Abstract Book

S1045

ESTRO 37

influence of individual treatment planner skill and knowledge, based on a large library of treatment plans. Corollary is to reduce the planning time as minimum as reasonably achievable. Material and Methods An ensemble of 120 VMAT based cranial stereotactic plans were categorised on basis of following eight parameters ,PTV dose coverage OAR challenged or unchallenged, prescription dose, number of PTVs, laterality (left /right), tumour volume, shortest distance between OAR-PTV, centre to centre distance between OARs-PTV, lateral dimension of brain (fig-1). Performance of ensemble mapping technique was validated in existing library plan. Further for new patient a knowledge based planning (KBP) was created by coping most appropriate library plan with all parameters unchanged; optimization and dose calculation was carried out with no or very minimal changes in the constraint. Another independent treatment plan (IP) using the same arc configuration was generated by an experience dosimetrist and compared with KBP for PTV_V98%,conformity index (PCI),50% and 20% dose spillage (I50%, I20%) and OAR doses. Results Validation result for ensemble mapping technique shows that for an OAR challenged PTV dose patient it is appropriately picking up the library plan. However for patent having an OAR unchallenged PTV it is not very accurate and may pick up a nearby plan as well. Independent plan (IP) was better than the knowledge based plans (KBP) in PTV coverage and dose conformity. PTV volume receiving 98%dose (V98%) was 98.7±1.1% and 97.5±1.3% for IP and KBP respectively. For OAR challenged PTV’s conformity index was slightly high in IP (0.712) than KB plans (0.693). If collimator angle optimization is not done for the OAR unchallenged PTV’s PCI is higher than KBP and variation was statistically significant (p<0.04). For the largest prescription dose group (12Gy in 1#, 64 patients) brainstem 0.5 cc Volume exhibit a mean dose of 873.1±134.2 cGy and 854.5±122.4 cGy for IP and KBP respectively. Mean 0.2 cc optic chiasma dose were 734.0±67.8 cGy and 690.1±78.3 cGy for KBP and IP respectively. MU difference was very nominal with IP shows a mean excess MU of 3.7% over KBP. IP and KBP requires on 3.5 and 1.5 runs respectively costing about 3.5-5 and 1.5-2 hrs. Conclusion KBP plans validation result indicate multidimensional ensemble mapping mechanism can pick up the library plan accurately. KBP plans, although marginally inferior in the dosimetric quality, they fulfil all the required clinical condition and dose constraints. KBS plans save a considerable planning time and almost independent of the treatment planner skill and knowledge. KBP works well with Monte Carlo planning system like MONACO. EP-1923 Integration of microscopic spread and geometric uncertainties into a single target volume expansion E. Sterpin 1 , K. Haustermans 1 , M. Lambrecht 1 , X. Geets 2 , T. Mackie 3 , V. Gregoire 2 1 Katholieke Universiteit Leuven, Oncology, Leuven, Belgium 2 Universite catholique de Louvain, Molecular Imaging- Radiotherapy and Oncology, Brussels, Belgium 3 University of Wisconsin, Medical Physics, Madison, USA Purpose or Objective In external beam radiotherapy, the definition of several target volumes follows a conventional GTV-CTV-PTV margin expansion formalism. The magnitude of the GTV to CTV expansion is based on histology studies or patterns of recurrence over patient populations and typically determined by a clinician. The CTV volume is further corrected manually to exclude zones in which tumor

expansion is made impossible due to anatomical barriers. Finally, a PTV margin is added to the CTV and is computed from the statistics linked to geometric errors (patient setup, inter- and intra fraction motion). The PTV margin is also large enough in order to encompass the CTV with a sufficient amount of systematic geometric errors, typically 90-95%, as determined by the physics team. Although such approach has considerable pragmatic merits, it is fundamentally statistically biased because margin expansions for microscopic infiltration probability and systematic geometric errors should not be added linearly but in quadrature, as they can both be understood as a geometric probability of systematically missing tumor cells. We have implemented here a novel approach for computing the PTV directly from the GTV based on a more statistically sound methodology that, for a given confidence interval, mitigates excess planning target volumes compared to the conventional approach. Material and Methods We have implemented in MATLAB an algorithm that directly expands the GTV to the PTV, for the purposes of this study called the PTVc (PTV Combinatorial). The main inputs are the probability of microscopic infiltration, derived from generic distributions or published data, and the standard deviations for geometric systematic errors. For every point at the surface of the GTV, the probability of microscopic infiltration is corrected for anatomical barriers using editions of the clinician. The latter step could be automatized eventually. A local margin is then defined according to a probability threshold defined for the combined probability distribution. The volume is further dilated for geometric random errors leading to the final PTVc. It is important to mention that our methodology only applies to a conventional GTV-CTV-PTV expansion scheme with a CTV margin linked to a probability of microscopic infiltration; elective CTVs are not considered. We have tested our methodology for 5 lung, 3 H&N and 1 glioblastoma patients. Results The PTVc equals the conventional PTV for parts of the volumes where the physician has estimated that infiltration was impossible, as illustrated on the CT image. Volume mitigation ranges from 8 to 34% (see table).