001] The present invention relates to a system and method for estimating lung volume from radiographic images. [Background technology]

[0002] X-rays are a standard tool for tracking lung status in intensive care units (ICUs). Often, X-rays are taken every other day to check for lung changes, especially pleural effusion and atelectasis. Recently, it has been proposed to move to a quantitative assessment of X-ray images for the total air volume in the lungs to provide more reliable information on the patient's condition. Estimation of lung volume from a single X-ray image requires knowledge of the projection geometry. However, X-rays are often taken with a mobile detector placed under the patient in or under the bed, and therefore there is no constant source-detector distance. Furthermore, the patient's position may change due to other requirements, such as for example to ventilate the patient in prone or supine position. Such changes in projection geometry pose a difficult challenge for accurate lung volume estimation. Summary of the Invention [Problem to be solved by the invention]

[0003] The present invention addresses one or more of the concerns set forth above. [Means for solving the problem]

[0004] To better address one or more of these concerns, in a first aspect of the invention there is provided a computer-implemented method for estimating lung volume from radiographic images, the method comprising the steps of: registering a two-dimensional radiographic image of a patient's chest to a three-dimensional radiographic image of the patient's chest to estimate data describing a projection geometry of an imaging device setting used to capture the two-dimensional radiographic image, estimating at least one radiographic magnification factor for the imaging device setting using the projection geometry, and calculating an estimated lung volume using the two-dimensional radiographic image and the at least one radiographic magnification factor.

[0005] The step of registering the two-dimensional (2D) radiographic image and the three-dimensional (3D) radiographic image can be performed using at least one trained machine learning model. Additionally or alternatively, the registration can be performed using at least one computational or geometric registration algorithm.

[0006] At least one trained machine learning model can be used to align the 2D image to the 3D image and estimate data describing the projected geometry.

[0007] In one example, the geometric registration algorithm comprises an intensity-based registration algorithm that compares intensity patterns in two images via one or more correlation measures. For example, the geometric registration algorithm may be based on minimizing an intensity similarity measure between both images while varying a pose parameter. Additionally or alternatively, the registration algorithm may comprise a feature-based registration algorithm that finds matches between one or more image features, such as points, lines, contours, etc.

[0008] The registration may register the 2D and 3D images as a whole, or may use sub-images of one or both of the 2D and 3D images. Thus, the registration may include forming sub-images from one or both of the 2D and 3D images before performing the registration. When the sub-images are registered, they may depict at least one anatomical structure, i.e., at least one object of interest, to be used for the registration. In one specific, non-limiting example, the at least one anatomical structure includes one or more vertebrae. Additionally or alternatively, the at least one anatomical structure may include one or more ribs. The sub-images may be formed or generated by segmenting the 2D and 3D images to isolate at least one anatomical structure, e.g., segmenting at least one vertebra. Thus, the 2D image may be segmented to identify a region of interest depicting the anatomical structure, while the 3D image may be segmented to identify a volume of interest depicting that same anatomical structure.

[0009] The registration algorithm may include an iterative registration algorithm. Thus, in one example, comparing the intensity patterns in the two images may include using an iterative registration algorithm that iteratively performs incremental changes to the estimated projection geometry and calculates the correlation between the two images that are registered using the incrementally changed projection geometry until the correlation converges to a steady-state value. The steady-state value may be a maximum or optimal value. Thus, the method may include using an optimization algorithm to maximize the correlation. To compare a 2D image with a 3D image, the method may include first reconstructing the 2D image from the 3D image, for example, using the estimated projection geometry selected for the current iteration. The method may include subtracting the reconstructed 2D image (calculated from the 3D image) from the 2D image before calculating the correlation. The iterative algorithm may use a predetermined starting estimate of the projection geometry, a predetermined increment size, and / or a predetermined order in which the parameters of the projection geometry are increased. The correlation measure may include a similarity measure that quantifies the similarity between the intensity patterns in the two images. In one specific, non-limiting example, the similarity measure describes the structure of a difference image obtained by subtracting one of the two registered images from the other. Other examples of suitable similarity measures include cross-correlation, mutual information, sum of squared intensity differences, and ratio image uniformity.

