Method and system for image reconstruction

Abstract

A method for X-ray image reconstruction used in computed tomography (CT), comprising: receiving a projection image that represents a measurement obtained from a detector 105 when X-rays emitted by a s

Description

Field Embodiments of the present invention described herein relate to methods and systems for X-ray image reconstruction. More specifically, the present invention relates to a computer-implemented method, apparatus, computer program, computer-readable medium, and a system for X-ray image reconstruction. Background Tomography is a non-invasive imaging technique, which allows visualisation of the internal structures of an object without the superposition of other structures that often afflict conventional projection images. For example, in a conventional chest radiograph, the ribs, heart, and lungs may be superimposed upon the same film, whereas a computed tomography (CT) slice is able to capture each body part in its three-dimensional position. Tomography has been applied across a range of disciplines, including medicine, physics, chemistry, and astronomy. X-ray CT is an example of tomography. X-ray CT is used in medical imaging and treatment, for example. Mathematically, the raw data acquired by a tomographic detector consists of multiple "projections" of the object being scanned. Reconstruction algorithms may then be used to reconstruct a three-dimensional (3D) object from its projections. Example reconstruction algorithms include filtered back projection algorithms and Fourier-domain reconstruction algorithms. X-ray CT, especially Cone Beam Computer Tomography (CBCT), may be used for imaging in Radiotherapy (RT). Radiotherapy exploits the differential response to ionizing radiation of healthy and cancerous cells. RT treatments are delivered over a course of several fractions, generally ranging from 3 to 40 fractions. Localizing the patient over the different fractions, especially the relative position of the lesion, also known as the target, to nearby organs at risk (OARs), and directing the treatment beam accordingly, is crucial to maximize tumour control while minimizing side effects. Modern RT systems enable this by using onboard imaging, such as CBCT. In the context of medical imaging and radiotherapy, CBCT imaging is performed by acquiring multiple projections in an arc around a patient and performing reconstruction to generate a 3D (i.e. volumetric) image. The acquisition time may be comparable to the timescale of patient and / or organ motion such that motion artifacts are introduced. Those artifacts may compromise the accurate assessment of the position of a tumour or OAR. Therefore, there is a need for improved methods and systems for X-ray image reconstruction. Brief Description of the Drawings Embodiments of the invention will now be further described by way of example only and with reference to the accompanying drawings, wherein like reference numerals refer to like parts, and wherein: Figure 1 (a) illustrates an arrangement for cone beam projection image acquisition; Figure 1 (b) is a schematic illustration of a dual-layer detector; Figure 2 schematically illustrates an arrangement for image reconstruction; Figure 3 is a flow chart illustrating a method embodiment; Figure 4 is a flow chart illustrating a method embodiment; Figure 5 is a flow chart illustrating a method embodiment; Figure 6 is a flow chart illustrating a method of X-ray image reconstruction according to an of X-ray image reconstruction according to an of X-ray image reconstruction according to an of X-ray image reconstruction according to an embodiment; Figure 7 (a) is a flow chart illustrating a method of X-ray image reconstruction according to an embodiment; Figure 7 (b) is a flow chart illustrating a method of X-ray image reconstruction according to an embodiment; Figure 8 is a schematic illustration of a reconstruction model for X-ray image reconstruction; Figure 9 is a system suitable for implementing image reconstruction according to embodiments; and Figure 10 is a system suitable for implementing image reconstruction according to embodiments. Detailed Description In the context of medical imaging and radiotherapy, a CBCT volumetric image is obtained by acquiring multiple projections in an arc around the patient, each projection being from a different viewpoint, and then applying a reconstruction algorithm to these projections (which are, e.g., 2D images) to generate a 3D (volumetric) image. The acquisition time on a conventional RT system (e.g. a medical Linac) may be of the order of a minute. Such an acquisition time may delay the overall treatment workflow and may introduce motion artifacts if the anatomy moves during the acquisition. The methods and systems described herein relate to generating instantaneous CBCT images from a measurement (e.g. one single measurement). The measurement is represented by a projection image that comprises two components. Since the two components are obtained from a measurement, they may be thought of as two different ‘flavours’ of a single physical measured projection image. For example, the two components represent the region of interest when viewed at substantially the same viewpoint. However, the two components provide different information about objects in the region of interest (when viewed from substantially the same viewpoint). For example, one component may provide information from deflected or scattered X-rays while the other may provide information from undeflected X-rays. In another example, one component may provide information from X-rays within a different energy band than the other. How the two components are obtained is described below. The use of one measured projection image (that contains two components) at the same viewpoint renders the acquisition time negligible, reduces the imaging dose, and reduces motion artifacts. The fact that the two components contain different information may improve the image reconstruction. Furthermore, many sequential 3D images can be acquired during the treatment delivery itself for motion management, real-time dose calculations, and eventually can form the basis for real-time adaptive radiotherapy. Composite 3D images representing finite time intervals and 4D images over representative breathing cycles may be flexibly generated depending on the clinical application. The invention is defined in the independent claims, to which reference should now be made. Further features are set out in the dependent claims. According to a first aspect, there is provided a computer-implemented method for X-ray image reconstruction, the method comprising: receiving a projection image that represents a measurement obtained from a detector when X-rays emitted by a source pass through a region of interest and are detected at the detector, wherein the projection image comprises a first component and a second component; and, generating, using a reconstruction model, a volumetric image of the region of interest using the first component and the second component. In an embodiment, the first component and the second component provide different information about the region of interest. In an embodiment, the first component is a scatter component and the second component is a primary component. In an embodiment, the primary component is obtained by removing the scatter component from the projection image. In an embodiment, the first component is a scatter component and the second component is approximated by the projection image. In an embodiment, the scatter component is estimated from the projection image using a scatter estimation algorithm. In an embodiment, the first component corresponds to X-rays within a first energy band and the second component corresponds to X-rays within a second energy band. In an embodiment, the first energy band is lower than the second energy band. In an embodiment, the detector comprises a photon-counting detector, the photon-counting detector being operable to output an energy spectrum, the method comprising: receiving the projection image; and determining the first component and second component from the received projection image. In an embodiment, the detector comprises a dual-layer panel, the dual-layer panel comprising a front layer and a back layer, each layer operable to generate a signal, the method comprising: obtaining the first component from the front layer; and obtaining the second component from the back layer. In an embodiment, the reconstruction model is trained using: one or more reference volumetric images; and corresponding one or more first components and second components. In an embodiment, the projection image has not been scatter corrected. According to a second aspect, there is provided a computer-implemented method for X-ray image reconstruction, the method comprising: receiving a projection image that represents a measurement obtained from a detector when X-rays emitted by a source pass through a region of interest and are detected at the detector; generating, using a reconstruction model, a volumetric image of the region of interest using the projection image, wherein the projection image has not been scatter corrected. In an embodiment, the method comprises: receiving two or more projection images at different predetermined viewpoints, each projection image comprising a first component and a second component; and generating, using the reconstruction model, a volumetric image of the region of interest using the first components and the second components. In an embodiment, the projection image is obtained at a predetermined viewpoint, and the reconstruction model is operable to generate a volumetric image at the predetermined viewpoint. In an embodiment, the method comprises generating a composite volumetric image based on a plurality of volumetric images, each volumetric image generated according to the method of any preceding claim. In an embodiment, the composite volumetric image is obtained by deforming the plurality of volumetric images using a deformation vector field. In an embodiment, generating the composite volumetric image comprises: sorting the plurality of volumetric images into temporal bins and generating a volumetric image for each bin. According to another aspect, there is provided an apparatus comprising: a memory storing computer-executable instructions, and a processor configured to execute the instructions to carry out the methods above. According to another aspect, there is provided a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the methods above. According to another aspect, there is provided a computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the methods above. According to another aspect, there is provided a system comprising: A detector; and the apparatus according to embodiments. Optionally, the image reconstruction method may be applied in adaptive radiotherapy, which enables a patient’s treatment to be changed, or adapted, to respond to a signal that additional information is known about the patient or that the patient has changed from the original state at the time of planning. Embodiments of another aspect include a data processing apparatus comprising a memory storing computer-readable instructions and a processor. The processor (or controller circuitry) is configured to execute the instructions to carry out the image reconstruction method. Embodiments of another aspect include a computer program comprising instructions, which, when executed by computer, causes the computer to execute the image reconstruction method. Embodiments of another aspect include a non-transitory computer-readable storage medium comprising instructions, which, when executed by a computer, cause the computer to execute the image reconstruction method. Other features of the disclosure are described below and recited in the appended claims. The invention may be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in combinations thereof. The invention may be implemented as a computer program or a computer program product, i.e., a computer program tangibly embodied in a non-transitory information carrier, e.g., in a machine-readable storage device or in a propagated signal, for execution by, or to control the operation of, one or more hardware modules. A computer program may be in the form of a stand-alone program, a computer program portion, or more than one computer program, and may be written in any form of programming language, including compiled or interpreted languages, and it may be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a data processing environment. The invention is described in terms of particular embodiments. Other embodiments are within the scope of the following claims. For example, the steps of the invention may be performed in a different order and still achieve desirable results. Elements of the invention have been described using the terms “processor”, “input device” etc. The skilled person will appreciate that such functional terms and their equivalents may refer to parts of the system that are spatially separate but combine to serve the function defined. Equally, the same physical parts of the system may provide two or more of the functions defined. For example, separately defined means may be implemented using the same memory and / or processor as appropriate. Figure 1 (a) illustrates an arrangement 100 for cone beam projection image acquisition. The arrangement 100 may be used for X-ray image reconstruction according to embodiments. For simplicity, only the x-y plane is illustrated; it will be appreciated that the arrangement extends into a z-axis (not shown), orthogonal to the x-y plane. The image acquisition set up 100 captures projection images for a region of interest 103. In the context of radiotherapy, the region of interest represents the region where a patient is positioned. The region of interest may contain objects such as a target and / or an OAR. An X-ray source 101 generates and emits X-rays towards the region of interest 103. In cone beam arrangement (such as used for CBCT), X-rays may be considered as a beam of rays, emitted from a point source. Detector (or detectors) 105 captures projections, which are sets of line integrals along paths that radiate from the source 101. The X-rays passing through the region of interest may either pass through and reach the detector without being scattered or deflected (shown as solid line 109), or they may be deflected from their original path due to interaction with object(s) in the region of interest (shown as dashed line 107). The undeflected ray(s) 109 contribute to a primary component of a projection image while the deflected rays 107 contribute to a scatter component of a projection image. The undeflected 109 and deflected 107 X-rays are detected at the detector 105 where a signal is generated. For example, at the detector, the received X-rays are converted to electrical charges which are then read out using a thin film transistor (TFT) array (which generates a signal). In some detectors, the conversion of received X-rays to electrical charges is as follows: a phosphor layer is provided in front of the TFT array. X-rays are converted to visible light in the phosphor layer, and the visible light is converted to electrical charges which are then read out using a TFT. The signal may indicate the attenuation of the X-rays 107, 109 as they pass through the region of interest 103. The signal is the output of the detector. The projection image Ptot corresponds to the output of the detector, which represents a measurement obtained from the detector. Since both undeflected rays 109 and deflected rays 107 are detected, the generated signal, and hence the projection image, comprises a primary component Pp and a scatter component Ps. The degree to which any of Pp or Ps is present in the projection image Ptot may depend on the nature of the object(s) in the region of interest 103 and / or the arrangement 100 of the image acquisition set up. In Figure 1 (a), the source 101 and the detector 105 are shown at a particular orientation (also referred to as viewpoint) in relation to the region of interest. The projection image acquired from this viewpoint is specific to that viewpoint. Multiple projections of the image may be acquired from different angles (viewpoints) by rotating the source 101 and detector 105 around the centre 110 of the image acquisition set up 100. In the present example, the source 101 and detector 105 may be rotated arcuately along orbital path 112. In this way, the x-y plane may be rotated counterclockwise around the point of origin (or centre of the image 110) in a manner that keeps the mutual positional relationship between source 101 and detector 105 when passing through the orbital path 112. It will be appreciated that other configurations of imaging systems are feasible; for instance, the source may be configured to rotate and a complete ring of detectors may be configured to capture projections, or there may be multiple sources arranged circumferentially around a complete ring of detectors. Further, the imaging system may rotate the source along a helical path, so as to capture projection data along the axis of the region of interest 103 (along the z-axis in the present example). The detector 105 may comprise underlying detector elements (such as pixels or bins of pixels), rather than a single active region. The output of the detector may then be a function of underlying signals generated by detector elements of the detector. This may reduce the impact of random noise compared to a signal from a single detector element. For example, the generated signal may be an average (mean, mode or median) of the underlying signals from the single detector elements. Note, that whilst Figure 1 (a) shows a curved detector 105, it will be appreciated that the detector may alternatively be a flat panel detector (FPD). Optionally, the arrangement 100 further comprises an anti-scatter grid (not shown in the Figure). The anti-scatter grid would be mounted on the detector 105 such that it lies between the detector and the region of interest 103. An anti-scatter grid selectively suppresses the scatter component while transmitting the primary component such that the acquired projection image contains more primary component compared to scatter. This arrangement may improve image quality. On the other hand, optionally, the arrangement 100 is intentionally configured to not include any scatter removal means such as anti-scatter grid. This arrangement preserves the scatter component which can be advantageous. Additionally and optionally, the detector comprises a photon counting detector (PCD). Photon counting detectors may be more sensitive, have higher resolution, and be less noisy. Advantageously, the photon counting detector may be used to obtain different types of information about the region of interest, from a single measured projection. This is because the photon counting detector may be operated such that it detects an energy spectrum. In a PCD, when a photon is absorbed, the electrical signal is roughly proportional to the photon energy. By comparing the generated signal with different thresholds and grouping each absorbed photon into a corresponding energy bin, an energy spectrum may be obtained. Thus, the generated signal represents the energy spectrum of the received X-rays. (In particular, each pixel provides an energy spectrum). The generated signal is the output of the PCD. The projection image corresponds to the output of the PCD. Different bands of the energy spectrum may then be taken as separate projection components which convey different information. The energy bands may be overlapping or non-overlapping. The two projection components may be used as separate images for reconstruction purposes. The projection components may be determined from the projection image by considering the signal within particular bands (e.g. by filtering). Further advantageously, optionally, the PDC is used to differentiate between primary and scatter signals from a projection image. The primary and scatter may have different energy spectra, and spectral information provided by the photon counting detector may be used to obtain the primary and scatter component (e.g. by filtering). The scatter component and the primary component, or the scatter component and the measured projection, may be used as input for reconstruction purposes. Alternatively and optionally, the detector comprises a dual-layer panel. An example of a duallayer panel is described in more detail in Figure 1 (b). The dual-layer panel comprises a front layer and a back layer. With reference to Figure 1 (a), the front layer is located closer to the source 101 than the back layer. Each layer generates a different signal that corresponds to a different component of the projection image. The output of the DLP comprises the signal from each layer. The output of the DLP represents a measurement that corresponds to the projection image (i.e. the projection image comprises a signal from each layer). Note that, advantageously, both components correspond to the same measurement. The front layer may preferentially capture lower energy X-rays while the back layer