CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application Serial No. 61/317,590, filed on March 25, 2010, and entitled "System and Method for Frameless Stereotactic Radiosurgery of Arteriovenous Malformations."

BACKGROUND OF THE INVENTION

[0002] The field of the invention is systems and methods for stereotactic surgery, radiosurgery, and radiotherapy. More particularly, the invention relates to systems and methods for frameless stereotactic surgery, radiosurgery, and radiotherapy of arteriovenous malformations.

[0003] Stereotactic radiosurgery ("SRS") is an effective alternative to microsurgical resection or embolization for the treatment of arteriovenous malformations ("AVMs"). Digital subtraction angiography ("DSA"), with its superior vascular resolution and dynamic depiction of contrast flow, is the standard for diagnosis and anatomic characterization of AVMs. In some cases, it is the only means available for distinguishing between the vascular network between arteries and veins, referred to as the "nidus," and surrounding vasculatures. With DSA as the primary imaging modality utilized in the characterization of AVMs, other imaging modalities, including x-ray computed tomography ("CT"), CT angiography ("CTA"), and magnetic resonance angiography ("MRA"), are useful in supplying complementary information for surgical or radiation treatment planning.

[0004] Traditionally, stereotactic photon radiosurgery implies the fixation of a physical stereotactic frame to the patient's skull to serve as a Cartesian reference. Several frames have been developed for this purpose, including the Leksell frame, Brown-Roberts-Wells ("BRW") frame, and Fisher frame, among others. During angiography a localizer box is attached to the frame and two-dimensional images of the patient are obtained, in which the nidus can be readily identified. The two-dimensional projected nidus in the DSA images is transferred to the stereotactic frame's three- dimensional coordinate system. During subsequent CT imaging, a CT localization device is attached to the stereotactic frame, so that the obtained CT images are correlated to the stereotactic frame. An AVM can then be contoured in the CT images and a surgical or radiation treatment procedure planned. During radiation treatment, the frame is attached to a stand such that the center of the AVM is accurately placed in the isocenter of the treatment system. The technique allows for precise radiation treatment; however, using a frame is uncomfortable for the patient and limits the available radiosurgical options.

[0005] An image-guided photon radiosurgery system, such as the CyberKnife® system manufactured by Accuray, Inc. (Sunnyvale, California), is said to be a so-called "frameless" system. With a frameless, image-guided system, the invasive stereotactic frame and attached localizer box are no longer needed either during CT imaging, or radiation treatment of the patient. For brain diseases, the target treatment area can be determined on CT images, which may be fused with images obtained with other imaging modalities. A treatment is planned with the target contoured on multiple image slices. During treatment, the targeting of the radiation relies on comparing two-dimensional orthogonal real-time radiographs with two-dimensional digital reconstructed radiographs ("DRRs"] produced from the obtained CT images, as described, for example, in U.S. Patent No. 7,623,623. This comparison is based on anatomical structures or fiducial markers in both images. By comparing these two-dimensional images, information regarding the translations and rotations necessary to align the two images can be determined. Using such a method, there is no need for system transformations since the settings for the x-ray sources and the detectors are known, and since the identical coordinate space is shared by the two systems. Thus, the purpose of this method is to move the patient so that the on-line radiography images match the produced DRRs.

[0006] For proton and heavy charged particle treatment, it is highly desirable to reduce the number of devices that intersect the treatment beam trajectory to a minimum in order to minimize unwanted attenuation of the treatment beam. Therefore, a frameless setup is used for the treatment of intracranial diseases with proton and heavy charged particle treatment systems. In these frameless setups, for stereotactic treatment of patients, at least three fiducials are implanted into the patient's skull, after which, positioning is guided by digitized orthogonal skull radiographs that depict the fiducials. CT, CTA, and MRI may also be used for imaging the patient.