[0010] The method may include pre-processing one or both of the two images before performing the registration, for example one or both of the images may be pre-processed to adjust for greyscale variations, for example by removing the mean grey value from regions outside the object of interest and / or by applying one or more scaling factors.

[0011] The projection geometry can be described by a transformation, i.e., a geometric transformation, that can relate different coordinate systems used by, for example, 2D and 3D images to map one image to the other. The transformation can be represented by one or more transformation parameters, such as a rotation parameter and a translation parameter. In one example, up to nine parameters can be used to represent the projection geometry: three for the source position, three for the detector position, and three for the detector or detector plane orientation. Constraints imposed by the imaging device setup, such as a fixed source-detector distance or a source mounting that can rotate around fewer than three axes, may allow fewer than nine parameters (i.e., a subset of parameters) to be used to fully represent the projection geometry.

[0012] In one example, the expansion factor for at least one lung can be estimated using at least the source-lung distance. More specifically, the expansion factor for the lung can be calculated as the ratio of the source-lung distance to the source-detector distance (i.e., the sum of the source-lung distance and the lung-detector distance). Even more specifically, the expansion factor for the lung can be calculated as:

number

number

number

[0013] In another example, the magnification factor of at least one lung may be estimated using at least the source-anatomical structure distance (e.g., the distance between the X-ray source and the anatomical structure used for the registration) together with the lung-anatomical structure distance (e.g., the distance between the lung or lungs and said anatomical structure). Using the lung-anatomical structure distance in the calculation may include using knowledge of the position in which the patient was lying when the 2D image was captured (e.g., prone or supine), for example, to select an appropriate estimation method. In other words, the method may include applying the lung-anatomical structure distance differently depending on whether the patient is in supine or prone position. More specifically, the magnification factor for the lung or lungs may be:

number

number

number

number

number

[0014] The step of estimating at least one radiographic magnification factor may include estimating a magnification factor for at least one lung (e.g., for one or both lungs) if the same magnification factor is applied for both, or separately for each lung if different magnification factors are applied. The magnification factor may be different for the left and right lung if the patient is positioned obliquely in the beam, i.e., if the midpoint or center of the two lungs is not equidistant from the detector plane. In the above calculations, if the volume of only one lung is estimated, a single source-lung distance

number

number

number

number

[0015] In some examples, the vertebral magnification coefficients can be calculated either as global magnification coefficients (e.g., by taking the results of a single vertebra or averaging across multiple vertebrae) or as spatially varying magnification coefficients that vary along the cranio-caudal direction by interpolating magnification coefficients determined for different vertebrae.

[0016] Distances described herein, such as source-lung, lung-detector and lung-anatomical structure distances, can be calculated using a 3D model constructed using the determined projection geometry data. The 2D / 3D registration allows the positions and orientations of objects, such as source, detector and anatomical structures, to be described in a single coordinate system, making the object distances easily determinable. When calculating distances to or from anatomical structures, the mid-planes or center points of these structures can be taken as reference points, e.g., the center point of the lung can be used when calculating the distance from the source or detector to the lung.

[0017] Once the expansion factors are determined, any suitable method can be used to estimate the lung volume based on the expansion factors. In an example where the estimation of the lung volume is based on lung regions determined from a 2D image, the expansion factors for the lung or lungs can be used to scale the lung regions before estimating the lung volume using the scaled lung regions. In other words, the step of calculating the estimated lung volume includes a step of scaling the lung regions determined using the two-dimensional radiographic image using at least one lung expansion factor. If separate expansion factors are estimated for the left and right lungs, these expansion factors can be used to scale the regions of each lung before estimating the lung volume. That is, the method can include a step of scaling the lung regions of the left and right lungs separately using the respective expansion factors determined for the left and right lungs. In this way, different scaling can be applied to the left and right lungs. The lung regions can be scaled by 1 / M 2where M is the expansion factor for each lung. Calculating the estimated lung volume may include integrating pixel intensity values ​​over the scaled lung region. Similarly, the expansion factor may be used to scale the volumes calculated for individual pixels, their sum, or the water equivalent path length.