captures higher energy X-rays. Accordingly, the front layer component represents information from lower energy X-rays while the back layer component represents information from higher energy X-rays. Thus, the two components contain different information; the two components may be used as separate images for image reconstruction. Further, the primary component generally has higher energy than the scatter component. The front layer projection component comprises a larger proportion of the scatter than the back layer (which comprises a larger proportion of the primary). Thus, additionally and optionally, the front layer component may be used to estimate the scatter component while the back layer projection image may be used to estimate the primary component. The scatter component and the primary component, may be used as input for reconstruction purposes. Additionally and optionally, the source emits kV beams. kV X-ray beams are preferred over MV beams for imaging purposes as they provide better soft tissue contrast and lower patient dose. Thus, kV imaging allows more frequent imaging and better imaging quality. Figure 1 (b) shows a dual-layer panel (DLP) 1050 according to an example. The detector of the image acquisition arrangement shown in Figures 1 (a), 2, 8 and 10 may comprise a DLP. The DLP comprises a front layer 1052, a filter 1054, and a back layer 1056, wherein the filter 1054 is positioned between the front layer 1052 and the back layer 1056. The dual layer panel is generally arranged such that incident X-rays 1051 enter the detector by passing through the front layer 1052 first, then through the filter 1054 and on to the back layer 1056. The front layer is configured to preferentially absorb lower energy X-rays while higher energy X-rays penetrate through the front layer 1052 and the filter 1054 and are absorbed in the back layer 1056. Thus, the DLP enables two signals to be captured for each exposure. The two signals are: a lower energy signal from the front layer and a higher energy signal from the back layer. Both signals are acquired at the same viewpoint. The two signals (or components) may be used in different ways, as will be described below. The front layer and back layer may each comprise an amorphous-silicon panel onto which a Csl scintillator layer has been deposited. The scintillator layer for the front layer is thinner than the scintillator layer for the back layer. For example, the scintillator layer for the front layer is 200 pm and the scintillator layer for the back layer is 500 pm. The filter layer 1054 is a Cu filter (e.g. 1mm thick) that further improves the spectral separation between the lower energy and higher energy X-rays. The arrangement of Figure 1 (b) (e.g. materials, layers and thicknesses) is an example and it will be appreciated that different configurations are possible. Figure 2 illustrates the geometry of a cone beam projection acquisition setup, as incorporated into a medical CT scanner 200. The figure shows different possible positions 204 of radiation source 204 and different possible positions 208 of a detector. The sources emit beams of X-ray radiation, which may pass through the patient supported on a couch within the CT scanner (not shown). The radiation source 204 and detector 208 may correspond to the source 104 and detector 105 of Figure 1 (a). The radiation source 204 and detector 208 may be configured to rotate within a gantry of the CT scanner, so as to acquire projection images at a series of different angles or positions (viewpoints). As shown, the angular separation between different positions need not be equal (although a uniform angular separation is also possible). As explained above, the rotation of the sources and detectors to acquire of multiple projection images may add time to the imaging workflow. Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 (a) and Figure 7 (b) show methods of X-ray image reconstruction according to embodiments. The methods are applicable to the cone beam image acquisition arrangement of Figure 1 (a), Figure 1 (b), Figure 2, Figure 9 and Figure 10. The methods may be computer- implemented methods. Unlike CT methods which require the acquisition of multiple projection images at different viewpoints, the methods of Figures 3 to 7 (b) require one measured projection image. Thus, the reconstruction is nearly instantaneous (due to the fast acquisition time). Additionally, the method offers the benefit of a reduced imaging dose (since exposure at a single viewpoint is required rather than multiple exposures at different viewpoints), and reduced motion artifacts (due to the fast acquisition time). The methods exploit the fact that the acquired projection image comprises two components, where each component provides different information about the region of interest. One of the components may include contributions from deflected rays 107. This component (referred to as scatter component) may be thought of as a weighted contribution of the X-rays of radiation from all voxels in the region of interest. This is because, as illustrated in Figure 1 (a), the deflected rays 107 will have interacted with different voxels in the region of interest (by virtue of being deflected or scattered). Alternatively, the components may provide information from X-rays within different energy bands. Turning to Figure 3, the method comprises the following steps. At step 302, a projection image is received. The projection image represents a measurement obtained from a detector (e.g., the output of the detector), when X-rays emitted by a source pass through a region of interest and are detected at the detector. The projection image is acquired from one physical measurement. The projection image comprises two components; a first component and a second component. The two components may be separable. The components are 2D images and have the same size as the projection image (which is also a 2D image). The components may be referred to as projection components. The two projection components correspond to substantially the same viewpoint so they have substantially the same view of the region of interest. However, the two components are configured to provide different information about the region of interest. At step 304, a reconstruction model is used to generate a volumetric image of the region of interest, using the first and second components as input. The reconstruction model is operable to generate the volumetric image from the projection images at the viewpoint at which the projection is acquired. Note that receiving a projection image that represents a measurement (e.g. one physical measurement) means that the projection image represents one physical state. By one physical state, it is meant that the position of the source and / or detector relative to the region of interest is substantially constant (for example, the viewpoint is substantially the same). This also assumes that the region of interest has not changed (e.g. due to motion) - this assumption generally holds if the acquisition time is small compared to the timescale of the motion. In practice, there may be some changes to the physical state, due to e.g. mechanical drift, other inaccuracies or anatomical motion. Further, in practice, multiple exposures may be carried out to obtain multiple signals and those signals averaged to provide one projection image (this may be used to reduce noise, e.g.). The term ‘one physical measurement’ is intended to cover those scenarios. Turning to step 302, the first component and second component provide different information about the region of interest. The components may be obtained by, e.g., receiving the projection image and processing the image to obtain the two components that contain different information. Such approaches are as described in relation to Figure 4, 5 and 7 (b). Alternatively, detectors operable to provide two projection components, with different information, at the same viewpoint, may be used, as described in relation to Figure 7 (a). Turning to step 304, the reconstruction model will be described in more detail. The reconstruction model is a model that is trained using a training dataset that comprises one or more reference volumetric images, together with corresponding reference first components and reference second components. The reference volumetric images, reference first components and reference second components contain information similar to the type of image that the reconstruction model is expected to receive and generate in use. For example, the training dataset corresponds to the same anatomical region that the model is intended to be used for. Reference volumetric images may be planning CT images, for example. From the reference volumetric images, corresponding reference projection images and corresponding reference components may be determined. In general, a physics model that converts a 3D volume (reference volumetric image) into the two 2D projections (projection images or components) may be used. For example, a Monte Carlo algorithm that models the ionizing radiation (also referred to as photons or particles) interactions from the source, through the 3D volume and onto the detector may be used. It can be tracked onto which ‘pixel’ of the detector each particle / photon falls in, and this forms the 2D projection image or 2D projection component. Potentially, the photons / particles may be tracked into the detector to model the image formation process of the specific detector to form an image. The Monte Carlo approach provides rich information about particles (ionizing radiation), including whether the particles have been scattered or not, or their energies. This enables suitable reference projection images or reference projection components to be determined from the reference volumetric image. How the corresponding reference projection images and corresponding reference components are used is described further below, in relation to Figure 4 to 7 (b). Note that the reconstruction model may be patient specific. In this case, the training dataset is patient specific. It comprises one or more reference volumetric images of the patient that are augmented by applying a series of translations, rotations and / or deformations. Optionally, the augmented set of reference volumetric images comprises at least 1000 images. From the augmented set of reference volumetric images, first components and second components are obtained as above to form the training dataset. Alternatively, the reconstruction model may be trained on a population. In this case, several reference volumetric images, at least 30 -50 (and preferably at least 100), corresponding to the same organs and / or anatomical areas of interest and obtained from different patients, are used. Augmentation is optionally used to increase the size of the training set. From these reference volumetric images, corresponding reference first components and second components are determined and used as the training dataset. In use, the reconstruction model in step 304 may receive two projection components derived from a projection image obtained at a predetermined viewpoint. The reconstruction model is operable to generate a volumetric image at the predetermined viewpoint. To account for use at different potential viewpoints, a number of reconstruction models may be available, where each model is configured to generate a volumetric image at one particular viewpoint. Accordingly, each model is trained using a training dataset obtained by generating reference projection images (and corresponding reference projection components) at the particular viewpoint. For example, 360 viewpoints (one per degree over a full arc) may be predefined, and thus, 360 reconstruction models may be available. In use, the viewpoint (i.e the angle) at which the projection image is acquired is provided as an input so that the corresponding reconstruction model may be selected. The reconstruction model (or reconstruction models) may be implemented using a network architecture as described in relation to Figure 8. Turning to Figure 4, the method 400 comprises the following steps. At step 402, a projection image (Ptot) is received. The projection image may be acquired at a particular viewpoint. As described above in relation to step 302, the projection image comprises a first component and a second component. In step 402, the first component is a scatter component (Ps) and the second component is a primary component (Pp). At step 406, a reconstruction model is used to generate a volumetric image of the region of interest, using the scatter component and the primary component as inputs. The reconstruction model is operable to generate the volumetric image from the scatter component and primary component at that particular viewpoint. The scattering of X-rays may create noise and artifacts and reduce image contrast, and hence, conventional approaches have sought to suppress the scatter component. Unlike those conventional methods, the methods described herein seek to exploit the information contained within the scatter component. Returning to step 402, optionally, the scatter component is estimated as follows. The received projection image is fed into a scatter prediction model which generates a scatter component as output. Once the scatter component has been estimated, additionally and optionally, the primary component is obtained by removing the scatter component from the projection image (e.g. by subtraction). The primary component represents the signal from X-rays that pass through the region of interest and reach the detector without being deflected or scattered. The scatter component represents the signal generated from X-rays that pass through the region of interest and reach the detector, and that have been deflected from their original path due to interactions with one or more objects in the region of interest. The scatter prediction model is a trained model that is trained using pairs of reference projection images and simulated scatter components. The reference projection images are derived from e.g. one or more planning CTs. In the context of radiotherapy, a planning CT refers to a volumetric image acquired before the radiotherapy treatment. The planning CT may correspond to a single patient, or when multiple planning CTs are used, the planning CTs may correspond to a population. The reference projections images are obtained at one or more predetermined viewpoints. Further, the reference projection images are produced by using a geometry (e.g. relative positions between source, detector, and region of intertest) that corresponds to the image acquisition arrangement. Each reference projection image is then analysed to simulate a corresponding reference scatter component using a Monte Carlo (MC) algorithm, which produces scatter estimations. Alternatively, a Boltzmann solver is used instead of the MC algorithm. The scatter model may be a ll-net convolutional neural network. An example of the scatter prediction model is described in US20240249451A1, for example. Turning to step 406, the reconstruction model will be described in more detail. The reconstruction model is trained in a similar manner as for the model used in step 304. However, here, the reference scatter components and corresponding reference primary components are used instead of reference first and second components. From the reference volumetric images, corresponding reference scatter components and reference primary components may be determined. In particular, the reference scatter components may be obtained using a physics model based on a MC algorithm as described in relation to method 300. For example, the MC algorithm enables the modelling of interactions of the ionizing radiation (also referred to as photons or particles) as it travels from the source, through the region of interest (depicted in the reference volumetric image) and on to the detector. It can be tracked onto which ‘pixel’ of the detector each particle / photon falls in, and this forms the 2D projection image or 2D projection component. Photons that have suffered at least one scatter event can be tagged as scattered photons. Photons that are tagged as scattered are accumulated and form the scatter component. Conversely, photons that are not tagged as scattered are accumulated and form the primary component. Alternatively, the primary component may be derived from the reference volumetric image using a ray-tracing algorithm. With reference to step 402 of Figure 4, in an additional and optional example, when the detector comprises a photon-counting detector as described herein, a first component is obtained by considering captured X-rays that have energies within a first energy band and a second component is obtained by considering X-rays that have energies within a second band. The first projection component and the second projection component are obtained in the same measurement. The primary component and scatter component have different energy spectra and by setting the first energy band and the second energy band accordingly, the primary and scatter components may be estimated from the first and second components. For example, the scatter component has a lower energy spectrum than the primary component and by setting the second energy band to be lower than the first energy band, the scatter component is obtained from the second component while the primary component is obtained from the first component. For example, the scatter component is estimated from the second projection component using the scatter prediction model described herein. The primary component is approximated by the first projection image (in which the scatter component has been filtered by the selection of an appropriate energy band). Lastly, at step 406, the volumetric image is generated using the reconstruction model as described above. Turning to Figure 5, the method 500 comprises the following steps. At step 502, a projection image (Ptot) is received. Step 502 corresponds to step 402 of Figure 4, except that, at step 502, the first component is a scatter component (Ps) and the second component is a total component. At step 506, a reconstruction model is used to generate a volumetric image of the region of interest, using the scatter component and the total component as input. The total component represents the contribution from both scattered rays 107 and undeflected rays 109. Similar to the method 400, the method 500 uses two components that provide different information about the region of interest. At step 502, optionally, the scatter component is estimated is the same manner as for step 402, while the total component is approximated by the received projection image itself. At step 506, a reconstruction model is used to generate a volumetric image of the region of interest, using the scatter component and the total component (approximated by the received projection image) as inputs. The reconstruction model is operable to generate the volumetric image from the scatter component and projection image at the viewpoint at which the projection image is acquired. The reconstruction model used in step 506 is similar to the reconstruction model of step 406 except that the inputs to the model are a scatter component and a total component (instead of a scatter component and a primary component). Accordingly, the second reconstruction model is trained using a dataset comprising reference volumetric images, reference scatter components and reference total components. The reference volumetric images and reference scatter components are the same as in step 406. The reference total components are approximated by reference projection images (which may be derived from the volumetric image, in the same way as described in step 406). With reference to Figure 6, the method 600 comprises the following steps. At step 602, a projection image is received. Step 602 is similar to step 402 of Figure 4 with the additional feature that the projection image is not scatter corrected. The projection image still contains information from two components (e.g. scatter component and primary or total component) however, those components are contained within the projection image and not separated. At step 604, a reconstruction model is used to generate a volumetric image of the region of interest, using the received projection image as input. The reconstruction model is operable to generate the volumetric image from the projection image at the viewpoint at which the projection image is acquired. Unlike conventional approaches that seek to suppress the scatter component by applying scatter correction, the method 600 seeks to exploit the information contained within the scatter component, by intentionally not applying any scatter correction. Turning to step 602, scatter correction refers to either using a scatter correction grid (as described herein) and / or using a scatter correction algorithm to reduce the scatter component from the projection image (Ptot). Not scatter correcting the projection image means that a scatter grid is not applied, or that a scatter correction algorithm is not applied, or that neither a scatter grid nor a scatter correction algorithm is applied. An example