WO2006130771A2

WO2006130771A2 - Four-dimensional volume of interest

Info

Publication number: WO2006130771A2
Application number: PCT/US2006/021291
Authority: WO
Inventors: Hongwu Wang; John R. Dooley; Jay B. West
Original assignee: Accuray Incorporated
Priority date: 2005-06-02
Filing date: 2006-06-01
Publication date: 2006-12-07
Also published as: US7352370B2; WO2006130771A3; US20060274061A1

Abstract

A method and apparatus for a four-dimensional volume of interest is described.

Description

FOUR-DIMENSIONAL VOLUME OF INTEREST TECHNICAL FIELD

[0001] This invention relates to the field of radiation treatment

planning and, in particular, to a volume of interest applied to treatment

planning.

BACKGROUND

[0002] A non-invasive method for pathological anatomy (e.g., tumor,

legion, vascular malformation, nerve disorder, etc.) treatment is external beam

radiation therapy. In one type of external beam radiation therapy, an

external radiation source is used to direct a sequence of x-ray beams at a

pathological anatomy site from multiple angles, with the patient positioned so

the pathological anatomy is at the center of rotation (isocenter) of the beam.

As the angle of the radiation source is changed, every beam passes through

the pathological anatomy site, but passes through a different area of healthy

tissue on its way to the pathological anatomy. As a result, the cumulative

radiation dose at the pathological anatomy is high and the average radiation

dose to healthy tissue is low. The term radiotherapy refers to a procedure in

which radiation is applied to a target region for therapeutic, rather than

necrotic, purposes. The amount of radiation utilized in radiotherapy

treatment sessions is typically about an order of magnitude smaller, as compared to the amount used in a radiosurgery session. Radiotherapy is

typically characterized by a low dose per treatment (e.g., 100-200 centi-Gray

(cGy)), short treatment times (e.g., 10 to 30 minutes per treatment) and

hyperfractionation (e.g., 30 to 45 days of treatment). For convenience, the

term "radiation treatment" is used herein to mean radiosurgery and/or

radiotherapy unless otherwise noted by the magnitude of the radiation.

[0003] Traditionally, medical imaging was used to represent two-

dimensional views of the human anatomy. Modern anatomical imaging

modalities such as computed tomography (CT) are able to provide an

accurate three-dimensional model of a volume of interest (e.g., skull or

pathological anatomy bearing portion of the body) generated from a

collection of CT slices and, thereby, the volume requiring treatment can be

visualized in three dimensions. More particularly, in CT scanning numerous

x-ray beams are passed through a volume of interest in a body structure at

different angles. Then, sensors measure the amount of radiation absorbed by

different tissues. As a patient lies on a couch, an imaging system records x-

ray beams from multiple points. A computer program is used to measure the

differences in x-ray absorption to form cross-sectional images, or "slices" of

the head and brain. These slices are called tomograms, hence the name

"computed tomography." [0004] During treatment planning, a volume of interest (VOI) from

anatomical (e.g., CT) and/or functional imaging is used to delineate structures

to be targeted or avoided with respect to the administered radiation dose. A

volume of interest (VOI) may be defined as a set of planar, closed polygons, as

illustrated in Figure IA. The coordinates of the polygon vertices are defined

as the x/y/z offsets in a given unit from the image origin. Once a VOI has

been defined, it may be represented as a bit wise mask overlaid on the

functional and/or anatomical image (so that each bit is zero or one according

to whether the corresponding image volume pixel (voxel) is contained within

the VOI represented by that bit), or a set of contours defining the boundary of

the VOI in each image slice. Conventional VOI imaging architectures may

utilize a three-tier representation structure: VOI-contourslice-contour. Figure

IB illustrates the three-tier VOI structure in a Unified Modeling Language

(UML) graph with a sample VOI.

[0005] One problem encountered in external beam radiation treatment

is that pathological anatomies (e.g., a tumor) may move during treatment,

which decreases accurate target localization (i.e., accurate tracking of the

position of the target). Most notably, soft tissue targets tend to move with

patient breathing during radiation treatment delivery sessions. Respiratory

motion can move a pathological anatomy in the chest or abdomen, for

example, by more than 3 centimeters (cm). In the presence of such respiratory motion, for example, it is difficult to achieve the goal of precisely and

accurately delivering the radiation dose to the target, while avoiding

surrounding healthy tissue. In external beam radiation treatment, accurate

delivery of the radiation beams to the pathological anatomy being treated can

be critical, in order to achieve the radiation dose distribution that was

computed during the treatment planning stage.

[0006] Conventional methods for tracking anatomical motion utilize

external markers and/or internal fiducial markers. Such conventional

methods do not enable modeling of the anatomical change due to the

respiratory cycle using conventional VOI imaging architectures. Moreover,

such conventional methods do not take into account non-rigid motions and

deformations of surrounding anatomy, as a function of motion cycle.

BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The present invention is illustrated by way of example, and not

by way of limitation, in the figures of the accompanying drawings.

[0008] Figure IA illustrates a volume of interest defined by a stack of

planar closed polygons.

[0009] Figure IB illustrates a conventional three-tier VOI structure in a

Unified Modeling Language (UML) graph with a sample VOI.

[0010] Figure 2 illustrates one embodiment of the acquisition of pre-

treatment images (e.g., CT scans) of a changing target within a patient's

anatomy.

[0011] Figure 3 illustrates one embodiment of a 4D VOI architecture,

where the fourth dimension in the VOI architecture is a time dimension.

[0012] Figure 4 illustrates one embodiment of generating a VOI using a

4D VOI architecture.

[0013] Figure 5 illustrates of medical diagnostic imaging system

implementing one embodiment of the present invention.