[0018] In one example, calculating the estimated lung volume may include estimating, for each pixel in the lung region of the 2D image, the volume that contributes to its intensity value, the volume being approximately equal to the pixel size times the water equivalent path length, further corrected for magnification. More specifically, the volume is V=lS / M 2 where 1 is the water-equivalent path length, S is the pixel size, and M is the expansion factor of the lung. The water-equivalent path length can be estimated using any suitable known method.

[0019] To account for the local tilt angle of the x-rays onto the detector, estimating the volume for each pixel may include determining an effective pixel size for use in the above calculations instead of the actual pixel size. If the x-rays strike the detector perpendicularly, the effective pixel size S eff can be made equal to the physical pixel size S. Otherwise, for tilt angles α ≠ 90°, the (smaller) effective pixel size, i.e., S eff = Scos(90°-α). The tilt angle for each pixel can be determined using the 3D model described above.

[0020] In another example, calculating the estimated lung volume includes segmenting the two-dimensional radiographic image to identify the left and right lungs, modifying the two-dimensional radiographic image to replace image data in regions corresponding to the left and right lungs with predefined pixel intensity values, thereby forming a chest mask image, and subtracting the chest mask image from the original two-dimensional radiographic image to form a lung-only image that provides lung volume information. The predefined values ​​may correspond to pixel intensity values ​​expected for soft tissue or water, for example. Calculating the estimated lung volume may further include integrating pixel intensity values ​​over the lung-only image based on a known correlation between the area and intensity of pixels in the regions corresponding to the left and right lungs and the total lung volume. The region over which the intensity values ​​are integrated may be a scaled lung region determined based on the expansion factor of the lung or lungs.

[0021] According to a second aspect, there is provided a computer system configured to perform any of the methods described herein.

[0022] According to a third aspect there is provided a computer program product comprising instructions which, when executed by a computer system, enable or cause the computer system to perform any of the methods described herein.

[0023] According to a fourth aspect, there is provided a computer readable medium comprising instructions which, when executed by a computer system, enable or cause the computer system to perform any of the methods described herein.

[0024] The systems and methods disclosed herein provide improved lung volume estimation from radiographic images. Accurate estimation of lung volume provides valuable quantitative information about the patient's condition in the ICU. However, quantitative assessment requires knowledge of the system geometry and the patient's position within the system, which is typically not available in ICU settings that use mobile X-ray systems. According to the present disclosure, this problem is overcome by estimating the projection geometry by using previously acquired CT data.

[0025] A "two-dimensional radiographic image" as used herein may include an X-ray image or a radiographic image, in particular a projection radiographic image, more particularly one captured using a moving detector, such as a flat panel detector. A two-dimensional radiographic image, as a planar image, comprises an array of pixels representing the spatial intensity distribution of incident x-rays.

[0026] A "three-dimensional radiographic image" as used herein may include, for example, a computed tomography (CT) image or a radiographic image that includes a volume of voxels that represent the spatial intensity distribution as a volumetric image.

[0027] The present invention may include one or more aspects, examples or features, whether or not specifically disclosed in combination or in isolation, taken alone or in combination. Any optional feature or sub-aspect of one of the above aspects also applies to any of the other aspects, as appropriate.

