WO2006130659A2

WO2006130659A2 - Methods, program product and system for enhanced image guided stereotactic radiotherapy

Info

Publication number: WO2006130659A2
Application number: PCT/US2006/021070
Authority: WO
Inventors: Almon S. Shiu; Congjun Wang; Jim Mingfwu Lii; Eric L. Chang
Original assignee: Board Of Regents, The University Of Texas System
Priority date: 2005-05-31
Filing date: 2006-05-31
Publication date: 2006-12-07
Also published as: WO2006130659A3

Abstract

Methods, program product (21), and systems (20) to enhance accuracy of radiotherapy treatment delivery are provided. An embodiment of method, for example, can include registering a three-dimensional (3D) image (53, 55) of a frame (42) fixedly connected to a rigid treatment couch (41) responsive to maximization of mutual information (MMI). The mutual information (MI) is a measure of statistical dependence between image intensities of corresponding voxels of two sets of images. The embodiment of the method also includes registering a 3D image (52, 54) of a treatment plan of anatomy after registration of the frame (42), and cropping the 3D images (52, 54), (53, 55) of each of the frame (42) and the treatment plan to first and a second preselected sizes. One of either the first or second size is larger than the other. The embodiment of the method further includes fitting by aligning the smaller image to the larger image and removing systematic error.

Description

METHODS, PROGRAM PRODUCT AND SYSTEM FOR ENHANCED IMAGE GUIDED STEREOTACTIC RADIOTHERAPY

BACKGROUND OF THE INVENTION 1. Related Applications

This application is related to and claims priority to and the benefit of U.S. Provisional Application No. 60/685,894, filed on May 31, 2005, incorporated herein by reference in its entirety.

2. Field of the Invention [0001] The present invention relates to the radiotherapy industry and, more particularly, to methods, program product, and systems using image-guided stereotactic radiotherapy.

3. Description of the Related Art

[0002] The spine of a human body is the most common site for bone metastasis. These metastasis, can progress or recur due to insufficient dose used during radiation therapy. Such insufficient doses, for example, can be due to constraints imposed by spinal cord tolerance or other factors, and re-irradiation is usually not feasible. Due to proximity of the spinal cord and other critical structures, effective delivery of high doses per fraction to paraspinal tumors requires highly conformal dose distributions, precise and accurate patient setup, setup verification, and monitoring patient intrafractional external movement.

[0003] Intracranial stereotactic radiotherapy has widely been used for patient with brain metastases to enhance survival and tumor control and also permitting re- irradiation. The implementation of extracranial stereotactic radiotherapy, however, can be difficult due to immobilization of these extracranial regions of a patient and verification of patient treatment position, for example.

[0004] Stereotactic extracranial radiotherapy or radiosurgery, for example, can play a complementary role to surgery in the management of spinal tumors and other extracranial tumors located in the human body. The success of this procedure, however, can largely depend on the accurate delivery of a highly conformal dose to a planning target volume and on sparing the spinal cord and other surrounding critical structures from radiation damage. An example of a previously reported spinal radiosurgery technique involved the use of an invasive skeletal-fixation procedure to immobilize the vertebral bodies, see Hamilton A.J., LuLu Ba, "A Prototype Device For Linear Accelerator-Based Extracranial Radiosurgery, " Acta Neurochir 1995;63:40-43. In this example, small clamps were applied to the spinous processes to yield a placement error of 2 mm in the axial plan and <4 mm longitudinally. Not only was the procedure invasive, but also, because the patient was in the prone position, the setup accuracy was compromised by the vertebral body motion associated with patient breathing. [0005] Several other spinal procedures reportedly avoid the invasive approach of spinal irradiation. The CyberKnife™, commercially available from Accuray, Inc., of Sunnyvale, CA, uses a robotic arm to deliver radiation in a wide range of beam orientations, except in the posterior region of the patient where two amorphous silicon (aSi) flat panel digital detectors are located. The CyberKnife™ relies on co- registration of digitally reconstructed radiographs (DRRs) generated from computed tomography (CT) images and x-ray projections captured before each node, e.g., a tumor, is treated with irradiation. Matching mutual information (MI) from both data sets assumes that the treatment-planning geometry is reproduced at the time of each node treatment. Fiducial markers were implanted in the patient so that the treatment target could be determined based on the markers in the 2D x-ray images. Co- registration of DRRs with the 2D x-ray images to achieve the suhmillimeter setup accuracy, however, does not fully account for all types of motion because rotational errors are difficult to detect, hi a study by Yin et al. (see, Yin F., Ryu S., Ajlouni M., et al., "A technique of intensity-modulated radiosurgery (IMRS) for spinal tumors, " Med. Phys. 2002; 29:2815-2822.), who used a spinal radiosurgery technique that was a simplified version of the CyberKnife approach, the isocenter setup accuracy was within 2 mm.

[0006] A stereotactic paraspinal treatment that relies entirely on daily CT guidance was reported in Yenice K.M., Lovelock D.M., et al., "CT image-guided intensity-modulated therapy for paraspinal tumors using stereotactic immobilization, "

Int. J. Radiat Oncol. Biol. Phys. 2003; 55:583-593. This technique acquires planning CT images of 3 mm slice thickness and pretreatment CT images of 2 mm slice thickness. The position of the target and other anatomic landmarks are manually identified on the closely matched CT slices from both scans, and then their coordinates are computed in their frame's independent coordinate system. The new coordinates of all landmarks are compared with those determined from the planning CT study, and a daily setup deviation is assessed. According to this technique, a new isocenter position is determined on the basis of the average differences for all the landmarks used in the evaluation. The accuracy of reproducing the treatment- planning geometry heavily depends on the accuracy of manually identifying the same anatomic landmarks and the position of the target. The patient, positioned inside a stereotactic body frame (SBF), is then transferred to the treatment table using a rail system. Rotation around yaw can occur when the SBF is transferred from the CT couch to the treatment table. The pitch may also vary owing to the different structures of the CT couch and the treatment table. Neither yaw nor pitch was addressed. Patient position is verified by comparing DRRs with the portal images before the treatment is delivered.

