Chapter 6

Lesson 22

Flat Panel a-Si EPID Image Processing Digital vs Film EPI Library Prostate TargetingChapter 1 New Developments

Electron Beam VerificationLesson 22 MegaVoltageConeBeam CT

Lesson 22

Mega Voltage Cone Beam Reconstruction *



Info Box

Video HEAD presentation


Portal images, taken through the radiation aperture at time of treatment allow the physician to visualize and quantify the position of anatomical structures within the radiation field. Verification of the proper delivered dose in terms of field shape and localization with respect to patient anatomy is necessary to assure the safety and quality of complex radiation therapy treatments such as intensity-modulated radiotherapy (IMRT). This role of Electronic Portal Imaging Devices (EPID) for on-line two-dimensional treatment verification is well established. However, the use of 3D conformal therapy or IMRT requires precise 3D dose verification. Recently, we have shown that the new generation of EPIDs based on an amorphous-silicon active matrix flat panel imager can acquire multiple image projections with a quality sufficient to perform 3D Mega-voltage CT (MVCT) cone beam reconstruction with a small patient dose.  This 3D reconstructed MVCT can then be correlated with the planning CT for patient alignment purposes. This provides a direct comparison of the planned dose distributions based on the planning CT, to the actual patient anatomy, moments before dose delivery.

Proof of Concept

We have performed a proof-of-concept to demonstrate the feasibility of 3D cone beam reconstruction using a flat-panel detector. The breakthrough comes from the combination of the ability to tune the linear accelerator to deliver a tiny amount of radiation and the capacity to acquire images with a small fraction of the dose required by earlier generations of detectors. Three key requirements must be fulfilled to ensure clinical applicability. They are:


Acquisition of a large number of images with a very small exposure to the patient,


Patient positioning assessment, including image acquisition, 3D
reconstruction and correlation, in a short time,


Good resolution on anatomical information to allow set-up verification.

An amorphous silicon flat panel electronic portal imaging device (a-Si EPID) was used for image acquisition. The 41 x 41 cm active detector area (1024 x 1024 pixels) was located 133 cm from the source. For each gantry angle, the portal images (PI) were acquired by using a beam pulse trigger mode: the beam signal is integrated by the panel during irradiation and read-out after the beam is turned off. The Primus linac parameters were adjusted to produce low beam intensity able to deliver a small fraction of an M.U.

The head section of an anthropomorphic phantom (Rando) was placed at isocenter and images were acquired at different gantry angles in steps of two degrees for a complete rotation (180 images). The set of portal images was then used to perform the 3D cone beam reconstruction. The 1024 x 1024 images were re-binned to 512 x 512 images. A filtered backprojection type reconstruction was used for the cone beam reconstruction and a 256 cube was reconstructed. A Shepp Logan filter with a kernel of size 512 was used. Fiducial markers were placed on the accessory tray of the linac, and were used to correct the portal images for the lateral movement of the imaging panel. The placement of the phantom in the field of view of the EPID left bright areas in each image that correspond to x-ray transmission through air. The air intensity was used to scale/correct the EPID pixel intensities. The resulting MVCT slices could then be correlated with the regular CT scan obtained for dose planning purpose.

Each single image was exposed with 0.08 Monitor Unit (MU), equivalent to 0.08 cGy at isocenter. Despite the very small amount of radiation used for each image, the image contains sufficient information. The complete acquisition of the 180 images was performed with an isocenter dose of less than 15 cGy. The correlation of the MVCT slices with the planning CT slices demonstrated a rigid body transformation consisting of three translations and three rotations. Image blending between the two CT scans was also demonstrated. It should be noticed that the 3D reconstruction was performed with minimum correction for detector sagging or relative image intensities. MVCT cone beam reconstruction can can be performed with less than 15 MU and provide anatomical definition adequate for patient positioning verification.



Dose-Guided Radiation Therapy with Megavoltage Cone-Beam CT

J.Pouliot 1, J.Chen1, O.Morin1, M.Svatos2, F.Ghelmansarai2, M.Mitschke1, M.Aubin1, P.Xia1, C.Chuang1, K.Bucci1, M.Roach III1, P.Hernandez2, Z.Zheng2, D.Hristov2, A.Bani-Hashemi2 and L.Verhey1

1- Department of Radiation Oncology, University of California San Francisco, 1300 Divisadero Street, San Francsico, USA. 2- Siemens Oncology Care Systems, Concord, California


The ultimate strategy of image-guided radiation therapy extends beyond the geometrical verification of the patient anatomy. It would allow the verification and correction of the position of the patient 3D anatomy as well as the comparison of the dose delivered with the intended dose distribution. In this presentation, we describe the main steps of a Dose-Guided Radiation Therapy. We will show that Megavoltage Cone Beam CT (MV CBCT) can be acquired with a low enough patient exposure to make it practical for clinical use and with sufficient quality to register 3D MV CBCT images to kilovoltage planning CT for patient alignment purposes. Then we describe how the conebeam algorithm and the calibrated flat panel EPID are used to reconstruct the dose delivered at treatment time and to provide the information for the verification and correction of the planned dose distribution.

The flat panel EPID

The a-Si flat panel EPID used for this experiment has an active area of 41 x 41 cm2, and contains 1024 x1024 elements of 0.4 mm pitch [Perkin-Elmer RID-1640]. The fastest frame rate is 3.5 frames per second and the output data is 16-bit. For each angle, the portal images (PI) were acquired by using a beam pulse trigger mode.