[0028] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

[0029] The detailed description will now be given, by way of example only, with reference to the accompanying drawings, in which: [Brief description of the drawings]

[0030] [Figure 1]FIG. 1 is a flow chart illustrating a method for estimating lung volume from radiographic images. [Figure 2A] FIG. 2A illustrates one embodiment of the method of FIG. [Figure 2B] FIG. 2B illustrates another embodiment of the method of FIG. [Figure 2C] FIG. 2C illustrates another embodiment of the method of FIG. [Figure 2D] FIG. 2D illustrates another embodiment of the method of FIG. [Figure 2E] FIG. 2E illustrates another embodiment of the method of FIG. [Figure 2F] FIG. 2F illustrates another embodiment of the method of FIG. [Diagram 3] FIG. 3 shows the projection geometry in one X-ray imager setup. [Figure 4A] FIG. 4A illustrates the formation of a difference image as part of an example of 2D / 3D image registration. [Figure 4B] FIG. 4B illustrates the formation of a difference image as part of an example of 2D / 3D image registration. [Figure 4C] FIG. 4C illustrates the formation of a difference image as part of an example of 2D / 3D image registration. [Figure 4D] FIG. 4D illustrates the formation of a difference image as part of an example of 2D / 3D image registration. [Figure 5A] FIG. 5A shows an exemplary method for estimating the expansion factor of the lungs using the determined projection geometry. [Figure 5B] FIG. 5B shows an exemplary method for estimating the expansion factor of the lungs using the determined projection geometry. [Figure 6] FIG. 6 illustrates another exemplary method for performing lung volume estimation. [Figure 7] FIG. 7 illustrates an x-ray imager setup that may be used to capture an x-ray image. [Figure 8] FIG. 8 illustrates a computer system that may be used in accordance with the systems and methods disclosed herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0031] According to the present disclosure, lung volume estimation can be improved by estimating the projection geometry using previously acquired CT images, in particular by estimating the magnification factor to be applied to the X-ray imager settings using 2D / 3D registration of parts of the previously acquired CT images. Often such chest CT images are available for patients entering the ICU. FIG. 1 is a flow chart showing a method 100 for estimating lung volume from radiographic images according to the present disclosure. Broadly speaking, the method 100 comprises the following steps: In a first step 102, a two-dimensional radiographic image 12 and a three-dimensional radiographic image 14 of a patient's chest are received, the two-dimensional radiographic image 12 having been taken with an unknown projection geometry, and the two images 12, 14 are registered to estimate a projection geometry 16. In a second step 104, the projection geometry 16 is used to estimate at least one radiographic magnification factor 18 associated with the imager settings used to capture the two-dimensional radiographic image 12. In a third step 106 , an estimated lung volume 20 is calculated using the two-dimensional radiographic image 12 and at least one radiographic magnification factor 18 .

[0032] 2A-2F show an embodiment of the method of FIG.

[0033] FIG 2A shows a 3D image 14 of a patient's chest. The 3D image 14 is a CT image that has been segmented to identify relevant anatomical structures, in this case the spine and lungs. In particular, the sixth thoracic vertebra has been identified. The segmentation also provides the 3D relative positions of the relevant anatomical structures. For example, the arrows shown in FIG 2A point from the center of the sixth thoracic vertebra to the centers of the left and right lungs, respectively.

[0034] 2B shows the estimation of the projection geometry 16 by 2D / 3D registration of images for the sixth thoracic vertebra. The projection geometry data includes values ​​of nine parameters that uniquely describe the position of the X-ray source 200 and the position and orientation of the X-ray detector 202 in a vertebra-centered coordinate system.

[0035] FIG. 2C shows the optional step of performing a tilt correction to account for the effect that x-rays with non-normal incidence angles have on the effective pixel size. eff depends on the local tilt angle of the X-rays with respect to the detector 202. If the X-rays strike the detector 202 perpendicularly, the effective pixel size S eff is equal to the physical pixel size S. For tilt angles α ≠ 90°, the effective pixel size is smaller, i.e., S eff = Scos(90°-α). Therefore, the following calculations can optionally use the effective pixel size instead of the physical pixel size to perform the skew correction.

