Cone‐beam CT (CBCT) has been widely used in image guided radiation therapy (IGRT) to acquire updated volumetric anatomical information before treatment fractions for accurate patient alignment purpose. However the excessive x‐ray imaging doses (a few cGy per scan) delivered to patients from serial CBCT scans raise a clinical concern in most IGRT procedures. This fact has greatly limited exploitation of IGRT's maximal potential. The imaging dose can be effectively reduced by reducing the number of x‐ray projections and/or lowering mAs levels in a CBCT scan. The image quality reconstructed from conventional FDK‐type algorithms however will be highly degraded. Recently total variation (TV) method has demonstrated its ability to reconstruct CBCT images from a few number of noisy x‐ray projections. Nonetheless such a method can hardly be applied in real clinical environments due to its long computational time (a few hours). It is highly desirable to develop a fast reconstruction scheme to obtain high quality CBCT images from undersampled and noisy projection data so as to lower the imaging dose.Utilizing GPU to speed up the computationally intensive tasks in CBCT reconstruction problems has drawn a lot of attention recently. In this talk GPU‐based CBCT reconstruction algorithms will be reviewed with an emphasis on an iterative CBCT reconstruction algorithm via TV regularization. We have recently developed a GPU‐friendly version of the forward‐backward splitting algorithm to solve the TV‐based reconstruction problem. Multi‐grid technique is also employed. It is found that 40 x‐ray projections are sufficient to reconstruct CBCT images with satisfactory quality for IGRT patient alignment purpose. Phantom studies indicate that CBCT images can be successfully reconstructed with our algorithm under as low as 0.1 mAs/projection level. Comparing with currently widely used full‐fan head‐and‐neck scanning protocol of about 360 projections with 0.4 mAs/projection it is estimated that an overall 36 times dose reduction has been achieved. Moreover the reconstruction time is about 130 sec on an NVIDIA Tesla C1060 GPU card which is estimated ∼100 times faster than similar iterative reconstruction approaches. The high computational efficiency and satisfactory image quality make the iterative low dose CBCT reconstruction approach feasible in real clinical environments.Learning Objectives:1. Understand basic concepts of GPU‐based CBCT reconstruction.2. Understand main challenges in GPU‐based iterative CBCT reconstruction approach and how an iterative CBCT reconstruction problem is solved on GPU.Conflict of Interest: The research presented here is partially supported by NVIDIA.
Purpose: To develop a method to track lung tumors in rotational cone beam projections during rotational radiotherapy and cone beam CT scanning. Method and Materials: A multiple template based tracking algorithm was developed and used to track tumors in rotational cone beam projections. Templates were generated by creating DRRs of ten phases of 4DCT. These templates were generated for a sequential set of angles matching the projections of a CBCT scan. The position of the tumor in each template was derived from contours drawn on 4DCT. Shifting of the templates was used to allow for a greater tumor motion ranges. The mutual information between projections and templates was computed and used as one parameter in a probability function used to track tumor position. Tumor distance traveled and phase change between successive projections were also incorporated into the probability function. This output was compared to physician specified tumor locations on each cone beam projection. In addition to a patient study, the method was tested on a respiratory motion phantom programmed to exhibit sinusoidal motion in the SI direction. Results: In the phantom study, the SI motion of the tumor was tracked with a mean absolute error (MAE) ranging from 1.2mm to 1.6mm and a 95th percentile absolute error (P95) ranging from 3.2mm to 3.7mm. For the patient study, the SI motion was tracked with MAE ranging from 1.7mm to 1.9mm and P95 ranging from 3.4mm to 3.9mm. Conclusion: The algorithm has demonstrated the feasibility of tracking tumors in rotational x‐ray images. Further development is needed in order to achieve accuracy similar to that of fixed gantry fluoroscopic tracking.
Purpose: To quantify the correlation between patient surface motion and internal tumor motion. Method and Materials: Thoracic 4D‐CT scans are used for the study. The CT image at the end of inhale is set as the target image. On each CT slice of the target image, 201 points are evenly placed on the skin contour (excluding the patient's back). The other nine CT images, corresponding to nine different respiratory phases, are set as moving images. An optical flow deformable image registration algorithm is used to map the moving images to the target image. Then the surface points in the target image are transferred to the moving images using the deformation vector field. The motion trajectory of each surface point is quantified for the ten respiratory phases, and the correlation between each surface point motion and internal tumor movement in the superior‐inferior direction is computed. Results: Color maps of the surface point motion amplitude and correlation with internal tumor motion have been derived for each patient. Color maps highlight the high‐correlation and high‐amplitude motion regions on the surface. It has been observed that 88% of body surface has a correlation of 0.9 or higher with internal tumor motion in the SI direction but only 17% of surface points have motion amplitude bigger than 2mm. Conclusion: Color maps of the surface point motion amplitude and correlation with internal tumor motion have been derived for each patient from a 4D‐CT scan. These color maps can be used to identify patient specific surface regions of interest to be used as surrogates for external gating.
Purpose: The actual delivered dose to a moving tumor can deviate from prescribed dose not only during each fraction, but during each IMRT field. This deviation can be anticipated and incorporated into a treatment plan if the tumor specific motion probability density function (PDF) is accurately identified. In this study, a novel technique in determining the actual, tumor specific PDF of a moving gross tumor volume (GTV) is described and confirmed through phantom experiments using a quantitative approach of 4DCT imaging. Methods and Materials: We hypothesize that a PDF of the GTV can be obtained through weighted sums of phase‐sorted 4DCT images. Experimental validation is performed using a lung phantom attached to a programmable motion platform. The ground truth motion PDF of the targets was calculated through convolution of the motion pattern and the GTV dimension along the direction of motion. These PDF convolutions are compared to the normalized, reconstructed CT numbers obtained from averaging ten phase sorted 4DCT images after background subtraction. Results: The PDF reconstructed from averaged, weighted sets of 4DCT images was almost identical to the ground truth PDF in all comparisons made. Through sums of phase sorted 4DCT images, the resulting CT number represents a relative probability of finding some portion of the GTV in each geometric location (or voxel). In physical phantom reconstructions, the Pearson correlation coefficient between CT‐based PDF and the ground truth PDF was greater than .97 in all cases considered. An actual patient PDF was also evaluated in frontal and sagittal planes. Conclusions: Combining phase sorted CT images is an effective method in obtaining the actual tumor PDF as seen by an IMRT field. With the patient specific PDF, treatment planning to ensure proper target coverage can be achieved through intra‐field modulation, dose rate, and monitoring initial beam‐on phase.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.