GB2589929A

GB2589929A - Calibration of radiotherapy apparatus

Info

Publication number: GB2589929A
Application number: GB1918426.6A
Authority: GB
Inventors: Massai Marco; Andre Valentin
Original assignee: Elekta ltd
Priority date: 2019-12-13
Filing date: 2019-12-13
Publication date: 2021-06-16
Anticipated expiration: 2039-12-13
Also published as: GB2589929B; GB201918426D0

Abstract

A radiotherapy machine having an imaging apparatus mounted to a rotatable gantry (e.g. for IGRT) is calibrated as follows. The imaging apparatus is used to obtain a set of 2D images of a phantom positioned within the treatment volume, each image taken at a different gantry rotation angle. Based on a reference reconstruction of the phantom, an orientation indicator for at least one image of the set is determined, the orientation indicator showing how the image should be oriented with respect to the other images of the set during generation of a 3D image. An imaging calibration map is generated which provides the orientation indicator, and preferably also the location of a projected treatment isocentre, as a function of gantry rotation angle. The orientation indicator may be a location of an imaging rotation axis, or one or more image transformations applied to the image. Also claimed are a method of generating a 3D image using the image calibration map, and an image calibration map on a computer readable storage medium.

Description

CALIBRATION OF RADIOTHERAPY APPARATUS

Field

The present disclosure relates to methods of calibrating a radiotherapy machine comprising both treatment and imaging apparatus, for example a machine capable of providing Image Guided Radiotherapy "IGRT", so as to help clinicians in patient positioning.

Background

For the purposes of radiotherapy treatment, it is desirable to deliver a particular dose to a target region, as prescribed by a treatment plan, while minimising the dose to surrounding areas of healthy tissue. Radiotherapy systems typically deliver beams of MV radiation energy from different angles to the area to be treated, i.e. the target region. In this way, each portion of healthy tissue surrounding the target region is only exposed to the radiation beam at particular angles, whilst the target region is exposed to the MV radiation beam at every angle.

Some radiotherapy machines, e.g. machines capable of performing Image Guided Radiotherapy (IGRT), include imaging capabilities. The imaging capabilities may be provided by an imaging apparatus configured to provide images of the patient to assist with treatment planning and positioning of the patient. The imaging apparatus may for example comprise a source of kilovolt (kV) energy radiation, such as X-rays.

The imaging apparatus is typically mounted on a rotatable gantry of the radiotherapy device at an angle relative to a treatment apparatus. If images obtained via the imaging apparatus are to be used to inform the radiation therapy, then a relationship between the images and the alignment of the treatment apparatus must be obtained.

It is known that a 3D image, or a 3D reconstruction, of a patient may be obtained via obtaining 2D images taken at various angles around the patient. These 2D images may be described as projections, or projected images. As part of the reconstruction process, the resulting set of 2D projected images must be properly assembled to produce an accurate 3D reconstruction. This may involve aligning and/or registering the images with one another.

In an ideal system, the relative distance between each component of the imaging apparatus is fixed and stable, and the relative orientations of these components does not change. The projected location of the imaging apparatus axis of rotation at every angle of rotation is then represented by a central line along the imaging detector. This means that the projected location of the imaging apparatus axis of rotation is a central line in every image taken by the imaging apparatus, regardless of gantry rotation angle. This central line is perpendicular to a first set of image edges (e.g. top and bottom edges) and perpendicular to the other set of image edges (e.g. left and right side edges). This central line is used in some current methods as part of the alignment process when reconstructing a 3D image.

However, in a true' or 'real' system, the relative distances and orientations of the components of the imaging apparatus are not fixed and stable. As the imaging apparatus is rotated by the gantry of the radiotherapy machine, the imaging system, gantry, and the mechanical means via which the imaging system is coupled to the gantry may undergo small mechanical shifting or flexing' as a function of gantry angle. The flexing of the imaging apparatus components is compounded by the heavy treatment apparatus, which causes additional flexing of the gantry as a function of gantry rotation angle and which affects the relative positions of the components of the imaging apparatus.

A radiotherapy device with a rotatable source of radiation has an isocentre, and for an MV radiation source this may be referred to as an MV isocentre. It is important to position a patient within a radiotherapy machine with knowledge of the position of the MV isocentre, and its position with respect to the target region. In an ideal system, the isocentre can be thought of as the point in space intersected by the treatment beam at all angles of gantry rotation. However, due to the above-described mechanical effects, the position of the treatment apparatus (MV) isocentre may also vary slightly as the treatment apparatus is rotated.

In order to use the imaging capabilities to plan the radiotherapy treatment, it is necessary to ensure that the position of the MV isocentre relative to image(s) obtained by the imaging apparatus is accurately known. There is therefore a demand for calibration approaches that can accurately determine the position of the MV isocentre relative to images obtained by the imaging apparatus.

While prior art systems and methods are able to provide accurate 3D reconstructions, there is nevertheless a continual demand for improvement in image quality, and a demand for the high quality to be consistently achieved. It is known to use a software 'mapping' in an effort to address some of these issues. Existing mappings are capable of producing accurate 3D reconstructions which meet clinical and regulatory guidelines as well as provide the basis for safe radiotherapy.

However, despite their effectiveness, it has been appreciated by the present inventor(s) that prior methods of producing such a mapping contain flawed assumptions. To date, these assumptions have been considered to be essential in order to produce the mapping.

The present disclosure seeks to address these and other disadvantages in the prior 20 art

Summary

According to an aspect of the present disclosure, a method of calibrating a radiotherapy machine is provided. The radiotherapy machine comprises a treatment apparatus and an imaging apparatus both mounted to a rotatable gantry. The treatment apparatus is configured to direct a treatment beam of therapeutic radiation towards a treatment volume of the radiotherapy machine. The method comprises performing a calibration process. The calibration process comprises obtaining, using the imaging apparatus, a set of 2D images of a phantom positioned within the treatment volume, each 2D image of the set of 2D images taken at a different gantry rotation angle of a set of gantry rotation angles. The method comprises determining, based on a reference reconstruction of the phantom, an orientation indicator for at least one 2D image of the set of 2D images. The orientation indicator is indicative of how the at least one 2D image should be oriented with respect to the other 2D images of the set of 2D images during generation of a 3D image. The method also comprises generating an imaging calibration map which provides the orientation indicator as a function of gantry rotation angle.

The method may further comprise obtaining the location of a projected treatment apparatus isocentre at each gantry rotation angle of the set of gantry rotation angles, and the imaging calibration map may be generated such that it additionally gives the location of the projected treatment apparatus isocentre as a function of gantry angle.

According to another aspect, a method of generating a 3D image using the imaging calibration map is provided. The method comprises receiving an input set of 2D images, each of the input 2D images having been taken at a different gantry rotation angle of a set of gantry rotation angles. The method comprises identifying, using the imaging calibration map, an orientation indicator for each gantry rotation angle of the input set of 2D images. The method comprises orienting each image of the set of input 2D images based on the orientation indicators, and generating the 3D image based on the oriented input 2D images.

According to an aspect of the present disclosure, a radiotherapy machine comprising a treatment apparatus and an imaging apparatus both mounted to a rotatable gantry is provided. The treatment apparatus is configured to direct a treatment beam of therapeutic radiation towards a treatment volume of the radiotherapy machine. The machine further comprises, or is otherwise coupled to, a processor configured to perform any method or methods disclosed herein, including the aspects set out above.

According to an aspect of the present disclosure, there is disclosed a computer readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out any method or methods disclosed herein.

Also disclosed herein is a method of calibrating a radiotherapy machine. The radiotherapy machine comprises a treatment apparatus and an imaging apparatus both mounted to a rotatable gantry, and the treatment apparatus is configured to direct a treatment beam of therapeutic radiation towards a treatment volume of the radiotherapy machine. The method comprises performing a calibration process, which comprises obtaining, using the imaging apparatus, a set of 2D images of a phantom positioned within the treatment volume, each 2D image of the set of 2D images taken at a different gantry rotation angle of a set of gantry rotation angles.

The calibration process further comprises determining, based on a pre-acquired reference reconstruction of the phantom, a location of a projected imaging apparatus rotation axis in each 2D image, and obtaining the location of a projected treatment apparatus isocentre at each gantry rotation angle of the set of gantry rotation angles.

to The calibration process further comprises generating an imaging calibration map which provides the locations of the projected imaging apparatus axis and the treatment apparatus isocentre as a function of gantry angle.

Also disclosed herein is a method of calibrating a radiotherapy machine comprising a treatment apparatus and an imaging apparatus both mounted to a rotatable gantry is provided. The treatment apparatus is configured to direct a treatment beam of therapeutic radiation towards a treatment volume of the radiotherapy machine. The method comprises performing a calibration process which comprises obtaining, using the imaging apparatus, a set of 2D images of a phantom positioned within the treatment volume, each 2D image of the set of 2D images taken at a different gantry rotation angle of a set of gantry rotation angles. The method comprises determining, based on a pre-acquired reference reconstruction of the phantom, a location of a projected imaging apparatus rotation axis in each 2D image. The method comprises obtaining the location of a projected treatment apparatus isocentre at each gantry rotation angle of the set of gantry rotation angles. The method comprises generating an imaging calibration map which provides the locations of the projected imaging apparatus axis and the treatment apparatus isocentre as a function of gantry rotation angle.

Also disclosed herein is a radiotherapy machine comprising: a treatment apparatus and an imaging apparatus mounted to an annular gantry configured to rotate, wherein the treatment apparatus is configured to direct a treatment beam towards a location concentric to the gantry; and the imaging apparatus is configured to direct an imaging beam towards the location at an angle that is oblique to the treatment beam; a processor; and a computer readable storage medium comprising instructions which, when executed by the processor, cause the computer to carry out the method of the first aspect. The computer readable storage medium may also have stored thereon a digitised representation of an imaging phantom.

According to an aspect of the present disclosure, there is provided a computer readable storage medium storing an imaging calibration map which provides an orientation indicator and a location of a projected treatment apparatus isocentre as a function of gantry rotation angle.

The imaging calibration map may be generated according to the methods disclosed herein. The imaging calibration map may additional provide the location of the projected treatment apparatus isocentre as a function of gantry angle. The imaging calibration map may take the form of a look-up table.

Brief description of the drawings

Examples of the present disclosure will now be described with reference to the accompanying figures, in which: Figure 1 shows an Image Guided Radiotherapy (IGRT) machine comprising a treatment apparatus and an imaging apparatus; Figures 2a-2d show rotation of the IGRT machine of Figure 1; Figure 3 is an illustration showing how a centre of rotation of the imaging apparatus of Figure 1 can move relative to a patient being treated; Figure 4 is an illustration showing how a treatment beam produced by the treatment apparatus can move relative to a patient being treated; Figures 5a-5d are illustrations showing four 2D images obtained from the imaging apparatus of Figure 1; 10 15 20 25 Figure 6 is an illustration of a 3D image that comprises the 2D images of Figures 5a-5d.