[0036] FIG. 2D illustrates the estimation of the magnification factor 18 in step 104. As mentioned before, after 2D / 3D registration, the location of the centers of the left and right lungs relative to the center of the sixth thoracic vertebra is known. Using that information, the distance from the source 200 to the center of the right lung is calculated as

number

number

number

[0037] FIG. 2F illustrates the estimation of lung volume 20 in step 106. First, the water-equivalent path length through the lungs is estimated as known in the art. The next step is to estimate, for each pixel, the volume that contributed to that value. This volume is approximately equal to the water-equivalent path length multiplied by the effective pixel size additionally corrected for magnification. FIG. 2F illustrates the volume 204 that contributes to the value measured by one particular pixel of detector 202. The estimated water length corresponds to a larger volume if the corresponding volume is closer to the detector. The estimated volume V, corresponding to the estimated water length l, is then calculated as V=lS eff / M 2 The volumes estimated in this way for individual pixels are summed over the lung regions in the 2D image 12 to estimate the lung volume 20 of one or both lungs.

[0038] Returning to step 102, the registration of the 2D and 3D images 12, 14 to estimate the projection geometry 16 is described. In this example, the 2D image 12 is a recent X-ray image of the patient's chest to be processed to estimate the current lung volume, and the 3D image 14 is a previously acquired CT image of the patient's chest captured when the patient first enters the ICU. The registration is performed for one or more vertebrae that represent the anatomical structures that serve as the objects of interest. By registering the X-ray image 12 to the CT image 14, the positions of the vertebrae in the X-ray imaging setup can be estimated. The following example of a suitable 2D / 3D image registration process is based on that described in the paper "Voxel-based 2-D / 3-D Registration of Fluoroscopy Images and CT Scans for Image-guided Surgery" by WEESE et al., IEEE Transaction on Information Technology in Biomedicine, 1(4), 1997.

[0039] 3 shows the projection geometry. The X-ray image 12 is captured using an X-ray source 200 and a moving X-ray detector 202 that is in the projection plane. The X-ray image 12 is associated with a coordinate system (x,y,z) whose origin is at the center of the X-ray image 12. The projection plane is located in the xy plane of the coordinate system such that the source 200 is located on top of the projection plane in the z direction. The CT image 14 has its own coordinate system (x',y',z') with its origin at the center of the CT image 14, which may also be the center of a particular vertebra after image segmentation. The position and orientation of the CT image 14 with respect to the coordinate system of the X-ray image 12 can be described by the transformation x=R(ω)x'+t, where R(ω) is

number

number

[0040] The 2D / 3D registration algorithm determines the transformation parameters ω and t given the starting estimates ω and t by performing the following steps: 1) Segmentation of the vertebrae of interest in the CT image 14; 2) subtraction of the mean gray value of the tissue surrounding the vertebrae from the CT images 14; 3) Calculation of a reconstructed 2D image from the CT image 14 using the current values ​​of the rotation and translation parameters ω and t and taking into account only the CT volume with the segmented vertebrae. Since the mean grey value of the tissues directly adjacent to the vertebrae has been subtracted from the CT image 14, the reconstructed 2D image only shows the grey value changes due to the presence of the vertebrae; 4) Scaling the grey values ​​in the reconstructed 2D image by a predefined scaling factor I0 and subtracting it from the X-ray image 12 to form a difference image. With proper grey value scaling and correct position and orientation of the CT image 14, structures corresponding to the vertebrae in the X-ray image 12 will disappear and overall less structures will be visible after subtraction. FIG. 4A shows an example of an X-ray image 12 of a patient's spine, while FIGS. 4B-4D show three difference images 402-406 formed by subtraction of 2D images reconstructed from a CT volume for each of three different segmented vertebrae. The disappearance of each vertebra in the difference images 402-406 indicates a proper determination of the transformation parameters; 5) Calculation of a similarity measure that characterizes the structure of the difference image formed after subtraction of the gray value scaled reconstructed 2D image; 6) Optimization of the similarity measure with respect to the gray value scaling factor I0 and the parameters ω and t.