[0007] The frameless, image-guided approach is comfortable for the patient, and multi-fraction treatment can be routinely performed using this treatment planning approach. However, for patients with an AVM, without the stereotactic frame the standard method using DSA is no longer applicable because the location of the nidus cannot be readily and accurately transformed from the two-dimensional angiograms to the three-dimensional CT image volume. However, the inability to utilize DSA images yields a less-than-optimal result, and in some cases the patients cannot be treated.

[0008] Several efforts have been made to include DSA into a frameless treatment system since the early 1990s. For example, a method for frameless stereotactic radiosurgery for AVMs has been proposed by E. Coste, et al., in "Frameless Method of Stereotaxic Localization with DSA," Radiology, 1993; 189(3):828, in which four bony markers are implanted into the patient's skull and a localizer containing of four fiducial markers in a known geometric arrangement is positioned, without using a frame, near the top of the patient's skull. Two orthogonal images are then obtained for the patient with the localizer. The localizer is then removed and the patient is imaged under standard clinical conditions. There are several shortcomings with the system and method described by Coste. For instance, bony implants are required as an necessity, an additional pair of orthogonal images is obtained when the localizer is positioned near the top of patient's head, which requires additional care. Four fiducial markers, all lying in the same geometrical plane, are utilized, which requires knowledge of the distances from the x-ray sources to the detectors. Since, for modern angiography, these distances are not necessarily fixed, it is cumbersome to determine the distances accurately in a clinical environment for each imaging session.

[0009] Recently proposed approaches rely on three-dimensional digital rotation angiography ("3DRA"), such as the method described by J. Stancanello, et al., in "Development and Validation of a CT-3D Rotational Angiography Registration Method for AVM Radiosurgery," Medical Physics, 2004; 31:1363-1371. In such methods, the 3DRA image is registered with CT images. While there are certain advantages in these approaches, potential limitations may exist. For example, due to the timing characteristics of 3DRA, such as because the time required to complete the rotation needed for image acquisition is not comparable to fixed projections, the 3DRA method cannot be used for distinguishing between feeder and draining vessels. The registration between 3DRA and CT is challenging due to the residual distortion, mechanical instability of the C-arm imaging system, and the low degree of common information between CT and 3DRA. [0010] It would therefore be desirable to provide a system and method for frameless, stereotactic surgical or radiation treatment planning for arteriovenous malformations that can directly utilize highly accurate, two-dimensional images depicting the nidus of an arteriovenous malformation acquired with digital subtraction angiography.

SUMMARY OF THE INVENTION

[0011] The present invention overcomes the aforementioned drawbacks by providing a system and method for frameless stereotactic radiosurgery of arteriovenous malformations. In particular, digital subtraction angiography ("DSA") is utilized to obtain two-dimensional images, often referred to as angiograms or angiographs, that depict the location and extent of a nidus in a patient with an arteriovenous malformation ("AVM"). Without using an invasive stereotactic frame affixed to the patient, the location and extent of the nidus is transformed into three- dimensional coordinates that can be registered with an x-ray computed tomography ("CT") image volume of the patient. The determination of the nidus volume-of-interest ("VOI") from the two-dimensional angiograms is aided by the imaging of a localizer box with the same imaging system in the same source-detector setup as is used for obtaining the angiograms of the patient.

[0012] It is an aspect of the invention to provide a method for producing a stereotactic radiosurgery treatment plan to guide the treatment of an arteriovenous malformation in a patient. In particular, two-dimensional angiograms are obtained of the patient such that a nidus can be identified and a VOI containing it produced. This is achieved without the need for an invasive, uncomfortable stereotactic frame affixed to the patient's skull, as required by previous methods.