[0007] More recently, to initially address some of these concerns, assignee of the present application previously developed a computer tomography (CT) imaged- guided stereotactic radiotherapy for the treatment of spinal and paraspinal lesions. This technique uses an integrated CT-on-rails and linear accelerator (LINAC) treatment delivery and image verification system. The patient is lain on a couch or table, which is common to the CT-on-rails unit and the LDSfAC, and the couch can be rotated 180-degree between the CT scanner and the LINAC. Two different commercial software packages were used at the same time for image registration and DRR comparison, namely Pinnacle (Pinnacle, Elekta, Norcross, GA) and RT2 (Tyco/Radionics, Burlington, MA). Both software packages use a tale coordinate system, and a spreadsheet was developed by assignee for final data conversion. This technique demonstrated patient setup error within 1 mm. Three steps were required to position the patient's treatment isocenter exactly as planned: (1) target localization, (2) patient positioning, and (3) target position verification. This technique using these steps reduced some of the systematic errors caused by the uncertainty of the couch coordinate system. Systematic errors, however, remain, and this treatment process is still relatively slow.

[0008] Therefore, there is still a need to further enhance accuracy of extracranial stereotactic radiotherapy and to reduce the time required to plan radiotherapy treatments.

SUMMARY OF THE INVENTION

[0009] In view of the foregoing, embodiments of the present invention beneficially provide computer assist image-guided stereotactic patient alignment program product or software and methods in conjunction with a stereotactic body immobilization system and an image guided robotic patient positioning system to facilitate the accurate radiation treatment of patients having extracranial lesions such as spinal and paraspinal lesions. Embodiments of methods, program product, and systems of the present invention also beneficially provide enhanced setup speed for patient treatment plans. Embodiments of a computer assist image-guided stereotactic radiotherapy method and program product beneficially, for example, can also be extended to a non-rigid target with four-dimensional (4-D) computer tomography (CT) images.

[0010] For example, embodiments of a method and a program product according to the present invention can have three main functionalities: image registration, digital reconstructed radiograph (DRR) verification, and portal image verification. A method can include registering an image, digitally reconstructing a radiograph, and verifying the image such as by the use of a portal image.

[0011] Also, for example, an embodiment of a method of the present invention includes registering an image of a frame connected to a treatment couch responsive to mutual information (MI). The MI can be a measure of statistical dependence between image intensities of corresponding voxels of two sets of images. The method also includes registering an image of a treatment plan of anatomy after registration of the frame, cropping the images of each the frame and the treatment plan of anatomy to first and second preselected sizes, aligning the smaller image to the larger image with an image frame, and removing systematic error from the treatment plan. [0012] Another embodiment of a method to enhance accuracy and speed of radiotherapy treatment delivery according to the present invention includes registering a three-dimensional (3D) image of a frame fixedly connected to a rigid treatment couch responsive to maximization of mutual information (MMI). Mutual information (MI) can be a measure of statistical dependence between image intensities of corresponding voxels of two sets of images. The method also includes registering a 3D image of a treatment plan of anatomy after registration of the frame and cropping the 3D images of each the frame and the treatment plan of anatomy to first and second preselected sizes. One of either the first size or the second size can be larger than the other. The method further includes fittingly aligning the smaller image to the larger image within an image frame and removing systematic error from the treatment plan.

[0013] An embodiment of a program product stored in a tangible computer medium to be operable on a computer according to the present invention operates to perform the step of registering a 3D image of the frame of the stereotactic treatment system responsive to maximization of mutual information (MMI). The mutual information (MI) is a measure of statistical dependence between image intensities of corresponding voxels of two sets of images being displayed on the display. The product also operates to perform the further steps of registering a 3D image of a treatment plan of anatomy after registration of the frame, cropping the 3D images of each the frame and the treatment plan of anatomy to first and second preselected sizes, fittingly aligning the smaller image to the larger image, and removing systematic error from the treatment plan.

[0014] Additionally, for example, the present invention provides an embodiment of a radiotherapy system. The system includes a stereotactic frame system to substantially immobilize a patient and having a rigid treatment couch and a frame fixedly connected to the rigid treatment couch, an image scanner to acquire images of a patient's anatomy, a radiation source to deliver radiation to a preselected portion of the patient's anatomy, a display to display images of the patient's anatomy, and a controller in communication with the image scanner, the display, and the radiation source to control delivery of the radiation to the preselected portion of the patient's anatomy responsive to a predetermined treatment plan and to display images of the patient's anatomy on the display. The controller also includes memory having program product stored therein. The program product is operable to perform the steps of registering a 3D image of the frame of the stereotactic treatment system responsive to MMI. The MI can be a measure of statistical dependence between image intensities of corresponding voxels of two sets of images being displayed on the display. The steps further include registering a 3D image of a treatment plan of anatomy after registration of the frame, cropping the 3D images of each the frame and the treatment plan of anatomy to first and second preselected sizes, fittingly aligning the smaller image to the larger image, and removing systematic error from the treatment plan.

[0015] Beneficially, embodiments of methods, program product, and systems of the present invention also facilitate full automation of target localization and patient position verification. Further, for image registration, because rotational shifts tend to be lost in two-dimensional (2-D) image registration, 3D images registration can be used to achieve a desired accuracy in image- guided stereotactic radiotherapy, and MMI beneficially can be used as the criterion when performing image registration to further enhance such accuracy. Embodiments of methods, program product, and systems can be used to significantly increase the accuracy, e.g., less than 1 mm, in radiotherapy delivery, especially in extracranial applications.

BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In view of the features and benefits described above, some of the features and benefits of the present invention having been stated, others will become apparent as the description proceeds when taken in conjunction with the accompanying drawings, in which:

[0017] FIG. 1 is an environmental view of a stereotactic radiotherapy system according to an embodiment of the present invention;

[0018] FIG. 2 is a perspective view of a table having rails thereon of a stereotactic radiotherapy system and associated patient screen images according to an embodiment of the present invention; [0019] FIG. 3 is a front plan view of a split screen and corrections of a stereotactic radiotherapy system according to an embodiment of the present invention;

[0020] FIG. 4 is a front plant view of a graphical user interface of a display of a stereotactic radiotherapy system and of a program product according to an embodiment of the present invention;

[0021] FIG. 5 is a front plan view of a graphical user interface of a display of a stereotactic radiotherapy system and of a program product according to an embodiment of the present invention;

[0022] FIGS. 6A-6B are front plan views of screens of a stereotactic radiotherapy system according to an embodiment of the present invention;

[0023] FIGS. 7A-7D are front plan views of screens of a stereotactic radiotherapy system according to an embodiment of the present invention;

[0024] FIG. 8 is a graph time versus anatomy thickness of a stereotactic radiotherapy system, program product, and methods of embodiments of the present invention;

[0025] FIGS. 9A-9B are front plan views of screens of a stereotactic radiotherapy system according to an embodiment of the present invention;

[0026] FIGS. 10A- 1OB are front plan views of screens of a stereotactic radiotherapy system according to an embodiment of the present invention; [0027] FIG. 11 is a table of tissues, attenuation coefficients, and CT numbers as recognized by methods and program product according to embodiments of the present invention;

[0028] FIG. 12 is a table of cases examples of translations, degree of rotations, and time of a stereotactic radiotherapy system, methods, and program product according to embodiments of the present invention;

[0029] FIG. 13 is a flow chart of a method and program product according to an embodiment of the present invention;

[0030] FIG. 14 is a flow chart of a method and program product according to an embodiment of the present invention; [0031] FIG. 15 is a flow chart of a method and. program product according to an embodiment of the present invention;

[0032] FIG. 16 is a flow chart of a method and program product according to an embodiment of the present invention; and [0033] FIG. 17 is a flow chart of a method and program product according to an embodiment of the present invention.