Figure 7 is an illustration showing the location of a treatment beam isocentre within 5 the 3D image of Figure 6.

Figures 8a and 8b are flowcharts showing an image calibration method according to the present disclosure.

Figure 9 is a flowchart showing a calibration method for a multi-energy radiotherapy

device according to the present disclosure.

Figure 10 is a flowchart showing an IGRT workflow.

Figure 11 is a diagram showing how the sag of imaging apparatus can cause parts of a patient's anatomy to appear magnified on a image.

Figures 12a and 12b are diagrams showing the impact of sag from different gantry rotation angles.

Figure 13 depicts images taken at the gantry rotation angles shown in figures 12a and 12b, and a representation of an image constructed using these images.

Figures 14a-c show 2D slices through 3D images.

Specific description

To give a brief overview, the present disclosure relates to a method of calibrating a radiotherapy machine. As a result of a calibration process, an imaging calibration map is produced. Gravity acts on the heavy components of the imaging and treatment apparatus as they are rotated via the gantry, changing their relative positions and orientations. These components are said to undergo flexing'. The imaging calibration map of the present disclosure provides an orientation indicator and the location of a treatment apparatus isocentre as a function of gantry rotation angle. The calibration map can therefore be described as a map of the flex characteristics for a particular machine.

This calibration map may be used during the generation of 3D images on the same radiotherapy machine. In particular, the imaging calibration map describes how images taken at particular gantry rotation angles should be orientated or assembled with respect to other images taken at different gantry rotation angles in order to provide an accurate 3D reconstruction that accounts for mechanical flexing effects. Thus, the consistent accuracy of 3D images / reconstructions may be ensured.

By providing the position of a treatment apparatus isocentre as a function of angle, the imaging calibration map can be used to produce a 3D reconstruction which accounts for the slight variation of the MV isocentre with gantry angle. Such a 3D reconstruction may be used to ensure the patient is positioned accurately prior to radiotherapy treatment.

positioning of patients for IGRT and relevant context useful for understanding the presently disclosed methods Figure 10 shows a flowchart of a known IGRT workflow 1000. This workflow serves to demonstrate the advantages of improved quality of 3D reconstructed images prior to radiotherapy treatment, as can be achieved using calibrated machines and the methods of the present disclosure.

At 1001, one or more reference images are obtained. The images are 3D. The images are of a patient who is to undergo radiotherapy treatment. These reference images may be taken on a CT scanner and thus may have a high quality and high resolution. These images may form the basis of a radiotherapy treatment plan. These reference images may be taken some time before the radiotherapy treatment is to Occur.

At 1002, one or more pre-treatment images are obtained using the imaging apparatus of a radiotherapy device. The images are 3D. These images may be referred to as IGRT images. These images are taken just before treatment is to commence. CBCT techniques may be used to produce these images.

At 1003, the pre-treatment image(s) are compared with the one or more reference image(s). The images are aligned, for example using suitable features of the patient's anatomy which can serve as landmarks. Offset information is obtained. This offset information may be a three dimensional vector which indicates a shift required in order to align the pre-treatment images with the reference images. This process is known as registration and the skilled person in the field of radiotherapy will be familiar with registration techniques.

In an example, the clinical user with the aid of the IGRT software superimposes the 3D scan done by the IGRT system at 1002 on the 3D CT scan of the treatment plan, taken at 1001. When the IGRT image is perfectly superimposed on the treatment plan image, the software calculates how much the IGRT image had to be moved and in which direction. Then this amount is transformed in shifts to be sent to the table to position the patient at isocenter.

At 1004, the patient is positioned for radiotherapy treatment using the offset information. Typically, the patient positioning surface is re-positioned according to the determined offset. Thus, the patient is positioned accurately on the machine, and the MV isocentre of the machine is positioned at the region of the patient's anatomy as required by the treatment plan. Therefore, the precision of the radiotherapy treatment can be improved.

It will be appreciated from this workflow that the quality of the 3D images produced at step 1002 is important to ensure that accurate offset information is obtained, and thus to ensure the accuracy and precision of the subsequent radiotherapy treatment.

Figure 11 shows a sagging effect which the presently disclosed methods and calibration maps seek to address. For simplicity, diagrams 11-14c consider only the sag of the source of imaging radiation. In reality, the detecting panel also sags, and the present methods also account for this.

In figure 11, the position of a source of imaging radiation is shown without sag 1113, and with sag 1115. Radiation passes through objects 1102, 1104 at different angles depending on the level of sag displayed by the source of imaging radiation. Objects 1102, 1104 may be different regions of the patient's anatomy. The sag of the source of imaging radiation is towards the object. This happens when the source is at the top and the panel is at the bottom of the system. The effect of the sag is to magnify the objects and make them appear further apart in a resulting image. The opposite effect occurs when the imaging source is at the bottom of the gantry. It can be appreciated from the diagram that the apparent distance between the objects 1102 and 1104 is greater when the source sags (see arrows 1109, 1111).

Figures 12a and 12b show the effect of sag of the imaging source at gantry rotation angles of +90° and -90°. Th figures show an imaging panel 1205. Figure 12a shows a source position 1213 undergoing a small sag effect. Figure 12b shows a source at a position where it is undergoing a larger sagging effect 1215. Differences in the amount of sag may occur Lateral sag of the source in this way, when it is located at one side of the objects, creates a projection of the objects shifted from the true position. The sag could be different at a particular angle when compared to its opposite angle, as shown in figures 12a and 12b. For example, the direction of rotation of the gantry can create a difference in the magnitude of sag.

Figure 13 shows a projection image (furthest left image) of two spherical objects with a small distance between them as taken at the gantry position shown in figure 12a.

Figure 13 also shows the projection image (centre image) taken at the opposite gantry angle, when the source is experiencing a larger amount of sag, as might be taken at the gantry position depicted in figure 12b.

Finally, figure 13 shows a slice through a 3D image constructed using the two images. The centre of the images has been used as the intersection point. The sag has not been accounted for. In other words, no flexmap or calibration map has been used when creating the reconstructed image (right). The two distinct objects have become one single larger object. This is a blurred image/reconstruction.

The 3D spatial resolution, i.e. the ability to resolve two distinct near objects, decreases when the flex-map calibration is not optimal. This is demonstrated in figures 14a-c. Figure 14a shows a slide of the 3D CT scan when the object of interest is almost at centre of the image. This slide is from a CT scan taken at step 1001 of figure 10. Figure 14b shows a slide of the IGRT scan where the object of interest is higher on the right with respect to the centre of the image. This IGRT scan is taken at step 1002 of figure 10. During registration (e.g. at step 1003 of figure 10), the IGRT image is then moved down to the left until the images superimpose as much as possible.

However, figure 14c shows a slide from an alternative IGRT scan. This IGRT is of lower quality and shows a blurred and less accurate reconstruction of the same object of figure 14b. The registration of the image of 14c with the image of 14a will be more difficult compared to the registration of the image of 14b with the image of 14a, as important features (corners, edges) are missing or blurred. Hence the resulting positioning of the patient at step 1004 is likely to be less accurate.

It is, in part, an aim of the present disclosure to address problems which may arise in the positioning of patients in a radiotherapy workflow due to poor quality 3d reconstructions, such as those discussed with respect to figure 14c. Having described an IGRT patient positioning workflow and given relevant context for the present disclosure, flaws associated with existing flexmap methods will be discussed.

Flaws of existing flexmap methods While capable of producing 3D reconstructions which are accurate within clinical and regulatory guidelines, existing methods of producing imaging calibration maps, sometimes referred to in the art as 'flex maps', contain flawed assumptions. As described above, some current methods assume that the projected location of the imaging apparatus axis of rotation is represented by a central line in each image.

This central line is perpendicular to a first set of image edges (e.g. top and bottom edges) and perpendicular to the other set of image edges (e.g. left and right side edges). This central line is used in current methods as part of the alignment process when reconstructing a 3D image.

In another prior method of producing a flexmap, rather than using a central line for each image, the location of the imaging apparatus rotation axis is found for each of a set of gantry angles by making use of the MV isocentre location at each gantry angle.

In such prior methods, first, the location of the projected MV isocentre at each of the set of gantry angles is found. For each angle, it is assumed that the projected imaging apparatus rotation axis must pass through the projected MV isocentre. Determining the location of the projected imaging apparatus rotation axis is therefore simply a matter of first finding the MV isocentre location for each of a set of gantry o angles, and then assuming that the projected imaging apparatus axis passes through this location while being perpendicular to a first set of image edges (e.g. top and bottom edges) and perpendicular to the other set of image edges (e.g. left and right side edges).

This method assumes that the MV isocentre is always intersected by the axis of rotation of the imaging system, regardless of gantry angle, and that the imaging panel or detector of the imaging apparatus is perfectly aligned in a G-T direction (the G-T direction is parallel with the axis of rotation of the gantry in a co-planar arrangement).

However, due to mechanical flexing effects, these assumptions are not always necessarily true. For example, if the imaging detector! panel sags or flexes at a particular gantry rotation angle, then the projected imaging apparatus axis of rotation in an image taken at that gantry rotation angle may not be pass through the MV isocentre. Creating a flex map that makes use of this assumption means that any 3D reconstruction process that relies on the flexmap may not be accurate. To date, these assumptions have either not been identified, or it has been assumed that these assumptions are necessary in order to produce an imaging calibration map.

Overview In contrast to the prior methods, methods of the present disclosure make use of an optimisation process, and thereby find the location of the projected imaging apparatus rotation axis, or in fact any type of orientation indicator, in one or more of the projected images based on a known reference reconstruction of a phantom placed in the treatment volume of a radiotherapy machine.

The treatment volume may be thought of as a volume in which patients may undergo radiotherapy treatment. The treatment volume may be described as the volume in space in which the treatment beam may be directed. In turn, this is at least in part defined by the area swept out by the radiation treatment beam as the gantry is rotated.

The set of gantry rotation angles may comprise, for example, 6 different gantry rotation angles. Once the calibration process has been undergone, the alignment of subsequent (input) images taken at a particular gantry angle may be corrected or adjusted according to which of the 6 gantry rotation angles used in calibration the associated input image angle is closest to.

To obtain accurate 3D reconstructions of a patient using a rotatable imaging apparatus configured to obtain 2D images at different rotation angles, the location of a 'true' axis of rotation of the imaging apparatus may be found relative to each 2D image (i.e. at each of a plurality of gantry rotation angles). In particular, the location of the true axis of rotation of the imaging apparatus as projected onto each 2D image may be found. The true axis of rotation may therefore be referred to as the projected axis of rotation, or the projected true axis of rotation.

Using the projected true axis of rotation of the imaging apparatus, the 2D images can be correctly aligned with respect to each other to form a 3D image, i.e. so that the 2D images intersect each other in the 3D image along the projected true axis of rotation. Because the imaging apparatus may mechanically flex as it rotates (under the weight of the components of the imaging and treatment apparatus), the position of the true axis of rotation of the imaging apparatus may change as a function of rotation angle.