[0013] The foregoing and other aspects and advantages of the invention will appear from the following description. In the description, reference is made to the accompanying drawings which form a part hereof, and in which there is shown by way of illustration a preferred embodiment of the invention. Such embodiment does not necessarily represent the full scope of the invention, however, and reference is made therefore to the claims and herein for interpreting the scope of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] FIG. 1 is a schematic diagram of an exemplary stereotactic radiation surgery system including the stereotactic radiosurgery apparatus;

[0015] FIG. 2A is an exemplary localizer box for use when practicing embodiments of the present invention;

[0016] FIG. 2B illustrates an exemplary method for determining a point in the coordinate system associated with the exemplary localizer box of FIG. 2A;

[0017] FIG. 3 is a flowchart setting forth the steps of an exemplary method for producing a stereotactic surgical or radiation treatment plan for the treatment of an arteriovenous malformation in accordance with the present invention;

[0018] FIG. 4 is a flowchart setting forth the steps of an exemplary method for producing a volume-of-interest for a nidus identified in two or more two-dimensional angiograms; and

[0019] FIG. 5 is a flowchart setting forth the steps of an exemplary method for determining a position of a nidus in reliance on temporary fiducials that may shift between angiographic imaging and CT volume imaging of a patient.

DETAILED DESCRIPTION OF THE INVENTION

[0020] The following definitions and terms are provided to clarify the description of the present invention and to guide those of ordinary skill in the art in the practice of the present invention. As used herein, the term "radiosurgery" refers to the treatment of a patient through the application of radiation to a target region. As used herein, the term "radiation" refers to electrons; x-rays and other photons; protons; and heavy charged particles, such as elemental ions.

[0021] A block diagram of an exemplary frameless stereotactic radiosurgery system ("SRS") 100 is illustrated in FIG. 1. The frameless SRS 100 includes a stereotactic radiosurgery apparatus 102 that is controlled by controller 104 such that a radiation dose is delivered to a patient with high accuracy. Exemplary stereotactic radiosurgery apparatus include CyberKnife® systems manufactured by Accuray Inc. (Sunnvale, California); Gamma Knife® systems manufactured by Elekta AB (Stockholm, Sweden); and image-guided radiotherapy ("IGRT") systems such as the Novalis Tx™ stereotactic linear accelerator ("LINAC") systems manufactured by Varian Medical Systems, Inc. (Palo Alto, California); and proton and heavy charged particle treatment systems. The controller 104 receives a treatment plan from a treatment planner 106. An operator, such as a physician or medical physicist, interacts with the treatment planner 106 to produce an appropriate treatment plan that will effectively deliver radiation to the patient. The treatment planner 106 receives image data from an image processor 108, which receives images from one or more imaging systems. For example, the image processor may receive images from an x-ray imaging system 110 and an x-ray computed tomography ("CT"] imaging system 112.

[0022] A general procedure for producing a stereotactic surgical or radiation treatment plan for the treatment of an arteriovenous malformation is as follows. First, a patient has fiducial markers ("fiducials"), such as temporary fiducials or implanted fiducials, affixed to their skull. The patient is then positioned within an x-ray imaging system 110, and a series of radiographs of the patient are acquired as a contrast agent is administered to the patient and allowed to pass through the patient's vasculature. Included in this series of radiographs are acquired before the administration of the contrast agent. These so-called "mask images" are used to produce a series of patient angiograms using digital subtraction angiography ("DSA"] techniques. From the series of angiograms, a pair of DSA images that best depict an AVM is selected. After the patient is removed from the x-ray imaging system, and without adjusting the source- detector settings, a localizer box is imaged with the x-ray imaging system llOsuch that two orthogonal radiographs of the localizer box are produced. CT images of the patient are additionally acquired with an x-ray CT imaging system 112, either before or after acquisition of the angiograph. Other images of the patient can subsequently be acquired with different imaging modalities, such as magnetic resonance imaging ("MRI") and these other images fused with the image volume during treatment planning.