DETAILED DESCRIPTION

[0034] The present invention now will be described more fully hereinafter with reference to the accompanying drawings in which embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. In the description, many well-known functions and structures as understood by those skilled in the art are not described in detail for brevity and conciseness. Like numbers refer to like elements throughout.

[0035] FIGS. 1-17 illustrate embodiments of a system 20, methods and program product 21 to facilitate enhanced image-guided stereotactic body radiotherapy (SBRT) with high precision. Embodiments also facilitate full automation of target localization and patient position verification. For example, embodiments of a program product according to the present invention have three main functionalities: image registration, digital reconstructed radiograph (DRR) verification, and portal image verification in a near simultaneous computed tomography (CT) image-guided stereotactic radiotherapy system. As will be understood by those skilled in the art, trilinear interpolation can be used throughout in reconstructing images with subvoxel shifts and in rotation. For example, for image registration, because rotational shifts tend to be lost in 2D image registration. In these embodiments, 3D images registration can be used to achieve a desired accuracy in image-guided stereotactic radiotherapy. As will be understood by those skilled in the art, maximization of mutual information (MMI) can be used as the criterion when performing image registration. Mutual information (MI) is a measure of statistical dependence between the image intensities of corresponding voxels of two sets of images, either, the same modality or different modalities. When the two sets of images are geometrically aligned, for example, the MI can be assumed to be maximal. In embodiments of a program product 21 of the present invention, the daily shift of the localization box images with respect of that of the reference images can be identified first. This shift allows removal of the systematic errors caused by the uncertainty of a coordinate system associated with a couch 4D within a radiotherapy treatment system 20. Then, the anatomy registrations are performed. All the rotational shifts, i.e., roll, yaw, and pitch, are considered as well as all the translational shifts. The final translational and rotational shifts can be input into an operational computer 30 associated with a system 40, such as a HexaPOD robotic couch system, for auto setup.

[0036] As illustrated in FIG. 16, in DRR generation and comparison of embodiments of methods and program product 21, for example, DRRs are generated using ray- tracing (block 130) as understood by those skilled in the art. A ray that passes through a 3D CT volume of a patient P, for example, is divided into many segments (block 132). The CT number of each segment is taken as the value at the center of this segment (block 134), which is then converted to a total linear attenuation coefficient through a conversion curve (block 136). No image smoothing is performed so that the DRR at a system display 32 shows its true spatial resolution. When image registration is finished and the daily CT image set is transformed, DRRs from both the plan and the daily CT image sets are generated for comparison of target alignment. The updated daily isocenter for the body frame coordinate system can be input into a computer 30 associated with the HexaPOD robotic couch system to convert from the body frame coordinate system to HexaPOD coordinate system. The corrected and updated target isocenter can be set responsive to the updated daily isocenter within the HexaPOD system, for example, as understood by those skilled in the art. [0037] Portal image verification (see screen 60 in FIG. 4), for example, can be used such that when the updated target isocenter is aligned with the LINAC radiation Isocenter, AP and Lateral portal films are taken prior to treatment delivery. The portal images 62 and the corresponding DRRs are displayed in the split screen 50 of one or more displays 32 for final verification (see FIGS. 3-4).

[0038] Embodiments of a radiotherapy system 20 (see FIGS. 1-2), for example, can include a stereotactic frame system 40 to substantially immobilize a patient and having a rigid treatment couch 41 and a frame 42 fixedly connected to the rigid treatment couch 41, a camera system including cameras 22 and indicators 46 to detect positional changes in the patient P and frame 42, an image scanner 23, e.g., CT scanner, to acquire images of a patient's anatomy, a radiation source 24, e.g., LINAC, to deliver radiation to a preselected portion of the patient's anatomy, a display 32 to display images of the patient's anatomy, and a controller 25 in communication with the image scanner 23, the display 32, and the radiation source 24 to control delivery of the radiation to the preselected portion of the patient's anatomy responsive to a predetermined treatment plan and to display images of the patient's anatomy on the display.

[0039] As understood by those skilled in the art, the controller 25 can include memory 26 having program product 21 stored therein, the program product 21 being operable to perform the step of registering a three-dimensional (3D) image of the frame 42 of the stereotactic treatment system 40 responsive to maximization of mutual information (MMI). MMI can be used as the criterion as MMI is a good indicator of the alignment of two sets of images. It can be both powerful and robust as a criterion for image registration.

[0040] The mutual information (MI) can be a measure of statistical dependence between image intensities of corresponding voxels of two sets of images being displayed on the display.

[0041] The program product 21 can further operate to perform the steps of registering a 3D image of a treatment plan of anatomy after registration of the frame 42, cropping the 3D images of each the frame 42 and the treatment plan of anatomy to first and second preselected sizes, fittingly aligning the smaller image to the larger image, and removing systematic error from the treatment plan, as described in more detail below. Also, one of either the first size or the second size of the images beneficially can be larger than the other.

