Stereotactic body radiotherapy (SBRT) is an effective technique in lung cancer treatment and several prerequisites are essential in order to achieve good local control. These include precise imaging of the lesion before irradiation and accurate dose calculation to account for density heterogeneities in lung tissue. Both aspects were investigated within the framework of the thesis: a new approach in imaging with a conventional electronic portal imaging device (EPID) was investigated and the performance and limits of a new Monte Carlo (MC) calculation algorithm commercially available were studied.
More specifically, digitally reconstructed radiographs (DRR) of lung lesions were compared with MV portal images in a feasibility study to assess any displacement of the tumour. The precision of displacement results of three registration algorithms was tested when compared to a projection image of the tumour. The various algorithms were applied to test images, a lung simulation phantom and finally to patient data including 38 tumours and images of 113 fractions. Image guidance results of tested registration algorithms proved the accuracy in the lung phantom study whereas clinical patient data had successful registrations in about 59% of anterior-posterior (AP) and 46% of lateral projections, respectively. Excluding real patient data with a clinical target volume smaller than 10 cm3, successful registrations occurred in 90% of AP and 50% lateral projections.
With respect to dose calculation accuracy, experimental verification of a commercial Monte Carlo-based planning system was performed for high-energy photon beams. Several simple and complex treatment cases were calculated and compared with measurements in different phantom types. Besides ion chamber measurements, radiochromic films were irradiated to gain 2D dose distributions which were compared to calculations applying the gamma-index criterion. The dose calculation accuracy of the Monte Carlo algorithm implemented in two different treatment planning systems (TPS) was further compared to advanced kernel based methods, namely the collapsed cone (CC) convolution algorithm and the analytical anisotropic algorithm (AAA). 1D gamma-index of depth dose curves and profiles were investigated as well as 2D gamma values of axial planes for composite treatment plans. Concerning dose comparison of MC calculations to measurements, the average gamma value was 0.21 for all energies for simple test cases. Gamma results for depth dose curves in asymmetric beams were similar to symmetric beams. Simple regular fields showed excellent absolute dosimetric agreement to ion chamber measurements. A detailed analysis at tissue interfaces showed dose discrepancies below 3% for a 10x10 cm2 field at 18 MV energy. Advanced treatment techniques within a heterogeneous test phantom showed an average dose discrepancy of 0.5% 1.1% whereas relative dose results lead to a mean 2D gamma value of 0.44 0.07. With respect to advanced kernel dose calculations no large differences in absolute dosimetry were found when compared to the MC modules. In general, 1D gamma results obtained with both MC systems were similar and comparable with the CC algorithm whereas the results for the AAA algorithm were slightly worse.
2D gamma evaluation for four-field plans and IMRT plans revealed smaller mean gamma values for MC dose calculations compared to the advanced kernel algorithms.
In conclusion, 2D/3D EPID image registration for lung tumours is an attractive alternative for markerless online imaging and can complement planar kV imaging on state of the art linacs with cone beam computer tomography (CBCT). Size and location of the tumour are still the limiting parameters for an accurate registration process. Focusing on dose calculation accuracy of Monte Carlo-based treatment planning systems, tested modules led to accurate dosimetric results under clinical test conditions. The TPS with MC algorithms even performed best in heterogeneous conditions but the difference between the CC and the MC algorithms was found to be very small.