In particular, where the ideal axis of rotation of the imaging apparatus is the axis of rotation in the absence of any mechanical flex in the system, the true axis of rotation will move relative to the ideal axis of rotation as a function of gantry rotation angle, due to the mechanical flexes. Therefore, by constructing 3D images without regard to mechanical flexes in the system (i.e. on the implicit assumption that the true axis of rotation corresponds to the ideal axis of rotation for the imaging apparatus), the 3D images will not produce accurate 3D reconstructions of a patient, because the 2D images will be misaligned with respect to each other.

With regards to calibrating radiotherapy machines to find a position of the MV (treatment) isocentre relative to images obtained by an imaging apparatus, mechanical flexes (of the treatment apparatus and of the imaging apparatus) can also have an effect. In particular, mechanical flexes in the treatment apparatus and in the imaging apparatus, as well as changes in the beam profile at different rotation angles and at different MV beam energies, will mean that the treatment beam isocentre will not necessarily coincide with the axis of rotation of the imaging apparatus, and indeed will often be displaced from the axis of rotation of the imaging apparatus. The displacement will vary as a function of gantry angle. Therefore, performing calibration based on an assumption that the treatment beam isocentre and the axis of rotation of the imaging apparatus always coincide with each other will give rise to an inaccurate calibration, thus preventing a clinician from properly positioning a patient relative to the treatment beam isocentre using the imaging apparatus.

The beam profile is the intensity distribution of the MV radiation on a plane perpendicular to the radiation beam centre. The beam profile, usually flat, is continuously monitored by the hospital with Quality Assurance protocols. However, the flatness is claimed by tolerances and any beam profile variation inside these tolerances can affect the imaging calibration.

The imaging calibration map may be referred to as a flexmap'. The "flexmap" may be considered as a calibration map that accounts for mechanical flexes in the imaging and/or treatment apparatus. The flexmap may record the true axis of rotation of the imaging apparatus as projected onto each 2D image.

The rotatable gantry may be configured to rotate about a gantry axis of rotation, and may have a central opening for receiving a patient. The gantry axis of rotation may be the same as the ideal axis of rotation of the imaging apparatus. The gantry may be capable of rotating through a full 360 degrees. The angle-dependent imaging calibration map may be useable in constructing an accurate 3D image from 2D images obtained by the imaging apparatus. The treatment apparatus and the imaging apparatus may each be mounted at fixed positions on the gantry (but nonetheless susceptible to flexing under the weight of their componentry).

As discussed above, it is important to accurately determine the projection of the true axis of rotation of the imaging apparatus on each 2D image obtained by the imaging apparatus, in order for an accurate 3D image/reconstruction to be created from the 2D images. In an example method of assembly, the 2D images are assembled such that they intersect each other as close to the true axis of rotation of the imaging apparatus as possible.

By determining the projected true axis of rotation of the imaging apparatus in each 2D image obtained by the imaging apparatus (e.g. by determining the projected true axis of rotation of the imaging apparatus as a function of gantry rotation angle), calibration errors that would otherwise be introduced by assuming that the true axis of rotation of the imaging apparatus coincides with the true axis of rotation of the imaging apparatus, are avoided. 3D images of consistently high quality can thereby be generated.

Present methods may determine the location of a projected imaging apparatus rotation axis or other orientation indicator in one or more 2D images based on a pre-acquired reference reconstruction of a phantom. This is contrary to prior methods, which rely on a false assumption that the location of the projected MV isocentre at a particular angle of gantry rotation coincides with the location of the projected imaging apparatus axis of rotation. In the present method, the imaging calibration map is generated directly from the 2D images obtained by the imaging apparatus. In other words, the determination of the location of the projected imaging apparatus rotation axis in each 2D image is decoupled from the determination of the location of the MV isocentre in each image.

By calculating the imaging calibration map directly from the 2D images obtained by the imaging apparatus as opposed to calibrating the imaging apparatus based on the position of the treatment beam isocentre as determined using the treatment beam apparatus, calibration errors introduced vis this assumption are avoided. Additionally, the imaging phantom need only be placed roughly at the centre of the gantry and therefore the calibration process is made simpler.

The method may comprise generating a first imaging calibration map by performing the method of the first aspect while an imaging panel of the imaging apparatus has a first lateral position relative to the imaging beam/imaging apparatus, and generating a second imaging calibration map by performing the method of the first aspect while the imaging panel of the imaging apparatus has a second lateral position, offset from to the first position. The method may further comprise generating a third imaging calibration map by performing the method of the first aspect while the imaging panel has a third lateral position, offset from the first and second positions. For example, the first position may be one in which a centre of the imaging panel is generally in line with the imaging beam (subject to mechanical flexes in the system), the second position may be one in which the centre of the panel is laterally offset from the imaging beam by a first amount, and the third position may be one in which the centre of the imaging panel is laterally offset from the imaging beam by a second amount.

The method may (further) comprise generating a clockwise imaging calibration map by rotating the gantry in a clockwise direction; and generating an anticlockwise imaging calibration map by rotating the gantry in an anticlockwise direction.

In one example, the method may comprise generating six separate image calibration maps: a first, clockwise image calibration map obtained by rotating the gantry in the clockwise direction while the imaging panel has the first lateral position; a first, anticlockwise image calibration map obtained by rotating the gantry in the anticlockwise direction while the imaging panel has the first lateral position; a second, clockwise image calibration map obtained by rotating the gantry in the clockwise direction while the imaging panel has the second lateral position; and so on.

Accordingly, the multiple image calibration maps could be used to account for the different amounts of mechanical flex that can occur when a radiotherapy machine is rotated in different directions, and when the imaging panel is moved to different lateral positions. For example, when the radiotherapy machine is to be rotated in a clockwise direction with the imaging panel in the first position, the first clockwise flexmap could be used. While reference is made to separate flexmaps, it will be appreciated that a single flexmap may be generated, which takes as an input the direction of rotation of the gantry and/or the configuration of the imaging panel.

Generating the imaging calibration map may comprise: constructing a 3D image from the set of 2D images; optimizing the 3D image until it provides a 3D reconstruction (e.g. an accurate 3D reconstruction) of the imaging phantom (e.g. optimizing the alignment of one or more 2D image with respect to the other 2D images in the 3D image, until the 3D image provides a 3D reconstruction of the imaging phantom); recording a configuration of the 2D images that provides the 3D reconstruction; and generating the imaging calibration map based on the configuration of the 2D images that provides the 3D reconstruction. Optimising the alignment of one or more 2D images with respect to one another may comprise adjusting an orientation indicator for one or more of the 2D images until an accurate 3D reconstruction is achieved.

As used herein, an orientation indicator is any indication, e.g. an image transform or an image feature, which serves to give an indication of the manner in which a particular 2D image taken at a particular gantry rotation axis should be oriented with respect to other 2D images taken at other gantry rotation angles in order that an accurate 3D reconstruction is generated during the 3D reconstruction process. In other words, an orientation indicator is indicative of how a 2D image should be oriented with respect to the other 2D images of the set of 2D images during generation of a 3D image.

Generating the imaging calibration map may comprise an initial step of placing the imaging phantom at the location in the radiotherapy machine, e.g. approximately at a position where a patient would be located during treatment. Optimizing the 3D image may comprise at least one of translating, rotating, magnifying and demagnifying at least one 2D image of the 2D images which comprise the 3D image, until the 3D image provides an accurate or optimised 3D reconstruction of the imaging phantom. The optimizing may be done using a pre-existing/pre-stored digitized 3D representation, i.e. accurate representation, of the phantom for comparison with the constructed 3D image. In other words, the 2D image is at least one of translated, rotated, magnified, and demagnified, in such a way that a line along which it intersects the other 2D images in the 3D image is moved; until the line of intersection is such that the 3D image is an optimized representation of the imaging phantom (i.e. until the line of intersection corresponds to a projection of the true axis of rotation of the imaging apparatus on the 2D image). Each of the resulting image transformations determined by the optimisation process for a 2D image (translation, rotation, magnification, or a combination of any of these transforms) are saved as an orientation indicator for that image, and/or for the associated gantry rotation angle.

to Generating the imaging calibration map based on the optimized 3D image (i.e. based on the at least one of translation, rotation, magnification and demagnification of the 2D image) may comprise: for each 2D image, determining an intersection line along which it intersects the other 2D images comprising the 3D image to provide the (accurate and/or optimized) reconstruction of the imaging phantom, the intersection line corresponding to the projected true axis of rotation of the imaging apparatus rotation axis on the 2D image; and for each 2D image, recording a gantry rotation angle at which it was obtained, and a position and direction of the intersection line on the 2D image, to thereby generate the imaging calibration map.

The 3D image may be (initially) constructed by assembling the 2D images so that each 2D image initially intersects the other 2D images in the set at a longitudinal bisecting line thereof, i.e. a longitudinal line that passes through the geometric centre of the 2D image. The translating and/or rotating may then be done iteratively, starting from the initial intersection of the 2D image in the initially constructed 3D image. For example, the axis along which the 2D image intersects the other 2D images in the set may be iteratively translated and/or rotated by small, perturbative amounts, until a best reconstruction of the imaging phantom is found. This iterative translating and/or rotating may be repeated for each of the 2D images, until an optimized reconstruction of the imaging phantom is found.

The method may further comprise determining a projected position of a treatment beam isocentre relative to the true axis of rotation of the imaging apparatus. Determining the projected position of the treatment beam isocentre relative to the true axis of rotation of the imaging apparatus may comprise: obtaining a 2D image of an isocentre phantom at the treatment beam isocentre using the imaging apparatus; and determining a point in the 2D image at which the isocentre phantom is located. Determining the point may comprise determining a point concentric to the isocentre phantom as represented in the 2D image. The method may further comprise determining a displacement of the point from the projected actual axis of rotation in the 2D image. This displacement may correspond to the position of the isocentre of the treatment beam relative to the actual axis of rotation of the imaging apparatus.

Determining the projected position of the treatment beam isocentre relative to the actual location of the axis of rotation of the imaging apparatus may further comprise an initial step of positioning the isocentre phantom at the isocentre of the treatment beam. This may be done by imaging the isocentre phantom using the treatment apparatus. For example, the positioning may comprise placing the isocentre phantom where the isocentre is expected to be, e.g. at the location; taking an image of the isocentre phantom using the treatment apparatus, to thereby determine the position of the isocentre phantom relative to the treatment beam isocentre; moving the isocentre phantom towards the treatment beam isocentre; taking a further image of the isocentre phantom to determine its position relative to the treatment beam isocentre; and repeating the above steps if necessary, until the isocentre phantom is positioned at the isocentre of the treatment beam.