[0042] Note, the program product 21 can be in the form of microcode, programs, routines, and symbolic languages that provide a specific set for sets of ordered operations that control the functioning of the hardware and direct its operation, as known and understood by those skilled in the art. Although illustrated as stored in memory 26 of the controller 25, those skilled in the art will understand that the program product 21 can also or alternatively be distributed in memory of computer 30 or in a network server (not shown). [0043] In embodiments of a system 20, for example, a patient P is substantially immobilized in a stereotactic body frame system 40 to minimize intra-treatment movement and vertebral body motion such as associated with breathing. The use of pretreatment daily CT scans 53, 55, in conjunction with planning CT scans 52, 54, enable the radiation treatment source 24, for example, to accurately target a tumor regardless of daily setup variations in patient position within the immobilization components of the system 20. Embodiments of the system 20 can also include the stereotactic frame system 40 having a body frame system including the treatment couch 41. The treatment couch includes a base plate 44 to support a patient P, a body vacuum cushion 47, and fixation sheet 48 to secure the patent in a predetermined position, and a vacuum to remove air from the cushion 47 and between the cushion 47 and the fixation sheet 48 to substantially fix the patient P in the predetermined position. The stereotactic frame system 40 can be adapted to move from a scanning position to a treatment position when the patient is immobilized. The radiation source 24, for example, can include a linear accelerator (LINAC), and the image scanner 23, for example, can be selected from the group of: a computed tomography (CT) scanner and a magnetic resonant imaging (MRT) scanner. Although embodiments of the system 20 are described as a CT-on-rails system with a movable couch and stereotactic localizer, the present disclosure contemplates other systems as well as those employing other types of scanners or radiation delivery sources such as a cone beam CT system which makes the stereotactic component unnecessary. [0044] For example, a patient P is substantially immobilized in a predetermined position, for example, a supine position in a body cushion 47 supported by a carbon fiber base plate 44. The body cushion 47 is then custom-molded to the shape of the patient's body by removing air from the cushion via a vacuum as know to those skilled in the art. Fluoroscopic imaging, such as performed on a conventional simulator is then employed to verify or ensure that the patient is lying straight and flat on the patient's back. A plastic fixation sheet 48 is then placed on top of the patient P and sealed on the sides of the body cushion. The vacuum can then remove air surrounding and between the cushion 47 and sheet to confer the sheet 48 to the body of the patient P. Once a patient P is immobilized, a set of images, e.g., images 52, 54 (see FIGS. 2 and 3), can be acquired by the image scanner 23 during a planning phase as understood by those skilled in the art.

[0045] Embodiments of a program product or software 21 of the present invention, for example, can be programmed using MatLab (The Math Works, Inc., Natick, MA), version 6.5 in a personal computer {e.g., Dell Computer Corporation, Austin, TX ) with a 3.40 (3Hz) Pentium 4 CPU, 1.0 GB RAM and a 145 GB hard disk and can operate as a system controller, if desired, or supplement other system controllers. An EXACT targeting system (Varian Medical Systems, Palo Alto, CA) can be used for CT image acquisition and treatment delivery in embodiments of a system 20 of the present invention. An embodiment of a system 20 integrates a high speed CT scanner on rails (GE Medical Systems, Milwaukee, WI) and a Varian EX2100 series linear accelerator (LINAC) unit. The same number of CT slices (e.g., in the range of 90 to 100 slices) can be used for planning and daily verification. The resolution of the CT images can be 0.98 x 0.98 mm² and slice thickness can be 2.0 mm. The image size on each slice was 512 x 512 pixels. Other slice thicknesses can also be used in treatments, but for simplicity and easiness of comparison, all CT images described herein have the same voxel size, which is 0.98 x 0.98 x 2.0 mm³.

[0046] In highly conformal dose distributions such as in Intensity Modulated

Radiotherapy (IMRT), with critical organs such as the spinal cord residing so closely, the orientation change of the target caused by rotation can affect position change.

Rotational shifts of the target without any translational shifts can lead to misalignment. 3D image registration is desired to achieve the accuracy required in image-guided SBRT because rotational shifts tend to be lost in 2-D image registration.

[0047] In embodiments of methods of the present invention, the stereotactic localization frame 42 serves as the reference for the patient anatomy. For the daily CT images, e.g., images 53, 55 (see FIG. 3), if the frame 42 shifts, the anatomy will shift accordingly. Therefore, the first step is to determine the shift of the frame 42 in the daily setup by registering it with the plan setup. After the frame 42 is registered, registration of CT images of anatomy is performed. The registration results can be displayed on the user interface screen or display 32. In frame registration, only translational shifts are assumed because the frame 42 is fixed to a rigid treatment couch 41. For anatomy registration, however, translational shifts as well as three rotational shifts, roll, yaw and pitch are assumed.

[0048] In an embodiment of image-guided SBRT as shown in FIG. 13, for example, the goal of image registration is to register the target (block 80). To minimize the influence of the anatomy that is far from the target, plan and daily images both can be cropped (blocks 82, 84, 86). The reduced image size, for example, speeds up the registration process and also increases registration accuracy. The image size for the daily CT slab with 2 mm slice thickness, for example, is 101 x 101 x 33 pixels. This CT slab covers two whole vertebral bodies and spinous processes. The planning CT slab has a larger sample space of 171 x 171 x 53 pixels. The search space is the difference between the two image sets (block 88), which is 70 x 70 x 20 pixels. When the two CT slabs are co-centered, the search space in either direction is half of the total pixels. In this case, the search space can be denoted as ± 35 x ± 35 x ± 10 pixels, or ± 34.3 x ± 34.3 x ± 20 mm³ when converted to physical dimensions with voxel size of 0.98 x 0.98 x 2.0 mm³. This space, for example, should be more than enough to cover even the maximum shift of most patients. Another advantage of fitting a smaller image to a larger image for registration is to guarantee that the data sets used are always the same in calculating the MI. In this way, for example, the possible change in MI when image data sets change can be eliminated, therefore assuring a more accurate registration. [0049] The strategy of search used in embodiments of a method and program product is a result-guided over-search with variable searching step lengths, in which the search direction and the number of steps in that direction are guided by the search result. For example, as shown in FIG. 14, if MI value increases while searching in one direction, the search will continue (blocks 92 and 94). If it does not, the search will still go a few more steps and then decide whether to keep going in the same direction (block 96) or change to another direction (block 98). In the initial rough search, the search step used can be 1.0 pixel (block 90). After an MMI is determined (block 100), a fine search can be conducted around this point with a reduced search step of 0.05 of a pixel (block 102). This guarantees a subvoxel registration accuracy. Shift step of rotation can be set at 0.25 degree. The result-guided search strategy, for example, can save time by avoiding an exhaustive search, in which all the possible points in the whole search space is searched to find the global MMI. On the other hand, the over-search is desired to step over possible local maxima of ML This search strategy, for example, can offer good speed and consistent accuracy.