[0023] The pair of localizer radiographs; the pair of selected DSA images; and CT image are provided to the image processor 108, which produces a volume-of-interest ("VOI") containing a nidus of an AVM that is to be treated. A position of the nidus is determined on the pair of DSA images and, using the localizer radiographs, the nidus position is transformed into a three-dimensional coordinate space related to the localizer system. The fiducial positions are identified on the patient radiographs, such as the mask images, and in the CT image volume and a transformation between the localizer system coordinate space and the CT image volume coordinate space is calculated. A VOI containing the nidus in the localizer system is then transformed into the CT image volume coordinate space using the calculated transformation. In this manner, the VOI can be directly coregistered with the image volume. The registered VOI and the image volume are then provided to the treatment planning system 106 and a radiation treatment plan for treating the AVM is produced and provided to the controller 104 such that highly accurate radiation is delivered to the patient via the stereotactic radiosurgery apparatus 102.

[0024] An exemplary localizer box 200 for use in accordance with the present invention is illustrated in FIG. 2A. The localizer box 200 is formed of radiotransparent material, such as acrylic, and includes six faces: a left face 202, a right face 204, an anterior face 206, a posterior face 208, a superior face 210, and an inferior face 212. Disposed on four of the faces, such as the left 202, right 204, anterior 206, and posterior 208 faces, is a set of radiopaque markers 214. For example, sixteen markers 214 in total are utilized with four disposed on each of the left 202, right 204, anterior 206, and posterior 208 faces of the localizer box. The radiopaque markers 214 are positioned in a known arrangement and at known locations on the faces of the localizer box 200. While a localizer box 200 having sixteen markers 214 disposed on four different faces of the box 200 is disclosed and illustrated, the localizer box 300 may be constructed to contain a minimum of six total markers 314, in which at least three markers 314 are disposed on each of two different faces of the localizer box 300. It will be appreciated that other variations are possible as well.

[0025] As will be described in detail, the localizer radiograph images are utilized to transform the two-dimensional locations identifying a nidus of the AVM into a three- dimensional coordinate system related to the localizer system that can be further transformed to a three-dimensional image volume of the patient. In particular, these localizer radiograph images provide the ability to perform highly accurate, frameless stereotactic surgery, or radiation treatment, planning using two-dimensional angiograms obtained with DSA.

[0026] By way of example, an x-ray imaging system, such as an x-ray imaging system tailored for neuroangiography, is used to acquire distortion-free radiograph images of the localizer box and of the patient. The field size of the radiographs may be 32 centimeters in the posterior-anterior ("P-A") projection and may be 25 centimeters in the left-right ("L-R") projection. Several angiogram series are taken under standard conditions with positions of x-ray sources, detectors, and couch adjusted as clinically needed. Bay way of example, the source-to-image receptor distance ["SID"] may be around 100 centimeters and the distance from the source to the isocenter may be around 75 centimeters. The source-detector angles ("gantry angles") may vary within 0±15 degrees for the P-A projection and 90±15 degrees for the L-R projection, which gives flexibility for neurosurgeons. By way of example, radiographs contain pixels with a size of around 0.3 millimeters.

[0027] Once a series of angiograms is satisfactorily acquired, the patient leaves the couch of the x-ray imaging system and the localizer box is placed on the couch to be imaged. In the alternative, these radiographs of the localizer box may be obtained before the patient angiograms. Between the angiogram and the radiograph, the x-ray sources and detectors must remain fixed, but the couch position can be adjusted. The localizer box may be constructed from one-eighth inch thick Lucite plates forming a cube of, for example, 18x18x18 centimeters, as illustrated in FIG. 2 A. In the localizer box 200 illustrated in FIG. 2A, sixteen small radiopaque markers 214 are embedded in the plates. For example, the markers 214 are positioned at the corners of two high- precision 10x10 centimeter squares on the anterior and posterior sides, and two 8x8 centimeter squares on the left and right side. A coordinate system is defined with the origin at the center of the box and its three axes perpendicular to the sides. During radiography it is not necessary that the faces of the localizer box 200 be orthogonal or parallel to the detectors; however, the eight markers 214 in the anterior and posterior faces, and the eight markers 214 in the left and right faces, should all appear in the P-A and L-R projections. This can be done by adjusting the couch position, and is the reason for the optional ±15 degree restriction on the variation of the gantry angles.

