Douglas J. Moseley1, Michael B. Sharpe1,2, Jeffrey H. Siewerdsen1,2, Graham A. Wilson1, Stephen M. Ansell1, Tara Haycocks1,2, Thomas G. Purdie2, Mohammad Islam1,2 and David A. Jaffray1,2
1Princess Margaret Hospital/Ontario Cancer Institute, University Health Network, Toronto, Ontario, Canada
2Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
Advanced image-guidance techniques promise to revolutionize external beam radiation therapy of cancer through dramatic improvement in geometric precision and the possibility for novel dose escalation and fractionation techniques. A milestone critical to the adoption of image-guided radiotherapy into the clinical setting is the streamlining and optimization of the image acquisition and feedback to the delivery system. With an x-ray volume imaging system (flat-panel cone-beam CT) integrated with a medical linear accelerator combined with fast reconstruction and data transfer, this goal is being realized in our clinic using a pre-clinical research prototype. Building upon an existing protocol for image-guidance developed at our clinic over the last 5 years using implanted fiducial markers and MV imaging with an EPID system , the new process provides significant enhancement through the acquisition of a kV volumetric image following patient setup and immobilization. An “Acquisition Wizard” that streamlines image capture and reconstruction controls the imaging system. The volume reconstruction is automatically loaded by a commercial planning system for registration of soft tissue anatomy with the planned treatment volume, making the power and flexibility of such tools available on-line. The subsequent registration yields a 3D estimate of the necessary correction for manual adjustment of the treatment couch and this shift is verified using two orthogonal projections from an EPID system. The current pre-clinical prototype achieves a targeting accuracy for an unambiguous object of 1.20.3mm within a 20 minute treatment window.
Cone-beam CT, linear accelerator, image-guided radiation therapy
The success of the on-line approach depends on the issue of time: how long the target remains accurately localized and how quickly the combined imaging, planning, and delivery procedures can be performed. For example, the prostate remains in a 3mm margin for 20 minutes with a probability of 90% . Ideally, a real-time inverse plan for treatment delivery would be generated and delivered within moments of imaging. While technology exists to perform the inverse planning steps, it does not yet permit them to be performed in near real-time. This limitation should not prevent the pursuit of on-line treatment planning and delivery.
The cornerstone of the proposed image-guided system is a kilovoltage (kV) cone-beam computed tomography (CT) scanner that is integrated into the gantry of a medical linear accelerator [3,4,5]. This innovative combination permits tomographic imaging of the soft tissue anatomy and normal structures (Figure 1). The resulting images are high resolution (down to 250µm voxels), have isotropic spatial resolution, show good soft tissue contrast and best of all lie in the treatment reference frame (Figure 2).
A streamlined and optimized approach is proposed to permit on-line image-guided radiotherapy within a 20 minute treatment session. The captured 3D images of soft-tissue structures allow for the on-line measurement and correction of setup errors using a commercial planning system. Now the power and flexibility of a normally off-line tool can be used on-line. Physicians, physicists, and therapists received this design feature favourably since it provides a familiar set of tools for image interpretation and plan evaluation throughout the planning and delivery process. We also anticipate this standardized toolset will simplify staff training for image-guided procedures.
1) Construct a real process to implement image-guided radiotherapy in the clinic through integration of existing technologies.
2) Deliver a complete image-guided treatment session in less than 20 minutes.
3) Improve the existing delivery precision of <2mm of geometric uncertainty.
Figure 1: a). Medical linear accelerator used for acquisition of volumetric images. The kV imaging system is integrated into the gantry and allows a patient to be imaged in the treatment position immediately prior to treatment delivery. b) Patient setup for imaging/treatment.
1) No automatic decisions are made. All decisions regarding x-ray technique, target localization and couch movement require human approval.
2) Only 3D translations of the target volume are allowed (i.e. no rotations). The treatment couch only allows for pure translations to be corrected.
3) Streamline/optimize/simplify the process where possible. Minimize the input required through the use of wizards and intelligent defaults.
Results and discussion
The overall image-guided treatment system is comprised of four sub-systems (Figure 3). The sub-systems include the medical linear accelerator (red), the image acquisition system (yellow), the commercial planning system (cyan) and the radiation therapist (green) who interacts with all of the sub-systems and is the key decision maker.
The medical linear accelerator delivers the external beam treatment. Just prior to delivery, the linac also acquires the radiographic projections through means of a kV x-ray tube and AmSi flat-panel detector integrated into the gantry. The gantry is rotated around the patient acquiring some 285 projections through 360 degrees.
The image acquisition system consists of a high-end workstation which collects the projection images. After offset, gain and defective pixel correction the projections are reconstructed into a volumetric image (Figure 2). This volumetric image consists of 400x400x256 voxels that are 1mm on a side.
The planning system now serves a dual purpose. Initially, it is used off-line prior to the first treatment to contour the treatment volumes and plan the course of therapy. Then it is used on-line for targeting the soft tissue structures by means of matching the recently captured volumetric image to the treatment contours.
The final and key component in this process is the radiation therapist who makes decisions that control all the other components. The therapist is responsible for patient setup, initiating image acquisition, target localization, treatment couch adjustment and delivering the external beam radiotherapy fraction.