[0041] Within each iteration of the optimization, steps 3) to 5) are repeated. The transformation parameters are determined according to a predefined sequence, e.g., t x , ty , t z , ω x , ω y , ω z The optimization is performed iteratively until a predetermined number (e.g., 6) of successive maximizations do not further increase the similarity. The gray value scaling factor I0 may be adjusted by varying it in predetermined increments between 0 and a maximum value each time one of the other parameters is changed. Maximization with respect to a single parameter can be performed with an adjustable step size Δ, which is determined by multiplying N limit If successive steps increase the similarity, then (N limit If neither the forward nor backward steps increase the similarity, the step size Δ is the minimum value Δ min It will be reduced as long as it is greater than (max(Δ / N limit ,Δ min ) → Δ). For each parameter, the step size Δ and the minimum step size Δ min At the start of the optimization, the step size Δ is set to its minimum value Δ min It is initialized to a multiple of .

[0042] The similarity measure that characterizes the structuring can be implemented using a function of the difference image that assigns values ​​close to 0 to points in the vicinity of structures such as edges or lines in gray value, while assigning values ​​around 1 to points in areas that show only small changes in gray value. In one non-limiting example, the function is:

number

[0043] Another example of performing 2D / 3D image registration by geometrically determining pose parameters using landmarks is described in the article by YANG, H. et al., “Geometry Calibration method for a cone-beam CT system”. Med. Phys. 44(5), May 2017, pp. 1692-1706.

[0044] 5A illustrates another exemplary method for estimating the lung expansion factor 20 using the projection geometry 16 determined for a supine patient. The lung expansion factor is

number

number

[0045] 5B illustrates yet another exemplary method for estimating lung expansion factors 20 using the projection geometry 16 determined for a patient in an oblique position. In this example, separate lung-to-vertebrae distances for the left and right lungs are used to calculate the expansion factor for each lung.

number

number

[0046] 6 illustrates another exemplary method for performing lung volume estimation. Using an X-ray image 12 of a patient's chest, a lung mask image 602 is formed after segmenting the X-ray image 12 to identify the left and right lungs. The lung mask image 602 is then used to modify the X-ray image 12, replacing image data in the regions identified as belonging to the left and right lungs with pixel intensity values ​​expected for soft tissue, thereby forming a chest mask image 604. The chest mask image 604 is subtracted from the X-ray image 12 to form a lung-only image 606. At any point after the lungs are identified, the lung region is scaled using the magnification factor or factors determined for the lungs. The lung volume 20 is estimated by integrating pixel intensity values ​​in the lung-only image 606 over the scaled lung region.

[0047] Other suitable methods for estimating lung volume are described in further detail, for example, in U.S. Pat. No. 9,805,467 or U.S. Patent Application Publication No. 2018 / 0271465.

[0048] 7 shows an X-ray imager setup 700 that can be used to capture an X-ray image 12. The X-ray source 200 is repositionable to face a portable X-ray detector 202. In the illustrated example, the detector 202 is mounted on a stand 702 while the source 200 is provided on a rotatable mount 704, allowing the source to be pivoted to a horizontal position and directed towards the stand-mounted detector 202 for imaging of a patient in a standing position. As another example, imaging can be performed with the patient lying on a bed 706 and the detector 202 positioned above or below the bed 706. The X-ray image 12 is received by a computer system 800, further described below, for processing according to the methods described herein.

[0049] Many variations of the systems and methods described herein are envisaged, for example, although the examples given above relate to using vertebrae to form a correspondence between the X-ray image 12 and the CT image 14, in principle other anatomical structures such as bones, and in particular ribs, could also be used.

[0050] Alternative approaches to 2D / 3D registration, such as contour-based ones, can also be used.

[0051] The present invention has general application to mobile x-ray, such as for use in ICUs where clinical need is greatest, but is not so limited. The present invention is also applicable to quantitative dark field x-ray imaging, where lung volume estimation is used to estimate quantitative information about lung structure.