[0028] DSAs are obtained from the selected angiogram series. A pair of DSAs containing one P-A image and L-R image, where the AVM nidus can be seen most clearly, is chosen. Targets are identified in both the P-A and L-R views. Because the source and detector positions remain unchanged between angiography and radiography, the spatial information separately obtained from the different images can be used as if it were from one projection with the localizer and the patient physically overlapping. More specifically, the spatial information for the point targets can be taken from the DSA images. For the fiducials, the spatial information is taken from the angiograms before subtraction, and for the localizer markers, the spatial information is taken from the radiographs. All these images can be used in combination as if they were taken simultaneously.

[0029] Referring now to FIG. 2B, an exemplary method for determining the position of a point in a coordinate system associated with a localizer box, such as the localizer box illustrated in FIG. 2A. A point target T may be imaged inside of, or in the vicinity of the localizer box 200. The projection of the target on the P-A image 216 is identified in the DSA image 216 as point T' . The four markers on the posterior side of the localizer box 200 denoted as C , D , E, and F , and their positions in the localizer box 200 are known precisely. The projections of these markers on the P-A image 216 are denoted as C', D' , E' , and F' , The two-dimensional positions on the P-A image 216 for these four points and for point T' can be digitized. With this information, the intersection of the source-target ray with the posterior side of the localizer, point P, can be identified through projective geometry theory, which is described in detail, for example, by R. L. Siddon and N. H. Barth in "Stereotaxic Localization of Intracranial Targets," Int. J. Radiation Oncology, Biol, Phys., 1987; 13:1241-1246. Similarly, with the projection of the four anterior side markers on the P-A image 216, the intersection of the source-target ray with the anterior side, point A , can be identified. Line PA , the source-target ray, is, thereby determined in the localizer coordinate system. For this calculation no other parameters, including the source-detector positions are required; instead, they can be calculated where desired. For example, the source position in the

P-A projection can be determined by the intersection of the line, PA , and other source- target rays in the P-A projection.

[0030] The same method is applied to the L-R view. The intersection of the source-target ray with the left side, point L, and the right side, point R , can be identified. The line LR , thus, represents the source-target ray. The target, T , is the intersection of the lines PA and LR , determined as {x _T ,y _T ,z _T } in the three- dimensional localizer coordinate system. Even with the smallest digitizing errors, there will be instances where the two lines may not be coplanar and intersect exactly at one point; therefore, the target point, T , may be defined as the mid-point of the shortest segment connecting the two lines.

[0031] Following the same procedures, the fiducials implanted into the skull are also identified in the localizer coordinate system as:

[0032] 2, = { , ^ }, _{=1 N} (1);

[0033] where x _j ,y _j ,z _i are the three coordinates of the i' ^h fiducial, and N is the number of fiducials. It is important to notice that the couch position is not unique; the position can be readjusted as long as all the localizer markers still appear in the radiographs, in which case the point target and fiducial coordinates expressed in the localizer system will vary accordingly. However, the relative positions of all these points ( T and Q _{j t} i = 1, ... , N } remain unchanged.

[0034] After the target point has been identified in the localizer coordinate system, the next step is to establish a coordinate transformation from the localizer reference system to the CT, or image volume, coordinate system. The CT image volume can be taken either before or after the angiograms and radiographs. The fiducial positions in the CT coordinate system are identified as:

[0035] R, = {t„ _M „v,.} _{(=i N} (2);

[0036] where t _j , u _j ,v _j are the three coordinates of the i' ^h fiducial. With the coordinates identified in both localizer and CT coordinate systems, a transformation (Q— > R) such as a translation plus a complex rotation, between the two systems can be established. Via the transformation, Q R , any point target T determined in the localizer coordinate system can be transferred to the CT coordinate system.

[0037] Due to measurement errors, it is unlikely that there will be a perfect, error-free Q— » R transformation. The method employed here is based on the

Wolfgang Kabsch algorithm for calculating the optimal rotation matrix that minimizes the root mean squared deviation between the two sets of points.