Figure 2: Coronal mid-plane slice of first patient volume taken at Princess Margaret Hospital on the linac mounted Cone-Beam CT system. The image resolution is 1mm X 1mm with a 1mm slice thickness. The measured dose in phantom for this configuration is 1.1-1.7 cGy.
The proposed image-guidance system is currently functioning in a pre-clinical prototype form. Since each component is an autonomous system, the difficulty is integration. Interactions between sub-systems must be coordinated and made as fast and efficient as possible. Connection points are denoted by pentagons in Figure 3 and labeled with letters “A” through “G” (Table 1). For example, connection point “A” allows the transfer of the volume image set from the acquisition system to the commercial planning system. This is implemented with a Samba mount over a 1Gbit/s ethernet connection. Connection points “B” through “E” refer to dialog screens that interact with the therapist. These dialogs display information or prompt the therapist for decisions or actions. Connection point “F” is a display application, which continuously monitors the actual vs. desired couch position.
Figure 3: The overall workflow for the system. Each of the four columns refers to an autonomous sub-system. The vertical axis describes time. The process is broken into three distinct phases: planning, pre-treatment and treatment fraction. The connection points between the sub-systems (pentagons) are described in Table 1.
An example interaction between the treatment planning system and therapist is shown in Figure 4. This dialog box corresponds to connection point C in Table 1. Here, the therapist is presented the numerical results of the computed shift from the image registration and prompted for approval. If the shift is rejected, the therapist returns to the target localization window. At each juncture, the system requires the therapist to accept the current action or return to the previous step for revision.
The second objective was to deliver the radiotherapy fraction within a 20 minute appointment. The breakdown of times within a treatment is shown in Table 2. Since this system is pre-clinical, some times are estimated. Other time estimates come from the existing fiducial marker program. The largest portion of time (i.e. 4.5 minutes) is currently spent processing and reconstructing the projection set. This time will reduce as faster computers become available.
The geometric precision of the targeting system was evaluated but is not the focus of this paper. An initial calibration cycle followed by repeated image-guidance sessions over a three month period demonstrated that the system can be used to relocate an unambiguous object to less than 1 mm (0.8mm) of the prescribed location . Treatment delivery can now proceed within the mechanical accuracy and precision of the delivery system, which was verified independently via a pair of orthogonal portal images to be 1.2 mm (0.3mm).
Figure 4: Screen shot of the computed correction by the commercial planning system, which is connection point C in Figure 3. The proposed adjustment of the treatment couch is presented to the therapist for approval. The contoured planning target volume is used for alignment with the on-line cone-beam CT image.
A system has been designed to implement on-line image-guided radiation therapy by successfully integrating existing technologies. The system interfaces acquisition of cone-beam projections and CT reconstruction with image fusion and assessment of geometric targeting with a commercially available treatment planning system. The integrated system permits the imaging, guidance and treatment process to be executed in less than 20 minutes. The excellent spatial resolution and delivery precision of the system are ideal for clinical implementation of high-precision localization and treatment of soft-tissue targets.
Alasti H, Petric MP, Catton CN and Warde PR 2001 “Portal imaging for evaluation of daily on-linesetup errors and off-line organ motion during conformal irradiation of carcinoma of the prostate” Int. J. Radiat Oncol Biol Phys.49(3) 869-84
Ghilezan M, Jaffray DA, Siewerdsen JH, van Herk M, Shetty A, Sharpe MB, Jafri S, Vicini FA, Matter RC, Brabbins DS, Martinez AA 2003 “1-hour Intra-fraction prostate motion assessed with cine magnetic resonance imaging (cine-MRI)” Int. J. Radiat Oncol Biol Phys. (accepted)
Jaffray DA, Siewerdsen JH, Wong JW and Martinez AA 2002 “Flat-panel cone-beam computed tomography for image-guided radiation therapy” Int. J. Radiat Oncol Biol Phys. 53 1337-1349
Jaffray DA and Siewerdsen JH, 1998 “Cone-beam computed tomography with a flat-panel imager: initial performance characterization” Med. Phys. 25 1493–501
Jaffray DA, Drake D, Moreau M, et. al. 1999 “A radiographic and tomographic imaging system integrated into a medical linear accelerator for localization of bone and soft-tissue targets” Int. J. Radiat Oncol Biol Phys. 45 773-789.
Sharpe MB, Moseley DJ, Haycocks T, Siewerdsen JH and Jaffray DA 2003 “An integrated volumetric imaging and guidance system for targeting of soft-tissue structures in radiotherapy” Int. J. Radiat Oncol Biol Phys. 57 (2 Suppl):S183
Transfer reconstruction to planning system via a Samba mount on a 1GBit/sec Cu Ethernet connection. The data rate is 64 Mbytes/sec which requires 16 seconds to transfer a 5123 volume image
Register image: allows therapist to manually match the outlined target to the soft-tissue anatomy. Translations in all three planes are allowed. No rotations are permitted since these cannot be corrected.
Present quantitative results of image match: 3D translation value in mm.