The process of positioning a second (isocentre) phantom at the MV isocentre may be done once, and this phantom position may then be used for each gantry rotation angle such that a set of 2D images is obtained of the phantom. The resulting images can be used to determine the projected location of the second phantom at each gantry angle. Alternatively, the positioning step may be conducted for each gantry angle for which an image will be taken. The resulting flexmap then takes into account the small angular variance of the MV isocentre.

The treatment beam may comprise Megavolt (MV) radiation, e.g. may comprise MV x-rays. The imaging beam may comprise kilovolt (kV) radiation, e.g. may comprise kV x-rays.

The first (imaging) phantom may have an asymmetric shape, e.g. may have a surface profile resembling that of a human torso, or any other suitable asymmetric 3D object with a known surface profile. The second (isocentre) phantom may be spherical, e.g. may be a ball-bearing.

The calibration may be performed at multiple treatment beam energies, to thereby generate a calibration map for each treatment beam energy. For example, where the radiotherapy machine is configured to operate at one or more of a first treatment beam energy and a second treatment beam energy, the step of determining the projected position of the treatment beam isocentre relative to the actual location of the axis of rotation of the imaging apparatus at each gantry rotation angle may be performed at each of the first treatment beam energy and the second treatment beam energy.

The treatment apparatus may comprise a treatment beam source mounted on the gantry, and it may also comprise a target mounted on the gantry. The treatment beam source may be configured to direct the treatment beam radiation towards the location. The target may be configured to receive the treatment beam radiation, once it has passed through the location. In other words, the target may be mounted on the gantry, at an opposite side of the gantry from the treatment beam source. The target may be configured to capture images of the treatment radiation incident upon it, i.e. it may comprise an electronic portal imaging device (EPID). The treatment beam source may be attached to the gantry by an arm that extends laterally away from the gantry, i.e. extends away from the gantry at an angle that is parallel to, or oblique to, the axis of rotation of the gantry.

The imaging apparatus may comprise an imaging beam source mounted on the gantry, and it may also comprise an imaging panel mounted on the gantry. The imaging beam source may be configured to direct the imaging beam radiation towards the location. The imaging panel may be configured to receive the imaging beam radiation once it has passed through the location. In other words, the imaging panel may be mounted on the gantry, at an opposite side of the gantry from the imaging beam source. The imaging panel is configured to capture images of the imaging radiation incident upon it, i.e. it may comprise an EPID. The imaging beam source may be attached to the gantry by an arm that extends laterally away from the gantry.

The treatment beam may be substantially perpendicular to the imaging beam. In other words, the treatment beam source may be mounted to the gantry so as to direct radiation towards the location at a first angle relative to the gantry, and the imaging beam source may be mounted to the gantry so as to direct radiation towards the location at a second angle relative to the gantry, wherein the first angle is substantially perpendicular to the second angle.

Description of a radiotherapy machine

Fig. 1 shows an Image Guided Radiotherapy (IGRT) machine 100. IGRT machine 100 comprises a rotatable gantry 102 to which are affixed a treatment apparatus 104 and an imaging apparatus 106. In this example, the treatment apparatus 104 and the imaging apparatus 106 are attached to the gantry, so that they are rotatable with the gantry, i.e. so that they rotate as the gantry rotates. Positioned generally along an axis 'X' central to the gantry is a couch 110 upon which a patient 112 lies during radiotherapy treatment.

Treatment apparatus 104 is configured to direct a treatment beam of therapeutic radiation towards a treatment volume of the radiotherapy machine. Treatment apparatus 104 comprises a treatment beam source 114 and a treatment beam target 116. The treatment beam source 114 is configured to emit or direct therapeutic radiation, for example MV energy radiation, towards patient 112. As the skilled person will appreciate, the treatment beam source 114 may comprise an electron source, a linac for accelerating electrons toward a heavy metal, e.g. tungsten, target to produce high energy photons, and a collimator configured to collimate the resulting photons and thus produce a treatment beam. For reasons of clarity and brevity, these components are collectively referred to as the treatment beam source. Once the treatment radiation has passed from the source 114 and through the patient 112, the radiation continues towards treatment beam target 116, where it is blocked/absorbed. The treatment beam target 116 may include an imaging panel (not shown). The treatment beam target may therefore form part of an electronic portal imaging device (EPID). EPIDs are generally known to the skilled person and will not be discussed in detail herein.

Imaging apparatus 106 comprises an imaging beam source 118 and an imaging panel 120. The imaging beam source 118 is configured to emit or direct imaging radiation, for example X-rays and/or kV energy radiation, towards the patient 112 As the skilled person will appreciate, the imaging beam source 118 may be an X-ray tube or other suitable source of X-rays. The imaging beam source 119 is configured to produce kV energy radiation. Once the imaging radiation has passed from the o imaging beam source 118 and through the patient 112, the radiation continues towards imaging panel 120. The imaging panel 120 may be described as a radiation detector, or a radiation intensity detector. The imaging panel 120 is configured to produce signals indicative of the intensity of radiation incident on the imaging panel 120. In use, these signals are indicative of the intensity of radiation which has passed through a patient 112. These signals may be processed to form an image of the patient 112. This process may be described as the imaging apparatus 106 and/or the imaging panel 120 capturing an image. By taking images at multiple angles around the patient it is possible to produce a 3D image of the patient, for example using tomographic reconstruction techniques.

In the illustrated example, the treatment apparatus 104 and imaging apparatus 106 are mounted on the gantry such that a treatment beam travels in a direction that is generally perpendicular to that of the imaging beam.

Because the gantry 102 is rotatable, the treatment beam can be delivered to a patient from a range of angles. Similarly, the patient can be imaged from a range of angles. See for example Figures 2a-2d, each of which shows the gantry 102 of Fig. 1 at a different rotation angle. In Fig. 2a, the gantry is positioned at a 'first' gantry rotation angle, in which the treatment source 114 directs the treatment beam towards the patient in a vertical/downwards direction and in which the imaging source 118 directs the imaging beam towards the patient in a horizontal/right-to-left' direction. In Fig. 2b, the gantry has been rotated 45-degrees clockwise, into a 'second' rotation angle. In Fig. 2c, the gantry has been rotated a further 45-degrees clockwise (i.e. 90-degrees clockwise relative to Fig. 2a) into a 'third' rotation angle, so that the treatment source 114 directs the treatment beam towards the patient in a horizontal/'right-to-left' direction and in which the imaging source 118 directs the imaging beam towards the patient in a vertical/upwards direction. Finally, in Fig. 2d, the gantry has been rotated a further 45-degrees clockwise (i.e. 135-degrees clockwise relative to Fig. 2a) into a 'fourth' rotation angle.

As the skilled person will appreciate, the gantry 102 can be rotated to any of a number of discrete angular positions relative to a patient. The treatment apparatus 104 may direct radiation toward the patient at each or a number of these discrete angular positions, according to a treatment plan. The treatment apparatus 104 may even be used to continuously irradiate a patient at all rotation angles as it is rotated by the gantry 102. The angles from which radiation is applied, and the intensity and shape of the therapeutic beam, may depend on a specific treatment plan pertaining to a given patient. For reasons of clarity, the four angular positions of Figures 2a-2d will be discussed below. But as the skilled person will appreciate, the invention is not limited to the use of any particular number of angular positions, and is certainly not limited to the use of only four discrete gantry rotation angles.

Where a target region, which may comprise a tumour, in a patient is to be treated, a clinician will position the patient 112 on the couch 110, and/or will configure the treatment apparatus 104, such that the treatment beam is directed toward the target region at each gantry rotation angle. The target region is thus irradiated at each gantry rotation angle, and so will receive a dose of radiation according to the treatment plan. Areas of healthy tissue surrounding the target region will briefly be irradiated, but only at certain angles of rotation. Unlike the target region, a particular area of healthy tissue will not be irradiated from multiple angles, and therefore the particular area of healthy tissue will receive a reduced (and safe) dosage of radiation relative to the target region. In short, by correctly configuring the IGRT machine and correctly positioning the patient, irradiation of the target region can be performed while irradiation of healthy tissue surrounding the target region can be minimised.

Reference is made herein to the isocentre of the treatment apparatus 104. The isocentre may be described as a location in space that the treatment beam passes through at all gantry rotation angles. The therapeutic beam produced by the treatment beam source 114 may be conical, and in this case the isocentre may be described as the point or location in space through which the central radiation beam passes at all gantry rotation angles. In a hypothetical 'ideal system', in which flexing effects do not exist, the isocentre is a point in space. However, due to small mechanical movements or flexes, for example flexes in a connecting arm via which the treatment beam source 114 is connected to the gantry, the isocentre may actually be defined by a small region, area, or volume in space.

The imaging apparatus 106 is used to create a 3D image of the patient which incorporates the target region. Based on this 3D image, and a knowledge of the location in the 3D image at which the treatment beam isocentre lies, a clinician can identify a projected position of the isocentre relative to the target region. The couch 110 can then be moved, and/or the configuration of the treatment beam source 114 (e.g. the configuration of a beam collimator in the treatment beam source 114) can be modified, as necessary to ensure proper positioning of the target region relative to the isocentre.

It is therefore important that the location of the treatment beam isocentre in the 3D image is accurately known. For the clinician to accurately configure the apparatus and position the patient as above, it is also important for the 3D image to be an accurate 3D reconstruction of the patient, and for the projected position of the treatment beam isocentre in the 3D image to be known to a high degree of accuracy.

Mechanical flexing of the machine components As illustrated in Figure 3, the weight of the imaging beam source 118 and of the imaging panel 120 cause mechanical movements, or 'flexes', which occur as a function of gantry rotation angle. In particular, the imaging beam source 118 and the imaging panel 120 may be coupled or affixed to the gantry via mechanical means such as via connecting arms. The connecting arms or other mechanical means may sag, flex, or bow under the action of gravity due to the weight of the imaging panel and imaging beam source 118. The gantry itself may also flex and sag to differing degrees as it rotates, as the weight distribution of the various components attached and / or coupled with the gantry changes.

In Fig. 3, due to this effect, the imaging beam source 118 has become displaced from an ideal (unflexed) position 118, into a true position 118' relative to patient 112 having a target region 113. Similarly, imaging panel 120 has become displaced from an ideal (unflexed) position, into a true position 120' relative to the patient 112. The displacements are exaggerated in Fig. 3 for illustrative purposes. The amount by which the components are displaced relative to the patient, and/or relative to their 'ideal' positions, changes as a function of gantry rotation angle. Because of this, the true axis of rotation 300' of the imaging apparatus 106 is displaced from the ideal (unflexed) axis of rotation 300 of the imaging apparatus. As discussed in respect of Figs. 5-6 (below), adjustments must be made to account for this displacement in the true axis of rotation 300'.