[0038] For a small sized nidus, a point target is considered to be in the center of the nidus identified in both DSA views, and then determined in the localizer system. Subsequently, the point can be transformed directly to the CT coordinate system by the Q— R transformation and treated with a small margin. Generally, however, the nidus is more accurately characterized as a target volume. The method for point targets is the basis for solving this volumetric problem, but cannot be extended directly. This is because of the intrinsic uncertainties introduced in the process due to the two- dimensional nature of the DSA images. The strength of this method is based on the identification of identical points in each view, but individual points can seldom be identified in three-dimensional volumes. Using an epipolarity principle-based method, such as the one described by R. Foroni, et al., in "Shape Recovery and Volume Calculation from Biplane Angiography in the Stereotactic Radiosurgical Treatment of Arteriovenous Malformations," Int. J. Radiat. Oncol. Biol. Phys., 1996; 35:565-577, an encapsulating target contour is modeled by triangulation of a stack of almost parallel ellipses and superimposed on the corresponding CT image slice-by-slice, resulting a quadrangle-like area. This area covers but also overestimates the target due to limits of the two-projection, therefore, for a compromise a pseudo-ellipse that fits into the area is considered as the target contour.

[0039] Referring now to FIG. 3, a flowchart setting forth the steps of an exemplary method for producing a stereotactic surgical or radiation treatment plan for the treatment of an arteriovenous malformation is illustrated. A series of radiograph images of a patient are obtained, as indicated at step 302. Included in this series of radiograph images are mask images acquired before the administration of a contrast agent to the patient, and radiographs of the patient acquired after the administration of the contrast agent and as it passes through the patient's vasculature. Subsequently, DSA is performed such that angiograms of the patient are produced, as indicated at step 304. A pair of patient angiograms that best depict an AVM in the patient are then selected for subsequent processing, as indicated at step 306. For example, two orthogonal angiograms are selected, one sagittal angiogram and one coronal angiogram.

[0040] A pair of orthogonal radiograph images of a localizer box are also obtained, as indicated at step 308. For example, two orthogonal radiograph images are obtained, one sagittal image and one coronal image. The localizer box is not attached to a frame that is affixed to the patient, as is done with previous systems and methods. Instead, the localizer box is imaged independent from the patient, using the same imaging system and source-detector setup as used to acquire angiograms of the patient. While the source-detector setup remains fixed, the couch position may be adjusted between imaging the patient and the localizer box. These orthogonal radiograph images may be obtained in parallel with the acquired patient images described below.

[0041] As noted above, the angiograms are obtained with the same x-ray imaging system that is used to obtain the localizer radiographs. Moreover, the same imaging setup is utilized when obtaining both the localizer radiographs and the patient angiograms. That is, the setup for the x-ray sources and detectors remains unchanged between the localizer radiograph acquisition and patient angiogram acquisition.

[0042] Prior to obtaining the angiograms, a series of fiducials are fixed to the patient. These fiducials will remain on the patient during the acquisition of the CT image volume as well. In general, a minimum of three such fiducials are fixed to the patient; however, four or more such markers can readily be employed for redundancy. Exemplary fiducials include bony implants, such as gold or stainless steel "BBs"; temporary fiducials placed on the patient's skin, so as to maintain comfort for the patient as discussed in detail below.

[0043] Before or after the angiograms are obtained, the patient is imaged with an x-ray CT imaging system in order to acquire a plurality of CT images corresponding to a three-dimensional image volume, as indicated at step 310.

[0044] The localizer radiographs; a pair of patient radiographs, such as the mask images; selected pair of patient angiograms; and CT image volume described above are all stored and utilized for subsequent processing for the generation of a stereotactic surgical or radiation treatment guidance plan. Next, a volume-of-interest ("VOI") containing the identified nidus is produced, as indicated at step 312 and described, for example, below in detail.