Low Output Beam

MV CBCT experiments were performed with the 6 MV beam of a clinical Siemens PRIMUS accelerator calibrated to deliver 1cGy/MU under standard calibration conditions. The linac parameters and firmware were adjusted to allow the delivery of a small fraction of a Monitor Unit (MU) per image. Three specially designed firmware proms were installed to window the dose pulse rate so that only a fraction of the number of pulses were delivered. Thus an angular dose as low as 0.01 MU per degree of gantry rotation could be achieved. Several consecutive measurements of the beam dose rate with an ion chamber showed a beam dose reproducibility of 3% in this special low output mode.

Low MU Conebeam CT of a Head & Neck IMRT patient

An IMRT Head and Neck patient being treated at UCSF gave informed consent to undergo a cone beam CT. The patient was setup on the Linac equipped with the flat panel detector in the patientŐs-regular treatment position using a head holder and an Orfit mask. A cone beam acquisition was performed with a 180 degree rotation with one projection image per degree exposed to 0.08 M.U. for a total dose of 15 cGy delivered to the patient. Common anatomical landmarks used for the alignment of head and neck patients, such as the sphennoid sinus, nasopharynx, external auditory canals, mastoid air cells, and spinal canal are clearly visible.

2D and 3D MV CBCT to kVCT Image registration

The MV CBCT provides a 3D patient volume that can be very tightly aligned to the planning CT, referred to as kV CT. The MV CBCT image coordinate system is directly related to the linac isocenter coordinates without the use of an intermediate frame of reference. Therefore, MV CBCT provides the patient anatomy in the actual treatment position, relative to the treatment isocenter, moments before the dose delivery, allowing verification and correction of the patient position. Two methods of correlating a MV CBCT image set to a conventional kVCT were used, one manual and one automatic. For each MV CBCT experiment, a kVCT scan was first acquired during simulation using a Siemens Somatom Emotion CT scanner.

The registration process is a series of rigid body transformations consisting of translations and rotations. A commercial software tool available in the SyngoTM imaging workstation, is used to perform the 2D registration. The software displays slices of the MV CBCT from three orthogonal planes: transverse, sagital and coronal. The corresponding slices of the planning CT are superimposed. The system can display each CT with a different color scheme and the transparency levels can be adjusted to visualize either CT, or a blended image of both CTs superimposed.

Figure 1: MV Conebeam CT with Planning CT Registration of an IMRT Head&Neck patient. Left: before registration. Center: After registration. Right: Sagital MV conebeam image superimposed on planning CT image and target contours.

Dose Reconstruction from MV Conebeam CT

Prior to patient imaging, a one-time calibration is performed using a unique phantom. A set of geometrical transformation matrices for each projection is derived to convert the 2D image back to the known 3D phantom. These transformation matrices used for cone beam reconstruction are also ideally suited to reconstruct the dose distribution from portal images. They provide a unique pixel (EPID) to voxel (reconstruction volume) relation with complete geometrical information, allowing the exit dose measured by the panel to be back-projected at the correct location. For dosimetry, in order to calibrate the cone beam image and provide the attenuation coefficients, an additional calibration is performed consisting of a cone beam image of a CT phantom with various density inserts. Portal images of the treatment beams are acquired to measure the delivered fluence. The panel needs to be calibrated for exit dose measurements.

The delivered dose to a sheep head was reconstructed by back-projecting the energy fluence from each treatment beam into the reconstruction volume using the transformation matrices and spreading the energy released using a 3D dose spread kernel.

Figure 2: Delivered dose reconstructed for one to 5 beams.

By registration of the MV CBCT image with the planning CT, the actual and desired isodose distributions can be compared. A visual inspection shows a general good agreement between the isodose contour plots of the reconstructed and planned dose distributions. A more quantitative comparison is in progress.

Dose-Guided Radiation Therapy

The Figure below presents the main steps, and simultaneously provides our definition of Dose Guided Radiation Therapy using Megavoltage Cone-Beam Computerized Tomography (MV Cone Beam CT). After the patient is aligned on the treatment couch, a complete set of portal image projections is acquired (1) by rotating the gantry 190° around the patient. This part is completed in approximately 45 seconds. As soon as the first image is acquired, the cone beam reconstruction is started (2). Each image is pre-processed and the 3D MVCT image is produced. Verification of patient positioning can then be performed by registering (3) the 3D MVCT image with the planning CT. This can be accomplished manually by matching the 2D images on 3 orthogonal views or automatically by using a mutual information algorithm. Patient mis-alignment is then compensated for by modifying the couch position (4). The patient is then correctly aligned for treatment delivery with IMRT or 3D conformal techniques (5). For each beam angle (or each beam segment for IMRT), a portal image is acquired and the dose reconstruction is initiated using the treatment beam and the transformation matrix (6). By comparing the delivered and planned dose distributions (7), under-dosed regions inside the target volume or overdosed regions within critical organs can be detected early in the course of treatment. An adapted IMRT plan can be developed (8) to compensate for the dose deficit inside the tumor volume and to reduce the excess dose in the critical organs in the remaining treatment fractions.

Figure 3: Main Steps of Dose-Guided Radiation Therapy


A strategy for the clinical implementation of low exposure MV Conebeam CT on a conventional linac has been described. The same detector is used to determine the 3D patient anatomy and to derive the delivere dose in the actual treatment position. The anatomy is obtained with MV Conebeam imaging while the dose reconstruction is performed from the portal images acquired during the treatment. In this scenario, future treatment fractions could be modified based on the dose already delivered. This dose-guided radiation therapy is a particular form of adaptive radiation therapy where dosimetric considerations would constitute the basis of treatment modification.

Other site and publication on Mega Voltage Cone Beam CT

bOther sitebbbbbbb

Back Lesson 21 Next Conclusion