of a scatter correction algorithm comprises estimating the scatter component (using a scatter prediction model such as that described in relation to method 400), and then removing the estimated scatter component from the projection image (e.g. by subtracting). Turning to step 604, the reconstruction model is similar to the reconstruction model of step 406 except that the input to the model is a projection image (that has not been scatter corrected), instead of a scatter component and a primary component. Accordingly, the reconstruction model is trained using a dataset comprising reference volumetric images and reference projection images. The reference projection images are obtained from a reference volumetric image, in a manner similar to step 406, using e.g. an MC algorithm, and accumulating both unscattered and scattered photons to form the reference projection images. With reference to Figure 7 (a), the method 700a comprises the following steps. The method 700a relies on a detector comprising a DLP. The DLP is as described herein. At step 702a, a projection image is received. The projection image represents a measurement obtained from the DLP when X-rays emitted by a source pass through the region of interest and are detected at the detector. The DLP comprises a front layer and a back layer and each layer is operable to generate a signal. The projection image comprises a first component and a second component, similar to method 300. The first component may be obtained from the front layer while the second component may be obtained from the back layer. Further, as described herein, optionally, the front layer captures photons within a different energy band that the back layer. (E.g. the front layer signal corresponds to lower energy photons while the back layer signal corresponds to a higher energy signal). The two projection components provide different information about the region of interest from the same viewpoint. At step 704a, a reconstruction model is used to generate a volumetric image of the region of interest using the first and second components. The reconstruction model will be described in more detail next. The reconstruction model is a model that is trained using a training dataset that comprises one or more reference volumetric images, together with corresponding reference second components (from the back layer) and corresponding reference first components (from the front layer). The back-layer and front-layer reference components are obtained using a physics model (based on a MC algorithm) as described in relation to methods 300 and 400. Here, the physics model also accounts for the interaction of the ionising radiation as it passes through the layers of the DLP, and enables a back-layer component and front-layer component to be derived. As described herein, reference volumetric images may be planning CT images, for example. Further, the reconstruction model may receive projection images at a predetermined viewpoint. Different potential viewpoints are accounted for using the same approach as described in relation to method 400. With reference to Figure 7 (b), the method 700b comprises the following steps. The method 700b relies on a detector comprising a PCD. The PCD is operable to output an energy spectrum, as described herein. The projection image represents a measurement obtained from the PCD. At step 702b, a projection image is received. As for method 300, the projection image comprises a first component and a second component. Here, the first component corresponds to a first energy band and the second component corresponds to a second energy band. The first and second components are separable according to their energies. At step 704b, a reconstruction model is used to generate a volumetric image of the region of interest using the received components as input. The reconstruction model will be described in more detail next. The reconstruction model is a model that is trained using a training dataset that comprises one or more reference volumetric images, together with corresponding first components (corresponding to a first energy band) and second components (corresponding to a second energy band). The first and second components are obtained using a physics model (based on a MC algorithm) as described in relation to methods 300 and / or 400. Here, the physics model accounts for the energy of the photons and assigns photons to the first component or the second component according to their energies. As described herein, reference volumetric images may be planning CT images, for example. Further, the reconstruction model may receive projection images at a predetermined viewpoint. Different potential viewpoints are accounted for using the same approach as described in relation to methods 300 and / or 400. In relation to any of the methods, 300, 400, 500, 700a and 700b, additionally and optionally, the received projection image, Ptot, is not scatter corrected (where not scatter correcting is as described below in relation to the method of Figure 6). Although the methods 300, 400, 500, 700a and 700b refer to the projection image comprising a first component and a second component, it will be understood that the methods may be used with more than two components. For example, more than two components may be extracted from the projection image. Accordingly, the reconstruction model may be adapted to receive more than two components as input. For example, in relation to method 700b, more than two components may be obtained by considering more than two energy bands of the energy spectrum. In relation to methods 300, 400, 500, algorithms may be used to obtain more than two components from the projection image. The methods describe in relation to Figure 3, Figure 4, Figure 5, Figure 6, Figure 7(a) and Figure 7 (b) may be used for generating volumetric images in a near-instantaneous manner (that reduces motion artefacts) that also reduces imaging dose. The methods rely on receiving a projection image that represents a measurement (e.g. one physical measurement) obtained from a detector. Although the methods refer to obtaining a measurement, it will be understood that the methods may be adapted to obtain more than one measurement (e.g. between 2 to 10 measurements). Each measurement may be at a different predetermined viewpoint. From each measurement, two or more components may be obtained. The reconstruction model may be adapted to receive multiple components (2 or more components for each of the obtained measurement) as input, and to generate a volumetric image using those multiple components. Further, the reconstruction would be operable to generate a volumetric image from measurements made at those different predetermined viewpoints. The reconstruction model would be trained in a similar manner except that the training dataset would now comprise reference volumetric images and reference components representing components at each of the different predetermined viewpoints. The benefits of fast acquisition time are still retained, especially when compared to conventional CBCT, as fewer measurements are likely to be needed (since the two or more components obtained per measurement provide different information about the region of interest). In relation to any of the methods described herein, additionally and optionally, a plurality of volumetric images may be generated, at closely spaced time intervals (due to the near-instantaneous generation), by repeatedly applying those methods. The plurality of volumetric images may be turned into composite volumetric images. A composite image may be formed by deforming the plurality of volumetric images using a deformation vector field (DVF) as follows. Say X 3D images are generated with different timestamps. The image at time t1 may be taken as the reference image. A deformation vector field (DVF) between a reference image and the other (X-1) images may be determined. Each of the other images may be deformed to match the reference image, and the average of the deformed images and the reference image may then determined. The average is then the composite image. Alternatively, the composite image may comprise a 4D image. Forming the 4D image comprises: generating a plurality of 3D images, separating a breathing phase into a number of bins, finding which bin each of the generated 3D images belongs to, and averaging all of the 3D images in each particular bin. Each bin will be represented by one 3D image. Figure 8 is a schematic illustration of a reconstruction model. The reconstruction model may be used for the methods described in relation to Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 (a) and Figure 7 (b). The reconstruction model is a deep neural network formulated as an encoder-decoder network. One or more projection images (2D) or projection components (2D) is provided as input 802. The projection image is a view of a region of interest, as viewed from a particular viewpoint. The size of an input may be A / x128x128, where 128x128 represents the size of a 2D projection image or 2D projection component, and / V represents the number of projection images or components that are inputted to the model. The high-dimensional input data is then encoded into a feature representation by a representation network 804 (encoder network). The representation network comprises a stack of five 2D convolutional residual blocks with different number and size of convolutional filters. The number and size of each residual block is selected such that the input image (A / x128x128) is mapped to the following feature maps as data flows through the different blocks: 256x64x64 -> 512x32x32 ->1024x16x16 ->2048x8x8 ->4096x4x4. The feature representation output by the representation network 804 is a tensor of size 4096x4x4. The feature representation from network 804 is then applied to a transformation module 806. The transformation module 806 maps the feature representation (which represents features from a 2D image) into representative features for a 3D (volumetric image). The transformation module comprises a sequence of: (i) 2D convolution or fully connected layer (kernel size 1 and stride 1) followed by ReLLI activation, (ii) a reshaping function, and (iii) a 3D deconvolution layer with kernel size of 1x1x1 and sliding stride of 1x1x1. As the 4096x4x4 tensor flows through stages (i) to (iii), it is transformed to 4096x4x4, 2048x2x4x4 and 2048x2x4x4. Thus the output of the transform module is a 2048x2x4x4 vector. The output of the transform module is then applied to the generation network 808. The generation network 808 (generator network) is configured