Registration of synthetic tomographic projection data sets using cross-correlation.

Tomographic registration, a method that makes possible accurate patient registration directly from projection data, consists of three processing steps: (i) manual coarse positioning, (ii) tomographic projection set acquisition, and (iii) computer mediated refined positioning. In the coarse positioning stage, the degree of patient alignment is comparable with that achieved with the standard radiotherapy set-up. However, the accuracy requirements are somewhat more relaxed in that meticulous alignment of the patient using external laser indicators is not necessary. Instead, tomographic projection sets are compared with planning CTs in order to achieve improved patient set-up. The projection sets are cross-correlated to obtain the best-fit translation and rotation offsets. The algorithm has been tested on synthetic data with the incorporation of varying amounts of Gaussian pseudo-random noise. These tests demonstrate the algorithm's stability and also confirm that alignment can be achieved with an accuracy of less than one projection pixel.

Maximum likelihood as a common computational framework in tomotherapy.

Tomotherapy is a dose delivery technique using helical or axial intensity modulated beams. One of the strengths of the tomotherapy concept is that it can incorporate a number of processes into a single piece of equipment. These processes include treatment optimization planning, dose reconstruction and kilovoltage/megavoltage image reconstruction. A common computational technique that could be used for all of these processes would be very appealing. The maximum likelihood estimator, originally developed for emission tomography, can serve as a useful tool in imaging and radiotherapy. We believe that this approach can play an important role in the processes of optimization planning, dose reconstruction and kilovoltage and/or megavoltage image reconstruction. These processes involve computations that require comparable physical methods. They are also based on equivalent assumptions, and they have similar mathematical solutions. As a result, the maximum likelihood approach is able to provide a common framework for all three of these computational problems. We will demonstrate how maximum likelihood methods can be applied to optimization planning, dose reconstruction and megavoltage image reconstruction in tomotherapy. Results for planning optimization, dose reconstruction and megavoltage image reconstruction will be presented. Strengths and weaknesses of the methodology are analysed. Future directions for this work are also suggested.

Image/patient registration from (partial) projection data by the Fourier phase matching method.

A technique for 2D or 3D image/patient registration, PFPM (projection based Fourier phase matching method), is proposed. This technique provides image/patient registration directly from sequential tomographic projection data. The method can also deal with image files by generating 2D Radon transforms slice by slice. The registration in projection space is done by calculating a Fourier invariant (FI) descriptor for each one-dimensional projection datum, and then registering the FI descriptor by the Fourier phase matching (FPM) method. The algorithm has been tested on both synthetic and experimental data. When dealing with translated, rotated and uniformly scaled 2D image registration, the performance of the PFPM method is comparable to that of the IFPM (image based Fourier phase matching) method in robustness, efficiency, insensitivity to the offset between images, and registration time. The advantages of the former are that subpixel resolution is feasible, and it is more insensitive to image noise due to the averaging effect of the projection acquisition. Furthermore, the PFPM method offers the ability to generalize to 3D image/patient registration and to register partial projection data. By applying patient registration directly from tomographic projection data, image reconstruction is not needed in the therapy set-up verification, thus reducing computational time and artefacts. In addition, real time registration is feasible. Registration from partial projection data meets the geometry and dose requirements in many application cases and makes dynamic set-up verification possible in tomotherapy.

Delivery verification in sequential and helical tomotherapy.

Conformal and conformal avoidance radiation therapy are new therapeutic techniques that are generally characterized by high dose gradients. The success of this kind of treatment relies on quality assurance procedures in order to verify the delivery of the treatment. A delivery verification technique should consider quality assurance procedures for patient positioning and radiation delivery verification. A methodology for radiation delivery verification was developed and tested with our tomotherapy workbench. The procedure was investigated for two cases. The first treatment using a torus-shaped target was optimized for 72 beam directions and sequentially delivered as a single slice to a 33 cm diameter cylinder of homogeneous solid water. For the second treatment, a random pattern of energy fluence was helically delivered for two slices to a 9.0 cm diameter phantom containing inhomogeneities. The presented process provides the energy fluence (or a related quantity) delivered through the multileaf collimator (MLC) using the signal measured at the exit detector during the treatment delivery. As this information is created for every pulse of the accelerator, the energy fluence and state for each MLC leaf were verified on a pulse-by-pulse basis. The pulse-by-pulse results were averaged to obtain projection-by-projection information to allow for a comparison with the planned delivery. The errors between the planned and delivered energy fluences were concentrated between +/-2.0%, with none beyond +/-3.5%. In addition to accurately achieving radiation delivery verification, the process is fast, which could translate to radiation delivery verification in real time. This technique can also be extended to reconstruct the dose actually deposited in the patient or phantom (dose reconstruction).

Registration using tomographic projection files.

An algorithm has been developed and experimentally verified for tomographic registration--a patient positioning method using internal anatomy and standard external fiducial marks. This algorithm improves patient set-up and verification to an accuracy sufficient for tomotherapy. By implementation of this technique, the time-consuming reconstruction process is avoided. Instead, offsets in the x, y and z directions are determined directly from sinogram data by an algorithm that utilizes cross-correlations and Fourier transforms. To verify the efficiency and stability of the algorithm, data were collected on the University of Wisconsin's dedicated tomotherapy research workbench. The experiment indicates offset statistical errors of less than +/-0.8 mm for offsets up to 30 mm. With standard clinical techniques, initial patient offsets are expected to be less than 5 mm, so the 30 mm limitation is of no consequence. The angular resolution for the direction of patient translation is within the +/-2 degrees needed for tomotherapy.