Similarly, as shown in Fig. 4, the weight of the treatment beam source 114 and of the treatment beam target 116 mean that the treatment beam source 114 becomes displaced to position 114', and the treatment beam target 116 also becomes displaced to position 116'. Accordingly, the true path 400' of the treatment beam is displaced from the expected path 400 of the treatment beam. The amount by which the components are displaced relative to the patient changes as a function of gantry rotation angle. Furthermore, at any given angle, each of the treatment apparatus 104 and the imaging apparatus 106 will be displaced by different amounts. Therefore, there is no guarantee that the treatment beam isocentre will coincide with the centre of rotation of the imaging apparatus 106.

The assumption that the MV isocentre does lie on the axis of rotation of the imaging apparatus is a flaw in prior methods of calibration, and of constructing a 3D reconstruction of a patient.

Therefore, a projected position of the treatment beam isocentre relative to 3D images obtained by the imaging apparatus must be found (in order for 3D images obtained by the imaging apparatus to be useable in positioning a patient for radiotherapy treatment). This is discussed in respect of Fig. 7 below. Furthermore, the angular dependence of the MV isocentre must also be taken into account.

Fig. 5a shows a 2D image 500 as obtained when the gantry is oriented as shown in Fig. 2a. Fig. 5b shows a 2D image 502 as obtained when the gantry is oriented as shown in Fig. 2b. Fig. 5c shows a 2D image 504 as obtained when the gantry is oriented as shown in Fig. 2c. Fig. 5d shows a 2D image 506 as obtained when the gantry is oriented as shown in Fig. 2d. Also shown in each of images 5a-5d is an ideal rotation axis 300 of the imaging apparatus (i.e. an axis that would be expected in the absence of mechanical flexes). As shown, the ideal axis 300 for each 2D image is a longitudinal line that extends through the geometric centre of each image. In images 500, 502, 504, it happens (in this illustrative example) that the true axis of rotation 300' coincides with the ideal axis of rotation 300. However, in the case of 2D image 506, flexes in the imaging apparatus 106 when oriented as shown in Fig. 2D have caused the true axis of rotation 300' to become displaced from the ideal axis of rotation 300.

There are a number of ways to assemble or process the resulting 2D images in order to produce a 3D image. One way to process the images is to assemble them given knowledge of a shared axis in the images. Prior methods have found such a shared axis by finding the location of the MV isocentre in each image, for example by positioning a ball-bearing phantom at the MV isocentre during a calibration process and then taking images of the ball-bearing phantom at each of several gantry angles.

The location of the ball-bearing phantom in each image can be easily identified. Prior methods have then, on the assumption that the MV isocentre lies along the imaging system rotation axis, assumed that a line which passes through this identified MV isocentre location in each image may be used as a shared axis in each 2D image. In other words, a problem with current Flex-map algorithms / methods is the assumption that the MV isocentre is on the axis of rotation of the imaging system and that the panel is perfectly aligned in the G-T direction. So, once the pixel of group of pixels on the detector panel representing the isocentre is found, then the imaging system axis of rotation is the just the whole row of pixels where the isocentre is.

When initially constructing a 3D image from the 2D images of Figs. 5a-5d on the assumption that there is no flex in the imaging apparatus 106, the images are 'assembled' such that they intersect each other along the ideal axis 300, as illustrated in Fig. 6. However, as can be seen from Fig. 6, when the images 500-506 are assembled in this way, the displacement of the true axis of rotation 300 in image 506 has meant that the 3D image does not provide an accurate reconstruction of the patient. Image 506 has to be translated in direction T until it intersects the other images 500, 502, 504 along the true axis of rotation 300'. Only when image 506 has been translated in this way (i.e. so that it intersects the other 2D images along the true axis of rotation 300' of the imaging apparatus), does the 3D image provide an accurate 3D reconstruction. In practice, when an IGRT machine is first installed, the projected location of the true axis of rotation 300' of the imaging apparatus in each obtained 2D image is unknown. Calibration is required to determine the projected location of the true axis of rotation 300'. The calibration is described in detail with reference to Fig. 7.

Above, correction for a displacement of the components of the imaging apparatus has been described. However, as the skilled person will appreciate, torsional forces may also act on the imaging apparatus, causing the true axis of rotation 300' to be rotated relative to the expected axis of rotation 300 (rather than translated). In this case, 2D image 506 (for example) may have to be rotated, rather than translated, in order to obtain an accurate 3D reconstruction. And in other cases, forces may also cause the displacement between the image beam source 118 and the imaging panel 120 to increase or decrease. In this case, 2D image 506 (for example) may have to be magnified, or demagnified, in order to obtain an accurate 3D reconstruction.

Once the projected true axis of rotation 300' of the imaging apparatus in each 2D image obtained by the imaging apparatus has been found, such that accurate 3D reconstructions can be constructed from the 2D images, the next step is to determine where the treatment beam isocentre lies, as a projected position within the accurate 3D reconstructions. The position of the treatment beam isocentre relative to the IGRT machine may be determined by trial and error. In particular, a ball-bearing phantom is placed at the expected position of the treatment beam isocentre (i.e. along the geometric centre of the gantry), and images of the ball bearing phantom are acquired, using the treatment apparatus, for a full rotation of the gantry. From the images, it is determined whether the ball-bearing phantom is positioned at the treatment beam isocentre. If the ball bearing is not positioned at the isocentre, then it is moved slightly. This process is repeated until the ball-bearing phantom coincides with the treatment beam isocentre. Keeping the ball-bearing phantom in place, the imaging apparatus is then used to obtain 2D images for a full rotation of the gantry. The 2D images from the imaging apparatus 106 are then checked for an image of the ball-bearing, which corresponds to the projected position of the treatment beam isocentre in the 3D image. Fig. 7 illustrates where the projected treatment beam isocentre 700 may be located On this example, it is located in image 500).

Calibration according to the present disclosure

Following on from the above discussion, the calibration approach according to the

present disclosure will now be discussed.

Fig. 8a is a flowchart showing a calibration approach or process 800 according to the present disclosure. The calibration process may be performed upon installation of a radiotherapy machine. The process is split into two parts. The process may also be performed periodically after installation, for example every six months. Additionally, or alternatively, it may be performed whenever 3D image consistency falls below a required threshold. Once calibration has been completed, the resulting calibration map is used in subsequent operation using the IGRT machine. In particular, the calibration map is used when obtaining a 3D image of a patient. These 3D images are used during the registration procedure and when positioning a patient for treatment.

Calibration method 800 comprises two stages. The first stage is described with respect to steps 802-812 of the flowchart, and the second stage is described with respect to steps 814-818 of the flowchart. In the first stage, a location of a projected imaging apparatus rotation axis in each of a plurality of 2D images of a phantom positioned in the treatment volume is determined. This allows an imaging calibration map to be generated which provides the location of a projected imaging apparatus axis as a function of gantry rotation angle. This calibration map may be described as an 'intermediate' calibration map and may be used to provide more consistently accurate images. The second stage of the method 800 comprises generating an imaging calibration map which provides the locations of the projected imaging apparatus axis and the treatment apparatus isocentre as a function of gantry rotation angle. This calibration map may also be used to provide more accurate images, but the additional information regarding the location of the treatment apparatus isocentre as a function of gantry rotation angle additionally allows the more accurate planning of radiation therapy.

At step 802, a phantom is positioned / placed within a treatment volume of the radiotherapy apparatus. The phantom may be an imaging phantom which is visible in images taken via the imaging apparatus. The imaging phantom may be positioned / placed at, or near, the expected position of the treatment apparatus isocentre. This may be substantially at, or near, a geometric centre of the treatment volume of the IGRT machine 100. However, the exact positioning is not important. The imaging phantom may be positioned within the treatment volume at a location other than a treatment apparatus isocentre, and therefore the method may comprise positioning the phantom in the treatment volume without regard for the location of the treatment apparatus isocentre. This is in marked contrast to existing methods which are entirely reliant on the accurate positioning of a phantom at the MV isocentre.

Therefore, methods of the present disclosure may be considered to provide an easier or more simple calibration process.

The imaging phantom has a well-defined 3D shape, for which an accurate 3D reconstruction is known. Images will be taken at each of a plurality of gantry angles, and therefore the shape of the phantom may be an irregular and/or asymmetric shape to assist in ensuring the calibration is accurate. The shape of the imaging phantom may present a different 2D projected shape at each of the plurality of gantry angles. The shape of the first phantom may be described as anisotropic. For example, the first phantom may have a shape resembling that of a human torso. The first phantom may comprise any suitable material which will render the phantom visible to the imaging modality of the imaging apparatus.

At step 804, a set of 2D images is obtained. The set of 2D images is obtained using the imaging apparatus. The set of 2D images are of the phantom positioned within the treatment volume. Each 2D image of the set of 2D images is taken at a different gantry rotation angle of a set of gantry rotation angles. For example, a 2D image of the imaging phantom may be taken at each of the gantry rotation angles shown in Figs. 2a-2d. In another example, an image may be taken at each of the following gantry rotation angles: 0°, 60°, 1200, 1800, 240°, 300°. In this particular example, the set of 2D images would comprise 6 images. Use of the term 2D image is intended to describe an image, a projection, as well as projection or radiation intensity data, as would be understood by the person skilled in the art.

At step 810, an orientation indicator for at least one 2D image of the set of 2D images is determined. An orientation indicator for a particular 2D image gives an indication of how the image should be oriented with respect to the other 2D images during generation of a 3D image. Preferably, an orientation indicator is determined for each image of the set of 2D images, and this may comprise simply determining that the orientation of one or more of the images need not be changed.

As discussed above, an example of an orientation indicator includes a location of a projected imaging apparatus rotation axis in the at least one 2D image. The location of the projected imaging apparatus rotation axis is indicative of how the at least one 2D image should be oriented with respect to the other 2D images of the set of 2D images during generation of a 3D image. In some implementations, when the location of the imaging apparatus rotation axis is known for each image, and hence for each gantry rotation angle, this can be used as an alignment axis and images can be aligned along this shared or common axis in a manner described above in relation to figures 6 and 7.

Once the location of an alignment axis, which may be referred to as a common or shared axis, or an 'intersection line', has been identified in each image, it can then be determined how each image should be transformed in order for each 2D image to be properly aligned with the other 2D images during the 3D reconstruction process. For example, each image may be adjusted such that they align along the alignment axis. For example, each image may be adjusted such that they intersect one another along the alignment axis. To bring the identified common axis into alignment, one or more of the images may need to be transformed. For example, due to the flexing of mechanical components associated with the imaging panel at a particular gantry rotation angle, images taken at that angle may need to be rotated to allow them to be properly aligned with images taken from other gantry rotation angles.

In this way, the orientation indicator for the at least one 2D image may comprise a location of an axis, such as the projected imaging apparatus rotation axis, in that image. The orientation indicator for the at least one 2D image may additionally or alternatively comprise one or more image transformations which result in the at least one 2D image being properly orientated with the other 2D images. The image transformation may describe how the at least one 2D image should be transformed in order for the projected imaging rotation axis in the image to be aligned with the projected imaging apparatus locations in each of the other images. Suitable image transformations include any of a translation, a rotation, a magnification, and a demagnification. Preferably, an orientation indicator is determined for each 2D image of the set of 2D images such that each image can be properly oriented and/or aligned during the 3D image generation process.