to generate a 3D volumetric image based on the output of the transform module which is reshaped version of the features output by the representation network 804. The generation network 808 comprises a stack of five 3D deconvolutional residual blocks with different number and size of deconvolutional filters. The number and size of each block is selected such that the input tensor (2048x2x4x4) is mapped to the following feature maps as data flows through the generation network: 1024x4x8x8 -> 512x8x16x16 ->256x16x32x32 -> 128x32x64x64->64x64x128x128. At the end of the stack of blocks, the vector (64x64x128x128) is further fed into another transformation module that comprises a 3D convolutional layer and a 2D convolutional layer (kernel size 1), to obtain a 3D image of size Czx128x128, where Czrepresents the size of the output 810 in the z axis. Thus, the output 810 has size Czx128x128 and is a 3D volumetric image reconstruction of the one or more input projection image 802 (A / x128x128). The model 800 is trained using reference volumetric images and corresponding one or more reference first and second components as described in relation Figures 3, 4, 5,7(a) and 7 (b). The one or more reference first and second components may be a scatter component and a primary component as in Figure 4, a scatter component and a total component as in Figure 5, a front layer component (first component) and back layer component (second component) as in Figure 7 (a), or a first and second component separable according to their energies as in Figure 7 (b). To implement the reconstruction model of method 600, the model 800 is trained using reference volumetric images and corresponding reference projection images. The model 800 is trained to predict a volumetric image that is close to the reference volumetric image. The difference between the predicted volumetric image and the reference volumetric image is obtained using a mean square error loss function, and an Adam optimizer may be used to minimize the loss function and update the model parameters iteratively through back-propagation. A learning rate of 0.00002 may be used. While the above example refers to an input image size of Nx128x128 and other matrix sizes as the data flows through the network, it will be appreciated that different dimensions may also be used. Further, while Figure 8 illustrates one possible implementation of the reconstruction model, it will be understood that any of the following neural network architectures could also be used: convolutional neural networks (CNN) such as modified ll-net, generative adversarial networks (GANs), variational auto encoders (VAEs), or transformer networks. Hybrid approaches (CNN + VAE / GAN) could also be considered, where CNNs extract feature representations from a 2D projection, and a VAE or GAN reconstructs the 3D volume from those features. Another example of an architecture that could be used is described in Shen L, et al. Nat Biomed Eng. 2019 Nov, 3(11), 880-888. Figure 10 depicts a radiotherapy apparatus, suitable for acquiring projection data for image reconstruction according to embodiments. The cross-section through radiotherapy apparatus 1000 includes a radiation head 1010 and a beam receiving apparatus (detector) 1002, both of which are attached to a gantry 1004. The radiation head 1010 includes a radiation source 1012, which emits a beam of radiation 1006. The radiation head 1010 also includes a beam shaping apparatus 1018, which controls the size and shape of the radiation field associated with the beam. The beam receiving apparatus 1002 is configured to receive radiation emitted from the radiation head 1010, for the purpose of absorbing and / or measuring the beam of radiation. In the view shown, the radiation head 1010 and the beam receiving apparatus 1002 are positioned diametrically opposed to one another. The gantry 1004 is rotatable, and supports the radiation head 1010 and the beam receiving apparatus 1002 such that they are rotatable around an axis of rotation 1008, which may coincide with the patient longitudinal axis. The gantry provides rotation of the radiation head 1010 and the beam receiving apparatus 1002 in a plane perpendicular to the patient longitudinal axis (e.g., a sagittal plane). Three gantry directions xG,yG, zG may be defined such that the yG direction is perpendicular with the gantry axis of rotation. The yG direction extends from a point on the gantry corresponding to the radiation head 1010, towards the axis of rotation of the gantry. Therefore, from the patient frame of reference, the yG direction rotates around as the gantry rotates. The radiotherapy apparatus 1000 also includes a support surface or couch 1020 on which a subject (or patient) is supported during radiotherapy treatment or image acquisition. The radiation head 1010 is configured to rotate around the axis of rotation 1008 such that the radiation head 1010 directs radiation towards the subject from various angles around the subject in order to spread out the radiation dose received by healthy tissue to a larger region of healthy tissue while building up a prescribed dose of radiation at a target region. The radiotherapy apparatus 1000 is configured to deliver a radiation beam towards a radiation isocentre, which is substantially located on the axis of rotation 1008 at the centre of the gantry 1004 regardless of the angle at which the radiation head 1010 is placed. The rotatable gantry 1004 and radiation head 1010 are dimensioned so as to allow a central bore 1022 to exist. The central bore 1022 provides an opening, sufficient to allow a subject to be positioned therethrough without the possibility of being incidentally contacted by the radiation head 1010 or other mechanical components as the gantry rotates the radiation head 1010 about the subject. The radiation head 1010 emits the radiation beam 1006 along a beam axis 1024 (or radiation axis or beam path), where the beam axis 1024 is used to define the direction in which the radiation is emitted by the radiation head 1010. The radiation beam 1006 is incident on the beam receiving apparatus 1002, which may include at least one of a beam stopper and a radiation detector. The beam receiving apparatus 1002 is attached to the gantry 1004 on a diametrically opposite side to the radiation head 1010 to attenuate and / or detect a beam of radiation after the beam has passed through the subject. The radiation beam axis 1024 may be defined as, for example, a centre of the radiation beam 1006 or a point of maximum intensity. The beam shaping apparatus 1018 delimits the spread of the radiation beam 1006. The beam shaping apparatus 1018 is configured to adjust the shape and / or size of a field of radiation produced by the radiation source. The beam shaping apparatus 1018 does this by defining an aperture (also referred to as a window or an opening) of variable shape to collimate the radiation beam 1006 to a chosen cross-sectional shape. In this example, the beam shaping apparatus 1018 may be provided by a combination of a diaphragm and an MLC. Beam shaping apparatus 1018 may also be referred to as a beam modifier. The radiotherapy apparatus 1000 may be configured to deliver both coplanar and non-coplanar (also referred to as tilted) modes of radiotherapy treatment. In coplanar treatment, radiation is emitted in a plane that is perpendicular to the axis of rotation of the radiation head 1010. In non-coplanar treatment, radiation is emitted at an angle that is not perpendicular to the axis of rotation. In order to deliver coplanar and non-coplanar treatment, the radiation head 1010 may move between at least two positions, one in which the radiation is emitted in a plane which is perpendicular to the axis of rotation (coplanar configuration) and one in which radiation is emitted in a plane which is not perpendicular to the axis of rotation (non-coplanar configuration). In the coplanar configuration, the radiation head 1010 is positioned to rotate about a rotation axis and in a first plane. In the non-coplanar configuration, the radiation head is tilted with respect to the first plane such that a field of radiation produced by the radiation head is directed at an oblique angle relative to the first plane and the rotation axis. In the non-coplanar configuration, the radiation head 1010 is positioned to rotate in a respective second plane parallel to and displaced from the first plane. The radiation beam is emitted at an oblique angle with respect to the second plane, and therefore as the radiation head rotates the beam sweeps out a cone shape. In one configuration, the beam receiving apparatus 1002 may remain in the same place relative to the rotatable gantry when the radiotherapy apparatus is in both the coplanar and non-coplanar modes. Therefore, the beam receiving apparatus 1002 is configured to rotate about the rotation axis in the same plane in both coplanar and non-coplanar modes. This may be the same plane as the plane in which the radiation head rotates. In alternative configurations, the beam receiving apparatus 1001 may also rotate. The beam shaping apparatus 1010 is configured to reduce the spread of the field of radiation in the non-coplanar configuration in comparison to the coplanar configuration. The radiotherapy apparatus 1000 includes a controller 1030, which is programmed to control the radiation source 1012, beam receiving apparatus 1006 and the gantry 1002. Controller 1030 may perform functions or operations such as treatment planning, treatment execution, image acquisition, image processing, motion tracking, motion management, and / or other tasks involved in a radiotherapy process. Controller 1030 is programmed to control various components of apparatus 1000, such as gantry 1004, radiation head 1910, beam receiving apparatus 1002, and support surface 1020, so as to acquire projection data suitable for image reconstruction. Hardware components of controller 1030 may include one or more computers (e.g., general purpose computers, workstations, servers, terminals, portable / mobile devices, etc.); processors (e.g., central processing units (CPUs), graphics processing units (GPUs), microprocessors, digital signal processors (DSPs), field programmable gate arrays (FPGAs), special-purpose or specially-designed processors, etc.); memory / storage devices such as a memory (e.g., read-only memories (ROMs), random access memories (RAMs), flash memories, hard drives, optical disks, solid-state drives (SSDs), etc.); input devices (e.g., keyboards, mice, touch screens, mics, buttons, knobs, trackballs, levers, handles, joysticks, etc.); output devices (e.g., displays, printers, speakers, vibration devices, etc.); circuitries; printed circuit boards (PCBs); or other suitable hardware. Software components of controller 830 may include operation device software, application software, etc. The radiation head 1010 may be connected to a head actuator 1014, which is configured to actuate the radiation head 1010, for example between a coplanar configuration and one or more non-coplanar configurations, or for example to actuate the radiation source 1012 and / or detector 1002 in response to detection of flex. This may involve translation and rotation of the radiation head 1010 relative to the gantry. In some implementations, the head actuator may include a curved rail along which the radiation head 1010 may be moved to adjust the position and angle of the radiation head 1010. The controller 1030 may control the configuration of the radiation head 1030 via the head actuator 1014. The beam shaping apparatus 1018 includes a shaping actuator 1016. The shaping actuator is configured to control the position of one or more elements in the beam shaping apparatus 1018 in order to shape the radiation beam 1006. In some implementations, the beam shaping apparatus 1016 includes an MLC, and the shaping actuator 1016 includes means for actuating leaves of the MLC. The beam shaping apparatus 1018 may further comprise a diaphragm, and the shaping actuator 1016 may include means for actuating blocks of the diaphragm. The controller 1030 may control the beam shaping apparatus 1018 via the shaping actuator 1016. Figure 9 is a block diagram of an implementation of a radiotherapy system 900, suitable for executing methods for image reconstruction according to embodiments. The example radiotherapy system 900 comprises a computing system 910 within which a set of instructions, for causing the computing system 910 to perform the method (or steps thereof) discussed herein, may be executed. The computing system 910 may implement an apparatus for image reconstruction. The apparatus for image reconstruction (computing system 910) may be a standalone apparatus (that is separable from treatment device 950 and / or image acquisition device 940). The computing system 910 may also be referred to as a computer. In particular, the methods described herein may be implemented by a processor or controller circuitry 911 of the system 910. The computing system 910 shall be taken to include any number or collection of machines, e.g., computing device(s), that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methods discussed herein. That is, hardware and / or software may be provided in a single computing device, or distributed across a plurality of computing devices in the computing system. In some implementations, one or more elements of the computing system may be connected (e.g., networked) to other machines, for example in a Local Area Network (LAN), an intranet, an extranet, or the Internet. One or more elements of the computing system may operate in the capacity of a server or a client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. One or more elements of the computing system may be a personal computer (PC), a tablet computer, a set-top box (STB), a Personal Digital Assistant (PDA), a cellular telephone, a web appliance, a server, a network router, switch or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by that machine. The computing system 910 includes controller circuitry 911 and a memory 913 (e.g., read-only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM), etc.). The memory 913 may comprise a static memory (e.g., flash memory, static random access memory (SRAM), etc.), and / or a secondary memory (e.g., a data storage device), which communicate with each other via a bus (not shown). Memory 913 may be used to store or buffer projection data until required for image processing. Controller circuitry 911 represents one or more general-purpose processors such as a microprocessor, central processing unit, accelerated processing units, or the like. More particularly, the controller circuitry 911 may comprise a complex instruction set computing (CISC) microprocessor, reduced instruction set computing (RISC) microprocessor, very long instruction word (VLIW) microprocessor, processor implementing other instruction sets, or processors implementing a combination of instruction sets. Controller circuitry 911 may also include one or more special-purpose processing devices such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), network processor, or the like. One or more processors of the controller circuitry may have a multicore design. Controller circuitry 911 is configured to execute the processing logic for performing the operations and steps discussed herein. The computing system 910 may further include a network interface circuitry 915. The computing system 910 may be communicatively coupled to an input device 920 and / or an output device 930, via input / output circuitry 916. In some implementations, the input device 920 and / or the output device 930 may be elements of the computing system 910. The input device 920 may include an alphanumeric input device (e.g., a keyboard or touchscreen), a cursor control device (e.g., a mouse or touchscreen), an audio device such as a microphone, and / or a haptic input device. The output device 930 may include an audio device such as a speaker, a video display unit (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)), and / or a haptic output device. In some implementations, the input device 920 and the output device 930 may be provided as a single device, or as separate devices. In some implementations, the computing system 910 may comprise image processing circuitry 914. Image processing circuitry 914 may be configured to process image data 970 (e.g., images, imaging data, projections, projection data), such as medical images obtained from one or more imaging data sources, a treatment device 950 and / or an image acquisition device 940. Image processing circuitry 914 may be configured to process, or pre-process, image data 970. For example, image processing circuitry 914 may convert received image data into a particular format, size, resolution or the like. Image processing circuitry 914 may be configured to perform image reconstruction. In some implementations, image processing circuitry 914 may be combined with controller circuitry 911. In some implementations, the radiotherapy system 900 may further comprise an image acquisition device 940 and / or a treatment device 950. The image acquisition device 940 and the treatment device 950 may be provided as a single device. In some implementations, treatment device 950 is configured to perform imaging, for example in addition to providing treatment and / or during treatment. Image acquisition device 940 may be configured to acquire a 2D X-ray projection image as described herein. The image acquisition device 940 may correspond the detector described herein. Image acquisition device 940 may be configured to output image data 970, which may be accessed by computing system 910. Treatment device 950 may be configured to output treatment data 960, which may be accessed by computing system 910. Treatment data 960 may be obtained from an internal data source (e.g., from memory 913) or from an external data source, such as treatment device 950 or an external database. The various methods described above may be implemented by a computer program. The computer program may include computer code (e.g., instructions) arranged to instruct a computer to perform the functions of one or more of the various methods described above. For example, the steps of the methods described in relation to Figure 1 may be performed by the computer code. The steps of the methods described above may be performed in any suitable order. The computer program and / or the code for performing such methods may be provided to an apparatus, such as a computer, on one or more computer readable media or, more generally, a computer program product. The computer readable media may be transitory or non-transitory. The one or more computer readable media could be, for example, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, or a propagation medium for data transmission, for example for downloading the code over the Internet. Alternatively, the one or more computer readable media could take the form of one or more physical computer readable media such as semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read-only memory (ROM), a rigid magnetic disc, and an optical disk, such as a CD-ROM, CD-R / W or DVD. The instructions may also reside, completely or at least partially, within the memory 913 and / or within the controller circuitry 911 during execution thereof by the computing system 910, the memory 913 and the controller circuitry 911 also constituting computer-readable storage media. In an implementation, the modules, components and other features described herein may be implemented as discrete components or integrated in the functionality of hardware components such as ASICS, FPGAs, DSPs or similar devices. A “hardware component” is a tangible (e.g., non-transitory) physical component (e.g., a set of one or more processors) capable of performing certain operations and may be configured or arranged in a certain physical manner. A hardware component may include dedicated circuitry or logic that is permanently configured to perform certain operations. A hardware component may comprise a special-purpose processor, such as an FPGA or an ASIC. A hardware component may also include programmable logic or circuitry that is temporarily configured by software to perform certain operations. In addition, the modules and components may be implemented as firmware or functional circuitry within hardware devices. Further, the modules and components may be implemented in any combination of hardware devices and software components, or only in software (e.g., code stored or otherwise embodied in a machine-readable medium or in a transmission medium). Unless specifically stated otherwise, as apparent from the following discussion, it is appreciated that throughout the description, discussions utilizing terms such as “receiving”, “determining”, “comparing ”, “enabling”, “maintaining”, “identifying”, “obtaining”, “accessing”, or the like, refer to the actions and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices. While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel methods and apparatuses described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of methods and apparatus described herein may be made.