[0014] Figure 6 illustrates one embodiment of a 4D mask volume.

[0015] Figure 7 illustrates a 2-dimensional perspective of radiation

beams of a radiation treatment system directed at a target region according to

a treatment plan. [0016] Figure 8 is a flow chart illustrating one embodiment of

generating a treatment plan using a 4D mask volume.

DETAILED DESCRIPTION

[0017] In the following description, numerous specific details are set

forth such as examples of specific systems, components, methods, etc. in order

to provide a thorough understanding of the present invention. It will be

apparent, however, to one skilled in the art that these specific details need not

be employed to practice the present invention. In other instances, well-known

components or methods have not been described in detail in order to avoid

unnecessarily obscuring the present invention.

[0018] Embodiments of the present invention include various steps,

which will be described below. The steps of the present invention may be

performed by hardware components or may be embodied in machine-

executable instructions, which may be used to cause a general-purpose or

special-purpose processor programmed with the instructions to perform the

steps. Alternatively, the steps may be performed by a combination of

hardware and software.

[0019] Embodiments of the present invention may be provided as a

computer program product, or software, that may include a machine-readable

medium having stored thereon instructions, which may be used to program a

computer system (or other electronic devices) to perform a process. A

machine-readable medium includes any mechanism for storing or

transmitting information in a form (e.g., software, processing application) readable by a machine (e.g., a computer). The machine-readable medium may

include, but is not limited to, magnetic storage medium (e.g., floppy diskette);

optical storage medium (e.g., CD-ROM); magneto-optical storage medium;

read-only memory (ROM); random-access memory (RAM); erasable

programmable memory (e.g., EPROM and EEPROM); flash memory;

electrical, optical, acoustical, or other form of propagated signal (e.g., carrier

waves, infrared signals, digital signals, etc.); or other type of medium suitable

for storing electronic instructions.

[0020] Embodiments of the present invention may also be practiced in

distributed computing environments where the machine-readable medium is

stored on and/or executed by more than one computer system. In addition,

the information transferred between computer systems may either be pulled

or pushed across the communication medium connecting the computer

systems, such as in a remote diagnosis or monitoring system. In remote

diagnosis or monitoring, a user may utilize embodiments of the present

invention to diagnose or monitor a patient despite the existence of a physical

separation between the user and the patient.

[0021] Some portions of the description that follow are presented in

terms of algorithms and symbolic representations of operations on data bits

that may be stored within a memory and operated on by a processor. These

algorithmic descriptions and representations are the means used by those skilled in the art to effectively convey their work. An algorithm is generally

conceived to be a self-consistent sequence of acts leading to a desired result.

The acts are those requiring manipulation of quantities. Usually, though not

necessarily, these quantities take the form of electrical or magnetic signals

capable of being stored, transferred, combined, compared, and otherwise

manipulated. It has proven convenient at times, principally for reasons of

common usage, to refer to these signals as bits, values, elements, symbols,

characters, terms, numbers, parameters, or the like.

[0022] It should also be noted that the methods and apparatus are

discussed herein in relation to CT imaging only for ease of explanation. The

method and apparatus discussed herein may also be used to generate VOIs

from other types of medical diagnostic images (anatomical and/or functional),

for example, magnetic resonance (MR), ultrasound (US), nuclear medicine

(NM) PET/SPECT, etc. In addition, the "targets" discussed herein may be an

anatomical feature(s) of a patient such as a pathological or normal anatomy

and may include one or more non-anatomical reference structures.

[0023] Figure 2 illustrates one embodiment of the acquisition of pre-

treatment images (e.g., CT scans) of a changing target within a patient's

anatomy. In the illustrated embodiment, the target 211 (e.g., anatomical

feature volume) may move or undergo a deformation (which may be a non-

rigid deformation) during a patient's respiration, heartbeats, or other motion. While in embodiments herein the target 211 is described as moving

periodically while undergoing a non-rigid deformation, any other type of

motion (e.g. aperiodic motion) and any type of deformation or motion of the

target may be accounted for. Furthermore, while embodiments may be

discussed herein in regards to a CT scans, any other type of imaging modality

may be used, for example, magnetic resonance imaging (MRI), positron

emission tomography (PET), ultrasound, etc.

[0024] In one embodiment, CT scans (e.g., CT image 221, 222, 223) are

taken at different times points b within, for example, a breathing cycle P 270 of

a patient. The time points b correspond to different epochs in the patient

breathing cycle, with, for example, to < ti < t2. The cycle P 270 may be

monitored by an external sensor, for example, a breathing sensor, markers

placed on the chest, etc. The CT images 221, 222, and 223 are taken at time

points to, ti, and t2, respectively, containing the target 211a, 211b and 211c at

those respective time points. The epochs or time points within the breathing

cycle P 270 may be chosen to substantially encompass the overall dynamic

range of the periodic motion 275. For example, in one embodiment, the time

points may include: a time point ti corresponding to a trough of the cycle P; a

time point h corresponding to a peak of the cycle P; and a third time point to

disposed at an intermediate location between the peak and the trough of the

cycle P 270. In other embodiments, the time points selected for taking the CT images may include fewer or more than the three time points to, ti, and te

described above. Accordingly, fewer or more than three CT images may be

used.