[0052] 8 illustrates an exemplary computer system 800 that can be used in accordance with the systems and methods disclosed herein. The computer system 800 can form part of or comprise any desktop, laptop, server, or cloud-based computer system. The computer system 800 includes at least one processor 802 that executes instructions stored in a memory 804. These instructions can be, for example, instructions for performing a function described as being performed by one or more elements described herein or instructions for implementing one or more of the methods described herein. The processor 802 can access the memory 804 via a system bus 806. In addition to storing executable instructions, the memory 804 can also store conversational inputs, scores assigned to the conversational inputs, etc.

[0053] Computer system 800 also includes a data store 808 accessible by processor 802 via system bus 806. Data store 808 may include executable instructions, log data, etc. Computer system 800 also includes an input interface 810 that allows external devices to communicate with computer system 800. For example, input interface 810 may be used to receive instructions from an external computing device, a user, etc. Computer system 800 also includes an output interface 812 that interfaces computer system 800 with one or more external devices. For example, computer system 800 may display text, images, etc. via output interface 812.

[0054] It is contemplated that the external devices communicating with computer system 800 via input interface 810 and output interface 812 may be included within an environment that provides substantially any type of user interface with which a user may interact. Examples of types of user interfaces include graphical user interfaces, natural user interfaces, and the like. For example, a graphical user interface may receive input from a user using an input device such as a keyboard, mouse, remote control, and the like, and provide output to an output device such as a display. Furthermore, a natural user interface may allow a user to interact with computer system 800 without the constraints imposed by input devices such as a keyboard, mouse, remote control, and the like. Rather, a natural user interface may rely on voice recognition, touch and stylus recognition, gesture recognition both on and near the screen, air gestures, head and eye tracking, voice and speech, vision, touch, gestures, machine intelligence, and the like.

[0055] Moreover, although illustrated as a single system, it should be understood that computer system 800 can be a distributed system such that, for example, several devices can communicate over a network connection and jointly perform the tasks described as being performed by computer system 800.

[0056] Various functions described herein can be implemented in hardware, software, or any combination thereof. If implemented in software, the functions can be stored or transmitted as one or more instructions or code by a computer-readable medium. A computer-readable medium includes a computer-readable storage medium. A computer-readable storage medium can be any available storage medium accessible by a computer. By way of example, and not limitation, such computer-readable storage media can include flash storage media, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. A disk as used herein includes a compact disk (CD), a laser disk, an optical disk, a digital versatile disk (DVD), a floppy disk, and a Blu-ray disk (BD), where the disk typically reproduces data magnetically or optically using a laser. Additionally, propagated signals are not included within the scope of computer-readable storage media. Computer-readable media also includes communication media, including any medium that facilitates transfer of a computer program from one place to another. For example, a connection may be a communication medium. For example, if the software is transmitted from a website, server or other remote source using coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL), or wireless technologies such as infrared, radio wave, microwave, etc., the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio wave, microwave, etc. are included within the definition of communication media. Combinations of the above should also be included within the scope of computer-readable media.

[0057] Alternatively or in addition, the functionality described herein may be implemented, at least in part, by one or more hardware logic elements. For example, without limitation, exemplary types of hardware logic elements that may be used include field programmable gate arrays (FPGAs), program specific integrated circuits (ASICs), program specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), etc.

[0058] It will be understood that the circuits described above may have other functions in addition to those mentioned, and that these functions may be performed by the same circuit.

[0059] Applicant hereby discloses each feature described herein separately and combinations of two or more such features to the extent that such features or combinations could be implemented in accordance with the specification as a whole in light of the ordinary skill in the art, regardless of whether such features or combinations solve any problems disclosed herein, and without limitation to the scope of the claims. Applicant indicates that aspects of the invention may consist of any such individual feature or combination of features.

[0060] It should be noted that embodiments of the present invention are described with respect to different categories. In particular, some examples are described with respect to methods, while other examples are described with respect to devices. However, a person skilled in the art will understand from the description that, unless expressly stated otherwise, any combination of features belonging to one category as well as any combination between features relating to different categories is considered to be disclosed by the present application. However, all features can be combined to provide synergistic effects that go beyond a simple collection of features.