[0050] After image registration is finished, as shown in FIG. 15, the transformed daily CT images, e.g., images 53, 55, are displayed with the planning CT images, e.g., 52, 54, in the same picture frame 56 on the interface screen or display 32 (see FIG. 3) in split screen format (block 110), which has the top part of the plan CT image and the bottom part of the daily CT image joining together in one picture frame. In this format, any misalignment of the two images can be readily detected along the joining line 57 (block 112). The user can move the joining line 57 up and down to see the registration results in different areas of the same slice (block 114). hi addition, digital subtraction of the two DRR images can be displayed in the same frame (block 116). If the alignment is good (block 118), the intensity difference is uniform. When the alignment is poor (block 120), brightness or dark shadows may appear. The overall registration accuracy can be seen at one glance. Should it be necessary, any of the shifts can be modified manually (block 122). The image with the new transformation is displayed immediately after the new shift is entered (block 124). [0051] DRRs can be generated using a ray-tracing method (block 130), for example, as shown in FIG. 16. The ray that passes through the patient's 3D CT volume is divided into many segments (block 132). The CT number of each segment is taken as the value at the center of this segment (block 134), which is then converted to a total linear attenuation coefficient through a conversion curve (block 136). Interpolation can be used to find the CT number at the center of each segment of the ray (block 138). The treatment isocenter is set to be 100 cm from the x-ray source just as in real treatment, and the image plane for DRR can be set at 140 cm from the source. The image size of DRR can be 381 x 381 pixels corresponding to a physical area of 11.0 x 11.0 cm² at a distance 100 cm from the source. The spatial resolution of the DRR is therefore 0.289 x 0.289 mm². No image smoothing is performed so that the DRR at display shows its true spatial resolution. DRR at this size is smooth enough even magnified to a physical size of 20 x 20 cm² on a computer screen.

[0052] Measures can be taken to enhance the image quality of the DRR (block 140), which includes cropping the 3D CT volume to include only the CT slab of interest to generate the DRR with. With all the soft tissue, internal organs and bones such as ribs removed, the final DRR can have a much cleaner background (block 142). The result is a sharper image with good contrast. This is especially helpful when the target is in one of the low cervical or high thoracic vertebrae. In the lateral view DRR, the clavicles often make details of the vertebrae hard to see. By keeping the clavicles out, the image of the vertebrae becomes very clear. [0053] Conversion of CT numbers to the total linear attenuation coefficients can be performed according to the CT numbers of some typical tissues from the CT scanner used and the published total linear attenuation coefficient values (mass attenuation coefficient μ/p and density /?) such as illustrated in the table shown in FIG. 11. Because it gives good contrast of DRR, 30 keV x-ray can be used. The table in FIG. 11 is a list of the tissues used for conversion, their CT numbers, mass attenuation coefficients, and specific mass. Their total linear attenuation is calculated from those data. Then linear interpolation can be used to convert CT numbers to linear attenuation coefficient p for any given CT number. The conversion curve can be cut off at CT number of 2100 to avoid extreme brightness of metals that were found in patients who had had surgeries to remove tumors from their spine. [0054] When image registration is finished and the daily CT image set is transformed, DRRs from both the plan and the daily CT image sets are generated for comparison of target alignment. A user of embodiments of a system 20, method, or program product 21 of the present invention, for example, can either use the default CT image volume or choose a different one for DRR generation. Once generated, centimeter and quarter centimeter markers are added to the DRR images for easy comparison. Contrast and brightness of the DRRs can be manipulated individually. Manual adjustments can also be made in any of the translational or rotational shifts. If changes are made, daily DRRs can be regenerated to reflect the new shifts. For final comparison, the two sets of DRRs can be displayed in the new screen in the format of split screen and intensity difference. FIGS. 6A-6B show a pair of AP view DRRs in the formats of split screen and intensity difference. The shadow of the spinous process shown on the bottom center in FIG. 6B indicates lateral misalignment of the spine in patient's daily setup. This misalignment, for example, turned out to be 0.2 mm.

[0055] Coordinates of the daily isocenter that are converted to RT2 system are displayed on this screen according to the localization picture frame in use. These numbers are needed to mark the new isocenter position on the localization picture frame to align the patient on the LINAC side for portal image verification. [0056] As shown in FIGS. 3-4 and 17, for example, when the patient is aligned using the new isocenter position on the localization frame, orthogonal portal images 62 are taken immediately for final verification of target against the plan DRRs before treatment is delivered (block 150). This was previously done through side-by-side comparison of the portal images with the corresponding DRRs. To facilitate this process, portal image verification capability can be added in embodiments of a program product or software 21 to have the portal image and the corresponding DRR displayed in one split screen (see FIG. 4). The split screen 50 is more accurate and more convenient than the conventional side-by-side comparison (see also FIG. 3). Using this method, even misalignment of a fraction of a millimeter can be readily detected. [0057] As shown in FIG. 4, the portal image 62 and its corresponding DRR should have the same physical size and the same image size to be displayed in the one frame for alignment verification. To achieve this, for example, the portal image 62 can be first cropped to a physical size of 11 x 11 cm², and then resampled to an image size of 381 x 381 pixels (block 152) to match that of the DRR. The centimeter marks on the gradicule in the portal image 62 are used to center it and to crop it. It is possible that the physical size or the center of the portal image 62 may slightly differ from those of the DRR. In such cases, fine adjustments can be performed manually to more perfectly align a match (block 154). Due to the poor contrast of the megavolt portal image, enhancement functionality can be added for improving the image contrast. Also, a centered, physical sized matched portal image and DRR can be displayed in one frame in split screen format (block 156). Sliders are provided to view the alignment at different sections. Should any fine shift adjustment be needed (block 158), additional shifts can be entered manually (block 160) and the portal image will shift accordingly. This process can be repeated until the best alignment is made with the plan DRRs (block 162).

[0058] For example, accuracy of the image registration was tested using a head and neck phantom as illustrated in an image registration and verification screen 70 (see FIG. 5). After two sets of CT images of the phantom were registered using an embodiment of a program product or software 21, manual shift was made in the lateral direction. A section of the vertical wall of the head and neck phantom was chosen to demonstrate the registration accuracy in the lateral direction as shown in FIGS. 7A-7D. When the two sets of images were registered, the vertical wall of the phantom was very well aligned as one, as demonstrated in (FIG. 7A). With presence of misalignment, shift of the vertical wall where the two images are joined together can be seen. As an example, lateral shifts in the amounts of 0.1, 0.3 and 0.5 mm were introduced in FIGS. 7B, 7C, and 7D, respectively. Misalignment of the phantom wall inside the circles also can be seen. FIG. 7B shows that a shift as small as 0.1 mm can produce a misalignment detectable to the naked eyes. The perfectly looking alignment of the vertical phantom wall in FIG. 7B clearly demonstrates that the embodiment of the program product or software 21, system 20, and methods has a registration accuracy of better than 0.1 mm. [0059] Additionally, for example, a table in FIG. 12 illustrates a list of computer time taken by the image registration process in several example cases. Volume of data sets used for the data in the table of FIG. 12 is as follows: plan CT is 171 x 171 x 53, and daily CT is 101 x 101 x 33. The results show that the time for image registration is basically between 1-2 minutes. The average time is 97 seconds. In general, the total time for image registration increases with initial patient setup errors, but this increase is minuscule. This is because a considerable amount of the total time is spent on over-search, which is the search after an MMI is determined. Therefore, doubling the shift distances does not double the total registration time. This can be seen from case No. 11 in the table of FIG 12, where no shift is needed but the total time to finish the registration is still over 63 seconds. Although in most cases image registration time is doubled by the over-search, the extra time is justified. Because an embodiment of a program product 21 and a method of the present invention can use MMI as the criterion for image registration and local maxima of MI often exist, the search may trap into one of the local maxima if over-search is not conducted. If this happens, the registration result may be poor. An embodiment, for example, should be able to step out of the local maximum after the over-search step number was increased.