This is in contrast to known methods which may, for example, determine the location of the MV isocentre in an image and then, based on the false assumptions that a) the MV isocentre always lies along the KV imaging apparatus rotation axis, and b) the imaging panel is rigidly and fixedly attached to the radiotherapy apparatus and its relative position to the imaging source does not change, use the MV isocentre as a single reference point. Having found this reference point, prior methods assume that a line which contains the reference point and which is perpendicular/parallel to the image edges represents the location of the projected imaging apparatus rotation axis.

By removing the need to use these flawed assumptions by allowing an orientation indicator to comprise a rotation or a translation, the consistent accuracy of subsequent 3D reconstructions using the imaging calibration map is assured.

A further flawed assumption of previous methods is that the distance of the imaging panel from the imaging radiation source and from the phantom is assumed to be fixed and unchanging with gantry angle. However, again, it has been appreciated by the present inventor that this assumption is not correct. More accurate results can be achieved by allowing the orientation indicator to be a magnification or a demagnification.

It will be appreciated that orientation indicators which describe the location of a common or shared axis in the images may directly relate to a corresponding image transformation. For example, assigning an orientation indicator image of a projected imaging apparatus rotation axis which is not perpendicular/ parallel to the image edges for a particular 2D suggests that that image should be rotated with respect to the other 2D images during generation of a 3D image.

The determination of an orientation indicator for each 2D image is performed based on a pre-acquired reference reconstruction of the phantom, for example using an optimisation process. A suitable optimisation process may seek to iteratively adjust a trial orientation indicator (such as a trial location of the projected imaging apparatus rotation axis) for each 2D image, and assess a 3D trial reconstruction produced using the trial orientation indicators (e.g. the trial locations of the projected imaging apparatus). The assessment of the 3D trial reconstruction is performed based on the known reference 3D reconstruction.

The method may comprise constructing a trial 3D reconstruction from the set of 2D images using a trial orientation indicator for each 2D image, and performing an optimisation process which comprises iteratively adjusting at least one trial orientation indicator until the trial 3D reconstruction provides an accurate 3D reconstruction of the phantom. In particular, the method may comprise performing an optimisation process which comprises adjusting a trial location of the projected imaging apparatus rotation axis in each 20 image until the trial 3D reconstruction provides an accurate 3D reconstruction of the phantom. The accuracy of the trial 3D reconstruction is assessed based on the pre-acquired 3D reference reconstruction of the phantom.

With reference to the flowchart of figure 8a, in an example implementation, determining, 810, the orientation indicator for each 2D image may comprise constructing, 806, a trial 3D reconstruction from the set of 2D images, and performing, 808, an optimisation process based on a reference reconstruction of the phantom.

At step 806, a trial 3D reconstruction is constructed from the set of 2D images. The trial 3D reconstruction may be referred to as a trial 3D image. The trial 3D image may be constructed using a tomographic reconstruction technique. For example, the trial 3D image may be constructed by assembling the 2D images in the manner depicted in, and described with respect to, Fig. 6. The 'assembly' may be performed using known techniques such as digital geometry processing and computed tomography techniques.

As part of the process, the 2D images are assembled based on a trial orientation indicator assigned to each 2D image. A suitable trial orientation indicator may be a line which is perpendicular and parallel to the image edges and which passes through a geometric centre of the image (or, if the projection MV isocentre location is known for each gantry angle, which passes through the projected MV isocentre location). In this way the 2D images may be assembled so that each 2D image intersects the other 2D images along a longitudinal line passing through a geometric centre of the 2D image. Based on this alignment, a trial 3D image is created. However, as described with respect to Fig. 6, the trial 3D image may not be an accurate representation of the imaging phantom.

At step 808, the trial 3D reconstruction undergoes an optimisation process. The purpose of the optimisation process is to determine an orientation indictor for each 2D image, i.e. for each gantry angle of the set of gantry angles from which the images were taken. By using images of a known phantom for which an accurate reference reconstruction exists, it is possible to assign orientation indicators based on the known 3D reconstruction. The optimisation process is based on this 3D reference reconstruction of the phantom. The reference reconstruction may have been previously acquired. The reference 3D reconstruction of the phantom may be a stored digitized 3D representation of the imaging phantom. The reference reconstruction is created with knowledge of the shape of the imaging phantom. The reference reconstruction represents an accurate reconstruction of the first phantom.

The trial orientation indicators can be iteratively adjusted based on the known 3D reconstruction until an accurate trial 3D reconstruction of the phantom is produced. The resulting orientation indicators are then saved and can subsequently be used to determine how new input images taken at various gantry angles on the same radiotherapy machine / device should be orientated as part of the reconstruction process such that an accurate 3D image is produced.

The accuracy of the trial 3D reconstruction is assessed based on the reference reconstruction of the phantom. The optimisation process may seek to maximise the accuracy of the trial 3D reconstruction based on the reference reconstruction, i.e. to minimise the error in the trial reconstruction, over a pre-determined number of iterations. This optimisation process may be perturbative process.

In implementations in which the optimisation indicator comprises a projected location of the imaging apparatus rotation axis, step 808 may comprise iteratively updating a trial location of the projected imaging apparatus rotation axis in each 2D image. This optimisation process comprises adjusting a trial location of the projected imaging apparatus rotation axis in one or more 2D image of the set of 2D images obtained at step 804 until the trial 3D reconstruction provides an accurate 3D reconstruction of the phantom.

One way to optimise the trial 3D reconstruction is to adjust a trial location of the projected imaging apparatus rotation axis in each 2D image in an iterative process until the trial 3D reconstruction provides an accurate 3D reconstruction of the phantom. The assessment of accuracy may be performed in a number of ways. For example, the optimisation process may comprise adjusting the trial locations of the projected imaging apparatus rotation axes, based on the reference 3D reconstruction, until a stopping criterion has been met. The stopping criterion may comprise reaching a threshold number of iterations. Alternatively, the optimisation process may be stopped when the obtained 3D reconstruction at the end of each iteration reaches some accuracy specification. The skilled person will be aware of suitable iterative optimisation processes of this nature.

The optimisation process may comprise iteratively translating, rotating, magnifying, and/or de-magnifying one or more of the 2D images as part of the optimisation process, until the trial 3D image matches the reference 3D representation to a sufficient degree of accuracy. Considering fig 5, in particular starting with first 2D image 500, image 500 is iteratively adjusted as part of the optimisation process The adjustment comprises one or more of translation, rotation, magnification and demagnification, until a best fit to the 3D reference representation is achieved. Similarly, second 2D image 502 is iteratively adjusted (e.g. translated, rotated, magnified and/or de-magnified), until a best fit to the 3D reference representation is achieved. This process will be repeated for each 2D image comprising the trial 3D reconstruction until the best fit to the 3D reference reconstruction / representation is found, in other words until an accurate 3D reconstruction of the phantom is provided, where the accuracy of the trial 3D reconstruction at each iteration of the optimisation process is assessed based on the pre-acquired reference reconstruction of the first phantom.

In summary, the optimisation process finds an orientation indicator for one or more 2D images of the set of 2D images. The orientation indicator may be a line in each image, i.e. on each projection, which should be used as an alignment axis during the reconstruction process in order to provide an accurate 3D image of the phantom.

Optionally, at step 812, a calibration map is generated. The calibration map generated at block 812 may be used to provide more accurate images. This calibration map may be refer to as an 'intermediate' calibration map, as it does not yet contain any information about the MV isocentre and so cannot yet be used to plan radiotherapy treatments.

The generation at block 812 is based on the determined orientation indicators for each 2D image and can be used to provide the orientation indicator as a function of gantry angle. For example, the intermediate imaging calibration map can be used to provide the location of the projected imaging apparatus axis as a function of gantry angle. The intermediate calibration map may take the form of a look-up table comprising two columns; the first containing each gantry rotation angle of the set of gantry rotation angles, and the second containing the determined orientation indicator, e.g. the location of the projected imaging apparatus rotation axis, for that angle.

For each image, the projected imaging apparatus rotation axis may simply be a line in the image, for example a line of pixels. The location of the line is recorded and saved in memory. This may be in the form of a saved location of each pixel making up the line, or else the location of a starting and ending pixel which can be connected to form the line. The location of the line may be saved as a vector. The location of the projected imaging apparatus rotation axis is recorded alongside a gantry rotation angle at which the 2D image was obtained. By repeating this step for each 2D image obtained, an imaging calibration map comprising pairs of data points (e.g. vectors paired with gantry rotation angles) is created. The imaging calibration map can thus be stored in memory as a lookup table having two columns.

Having obtained the intermediate calibration map, when imaging a patient with the IGRT machine the intermediate calibration map may be used to ensure that 3D images constructed from the obtained 2D images provide an accurate 3D reconstruction of the patient which accounts for the flexing of the various components of the radiotherapy machine.

In some radiotherapy devices, the imaging panel 120 can be moved between each of a plurality of positions relative to the imaging beam. For example, the imaging panel may be moveable to one of three positions: a neutral/central position, a first laterally displaced position, and a second laterally displaced position. Therefore in some implementations of the method, steps 802-812 may be repeated for each of the three positions in order to generate three different intermediate calibration maps which are used depending on the position of the imaging panel. Furthermore, steps 802-812 may be repeated once in a clockwise direction, and again in an anticlockwise direction. Therefore, steps 802-810 may be repeated six times, to capture each permutation of rotation direction and imaging panel position. The resulting intermediate calibration maps may be saved in memory and each one is used to account for mechanical flexing depending on the position of the imaging panel and / or the direction of rotation of the gantry.

By performing calibration steps as set out above, consistently accurate patient images can be achieved. However, to plan a radiotherapy treatment using the imaging calibration map (flexmap), it is useful to perform the additional calibration steps as outlined below in order to generate an imaging calibration map which provides the location of not only the projected imaging apparatus axis as a function of gantry angle, but also the location of the treatment apparatus isocentre as a function of gantry rotation angle.

At step 814, a set of 2D images of a second phantom positioned at a treatment apparatus isocentre is obtained. The second phantom is spherical and may be e.g. a ball-bearing phantom. Step 814 comprises obtaining, using the imaging apparatus, a set of 2D images of the second phantom positioned at the treatment apparatus isocentre for each gantry angle of the set of gantry angles. The location of the treatment apparatus isocentre may be slightly different at each gantry rotation angle ad therefore the second phantom may be re-positioned before each image is obtained.

A suitable method of obtaining a set of 2D images of the second phantom, as required at block 814 if process 800, is depicted in the flowchart of figure 8b.