[0025] One way of providing a model for the continuous non-rigid

deformation of the anatomy as a function of the motion cycle involves

constructing a 4D mathematical model that morphs the CT image acquired at

one instant or time point in the motion cycle into another CT image acquired

at a subsequent instant or time point in the motion cycle. Any standard

software and/or algorithm that is known and may be commercially available

can be used to morph one image into another image, and to describe this in

terms of a mathematical model. The 4D mathematical model relates the 3D

locations of one or more reference structures (e.g., fiducials, skeletal structure,

etc.) with the 3D locations of the target, as a function of the instant in the

motion cycle. As such, if the deformation of the target is desired to be known

at an intermediate position between CT images e.g., ti.s, the 4D mathematical

model is applied to every voxel in one of the acquired CT images (e.g., CT

image 222) to map the deformation of the target in each voxel of the CT

image. However, such a 4D mathematical modeling may require a lot of

computing power and may be slower than desirable due the deformation

mapping of each voxel in the CT image. [0026] Described hereafter is a method and apparatus for tracking of a

changing (e.g., moving, deforming, etc.) target 211 using a four-dimensional

(4D) VOI, where the fourth dimension in the VOI architecture is a time

dimension. A 4D VOI may be represented as a set {Vo, Vi, . . . , Vk} where each

Vi is a 3D VOI representing a given point in time. In one embodiment, such a

4D VOI may be created by generating a 3D VOI (Vo) in a 3D image Io. In the

above example, the first and second dimensions of the VOI correspond to one

image slice in the 3D VOI and the third dimension corresponds to other slices

in the 3D VOI. It should also be noted that the definitions of the first three

dimensions could be the three axes of any coordinate system defined within

the 3D image space.

[0027] Then, given subsequent images {Io, Ii, . . . , Ik} taken at different

times (e.g., to model anatomical change due to a patient's respiratory cycle),

the corresponding VOIs could be formed by performing a non-rigid

registration to determine the deformation mapping of image L to image L,

and applying the deformation mapping to Vo to give an initial estimate of V.

In one embodiment, one or more additional refinement steps may be applied

to Vi, for example, a model-based refinement if Vi represents an anatomical

organ.

[0028] A VOI representing any given point in time may be formed by

direct interpolation between Vi and Vi+i, where i and i +1 are the VOIs immediately preceding and succeeding a desired point in time. This method

may be much faster than having to interpolate the underlying deformation

field itself and may result in a more efficient way of modeling tissue (e.g.,

organ) deformation over time. This is because the underlying deformation

field maps the space of image L to the space of image L+i, hence if the

deformation field is used directly to map between Vi and Vi+i the underlying

bit mask representation must be used (for example as described below in

relation to Figure 6). However, if interpolation is used between the contours

(e.g., of Figure IA and Figure IB) of Vi and Vi+i, it is possible to perform the

interpolation much faster. In other words, instead of having to look up values

of the deformation field for all the voxels within Vi and Vi+i, the points on the

contours of Vi and Vm may be directly interpolated to create a set of bounding

contours for the intermediate VOI.

[0029] In one embodiment, the 4D VOI architecture may be used with a

robotic based linear accelerator (LINAC) radiosurgery system (as discussed in

further detail below) to supplement or supplant the robot motion tracking

mechanisms that may already present in such a system. Because the

coordinate system in which each member of the VOI set is represented is

arbitrary, a coordinate system may be chosen that is invariant with respect to

the robotic-based LINAC. That is, the compensation for target change that

would have otherwise been determined by the treatment delivery system for the robotic controlled LINAC during the treatment delivery is already

predetermined by the treatment planning system using the 4D VOI

architecture.

[0030] Figure 3 illustrates one embodiment of a 4D VOI architecture,

where the fourth dimension in the VOI architecture is a time dimension. In

this embodiment, the VOI is represented using a three-tier structure (VOI-

contourslice-contour) in a UML graph with an example target. UML is a

graphical language for visualizing, specifying, constructing and documenting

artifacts of a software-intensive system. The UML offers a standard way to

write programming language statements, database schemas, and software

components. UML is well known in the art; accordingly, a more detailed

discussion is not provided herein. In one embodiment, each contour slice in

VOI architecture 200 may be restricted so that it contains only a simple (i.e.

closed boundary with no holes or intersections) contour. If this restriction is

present, contour slices may be created in non-adjacent slices, and

interpolation used to create the contours in the intermediate slices. In another

embodiment, VOI architecture 200 allows multiple contours to be defined for

each contour slice. In this case, VOIs with cavities, branches, and

unconnected bodies may be drawn. Alternatively, the methods described

herein may also be used with other tiered (e.g., 4 tier) VOI architectures. [0031] In this embodiment, VOI architecture 200 includes a contour

tier 210, a contour slice (L) tier 220, a VOI (Vi) tier 230. In this embodiment,

the 4D VOI architecture 200 may represented as a set {Vo, Vi, . . . , Vk) where

each of the Vi is a 3D VOI representing a given point in time. Only three Vi

(Vo 235a, Vi.5, 235b, and V>235c) are illustrated for ease of discussion purposes.

Continuing the example of Figure 2, the CT slices 221 and 223 are taken at

time points ti and t2, respectively, containing the target 211 in its state (e.g.,

position and deformation) at those respective time points 211b and 211c.

[0032] Figure 4 illustrates one embodiment of generating a VOI using a

4D VOI architecture. In this embodiment, the method may include receiving

at least two 3D images Ii 222 and h 223, step 405, and generating at least two

3D VOIs (e.g., Vi and V2) with the at least two corresponding 3D images Ii 222

and I2223, step 410. In step 420, a non-rigid registration of Vi 235b and V>

235c is performed. Registration may be performed using techniques known to

those of ordinary skill in the art; accordingly, a detailed description of

registration is not provided. In one embodiment, for example, a registration

technique as described in "Hybrid point-and-intensity based def ormable

registration for abdominal CT images", J. B. West, C. R. Maurer, Jr., and J. R.