[0060] FIG. 8 shows how DRR generation time varies with the thickness of the CT volume used to generate DRR at different DRR image sizes. The result shows that the total DRR generation time varies linearly with the thickness of the CT volume, and also linearly with the total image points of DRR. This can be easily explained. Because the DRR method can use ray tracing, doubling the number of voxels the ray passes through means doubling the amount of work for ray tracing, therefore doubling the DRR generation time. For the same reason, when the image size is increased from 128 x 128 to 256 x 256, that is when the total points of the image is quadrupled, the total time needed to generate the DRR is also quadrupled. An example of a typical thickness of CT volume used is about 100 pixels. FIG. 8 shows that the DRR generation time is about 27 seconds at image size of 381 x 381.

[0061] The average time to finish the process is 97 seconds for image registration plus 108 seconds of DRR generation (4 DRRs at 27 seconds each), which gives a total time of 205 seconds or roughly 3.5 minutes. In some of the example cases tested, image registration and DRR generation were finished in as little as 3 minutes. Even the most "time consuming" case (Case No. @3 in the table of FIG. 12) took only 4 minutes and 5 seconds. For comparison, the process using Pinnacle and RT2 took anywhere from 15 to 30 minutes or even longer. Embodiments of program product 21 and methods of the present invention can save time not only because the methods are faster, but more importantly because the process can be fully automated.

[0062] On the DRR comparison screen, the lateral view and AP view DRRs from the plan CT image sets can be displayed at the left, and the corresponding DRRs from the daily CT images are displayed on the right (see FIG. 4). Color coded centimeter and quarter-centimeter marks added to the DRRs help to locate the isocenter and other features on the DRRs more accurately in side-by-side comparison.

[0063] As mentioned earlier, rotational shifts can be interpreted as translational shifts on a 2D image such as DRR or portal image. They must be corrected before comparison is performed. FIGS. 9A- 1OB show a pair of DRRs generated from the same CT volumes with and without a 3° rotation in the axial plane. If the spinous processes (as indicated by the arrows) is used for alignment, the isocenter in FIG. 9B needs to move 2.5 mm to the right to match that in FIG. 9A. This means that the 30 rotation in the axial plane can cause an alignment error of 2.5 mm in the lateral direction. This error is unacceptable in SBRT, and it has to be corrected. Rotational shifts should be corrected if high accuracy of target localization is to be achieved.

[0064] Portal images, for example, portal images 62 shown in FIG. 4, are views of the treatment target from the eye of the radiation beam. Verification of portal images can be important to make sure the patient's treatment target is at the exactly the same position as planned. The isocenter of the treatment target in the portal images is compared with that in the DRRs generated from the plan CT images. The advantage of showing two images, e.g., plan and daily, in one frame in the format of split screen is that a very small shift can be detected more easily than side-by-side comparison. FIG. 1OA shows a daily portal image displayed in a split screen with the corresponding planning DRR. If judging from side-by-side comparison, the patient's daily position may look perfect. When they are displayed in a split screen, a misalignment of a fraction of a millimeter in the lateral direction can be found, as shown in the area within the oval circle. FIG. 1OB is the portal image after a 0.2 mm lateral shift, which shows a perfect alignment. Although the 0.2 mm shift may not be necessary in daily treatment, this example does show the accuracy a split screen is capable of in portal image verification. [0065] Using a phantom, the registration result showed that the accuracy of a registration is better than 0.1 mm. DRR comparison and portal image verification are capable of similar accuracy. Accuracy at this level should meet the positioning requirement of any image-guided radiotherapy or radiosurgery. To achieve the same accuracy with real patients, however, attention should be paid when aligning the patient. The human body, even the more rigid bone structure such as spinal vertebra column, is not as rigid as a phantom. Deformation may occur due to twist or bend in the patient's body position. Without any twist or bend, the image registration process can find the new target position by a transform of translation and rotation and substantially perfect registration of the whole target can be achieved. With twist or bend in the patient's body position, substantially perfect registration of the whole target is not possible, because a registration method assumes no deformation of the anatomy. The reason for this is not because the MI based image registration methods are not able to accommodate the deformation in human anatomy. Rather, it is because the treatment plan cannot be changed to conform to the deformed shape of the target. Therefore, great precaution and measures should be taken when aligning the patient to avoid patient body deformation.

[0066] Even with great precaution, some small deformation is almost inevitable in the daily patient setup. The deformation of the bone structure that the registration relies on can impact the final registration accuracy. For this reason, it is advantageous to use a small CT volume to include only the treatment target and the neighboring bone structure that is going to be used for alignment. In this way, the target volume can be accurately aligned even though other parts of the body may be deformed and out of position. This is also one of the reasons to crop the CT volume for registration in the first place. [0067] Embodiments of a program product or software 21 can integrate all the functionalities needed for image-guided SBRT in one package. As understood by those skilled in the art, the process can be automated and require only minimal user interaction. The result is that the time spent on aligning the patients is reduced from as long as 30 minutes or more to 4 minutes or less. The shortened alignment process increases alignment accuracy by reducing the possibility of patient motion. The fully automated process also can help eliminate possible human errors in data transfer from one system to another.

[0068] Comparison of corresponding DRRs generated from the plan and daily CT images is important in finding how good the overall alignment of the treatment target on the patient is. It is also the final verification of the registration result. Apart from displaying the DRRs side-by-side for comparison, embodiments of program product 21, system 20, and methods also provide the ability to view the corresponding DRRs in one frame in split screen format (see FIG. 3). This format of viewing is beneficially more sensitive to misalignments between the DRR pair. Misalignment of 0.1 mm can be readily detected. The intensity difference of two DRRs shows overall registration of the whole target. Another advantage of displaying two corresponding DRRs in one frame is that relative position change of the bone structure can be clearly seen when the two DRRs are displayed alternatively like in a slide show. This gives the user a very clear picture of the patient's daily position as compared with his/her plan position. [0069] It is more convenient and accurate to verify portal images 62 (see FIG. 4) while displaying them in the same frame as in DRR comparison. Misalignment as little as 0.2 mm can be detected. One issue that sometimes can become a problem about portal image is their low contrast resulting from the megavolt beam of LINACs. Although enhancements can be made, there can still be situations when further improvement of image quality is desired, especially in the lateral views where neck and shoulder are both included. Use of kilovolt cone beam CT, as understood by those skilled, in the art, for example, can improve the portal image quality.