[0045] After the nidus VOI has been produced, it is transformed into the same coordinate space as the CT image volume, as indicated at step 314. This transformation of the nidus VOI into the CT image volume coordinate space allows the co-registration of the nidus VOI with the anatomical CT image volume. Knowing the fiducial positions in both of the localizer, or x-ray imaging system, and the CT system coordinate space, a transformation between the two coordinate spaces can be established. The sets of at least three non-collinear points that represent the coordinates of the same number of fiducials in each system coordinate space are given by:

[0046] R = {R _i } = {t.,u _i ,v _i } (3);

[0047] in the CT system coordinate space, and

[0048] 0 = te} = { _/ > *<} (4);

[0049] in the localizer system coordinate space, where l≤i≤N and N > 3 is the number of such points in each coordinate space. The center-of-mass, A _R , for the set R is given as: 0051] and the center-of-mass, , for the set Q is given as:

[0053] The transformation Q— > R may therefore be established using the following transformation operator, T :

[0054] T (q) = V (q - A _Q ) + A _R (7);

[0055] where q = (x, y, z) is an arbitrary point in the localizer coordinate space that is to be transformed into the CT image volume coordinate space, and Ψ is a 3 x 3 rotation matrix determined by the Wolfgang Kabsch algorithm, which calculates the optimal rotation minimizing the root-mean-squared deviation between the two sets of points, R and Q . The Wolfgang Kabsch algorithm is described, for example, by W.

Kabsch in "A Solution of the Best Rotation to Relate Two Sets of Vectors," Acta Crystallographica, 1976; 32:922, and by W. Kabsch in "A Discussion of the Solution for the Best Rotation to Relate Two Sets of Vectors," Acta Crystallographica, 1978; 34:827- 828, which are herein incorporated by reference in their entirety. It will be appreciated that other methods are similarly suitable for determining the transformation, Q— > R .

[0056] With the establishment of the transformation between the localizer system coordinate space and the CT system coordinate space, the nidus VOI can now be transferred and contoured to the CT image volume with a standard method. The transformed nidus VOI and CT image volume can then be fused with other modalities and the final treatment volume can be determined, as indicated at step 316. This treatment volume can be exported in a DICOM-format file to the treatment planner so that a radiation treatment plan can be produced to guide the treatment of the AVM.

[0057] Referring now to FIG. 4, a flowchart setting forth the steps of an exemplary method for producing a VOI for an identified nidus is illustrated. As indicated at step 402, this process first includes identifying the localizer markers in the localizer radiographs, the fiducials in the patient radiographs, the fiducials in the CT image volume, and the position of one or more nidi in the patient angiograms. These are identified by a user and recorded, for example, via mouse clicks. Previously recorded features are optionally removed from the images, or new ones are added. The data entered or modified by the user is optionally saved to a file that can be reloaded for subsequent treatments of fractions delivered to the patient. Next, the spatial information related to the fiducials, markers, and nidi are combined to allow the transformation of the nidi positions into a coordinate space associated with the localizer system, as indicated at step 404.

[0058] As noted above, from the patient angiograms, a position of a nidus of an AVM in the patient is determined. In the alternative, a plurality of such positions are determined, the positions identifying a boundary of the nidus. Using information regarding the location of the identified localizer markers, the nidus position is transformed into the three-dimensional localizer system coordinate space, as indicated at step 406. An exemplary method for transforming an arbitrary point from its two orthogonal angiogram projections to these three-dimensional coordinates, as discussed above, is described in detail, for example, by R. L. Siddon and N. H. Barth in "Stereotaxic Localization of Intracranial Targets," Int. J. Radiation Oncology, Biol., Phys., 1987; 13:1241-1246. Optionally, the relative positions of the x-ray sources, patient, and the detectors can also be determined. While this information is not required, it can be beneficial for verification and adjustment of the fiducial positions if needed or desired. Once the three-dimensional nidus position has been determined, an envelope is formed around the nidus position to form the desired nidus VOI, as indicated at step 408.