Dooley, Proc. SPIE vol. 5747, pp.204-211, 2005, may be used. The output of

the registration of step 420 may be referred to as a deformation field. A

deformation field relating two images, A and B, is a set of vectors defined such that in any position x in image A, the corresponding element of the

deformation field, D(x), is a vector v such that the position (x+v) in image B

describes the same anatomical location as x in image A.

[0033] Then, in order to determine V1.5236 at a time in between ti and t2,

an interpolation may be performed, using the registration results of step 420,

on Vi 235b and V2235c, step 430. More generally, a VOI representing any

given point in time may be formed by direct interpolation between Vi and Vi+i,

where i and i +1 are the 3D VOIs immediately preceding and succeeding a

desired point in time. For example, a three-dimensional space deformation

model, as described in "D. Bechmann and N. Dubreuil. Animation through

space and time based on a space deformation model. The Journal of

Visualization and Computer Animation.4(3)165-184, 1993" may be used to

generate the interpolated VOI.

[0034] In one embodiment, one or more additional refinements, step

440, may be applied to V1.5236 such as a model-based refinement. With a

model-based refinement, a model of Vi.s236 is used to ensure that the

contours describing V1.5236 give a valid shape for the organ being described

by V1.5236. In one embodiment of a model-based refinement, the principal

modes of variation of the boundary of V1.5236 are stored as part of the model,

and the contours of V1.5236 are refined so that their principal modes of

variation are within given limits of those of the model. Refinement of a VOI may be performed either manually by the user (e.g., through a graphical user

interface) or through the use of an algorithm that operates on the VOI.

[0035] Once V1.5236 has been generated, it may be used to generate a

visualization of the VOI at time point ti.s, step 450. The generation of a

visualization from a VOI could be achieved by rendering the mask volume of

that VOI using volume rendering techniques as described in "Levoy, M., et.

al, Volume Rendering in Radiation Treatment Planning, Proc. First

Conference on Visualization in Biomedical Computing, IEEE Computer

Society Press, Atlanta, Georgia, May, 1990, pp.4-10/' or by directly rendering

the 3D geometrical structure of that VOI. Using 4D VOI architecture 200, the

visualization (e.g., images 222, 224 and 223) may be graphically displayed to a

user to animate changing structures (e.g., image 211) faster and more accurate

than may be possible when performing 4D mathematical modeling to map the

underlying deformation field itself.

[0036] Figure 5 illustrates one embodiment of medical diagnostic

imaging system in which features of the present invention may be

implemented. The medical diagnostic imaging system may be discussed

below at times in relation to CT imaging modality only for ease of

explanation. However, other imaging modalities may be used as previously

mentioned. [0037] Medical diagnostic imaging system 700 includes an imaging

source 710 to generate a beam (e.g., kilo voltage x-rays, mega voltage x-rays,

ultrasound, MRI, etc.) and an imager 720 to detect and receive the beam

generated by imaging source 710. In an alternative embodiment, system 700

may include two diagnostic X-ray sources and/or two corresponding image

detectors. For example, two x-ray sources may be nominally mounted

angularly apart (e.g., 90 degrees apart or 45 degree orthogonal angles) and

aimed through the patient toward the imager(s). A single large imager, or

multiple imagers, can be used that would be illuminated by each x-ray

imaging source. Alternatively, other numbers and configurations of imaging

sources and imagers may be used.

[0038] The imaging source 710 and the imager 720 are coupled to a

digital processing system 730 to control the imaging operation. Digital

processing system 730 includes a bus or other means 735 for transferring data

among components of digital processing system 730. Digital processing

system 730 also includes a processing device 740. Processing device 740 may

represent one or more general-purpose processors (e.g., a microprocessor),

special purpose processor such as a digital signal processor (DSP) or other

type of device such as a controller or field programmable gate array (FPGA).

Processing device 740 may be configured to execute the instructions for

performing the operations and steps discussed herein. In particular, processing device 740 may be configured to execute instructions to perform

the Boolean operations on the contour sets 241-244 to define VOI 231 as

discussed above with respect to Figure 3 and to generate a VOI mask volume

as discussed above with respect to Figure 5.

[0039] Digital processing system 730 may also include system memory

750 that may include a random access memory (RAM), or other dynamic

storage device, coupled to bus 735 for storing information and instructions to

be executed by processing device 740. System memory 750 also may be used

for storing temporary variables or other intermediate information during

execution of instructions by processing device 740. System memory 750 may

also include a read only memory (ROM) and/or other static storage device

coupled to bus 735 for storing static information and instructions for

processing device 740.

[0040] A storage device 760 represents one or more storage devices

(e.g., a magnetic disk drive or optical disk drive) coupled to bus 735 for

storing information and instructions. Storage device 760 may be used for

storing instructions for performing the steps discussed herein.

[0041] Digital processing system 730 may also be coupled to a display

device 770, such as a cathode ray tube (CRT) or liquid crystal display (LCD),

for displaying information (e.g., image slice, animation of the target using the

4D VOI, etc.) to the user. An input device 780, such as a keyboard, may be coupled to digital processing system 730 for communicating information

and/or command selections to processing device 740. One or more other user

input devices, such as a mouse, a trackball, or cursor direction keys for

communicating direction information and command selections to processing

device 740 and for controlling cursor movement on display 770 may also be

used.

[0042] It will be appreciated that the digital processing system 730

represents only one example of a system, which may have many different

configurations and architectures, and which may be employed with the

present invention. For example, some systems often have multiple buses,

such as a peripheral bus, a dedicated cache bus, etc.