[0070] Embodiments of a program-product or software 21 of the present invention facilitate image-guided SBRT where accurate target registration is required. The image registration methods are capable of image registration with an accuracy of better than 0.1 mm. The operation is automated. Image registration and DRR generation can be completed within 3 minutes. A distinctive feature, for example, is that the portal image is displayed with the corresponding DRR in the same frame to enhance verification (see FIG. 5). Sub-millimeter misalignment can be readily detected. Application of embodiments of a program product or software has demonstrated it is fast, accurate and the results are reliable. Although it was developed to facilitate our SBRT for spinal tumors, it can be a useful and convenient tool for any image-guided radiotherapy where high accuracy and fast speed are desired.

[0071] As illustrated in FIGS. 1-17, embodiments of the present invention also provide methods to enhance accuracy and set up speed of radiotherapy treatment delivery. A method includes registering a three-dimensional (3D) image, e.g., image

53, 55, of a frame 42, fixedly connected to a rigid treatment couch 41, responsive to maximization of mutual information (MMI). The mutual information (MI) can be a measure of statistical dependence between image intensities of corresponding voxels of two sets of images, e.g., taken AP and lateral. As perhaps best shown in FIG. 3, the method can also include registering a 3D image (or a 4D image), e.g., image 52,

54, of a treatment plan of anatomy after registration of the frame 42 and cropping the 3D images of each of the frame 42 and the treatment plan of anatomy to first and second preselected sizes. One of either the first size or the second size can be larger than the other. The method can further include fittingly aligning the smaller image, e.g., 52, 54, to the larger image, e.g., 53, 55, within an image frame 56, and removing systematic error from the treatment plan. The frame image, for example, can be larger than the plan image. The fittingly aligning can includes determining a shift to align the first (upper) and second (lower) images of each image frame 56, and the method can further include calculating an isocenter of the patient's anatomy in the second image responsive to the first image and the determined shift. The registering a 3D image (or a 4D image) of a frame 42 includes determining the isocenter responsive to the frame 42 fixedly connected to the rigid treatment couch 41, and the determining a shift includes calculating a translation of at least one of the first and second images in three dimensions. The fittingly aligning can further include selecting a plurality of landmarks in the first image, selecting substantially identical landmarks in the second image, aligning the selected landmarks of the first and second images, and fusing each voxel of the first image to each voxel of the second image after the selected respect landmarks are aligned.

[0072] Although the invention has been particularly shown as described with reference to embodiments herein, it will be understood by those skilled in the art that various changes in form and detail can be made therein without departing from the spirit and scope of the invention. For example, it will be appreciated that the concepts disclosed herein can be extended or modified to apply to other types of configuration constructs having different rules than the particular exemplary embodiments disclosed herein. In addition, although aspects of the present invention have been described with respect to a computer, a computer device, a computer system, or controller executing program product or software that directs the functions of embodiments of the present invention, it should be understood by those skilled in the art that present invention can be implemented as a program product for use with various types of data processing systems as well. Programs defining the functions of embodiments of the present invention, for example, can be delivered to a data processing system via a variety of signal-bearing media, which include, without limitation, non-rewritable storage media (e.g., CD-ROM, DVD-ROM), rewritable storage media (e.g., a floppy diskette, hard disk drive, CD-R, or rewritable ROM media), and communication media, such as digital and analog networks. It should be understood, therefore, that such signal-bearing media, when carrying or encoding computer readable instructions that direct the functions of embodiments of the present invention, represent alternative embodiments of the present invention.

[0073] The invention has been sufficiently described so that a person with average knowledge in the matter may reproduce and obtain the results mentioned in the invention herein. Nonetheless, any skilled person in the field of technique, subject of the invention herein, may carry out modifications not described herein, to apply these modifications to a determined structure, or in the manufacturing process of the same, requires the claimed matter in the following claims; such structures also shall be covered within the scope of the invention. In the drawings and specification, there have been disclosed embodiments of the invention and, although specific terms are employed, they are used in a generic and descriptive sense only and not for the purpose of limitation, the scope of the invention being set forth in the following claims.

Claims

THAT CLAIMED IS

1. A method to enhance accuracy of radiotherapy treatment delivery, the method characterized by the steps of: registering an image (53, 55) of a frame (42) responsive to mutual information (MI) being a measure of statistical dependence between image intensities of corresponding voxels of two sets of images; registering an image (52, 54) of a treatment plan of anatomy after registration of the frame (42); cropping the images (53, 55), (52, 54) of the frame (42) and the treatment plan of anatomy to first and second preselected sizes; aligning the images within an image frame (56); and removing systematic error from the treatment plan.

2. A method as defined in Claim 1, wherein one of either the first size or the second size of the images (53, 55), (52, 54) is larger than the other.

3. A method as defined in Claim 2, wherein the step of aligning the images includes fittingly aligning the smaller image to the larger image within the image frame (56).

4. A method as defined in any of Claims 1-3, wherein the frame (42) is fixedly connected to a rigid treatment couch (41).

5. A method as defined in any of Claims 1-4, wherein at least one of the images (52, 54), (53, 55) is selected from the group of: three-dimensional (3D) images and four-dimensional (4D) images.

6. A method as defined in any of Claims 1-5, wherein the step of registering an image (53, 55) of a frame (42) is also responsive to maximization of mutual information (MMI).

7. A method as defined in any of Claims 1-6, wherein the frame image (53, 55) is larger than the plan image (52, 54).

8. A method as defined in any of Claims 3-7, wherein the smaller image defines a first image ((52, 54) or (53, 55)), wherein the larger image defines a second image

((53, 55) or (52, 54)), and wherein the step of fittingly aligning includes determining a shift to align the first and second images (52, 54), (53, 55), and the method being further characterized by calculating an isocenter of the patient's anatomy in the second image responsive to the first image and the determined shift.