Claims

1. A computer-implemented method for X-ray image reconstruction, the method comprising: receiving a projection image that represents a measurement obtained from a detector when X-rays emitted by a source pass through a region of interest and are detected at the detector,wherein the projection image comprises a first component and a second component; and,generating, using a reconstruction model, a volumetric image of the region of interest using the first component and the second component.

2. The method of claim 1, wherein the first component and the second component provide different information about the region of interest.

3. The method of claim 1 or 2 wherein the first component is a scatter component and the second component is a primary component.

4. The method of claim 3, wherein the primary component is obtained by removing the scatter component from the projection image.

5. The method of claim 1 or 2 wherein the first component is a scatter component and the second component is approximated by the projection image.

6. The method of any of claims 1 to 5, wherein the scatter component is estimated from the projection image using a scatter estimation algorithm.

7. The method of claim 1 or 2, wherein the first component corresponds to X-rays within a first energy band and the second component corresponds to X-rays within a second energy band.

8. The method of claim 7, wherein the first energy band is lower than the second energy band.

9. The method of claim 7 or 8, wherein the detector comprises a photon-counting detector, the photon-counting detector being operable to output an energy spectrum, the method comprising:receiving the projection image; anddetermining the first component and second component from the received projection image.

10. The method of any of claims 1, 2, 7, or 8, wherein the detector comprises a dual-layer panel, the dual-layer panel comprising a front layer and a back layer, each layer operable to generate a signal, the method comprising:obtaining the first component from the front layer; and obtaining the second component from the back layer.

11. The method of any preceding claim, wherein the reconstruction model is trained using: one or more reference volumetric images; and corresponding one or more first components and second components.

12. The method of any preceding claim, wherein the projection image has not been scatter corrected.

13. A computer-implemented method for X-ray image reconstruction, the method comprising: receiving a projection image that represents a measurement obtained from a detector when X-rays emitted by a source pass through a region of interest and are detected at the detector;generating, using a reconstruction model, a volumetric image of the region of interest using the projection image,wherein the projection image has not been scatter corrected.

14. The method of any preceding claim comprising:receiving two or more projection images at different predetermined viewpoints, each projection image comprising a first component and a second component; andgenerating, using the reconstruction model, a volumetric image of the region of interest using the first components and the second components.

15. The method of any of claims 1 to 13 wherein the projection image is obtained at a predetermined viewpoint, and the reconstruction model is operable to generate a volumetric image at the predetermined viewpoint.

16. A method of generating a composite volumetric image based on a plurality of volumetric images, each volumetric image generated according to the method of any preceding claim.

17. A method according to claim 16 wherein the composite volumetric image is obtained by deforming the plurality of volumetric images using a deformation vector field.

18. A method according to claim 16 wherein generating the composite volumetric image comprises: sorting the plurality of volumetric images into temporal bins and generating a volumetric image for each bin.

19. An apparatus comprising:a memory storing computer-executable instructions, anda processor configured to execute the instructions to carry out the method of any preceding claims.

20. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method of any of claims 1 to 18.

21. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method of any of claims 1 to 18.

22. A system comprising:A detector; andThe apparatus of claim 19.

Images