[0043] One or more of the components of digital processing system 730

may form a treatment planning system. The treatment planning system may

share its database (e.g., stored in storage device 760) with a treatment delivery

system, so that it is not necessary to export from the treatment planning

system prior to treatment delivery. The treatment planning system may also

include MIRIT (Medical Image Review and Import Tool) to support DICOM

import (so images can be fused and targets delineated on different systems

and then imported into the treatment planning system for planning and dose

calculations), expanded image fusion capabilities that allow the user to treatment plan and view isodose distributions on any one of various imaging

modalities (e.g., MRI, CT, PET, etc.).

[0044] In one embodiment, the treatment delivery system may be an

image guided robotic based linear accelerator (LINAC) radiation treatment

(e.g., for performing radiosurgery) system, such as the CyberKnife® system

developed by Accuray, Inc. of California. In such a system, the LINAC is

mounted on the end of a robotic arm having multiple (e.g., 5 or more) degrees

of freedom in order to position the LINAC to irradiate the pathological

anatomy with beams delivered from many angles in an operating volume

(e.g., sphere) around the patient. Treatment may involve beam paths with a

single isocenter, multiple isocenters, or with a non-isocentric approach (i.e.,

the beams need only intersect with the pathological target volume and do not

necessarily converge on a single point, or isocenter, within the target).

Treatment can be delivered in either a single session (mono-fraction) or in a

small number of sessions (hypo-fractionation) as determined during

treatment planning. Treatment may also be delivered without the use of a

rigid external frame for performing registration of pre-operative position of

the target during treatment planning to the intra-operative delivery of the

radiation beams to the target according to the treatment plan.

[0045] Alternatively, another type of treatment delivery systems may

be used, for example, a gantry based (isocentric) intensity modulated radiotherapy (IMRT) system. In a gantry based system, a radiation source

(e.g., a LINAC) is mounted on the gantry in such a way that it rotates in a

plane corresponding to an axial slice of the patient. Radiation is then

delivered from several positions on the circular plane of rotation. In IMRT,

the shape of the radiation beam is defined by a multi-leaf collimator that

allows portions of the beam to be blocked, so that the remaining beam

incident on the patient has a pre-defined shape. The resulting system

generates arbitrarily shaped radiation beams that intersect each other at the

isocenter to deliver a dose distribution to the target. In IMRT planning, the

optimization algorithm selects subsets of the main beam and determines the

amount of time for which the subset of beams should be exposed, so that the

dose constraints are best met.

[0046] In other embodiments, yet other types of treatment delivery

systems may be used, for example, a stereotactic frame system such as the

GammaKnif e®, available from Elekta of Sweden. With such a system, the

optimization algorithm (also referred to as a sphere packing algorithm) of the

treatment plan determines the selection and dose weighting assigned to a

group of beams forming isocenters in order to best meet provided dose

constraints.

[0047] The 4D VOI architecture described herein may be used to

perform inverse planning. Inverse planning, in contrast to forward planning, allows the medical physicist to independently specify the minimum tumor

dose and the maximum dose to other healthy tissues, and lets the treatment

planning software select the direction, distance, and total number and energy

of the beams. Conventional treatment planning software packages are

designed to import 3-D images from a diagnostic imaging source, for

example, magnetic resonance imaging (MRI), positron emission tomography

(PET) scans, angiograms and computerized x-ray tomography (CT) scans.

These anatomical imaging modalities such as CT are able to provide an

accurate three-dimensional model of a volume of interest (e.g., skull or other

tumor bearing portion of the body) generated from a collection of CT slices

and, thereby, the volume requiring treatment can be visualized in three

dimensions.

[0048] During inverse planning, the VOI 230 is used to delineate

structures to be targeted or avoided with respect to the administered

radiation dose. That is, the radiation source is positioned in a sequence

calculated to localize the radiation dose into VOI 230 that as closely as

possible conforms to the target (e.g., pathological anatomy such as a tumor)

requiring treatment, while avoiding exposure of nearby healthy tissue. Once

the target (e.g., tumor) VOI has been defined, and the critical and soft tissue

volumes have been specified, the responsible radiation oncologist or medical

physicist specifies the minimum radiation dose to the target VOI and the maximum dose to normal and critical healthy tissue. The software then

produces the inverse treatment plan, relying on the positional capabilities of

radiation treatment system, to meet the min/max dose constraints of the

treatment plan.

[0049] The 4D VOI architecture 200 may be used to create a 4D mask

volume, as discussed in further detail below. Hence, beams may be enabled

or disabled depending the target's change over time. Although the change in

target during treatment delivery may be different than the change in the

target during treatment planning (e.g., due to differences in a patient's

respiration at those different times), certain gross changes may be assumed to

be similar. Accordingly, in one embodiment, the 4D VOI architecture 200 may

be used to supplement (or possibly supplant) the robot motion tracking

mechanisms that may otherwise be present in a robotic-based LINAC

radiation treatment system, with finer changes handled by dynamic tracking

capabilities of the treatment delivery system. Dynamic tracking is known in

the art; accordingly a detailed description is not provided. Alternatively, the

4D VOI architecture 200 may be used to supplant the robot motion tracking

mechanisms that may otherwise be present in a robotic-based LINAC

radiation treatment system.

[0050] Because the coordinate system in which each member of the set

is represented is arbitrary, a coordinate system that is invariant with respect to the robot. That is, the compensation for target motion is already taking into

account by the treatment planning software using the 4D VOI architecture

200.

[0051] Figure 6 illustrates one embodiment of a 4D mask volume. For

ease of discussion, 4D mask volume 600 is shown with two overlaid masks

635b and 636 on Vi 235b and Vi.s236, respectively, for times ti and ti.s,

respectively. The VOI mask volumes 635b and 636 are volume

representations of all user defined VOIs that are geometrically considered as a

cuboid composed of many small cuboids of the same size (i.e., the voxels). In

this embodiment, every voxel (e.g., voxels 650, 665, etc.) contains 32 bits.