9. A method as defined in any of Claims 3-8, wherein the smaller image defines a first image ((52, 54) or (53, 55)), wherein the larger image defines a second image ((53, 55) or (52, 54)), and wherein the step of registering an image (53, 55) of a frame (42) includes determining the isocenter responsive to the frame (42) being fixedly connected to the rigid treatment couch (41), and wherein the step of determining a shift includes calculating a translation of at least one of the first and second images (52, 54), (53, 55) in three dimensions.

10. A method as defined in any of Claims 3-9, wherein the smaller image defines a first image ((52, 54) or (53, 55)), wherein the larger image defines a second image

((53, 55) or (52, 54)), and wherein the step of fittingly aligning includes selecting a plurality of landmarks in the first image, selecting substantially identical landmarks in the second image, aligning the selected landmarks of the first and second images (52, 54), (53, 55), and fusing each voxel of the first image to each voxel of the second image after the selected respect landmarks are aligned.

11. A method of radiotherapy treatment delivery characterized by the steps of registering an image (52, 54), (53, 55), digitally reconstructing a radiograph, and verifying the image by use of a portal image (62).

12. A computer readable medium having program product (21) adapted to enhance accuracy of radiotherapy treatment delivery, the program product (21) characterized by operating to perform the operations of: registering a three-dimensional (3D) image (53, 55) of the frame (42) of the stereotactic treatment system (40) responsive to maximization of mutual information

(MMI), mutual information (MI) being a measure of statistical dependence between image intensities of corresponding voxels of two sets of images (52, 53), (54, 55) being displayed on a display (32); registering a 3D image (52, 54) of a treatment plan of anatomy after registration of the frame (42) ; cropping the 3D images (53, 55), (52, 54) of the frame (42) and the treatment plan of anatomy to first and second preselected sizes, one of either the first size or the second size being larger than the other; fittingly aligning the smaller image to the larger image; and removing systematic error from the treatment plan.

13. A computer readable medium as defined in Claim 12, wherein the frame image (53, 55) is larger than the plan image (52, 54).

14. A computer readable medium as defined in either of Claims 12 or 13, wherein the smaller image defines a first image ((52, 54) or (53, 55)), wherein the larger image defines a second image ((53, 55) or (52, 54)), and wherein the operation of fittingly aligning includes determining a shift to align the first and second images (52, 54), (53, 55), the operations being further characterized by calculating an isocenter of the patient's anatomy in the second image responsive to the first image and the determined shift.

15. A computer readable medium as defined in Claim 14, wherein the operation of registering a 3D image (53, 55) of a frame (42) includes determining an isocenter responsive to the frame (42) when fixedly connected to the rigid treatment couch (41), and wherein the operation of determining a shift includes calculating a translation of at least one of the first and second images in three dimensions.

16. A computer readable medium as defined in any of Claims 12, 13, 14, or 15, wherein the operation of fittingly aligning includes selecting a plurality of landmarks in the first image, selecting substantially identical landmarks in the second image and aligning the selected landmarks of the first and second images (52, 54), (53, 55), and fusing each voxel of the first image to each voxel of the second image after the selected respect landmarks are aligned.

17. A radiotherapy system (20) comprising a stereotactic frame system (40) to substantially immobilize a patient and having a rigid treatment couch (41) and a frame

(42) fixedly connected to the rigid treatment couch (41); an image scanner (23) to acquire images of a patient's anatomy; a radiation source (24) to deliver radiation to a preselected portion of the patient's anatomy; a display (32) to display images of the patient's anatomy; and a controller (25) in communication with the image scanner (23), the display (32), and the radiation source (24) to control delivery of the radiation to the preselected portion of the patient's anatomy responsive to a predetermined treatment plan and to display images of the patient's anatomy on the display (32), the controller (25) including memory (26) having program product (21) stored therein, the program product (21) characterized by being operable to perform the following steps: registering a three-dimensional (3D) image (53, 55) of the frame (42) of the stereotactic frame system (40) responsive to maximization of mutual information (MMI), mutual information (MI) being a measure of statistical dependence between image intensities of corresponding voxels of two sets of images (52, 53), (54, 55) being displayed on the display (32); registering a 3D image (52, 54) of a treatment plan of anatomy after registration of the frame (42); cropping the 3D images (53, 55), (52, 54) of each the frame (42) and the treatment plan of anatomy to first and second preselected sizes; fittingly aligning the smaller image to the larger image; and removing systematic error from the treatment plan.

18. A system (20) as defined in Claim 17, wherein the stereotactic frame system (40) comprises a body frame system including the treatment couch (41) comprising a base plate (44) to support a patient, a body vacuum cushion (47) and fixation sheet (48) to secure the patent in a predetermined position, and a vacuum to remove air from the cushion (47) and between the cushion (47) and the fixation sheet (48) to substantially fix the patient in the predetermined position.

19. A system (20) as defined in either of Claims 17 or 18, wherein the stereotactic frame system (40) is adapted to be moved from a scanning position to a treatment position when the patient is immobilized, and wherein one of either the first size or the second size of the images (52, 54), (53, 55) is larger than the other.

20. A system (20) as defined in any of Claims 17, 18, or 19, wherein the radiation source (24) includes a linear accelerator, and wherein the image scanner (23) is selected from the group of: a computed tomography (CT) scanner and a magnetic resonant imaging (MRI) scanner.

21. A system (20) as defined in any of Claims 17, 18, 19, or 20, wherein the frame image (53, 55) is larger than the plan image (52, 54).

22. A system (20) as defined in any of Claims 17, 18, 19, 20, or 21, wherein the smaller image defines a first image ((52, 54) or (53, 55)), wherein the larger image defines a second image ((53, 55) or (52, 54)), and wherein the step of fittingly aligning includes determining a shift to align the first and second images (52, 54), (53, 55), and the program product (21) being further characterized by being operable to perform the step of calculating an isocenter of the patient's anatomy in the second image responsive to the first image and the determined shift.

23. A system (20) as defined in any of Claims 17, 18, 19, 20, 21, or 22, wherein the smaller image defines a first image ((52, 54) or (53, 55)), wherein the larger image defines a second image ((53, 55) or (52, 54)), and wherein the step of fittingly aligning includes selecting a plurality of landmarks in the first image, selecting substantially identical landmarks in the second image and aligning the selected landmarks of the first and second images (52, 54), (53, 55), and fusing each voxel of the first image to each voxel of the second image after the selected respect landmarks are aligned.

24. A system (20) as defined in any of Claims 17, 18, 19, 20, 21, 22, or 23, wherein the step of registering a 3D image (53, 55) of a frame (42) of the program product steps of the controller (25) includes determining an isocenter responsive to the frame (42) when fixedly connected to the rigid treatment couch (41).