At 832, the second phantom is positioned at the isocentre for one of the gantry rotation angles of the set of gantry rotation angles. The set of gantry rotation angles at block 832 is the same set as referred to with respect to block 804. The second phantom may be positioned at the isocentre for a particular gantry rotation angles based on data and/or images produced by the detector of the treatment apparatus. The detector may be an Electronic Portal Imaging Device (EPID). As will be appreciated by the skilled person, it is possible to position the second phantom at the isocentre for a particular gantry rotation angle by positioning the phantom roughly at the isocentre, directing the treatment beam toward the phantom, and then repositioning the phantom in an iterative process based on the resulting treatment beam intensity data generated by the treatment beam detector. The initial positioning of the phantom may be assisted by lasers suitably positioned to direct visible light along the same path as the treatment beam. Accordingly, the method may comprise positioning the second phantom at the treatment apparatus isocentre for each gantry angle of the set of gantry angles by directing the treatment beam into the treatment volume and using data provided by the treatment beam detector.

Once the second phantom has been positioned at the treatment beam isocentre for a particular gantry rotation angle, a 2D image is obtained at block 834 using the imaging apparatus. At block 836 it is determined whether an image of the second phantom at each gantry rotation angle of the set of gantry rotation angles has been obtained. If so, the process continues to block 816 of process 800. If not, the process of blocks 832 and 834 is repeated for a different gantry rotation angle, until a 2D image of the second phantom at the isocentre is obtained for each of the gantry rotation angles of the set of gantry rotation angles.

Returning to discussion of the figure 8a, at block 816 a location of a projected treatment apparatus isocentre in each 2D image is determined. This determination is carried out based on the images obtained at block 814. The determination may be as simple as locating the centre of the visible second phantom as shown in the images. The location may be stored and saved as a single point in the image, for example using x,y co-ordinates.

At step 818, an imaging calibration map is generated. This imaging calibration map may be used to provide the locations of both the projected imaging apparatus axis and the treatment apparatus isocentre as a function of gantry angle.

The calibration map generated at step 818 may take the form of a look-up table comprising three columns: the first containing each gantry rotation angle of the set of gantry rotation angles, the second containing the determined location of the projected imaging apparatus rotation axis for images taken at that angle, and the third containing the location of the projected treatment apparatus for images taken at that angle.

The set of gantry rotation angles may comprise every available or possible gantry rotation angle, or a subset of the available gantry rotation angles. For this reason, providing the orientation indicator and the location of the projected treatment apparatus isocentre as a function of gantry rotation angle for a new image taken at a first gantry angle may comprise determining which gantry angle of the set of gantry angles is nearest to the first gantry angle, and then providing the orientation indicator and the location of the projected treatment apparatus isocentre for the nearest gantry angle. For example, an imaging calibration map may be produced using a set of gantry rotation angles as follows: 00, 60°, 120°, 180°, 240°, 300°. Subsequently, it is desired to use the calibration map to produce an accurate 3D image based on 2D images taken at gantry rotation angles of 10°, 60°, 120°, 180°, 240°, 285°. In this case, the first image is oriented with respect to the other 2D images using the orientation indicator for 0°, and the final image is oriented with respect to the other 2D images using the orientation indicator for 300°. This can be thought of as sorting incoming images into bins according to the gantry angle at which they were taken and based on the nearest gantry angle of the set of gantry rotation angles.

Having obtained the calibration map at block 818, it may be used to determine not only how images taken at different gantry angles should be adjusted and assembled in order to produce an accurate 3D reconstruction of a patient, but also where the treatment beam isocentre is located for each gantry angle. Account is taken of both the flexing of the imaging apparatus and the treatment apparatus as the gantry rotates. Therefore, more consistently accurate images may be obtained, and more accurate radiation therapy may be achieved via positioning patients based on the resulting images.

By basing the determination of the location of the projected imaging (KV) rotation axes on a pre-acquired reference reconstruction of a phantom rather than on a determined location of a treatment beam (MV) isocentre, the determination of the locations of the projected imaging rotation axes and the projected treatment apparatus isocentre may be decoupled. By decoupling these determinations, flawed assumptions are removed from the calibration process. The resulting imaging calibration map, or flexmap, is not dependent on the MV beam profile. This means that high quality images can be obtained consistently. Accordingly, use of the present calibration process improves the consistency of accuracy and quality of patient imaging and of patient positioning prior to radiotherapy treatment.

It will be understood that the above description of specific implementations is by way of example only and is not intended to limit the scope of the present disclosure. Many modifications of the described implementations, some of which are now described, are envisaged and intended to be within the scope of the present disclosure.

Steps 814 to 818 may be performed for each of the intermediate calibration maps described above, e.g. for each imaging panel position relative to the treatment beam and/or for both gantry rotation in a clockwise and an anti-clockwise direction. In some implementations, the process 800 may be repeated multiple times, to capture each permutation of rotation direction and/or imaging panel position.

Reference is made to the flowchart of figure 8b, which depicts a method via which block 814 of method 800 can be implemented. The flowchart of figure 8b describes re-positioning the second phantom for each gantry rotation angle. However, in an alternative implementation, rather than re-positioning the second phantom for each gantry rotation angle as shown in the flowchart of figure 8b, each image of the set of 2D images of the second phantom may be taken with the second phantom located at the same position. The position may be at an 'average' isocentre when taking into account the position of the isocentres at each gantry angle. This alternative implementation does not account for the small variations in the isocentre location as a function of gantry angle, which may be acceptable for IGRT machines 100 in which this variation is sufficiently small.

In this alternative implementation, the phantom may be placed at or substantially at the geometric centre of the treatment volume, and a full 360-degree scan is performed with the treatment apparatus 104. The imaging panel on the treatment beam target 116 is used to image the second (ball-bearing) phantom. If it is determined, from the images taken at the treatment beam target, that the ball-bearing phantom is not positioned at the treatment isocentre, then the ball-bearing is moved, and the full 360-degree scan is performed again. This process is repeated until it is determined, from the images taken at the treatment beam target, that the ball-bearing phantom is located at the treatment isocentre, or the 'averaged' treatment isocentre. After this process has been completed, the second phantom is imaged at each of the gantry rotation angles of the set of gantry rotation angle in order to obtain the 2D images referred to at block 814.

Some radiotherapy / IGRT machines are configured to produce a treatment beam at each of a plurality of treatment beam energies. For example, the machine may comprise a linac capable of producing treatment beams at multiple energies. During treatment using such machines, different treatment beam energies may be used depending on the dose called for by the treatment plan for a particular patient. However, the position of the treatment beam isocentre may change as a function of treatment beam energy. These small variations may be caused by a range of factors; for example, the effect of the steering magnets typically used in a linac will vary depending on the energy of the treatment beam. Therefore, in some implementations, the projected location of the treatment beam isocentre for each gantry angle is determined for each treatment beam energy at which the radiotherapy machine can operate. Therefore the disclosed method may further comprise performing the calibration process at each treatment beam energy, to thereby generate an imaging calibration map which provides the locations of the projected imaging apparatus axis and the treatment apparatus isocentre as a function of gantry angle and as a function of treatment beam energy.

Calibrating a multi-energy device and other variants Figure 9 is a flowchart showing a method 900 for calibrating a multi-energy radiotherapy device according to the present disclosure. The blocks of process 800, shown in figure 8a, which are affected by consideration of multi-energy isocentres are 814 to 818. Therefore, blocks 802 to 810 may not need to be performed for each different beam energy. Accordingly, method 900 may comprise the steps 802-810 or 802 to 812 as described above in relation to figure 8a.

At block 942, one of the available treatment beam energies is selected. At block 932, the second phantom is positioned at the isocentre for a particular one of the gantry rotation angles of the set of gantry rotation angles in the manner described above in relation to block 832. At block 934, a 2D image of the second phantom positioned at the isocentre for the particular one of the gantry rotation angles in the manner described above in relation to block 834. At block 936, it is determined whether an image of the second phantom has been obtained for each gantry rotation angle of the set of gantry rotation angles, and thus it is determined whether a set of 2D images of the second phantom positioned at a treatment apparatus isocentre has been obtained. This is performed in the same manner described above in relation to block 836.

At block 944, it is determined whether a set of images has been obtained for each available treatment beam energy. If not, the process returns to block 932 and a different treatment beam energy is selected. If yes, then a plurality of sets of 2D images has been produced, with each set of images being similar to the set of images described above in relation to block 814 but each being associated with a different treatment beam energy.

The process then proceeds in a manner similar to that described above in relation to to blocks 816 and 818, but an imaging calibration map is produced for each treatment beam energy. At block 916, a location of a projected treatment apparatus isocentre is determined in each 2D image of the plurality of sets of 2D images. Finally, at block 918, an imaging calibration map is generated for each available treatment beam energy by performing the method described above in relation to block 818 but for each set of 2D images.

While reference is made to producing an imaging calibration map for each image, it will be appreciated that the imaging calibration map may be a single calibration map, but which provides the location of the projected imaging apparatus axis and the location of the treatment apparatus isocentre as a function of both gantry rotation angle and treatment beam energy.

Such an imaging calibration map may take the form of a look-up table having four columns: the first containing each gantry rotation angle of the set of gantry rotation angles, the second containing the determined location of the projected imaging apparatus rotation axis for images taken at that angle, the third containing the location of the projected treatment apparatus for images taken at that angle; and a fourth containing the treatment beam energy.

Such an imaging calibration map may be used to provide accurate images and patient positioning for each treatment beam energy of a multi-energy radiotherapy device.

While reference is made in figure 9 to positioning the second phantom at the isocentre for each gantry rotation angle, it will be appreciated that an alternative implementation may use a single isocentre position for each treatment beam energy. This alternative implementation is described above in relation to a single treatment beam energy. To take into account the variation in location of the isocentre with treatment beam energy in this alternative implementation, each image of a set of 2D images of the second phantom for a particular treatment beam energy is taken with the second phantom located at the same position. The position may be at an 'average' isocentre for that treatment energy. As above, this alternative implementation may be acceptable for IGRT machines 100 in which the angular variation of isocentre position is sufficiently small.

In some implementations, MV isocentre information is retrieved from storage. Steps 814 and 816 of figure 8 and the accompanying discussion above describe a method of obtaining the location of the projected treatment apparatus isocentre at each gantry angle. However, it is not necessary to conduct the steps of positioning the second (isocentre) phantom and taking images at a number of gantry angles. In some implementations of the presently disclosed calibration process, the location of the projected treatment apparatus isocentre at each gantry angle may comprise retrieving pre-acquired locations from storage, e.g. from local memory or from central or server memory. The MV isocentre locations stored in memory may be the result of previous calibration activities performed on the radiotherapy machine or, where deemed appropriate, on other similar radiotherapy machines. The full steps of positioning the second (isocentre) phantom and taking images, as described with respect to figure 8b, may be conducted with less regularity than the steps 802 to 810 or 802 to 812. In this way, the radiotherapy device imaging calibration map for a particular radiotherapy machine may be updated by re-performing steps 802-810 or 802-812, and then relying on previously acquired treatment apparatus isocentre data saved in storage to generate a imaging calibration map.