Alternatively, other number of bit words may be used for a voxel. One bit, or

more, of a voxel (e.g., the ith bit) may be used to represent if the voxel is

covered by a VOI that is defined by the index of the bit. At every voxel

location (e.g., voxel 665), the bit value will be either a "1" or a "0" indicating

whether a particular voxel is part of the target. For example, a "1" bit value

may be used to indicate a voxel is contained within the VOI represented by

that mask position (as conceptually illustrated by the "1" for ith bit of voxel

665 of mask 635b). If, for example, the voxel bit is a "0" (as conceptually

illustrated by the "0" for the ith bit of voxel 667 of mask 636), the treatment

planning algorithm ignores the dose constraints for that corresponding voxel.

The VOI mask volume serves as an interface between the VOI structures and the rest of an imaging system's functions such as, for examples, a 4-D VOI

visualization and dose calculation in treatment planning. Using the 4D mask

volume 600, the radiation beams of a treatment delivery system may be

enabled or disabled based on a target's change in position.

[0052] The dose calculation process in the treatment planning

algorithm considers a set of beams that are directed at the target region 211.

In one embodiment, the treatment planning algorithm is used with a radiation

source that has a collimator that defines the width of the set of beams that is

produced. For each target 211, for example, the number of beams, their sizes

(e.g., as established by the collimator), their positions and orientations are

determined. Having defined the position, orientation, and size of the beams

to be used for planning, how much radiation should be delivered via each

beam is also determined. The total amount of radiation exiting the collimator

for one beam is defined in terms of Monitor Units (MU). Because the intensity

of the radiation source is constant, the MU is linearly related to the amount of

time for which the beam is enabled. The radiation dose absorbed (in units of

cGy) by tissue in the path of the beam is also linearly related to the MU. The

absorbed dose related to a beam is also affected by the collimator size of the

beam, the amount of material between the collimator and the calculation

point, the distance of the collimator from the calculation point, and the

distance of the calculation point from the central axis of the beam. [0053] Figure 7 illustrates a 2-dimensional perspective of radiation

beams of a radiation treatment system directed at a target region according to

a treatment plan. It should be noted that 3 beams are illustrated in Figure 7

only for ease of discussion and that an actual treatment plan may include

more, or fewer, than 3 beams. Furthermore, although the 3 beams appear to

intersect in the 2-dimensional perspective of Figure 7, the beams may not

intersect in their actual 3-dimensional space. The radiation beams need only

intersect with the target volume and do not necessarily converge on a single

point, or isocenter, within the target. In one embodiment, using 4D mask

volume 600 beams may be enabled or disabled based on the change in the

target over time.

[0054] Figure 8 is a flow chart illustrating one embodiment of

generating a treatment plan using a 4D mask volume. In one embodiment,

the treatment planning algorithm receives as input from a user, step 610, the

delineated target region 220 and any critical region 210 on one or more slices

of a CT image; and (2) dose constraints defining the minimum and maximum

doses for target region 220 and the maximum dose for the critical region 210.

It should be noted that additional dose constraints for additional regions may

also be provided. The delineation of the regions and the dose constraints

may be performed in any order. [0055] Then, the treatment planning algorithm performs beam

weighting of each one or more beams of the radiation treatment system to be

used in the treatment plan according to the inputs provided by the user

above. The user or the treatment planning algorithm assigns an arbitrary

weighting to each of one or more beams (e.g., beam 1, beam 2, beam 3 of

Figure 7) of the radiation treatment system. This weighting may be

determined using an algorithm designed to give a suitable "start point" for

planning, may be randomly chosen, or may simply be a constant weighting

for each beam.

[0056] In step 630, the 4D VOI is generated and, in step 640, the 4D

mask volume is generated with the methods discussed above in relation to

Figures 3 and 4. If a voxel bit from a 4D volume mask 600 is a "0", the

planning algorithm ignores the dose constraints for that corresponding dose

voxel for the particular point in time VOI. However, if a voxel bit from dose

contour mask 400 has a "\" bit value for a particular point in time VOI, then

determine whether any penalties should be accessed when performing beam

weighting based on the dose constraints for that dose voxel, step 650. In one

embodiment, in order to reduce dose to a given sensitive organ to minimal

levels, the 4D volume mask may be used so that if the 3D VOI at any of the

time points has a "1" bit value at any position intersected by a beam, that

beam is automatically set to have zero MU in the final plan. [0057] In one embodiment, the following algorithm may be used to

perform beam weighting. In this embodiment, to begin the beam weighting,

step 660, an assumption may be made that the size and trajectory of the beam

set has been defined. Let the beam set be {Bi; 1 < i < N}, where N » 500. Beam

1, beam 2, and beam 3 illustrate in Figure 7 have a respective weight 1, weight

2, weight 3 (i.e., a number of MU assigned to the beam, or how long a beam

will be maintained on) associated with it. The weight in MU of each beam is

designated by wi. The delineated regions are represented as objects Tj

(derived from the 4D mask volume 600), with corresponding minimum and

maximum allowed dose minj and maxj, and critical structures (critical region

210) Q, with corresponding maxj defined. Each region has an integer priority

PJ e [0,100] defining the relative importance of the dose constraints applied to

that region. For each beam, a 4D dose value mask is created. The 4D dose

value mask may be regarded as a set of 3D dose value masks, in the same way

that a 4D VOI is a set of 3D VOIs. Each 3D dose value mask provides a linked

list of floating point values and positions cL(r,t) at a given point in time, where

r is the position within the dose calculation volume, and di is the dose in cGy

delivered to r by beam i when wi is set to unity, and t is the time, represented

as a position in the respiratory cycle. Thus, the total dose at r, at position t in

the respiratory cycle is given by:

Hence the total dose at r, summed over the entire respiratory cycle is,

D(r)= ∑D(r,t) (1),

and the total dose for beam i, summed over the entire respiratory cycle is,

[0058] For each Bi, we define a beam value υ i, where

_ ∑ , ∑ r e T, d , {r) (2)

∑ d , ( r )

[0059] The beam value is the ratio of dose delivered into target region

220 to total dose delivered. To define the initial set of wi for optimization, we

set W₁ = υ_t,\/i . The maximum dose within the dose calculation volume, Dmax,

is computed and the beam weights renormalized so that the new maximum

dose is equal to the largest of the maximum dose constraints, maxj. Hence,

this provides:

w, = ϋ,. sup(max_y)/-D_max. (3)

[0060] At one iteration of the treatment planning algorithm, the

optimization process looks at all of the dose values in the dose volume and

determine if the target region 211 and a critical region 610 are within the dose constraints. For example, suppose the dose in the target region 211 is

specified to be equal to or greater than 2000 cGy and less than or equal to 2500

cGy. Suppose, the current dose value at grid location for voxel 665 of Figure 6

is 1800 cGy, then the optimization process determines that, at the current

beam weightings, the dose value at voxel 665 is 200 cGy short in order to

satisfy the treatment plan constraints.

[0061] Given the initial weights, the optimization process then alters

the beam weights so that the treatment solution is closer to meeting the

provided dose constraints. First, a set of Aw₁, the amount by which each

beam weight may be changed, is defined:

[0062] where s is the search resolution, having an initial value of 1.

[0063] The optimization process iterates through one or more of the

beams and for each of the beams, if a beam weight is increased or decreased

by a certain amount, determines the resulting dose distribution from such a

change (i.e., how such a change alters the amount of violation of the treatment

plan constraints). For example, an increase in one or more of the beam

weights may typically help in achieving the constraint in the target (e.g.,

tumor) region but, depending on the location of the beam, it may also hurt in the critical region due to a possible resulting increase of dose above the

maximum value in the critical region.

[0064] The optimization process traverses the volume of interest, adds

up all the penalties that are incurred by the increase in a beam weight, adds

up all the penalties that are incurred by the decreasing the beam weight (e.g.,

under-dosing the target region), and then provides a result. In one

embodiment, a multiplier may be used with each penalty to stress the

importance of one constraint (e.g., minimum dose value in the target region)

versus another constraint (e.g., maximum dose value in the target region).

For example, it may more important to achieve a minimum dose value than to

stay under the maximum dose value in the target region.

[0065] The optimization process then updates the dose and goes on to

the next beam and repeats the process until it has made its way through the

beam set. The optimization process then reaches a stage where it has looked

at all of the different weights for each of the beams at the different dose levels

and selects the beam weight that provides the optimal resulting dose values in

both the target region and critical region.

[0066] In one embodiment, an iterative optimization process is used as

follows: Iterate over the beams in decreasing order of υ, . For each beam Bj, calculate Pj and PJ , the relative penalties for respectively increasing or

decreasing Wj, that are defined as:

Aw_jd_j (r) and

[0067] where Vi is the volume in mm³ of region i. Hence, the penalty

for this beam is the sum of the additional amount of over-dosing and under¬

dosing that would be created by the change in the beam, weighted by the

priorities of the different regions and normalized according to the region

volumes. If PJ and Pj are both positive, Wj is kept the same, otherwise

change W_j = W_j ± Δw. according to whichever of PJ and Pj is smaller. If the

previous iteration moved Wj in the same direction as this iteration, the

following is set:

Aw _j - Aw _j + Δ⁽⁰⁾w _;. , /Ej\

else set:

Aw _j =A⁽⁰⁾W_j . (6) [0068] The change in dose according to Δw_; is computed and applied

to the dose volume before the optimization process moves on to a next beam,

because a correct decision on how to change the beam weight assumes an up-

to-date view of the dose including change sin previous wi. If all Wj remained

unchanged by the current iteration, s is reduced by a factor of 2.

[0069] In an alternative embodiment, the optimization algorithm may

perform convex optimization via, for example, the Simplex algorithm, in an

attempt to find an MU setting for all beams so that the dose constraints are

nowhere violated. The Simplex algorithm is known in the art; accordingly, a

detailed description is not provided. Alternatively, other iterative and non-

iterative optimization algorithms may be used.

[0070] The 4D VOI architecture 200 may also be used with a mixed

planning in which part of the treatment dose is generated by an isocenter

placed using forward planning and part generated by individual beams

during inverse planning.

[0071] It should be noted that although discussed at times herein in

regards to radiation treatment, the methods and apparatus described herein

are not limited for use solely in treatment planning but may also be used

independently for other applications, such as simulation and animation of

object changes (e.g., deformation) over time. In alternative embodiments, the methods and apparatus herein may be used outside of the medical technology

field, such as non-destructive testing of materials (e.g., motor blocks in the

automotive industry and drill cores in the petroleum industry) and seismic

surveying.

[0072] In the foregoing specification, the invention has been described

with reference to specific exemplary embodiments thereof. It will, however,

be evident that various modifications and changes may be made thereto

without departing from the broader spirit and scope of the invention as set

forth in the appended claims. The specification and drawings are,

accordingly, to be regarded in an illustrative sense rather than a restrictive

sense.