[0061] While the present invention has been illustrated and described in detail in the drawings and the foregoing description, such illustrations and descriptions are to be considered as illustrative and not restrictive. The present invention is not limited to the disclosed embodiments. Those skilled in the art can understand and effect other variations of the disclosed embodiments from a study of the drawings, the disclosure, and the appended claims.

[0062] The word "comprising" does not exclude other elements or steps.

[0063] The singular does not exclude a plurality. Moreover, as used herein, the singular should generally be construed to mean "one or more" unless expressly stated otherwise or unless it is clear from the context that a singular reference is made.

[0064] A single processor or other unit may fulfill the functions of several items recited in the claims.

[0065] The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0066] The computer program may be stored / distributed on a suitable medium, such as an optical storage medium or a solid-state medium provided together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless communication systems.

[0067] Any reference signs in the claims should not be construed as limiting the scope.

[0068] Unless otherwise specified or clear from the context, the phrases "one or more of A, B, and C," "at least one of A, B, and C," and "A, B, and / or C" as used herein refer to all possible permutations of one or more of the listed items. That is, the phrase "X has A and / or B" satisfies any of: X has A; X has B; and X has both A and B.

Claims

1. 1. A computer-implemented method for estimating lung volume from radiographic images, comprising: a registration step of registering a two-dimensional radiographic image of the patient's chest to a three-dimensional radiographic image of the patient's chest to estimate data describing the projection geometry of an imaging device setup used to capture said two-dimensional radiographic image; using the projection geometry to estimate at least one radiographic magnification factor for the imaging device setting; calculating an estimated lung volume using the two-dimensional radiographic image and the at least one radiographic magnification factor; 10. A computer-implemented method comprising:

2. 2. The computer-implemented method of claim 1, wherein the registering step comprises using at least one trained machine learning model to register the two-dimensional radiographic image to the three-dimensional radiographic image to estimate data describing the projection geometry.

3. 3. The computer-implemented method of claim 1, wherein the step of registering comprises, before performing the registration, forming a sub-image from one or both of the two-dimensional radiographic image and the three-dimensional radiographic image depicting at least one anatomical structure to be used in the registration.

4. The computer-implemented method of claim 3 , wherein forming the sub-images comprises segmenting each radiographic image to isolate at least one anatomical structure.

5. 3. The computer-implemented method of claim 1, wherein the step of aligning comprises using an iterative alignment algorithm that repeatedly performs incremental changes to an estimated projection geometry and calculates a correlation between two images aligned using the incrementally changed projection geometry until the correlation converges to a steady value.

6. The computer-implemented method of claim 5 , wherein the correlation measure comprises a similarity measure describing the structure of a difference image obtained by subtracting one of the two images to be aligned from the other.

7. 3. The computer-implemented method of claim 1, comprising estimating a magnification factor for the lung using at least a source-lung distance for the lung.

8. 8. The computer-implemented method of claim 7, further comprising calculating the magnification factor for the lung as a ratio of the source-lung distance to the source-detector distance.

9. 3. The computer-implemented method of claim 1, further comprising estimating the at least one magnification factor differently depending on whether the patient is in a supine or prone position.

10. 3. The computer-implemented method of claim 1, wherein calculating the estimated lung volume comprises estimating, for each pixel in a lung region of the two-dimensional radiographic image, the volume that contributed to its intensity value.

11. 11. The computer-implemented method of claim 10, comprising estimating a volume based on a pixel size times a water-equivalent path length corrected with the at least one estimated magnification factor.

12. The computer-implemented method of claim 11 , comprising determining the pixel size as an effective pixel size that accounts for non-normal x-ray incidence angles.

13. 3. The computer-implemented method of claim 1, wherein calculating the estimated lung volume comprises using expansion factors determined separately for the left and right lungs.

14. A computer system that performs the computer-implemented method of claim 1 or 2.

15. A computer program comprising instructions which, when executed by a computer system, cause the computer system to perform the computer-implemented method of claim 1 or 2.

Images