While the calibration process has been described primarily with reference to figure 8a in which the orientation indicators are acquired first (e.g. steps 802 to 810) and then the projected treatment apparatus isocentre positions are obtained (814 and 816), it will be appreciated by the skilled person that, unlike in prior methods, these determinations are not coupled. Accordingly, the calibration process may comprise first obtaining the projected treatment apparatus isocentre positions, and then determining the orientation indicators. In either variation of the method, an imaging calibration map is generated which provides the orientation indicator and the location of the projected treatment apparatus isocentre as a function of gantry rotation angle.

While the calibration process 800 has been described primarily as comprising two stages, it is also possible to perform them at the same time. To do this, the phantom used for the second stage (steps 814-818 of the flowchart) is placed in the treatment volume and is not moved as the gantry is rotated. While this phantom is located in the treatment volume, for each gantry angle, the phantom as used for the first stage (steps 802-812 of the flowchart) is positioned at the MV isocentre. An image is acquired at each gantry angle which includes both phantoms. The resulting images can form the basis of the optimisation process or other method of acquiring orientation indicators as discussed with respect to step 810 of the flowchart. The same images can be used to determine a location of a projected treatment apparatus isocentre in each 2D image at step 816. In other words, it is not necessary to obtain a first and a second set of images at steps 804 and 816, but the same set of images can be used by acquiring images which show both the first phantom, and the second phantom located at the isocentre.

While finding the location of the MV isocentre has been described as part of the calibration process, this need not be the case. Accordingly, a method is described herein of calibrating a radiotherapy machine comprising an imaging apparatus mounted to a rotatable gantry, the method comprising performing a calibration process which comprises: obtaining, using the imaging apparatus, a set of 2D images of a phantom positioned within the treatment volume, each 2D image of the set of 2D images taken at a different gantry rotation angle of a set of gantry rotation angles; determining, based on a reference reconstruction of the phantom, an orientation indicator for at least one 2D image of the set of 2D images, the orientation indicator being indicative of how the at least one 2D image should be oriented with respect to the other 2D images of the set of 2D images during generation of a 3D image; and generating an imaging calibration map which provides the orientation indicator as a function of gantry rotation angle.

The approaches described herein may be embodied on a computer-readable medium, which may be a non-transitory computer-readable medium. The computer-readable medium carrying computer-readable instructions arranged for execution upon a processor so as to make the processor carry out any or all of the methods described herein.

The term "computer-readable medium" as used herein refers to any medium that stores data and/or instructions for causing a processor to operate in a specific manner. Such storage medium may comprise non-volatile media and/or volatile media. Non-volatile media may include, for example, optical or magnetic disks. Volatile media may include dynamic memory. Exemplary forms of storage medium include, a floppy disk, a flexible disk, a hard disk, a solid state drive, a magnetic tape, or any other magnetic data storage medium, a CD-ROM, any other optical data storage medium, any physical medium with one or more patterns of holes, a RAM, a PROM, an EPROM, a FLASH-EPROM, NVRAM, and any other memory chip or cartridge.

It is to be understood that the above description is intended to be illustrative, and not restrictive. Many other implementations will be apparent to those of skill in the art upon reading and understanding the above description. Although the present disclosure has been described with reference to a specific example implementation, it will be recognized that the disclosure is not limited to the implementations described, but can be practiced with modification and alteration insofar as such modification(s) and alteration(s) remain within the scope of the appended claims.

Accordingly, the specification and drawings are to be regarded in an illustrative sense rather than a restrictive sense. The scope of the disclosure should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

Claims 1. A method of calibrating a radiotherapy machine comprising an imaging apparatus mounted to a rotatable gantry, the method comprising performing a calibration process which comprises: obtaining, using the imaging apparatus, a set of 2D images of a phantom positioned within the treatment volume, each 2D image of the set of 2D images taken at a different gantry rotation angle of a set of gantry rotation angles; determining, based on a reference reconstruction of the phantom, an to orientation indicator for at least one 2D image of the set of 2D images, the orientation indicator being indicative of how the at least one 2D image should be oriented with respect to the other 2D images of the set of 2D images during generation of a 3D image; and generating an imaging calibration map which provides the orientation indicator as a function of gantry rotation angle.
2. The method of claim 1, wherein the radiotherapy machine further comprises a treatment apparatus mounted to the gantry, and the calibration process further comprises: obtaining the location of a projected treatment apparatus isocentre at each gantry rotation angle of the set of gantry rotation angles; and generating an imaging calibration map which provides the orientation indicator and the location of the projected treatment apparatus isocentre as a function of gantry rotation angle
3. The method of any preceding claim, wherein the orientation indicator is a location of a projected imaging apparatus rotation axis in the at least one 2D image, and wherein determining an orientation indicator comprises determining, based on the reference reconstruction of the phantom, a location of a projected imaging apparatus rotation axis in each 2D image of the set of 2D images.
4. The method of claim 3, wherein determining the location in each 2D image of the projected imaging apparatus rotation axis comprises: constructing a trial 3D reconstruction from the set of 2D images; for each 2D image, determining an intersection line along which it must intersect the other 2D images comprising the trial 3D reconstruction to provide an accurate 3D reconstruction of the phantom, the accuracy of the trial 3D reconstruction being assessed based on the reference image of the phantom.
5. The method of claim 4, wherein generating the imaging calibration map comprises recording, for each gantry rotation angle of the set of gantry rotation angles, a position and/or a direction of the intersection line on the corresponding 2D 10 image.
6. The method of any of claims 3 to 5, wherein generating the imaging calibration map comprises recording, for each gantry rotation angle of the set of gantry rotation angles, the position of the projected treatment apparatus isocentre relative to the location of the projected imaging apparatus rotation axis.
7. The method of any preceding claim, wherein determining an orientation indicator for the at least one 2D image comprises determining an orientation indicator for each 2D image of the set of 2D images
8. The method of any preceding claim, wherein determining an orientation indicator for the at least one 2D image of the set of 2D images comprises: constructing a trial 3D reconstruction from the set of 2D images using a trial orientation indicator for each 2D image; and performing an optimisation process which comprises iteratively adjusting at least one trial orientation indicator until the trial 3D reconstruction provides an accurate 3D reconstruction of the phantom, the accuracy of the trial 3D reconstruction being assessed based on the reference image of the phantom.
9. The method of any preceding claim, wherein the orientation indicator for the at least one 2D image comprises an image transformation.
10. The method of claim 9, the image transformation comprising one or more of a translation, a rotation, a magnification, and a demagnification.
11. The method of claim 9 or 10, wherein determining the orientation indicator for the at least one 2D image comprises constructing a trial 3D reconstruction from the set of 2D images using a trial image transformation for each image; and performing an optimisation process which comprises iteratively adjusting at least one trial image transformation until the trial 3D reconstruction provides an accurate 3D reconstruction of the phantom, the accuracy of the trial 3D reconstruction being assessed based on the pre-acquired reference image of the phantom.
12. The method of any preceding claim, wherein the phantom is positioned within the treatment volume at a location other than a treatment apparatus isocentre, and/or wherein the method further comprises positioning the phantom in the treatment volume without regard for the location of the treatment apparatus isocentre.
13. The method of any preceding claim, wherein obtaining the location of the projected treatment apparatus isocentre at each gantry angle of the set of gantry angles comprises retrieving pre-acquired locations from storage.
14. The method of claim 2, wherein the phantom is a first phantom, and wherein obtaining the location of the projected treatment apparatus isocentre at each gantry rotation angle of the set of gantry rotation angles comprises: obtaining, using the imaging apparatus, a set of 2D images of a second phantom positioned at the treatment apparatus isocentre for each gantry rotation angle of the set of gantry rotation angles, each 2D image of the set of 2D images of the second phantom being taken at a different gantry rotation angle of the set of gantry rotation angles; and determining the location of the projected treatment apparatus isocentre at each gantry rotation angle of the set of gantry rotation angles based on a position of 30 the second phantom in each 2D image of the set of 2D images of the second phantom.
15. The method of claim 14, wherein the treatment apparatus further comprises a treatment beam detector, and the method further comprises, for each gantry rotation angle of the set of gantry rotation angles: positioning the second phantom at a trial position for the treatment apparatus 5 isocentre, directing the treatment beam into the treatment volume, and repositioning the second phantom at the treatment apparatus isocentre based on data provided by the treatment beam detector.
16. The method of any preceding claim, wherein the treatment beam is a beam of MV radiation and/or wherein the treatment apparatus comprises a source of imaging radiation, optionally wherein the imaging radiation is kV radiation.
17. The method of any preceding claim, wherein the treatment apparatus is configured to produce a treatment beam at each of a plurality of treatment beam energies, the method further comprising performing the calibration process at each treatment beam energy, to thereby generate an imaging calibration map which provides the orientation indicator and the location of the treatment apparatus isocentre as a function of gantry angle and as a function of treatment beam energy.
18. The method of any preceding claim, wherein the treatment apparatus comprises a treatment beam source attached to the gantry by an arm that extends laterally away from the gantry.
19. The method of any preceding claim, wherein the imaging apparatus comprises an imaging beam source and an imaging beam detector, and the treatment apparatus comprises a treatment beam source and a treatment beam detector, and wherein one or more of the imaging beam source, the imaging beam detector, the treatment beam source, and the treatment beam detector is attached to the gantry by an arm that extends laterally away from the gantry.
20. The method of any preceding claim, wherein the treatment apparatus and imaging apparatus are positioned such that, at any gantry rotation angle of the set of gantry rotation angles, the treatment beam is directed in a first direction which is different to a second direction in which an imaging beam is directed by the imaging apparatus; optionally wherein the first and second direction are substantially perpendicular to one another.
21. The method of any preceding claim wherein the imaging calibration map is a lookup table which maps gantry rotation angle to the orientation indicator and the location of the treatment apparatus isocentre.
22. A method of generating a 3D image using the imaging calibration map of any preceding claim, the method comprising: receiving an input set of 2D images, each of the input 2D images having been taken at a different gantry rotation angle of a set of gantry rotation angles; identifying, using the imaging calibration map, an orientation indicator for each gantry rotation angle of the input set of 2D images; orienting each image of the set of input 2D images based on the orientation indicators; and generating the 3D image based on the oriented input 2D images.
23. A radiotherapy machine comprising an imaging apparatus mounted to a rotatable gantry, the machine further being coupled to a processor configured to perform the method of any preceding claim.
24. A computer readable storage medium comprising computer-executable instructions which, when executed by a computer, cause the computer to carry out 25 the method of any of claims 1 to 22.
25. A computer readable storage medium storing an imaging calibration map which provides an orientation indicator as a function of gantry rotation angle.
26. The computer readable storage medium of claim 25, wherein the imaging calibration map has been generated according to the method of any of claims 1 to 22.