Stereotactic body radiotherapy (SBRT) is a promising treatment strategy for early–stage lung cancers. Conventional three–dimensional (3D) SBRT based on a static patient geometry is an insufficient model of reality, posing constraints on accurate Monte Carlo (MC) dose calculation and intensity–modulated radiotherapy (IMRT) optimization. Four–dimensional (4D) radiotherapy explicitly considers temporal anatomical changes by characterizing the organ motion and building a 4D patient model, generating a treatment plan that optimizes the doses to moving tissues, i.e., 4D dose (as opposed to the static 3D dose to tissue), and delivering this plan by synchronizing the radiation with the moving tumor. This thesis focuses on 4D robotic tracking lung SBRT.
By recalculating the conventional 3D plan on the 4D patient model using MC simulation, it was found that 4D moving dose distributions could detect increase of normal tissue doses and complication probabilities (NTCP), and decrease of tumor dose and control probability. For one patient, the risk of myelopathy was estimated at 8% and 18% from the 3D equivalent path–length corrected (EPL) and the 4D MC doses, respectively. Such increased NTCP suggests that better estimations of different dosimetric quantities using 4D MC dose calculation are crucial to improve the existing dose–response models.
Dosimetric error in 4D robotic tracking SBRT was found to be caused predominately by tissue heterogeneities, as assessed by the comparisons of the 4D moving tissue doses calculated using the conventional EPL and MC algorithms. At 3% tolerance level, our results indicated clinically significant dose prediction errors only in tumor but not in other major normal tissues. Furthermore, 4D tracking radiotherapy was found to have greater ability to limit the normal tissue volume receiving high to medium doses than the other advanced SBRT strategy combining volumetric–arc radiotherapy with 4D cone–beam CT verification.
Invariant target motion was found to be an unrealistic assumption of 4D radiotherapy from the analysis of probability motion function (pmf) of motion data. Systematic and random variations of motion amplitude, frequency, and baseline were found to reduce the reproducibility of pmfs, on average, to just 30% for the principal motion of 3400 seconds.
Experimental evaluations showed that systematic motion change reduced the gamma passing rate of radiochromic film measurements at 3mm distance–to–agreement and 3% dose difference criteria from 91% for 4D dose calculated with MCand EPL algorithms to 47% and 53% in the static object, respectively,. For moving target object, gamma passing rates of the 4D MC doses hardly changed with
reproducible and non–reproducible motion (95% vs. 93%), and barely differed between conventional 3D and 4D MC doses (95% vs. 95% with reproducible, and 96% vs. 93% with non–reproducible motions). Distortions due to image artifacts and registration errors were consistently observed in the 4D dose distributions but not the 3D dose distributions.
In conclusion, 4D Monte Carlo planning shall be considered for robotic target tracking only if robustness against uncertainties of patient geometry, and accuracy of 4DCT imaging and deformation registration are significantly improved. / published_or_final_version / Clinical Oncology / Doctoral / Doctor of Philosophy
|Creators||Chan, Ka-heng, 陳加慶|
|Publisher||The University of Hong Kong (Pokfulam, Hong Kong)|
|Source Sets||Hong Kong University Theses|
|Rights||Creative Commons: Attribution 3.0 Hong Kong License, The author retains all proprietary rights, (such as patent rights) and the right to use in future works.|
|Relation||HKU Theses Online (HKUTO)|
Page generated in 0.0033 seconds