[0059] For proton facilities, where implanted fiducials are traditionally routinely used, it is straightforward to implement the provided method. In addition, the fiducials are used for patient setup during treatment. For an image-guided treatment facility, implantation of fiducials may not be a necessary procedure; thus, for image-guided treatment systems that are based on anatomical structures, temporary fiducials are preferred. In some instances, it is contemplated that temporary fiducials may move between the CT and angiogram acquisitions, which may cause errors in transferring the target position from the localizer to the CT coordinate system. Therefore, in such instances, techniques different from those described above with respect to implanted fiducials may need to used.

[0060] Referring now to FIG. 5, it is possible to transfer the target nidus information to the CT image volume coordinate system directly by generating CT DRR pairs with all possible source-patient-detector positions and selecting the one that best matches the pair of angiograms based on the bony structure. Once this is accomplished, the 3D nidus can then be reconstructed directly in the CT system based on the target's angiogram projections, similar to the procedures discussed above. The difficulty with this approach, however, is that during angiography, the images are taken as clinically needed without the knowledge of the patient position relative to the angiogram. This is, neither the angiogram's source-detector distances and angles, nor their relative positions with respect to the patient are precisely known. Therefore, the required number of generated DRRs can be significantly large, thereby rendering such an approach impracticable. To reduce the complexity of this problem, using four temporal fiducials on the skull and the provided method, the source and detector positions can be calculated in the localizer system based on the projective geometry, and subsequently transferred to the CT system; thus, a pair of DRRs can be readily generated. Due to the possible movement of non-implanted fiducials by a few millimeters, and the digitizing errors, the pair of DRRs may not match the angiogram precisely. However, the DRRs can be matched by adjusting the source and detector positions within a small range, such as within a few degrees or several millimeters.

[0061] As indicated at step 502, this process first includes identifying the localizer markers in the localizer radiographs, the fiducials in the patient radiographs, the fiducials in the CT image volume, and the position of one or more nidi in the patient angiograms. These are identified by a user and recorded, for example, via mouse clicks. Then, the source and detector positions are identified in the localizer coordinate system, as indicated at step 504. By way of example, projective geometry may be used to determine the source and detector positions. These source and detector positions are then transferred into the CT image volume coordinate space, as indicated at step 506. These positions may be transferred from the localizer coordinate space to the CT image volume coordinate space using, for example, one of the methods described above. After the source and detector positions have been transferred into the CT image volume coordinate space, a plurality of different DRRs are generated in the determined source and detector positions and in the nearby areas with small deviations, as indicated at step 508. The generated DRRs are then co-registered with the patient angiograms, as indicated at step 510, and compared to identify the pair of DRRs that best match the patient angiograms, as indicated at step 512. An exemplary method for identifying the pair of DRRs that best match the patient angiograms is based upon the method described in U.S. Patent No. 7,522,779, modified to be applicable for identifying DRRs instead of beam adjustments. For example, bony structures in the DRRs and the patient angiograms may be compared. The nidus position can then be identified on the CT image volume coordinates directly, as indicated at step 514. The temporary fiducials can be removed after CT and angiography because they are not needed for delivering the subsequent image-guided radiosurgery treatment.

[0062] While the foregoing description provides an exemplary implementation of the present invention, it should be apparent to those skilled in the art that the present invention is readily applicable to other clinical applications. For example, a medical treatment plan, such as a surgical plan, can be produced and utilized to aid surgical procedures tasked with the resection or embolization of an AVM. Moreover, the location of the AVM need not be limited to intracranial AVMs; rather, the method of the present invention is readily applicable to guide the treatment of AVMs in other anatomical locations, such as in or around the spinal cord ("spinal AVM") and in the dura mater between a meningeal artery and a meningeal vein or dural venous sinus ("dural arteriovenous fistula").

[0063] The present invention has been described in terms of one or more preferred embodiments, and it should be appreciated that many equivalents, alternatives, variations, and modifications, aside from those expressly stated, are possible and within the scope of the invention.