Kalman filter framework to estimate 3D intrafraction motion from 2D projection
An iterative Kalman Filter method was developed to address the need for estimating randomly moving targets during cancer radiotherapy on a standard equipped linear accelerator. Extensive evaluation of this method using different treatment scenarios shows sub-mm accuracy and precision. In addition, the system and method allows the target (or surrogates of the target) to be monitored without the need of a learning arc, reducing additional imaging dose to the patient. In addition, the method and system performs robustly against imaging and segmentation noise.
1 . A method of estimating 3D target position during radiotherapy, the method comprising:
acquiring one or more two-dimensional (2D) image positions of one or more treatment targets or one or more surrogates from a (kilovoltage) kV imager on a linear accelerator;
implementing a Kalman filter framework using one or more computers to estimate a three-dimensional (3D) motion of the one or more treatment targets or the one or more surrogates from 2D image projection measurements in real-time during a radiotherapy treatment, wherein the Kalman filter is an iterative framework that allows for estimation of the measurement and estimation of a process error to be re-estimated from current and past measurements, wherein a population covariance of prostate motion is defined as:
Q
0
=
Q
k
=
(
0.3136
0.0114
-
0.0775
0.0114
1.882
1.5051
-
0.0775
1.5051
2.4733
)
;
updating a target motion covariance with every image frame based on a last observed position of the one or more treatment targets or the one or more surrogates on a previous kV image; and
outputting to a display during the radiotherapy treatment the updated target motion.
2 . The method of estimating 3D target position during radiotherapy according to claim 1 , wherein the method is implemented during a prostate cancer treatment.
3 . The method of estimating 3D target position during radiotherapy according to claim 2 , further comprising:
initializing an initial prostrate position of a patient on a day of treatment using image-guided radiation therapy;
estimating a current position of the prostate based on information of motion distribution of prostate motion up to a last 2D image frame;
estimating a current measurement error based on a previous distribution and expected distribution of prostate motion;
re-estimating the current position of the prostate given current projection information; and
estimating a posteriori covariance to be used for future prediction and update.
4 . The method of estimating 3D target position during radiotherapy according to claim 1 , further comprising:
initializing an initial position of the one or more treatment targets of a patient on a day of treatment using image-guided radiation therapy;
estimating a current position of the one or more treatment targets based on information of motion distribution of target motion up to a last 2D image frame;
estimating a current measurement error based on a previous distribution and expected distribution of target motion;
re-estimating the current position of the one or more treatment targets given current projection information; and
estimating a posteriori covariance to be used for future prediction and update.
5 . A method of monitoring movement of an organ or portion of an organ or one or more surrogates of the organ during irradiation, comprising:
directing radiation energy at, at least a portion of an organ in a body part of a human or an animal subject;
imaging multiple two dimensional images of the organ from varying positions and varying angles relative to the body part;
digitally processing at least a plurality of the multiple two dimensional images using a one or more computers with a software application running a Kalman filter algorithm, wherein the digital processing is initialized with a population covariance matrix of prostate motion defined as
Q
0
=
Q
k
=
(
0.3136
0.0114
-
0.0775
0.0114
1.882
1.5051
-
0.0775
1.5051
2.4733
)
;
and
displaying estimated three dimensional motion of the organ or portion of the organ in the body part based on output from the digital processing.
6 . The method of claim 5 wherein the multiple two dimensional images are obtained using a linear accelerator gantry mounted kilovoltage x-ray imager system.
7 . The method of claim 5 , further comprising:
initializing the digital processing with a population covariance matrix so as to avoid a learning period.
8 . The method of claim 5 wherein the organ is a prostate of a human subject, and wherein the body part is a pelvis of the human.