Claims

CLAIMSWhat is claimed is:

1. A volume of interest (VOI) architecture comprising a fourth dimension, where the fourth dimension is a time dimension.

2. The VOI architecture of claim 1, further comprising: first, second and third dimensions comprising three space dimensions within an image space.

3. The VOI architecture of claim 1, further comprising a plurality of VOIs, wherein each of the plurality of VOIs is a three-dimensional VOI representing a given point in time.

4. The VOI architecture of claim 3, wherein the three-dimensional VOI comprises three space dimensions within an image space.

5. A method, comprising: generating a first VOI including a target at a first point in time; generating a second VOI including the target at a second point in time; registering the first VOI with the second VOI; and interpolating a third VOI including the target at a third point in time based on the registering of the first VOI with the second VOI, the third point in time being between the first and second points in time.

6. The method of claim 5, further comprising generating a visualization of the third VOI.

7. The method of claim 6, wherein generating the visualization comprises rendering a mask volume of the third VOI.

8. The method of claim 7, wherein rendering comprises volume rendering.

9. The method of claim 7, wherein rendering comprises direct rendering of a three-dimensional geometrical structure of the third VOI.

10. The method of claim 5, further comprising: receiving a first image having the target at the first point in time; and receiving a second image having the target at the second point in time.

11. The method of claim 6, further comprising refining the third VOI.

12. The method of claim 11, wherein refining comprises a model-based refinement.

13. The method of claim 12, wherein the third VOI is changed with respect to at least one of the first and second VOIs.

14. The method of claim 5, wherein the target at the first time point is changed with respect to the target at the second time point based on motion adjacent the target.

15. The method of claim 14, wherein the motion comprises a periodic motion characterized by a cycle.

16. The method of claim 6, wherein the first and second images are acquired in an anatomical imaging modality.

17. The method of claim 16, wherein the anatomical imaging modality is computed tomography.

18. An article of manufacture, comprising a machine-accessible medium including data that, when accessed by a machine, cause the machine to perform operations comprising: generating a first VOI including a target at a first point in time; generating a second VOI including the target at a second point in time; registering the first VOI with the second VOI; and interpolating a third VOI including the target at a third point in time based on the registering, the third point in time being between the first and second points in time.

19. The article of manufacture of claim 18, wherein the instructions further cause the processor to perform the following comprising generating a visualization of the third VOI.

20. The article of manufacture of claim 19, wherein the instructions further cause the processor to perform the following comprising refining the third VOI.

21. The article of manufacture of claim 19, wherein generating the visualization comprises rendering a mask volume of the third VOI.

22. An apparatus, comprising: a storage device to store first and second images; and a processor coupled to the storage device to receive the first and second images, the processor to generate a first VOI including a target at a first point in time and a second VOI including the target at a second point in time, the processor further to register the first VOI with the second VOI and interpolate a third VOI including the target at a third point in time based on the registering, the third point in time being between the first and second points in time.

23. The apparatus of claim 22, wherein the processor is further configured to generate a visualization of the third VOI.

24. The apparatus of claim 23, wherein the processor is further configured to refine the third VOI.

25. The apparatus of claim 23, wherein the processor is further configured to render a mask volume of the third VOI.

26. The apparatus of claim 23, further comprising an imager coupled to the storage device, the imager to generate the first and second images.

27. A method of tracking a changing target within an anatomical region to deliver radiation to the target during motion of the anatomical region, the method, comprising: determining, at a time of treatment planning, a plurality of locations of the target over time using a four-dimensional volume of interest (VOI) having a fourth dimension being a time dimension; and determining, at a time of treatment planning using the four- dimensional VOI, one or more radiation beam trajectories to be delivered to the target within the moving anatomical region at a time of treatment delivery.

28. The method of claim 27, wherein determining comprises performing at least one of enabling and disabling of the one or more radiation beams depending whether the one or more radiation beams intersect the target at the plurality of locations of the target determined using the four-dimensional VOI.

29. The method of claim 27, further comprising generating a four- dimensional mask volume using the four-dimensional VOI, and wherein determining is performed using the four-dimensional mask volume.

30. The method of claim 29, wherein determining comprises: setting a bit in the four-dimensional mask volume to have a value indicating that a corresponding voxel is part of the target at a point in time; and setting a beam, which is to intersect the voxel at the point in time, to have a zero weight.

31. The method of claim 28, further comprising generating the four- dimensional VOI, wherein generating comprises: generating a first VOI including the target at a first point in time; generating a second VOI including the target at a second point in time; registering the first VOI with the second VOI; and interpolating a third VOI including the target at a third point in time based on the registering of the first VOI with the second VOI, the third point in time being between the first and second points in time.

32. The method of claim 31, wherein the target at the first time point is changed with respect to the target at the second time point based on the motion of the anatomical region.

33. The method of claim 32, wherein the motion comprises a periodic motion characterized by a cycle.

34. An apparatus, comprising: means for directly interpolating between a first contour of a first volume of interest (VOI) and a second contour of a second VOI using one or more points on the first and second contours; and means for generating an intermediate contour of an intermediate VOI based on the interpolating.

35. The apparatus of claim 34, further comprising means for generating a visualization of the intermediate VOI.

36. The apparatus of claim 34, further comprising means for refining the third VOI.

37. The apparatus of claim 34, wherein the means for directly interpolating comprises a four-dimensional VOI having a fourth dimension being a time dimension, and wherein the apparatus further comprises: means for determining, at a time of treatment planning, a plurality of locations of a target over time using the four-dimensional volume of interest (VOI); and means for determining, at a time of treatment planning using the four- dimensional VOI, one or more radiation beam trajectories to be delivered to the target within the moving anatomical region at a time of treatment delivery.