Validation of in-house knowledge-based planning model for predicting change in target coverage during VMAT radiotherapy to in-operable advanced-stage NSCLC patients

Abstract

Objectives: anatomical changes are inevitable during the course of radiotherapy treatments and, if significant, can severely alter expected dose distributions and affect treatment outcome. Adaptive radiotherapy (ART) is employed to maintain the planned distribution and minimise detriment to predicted treatment outcome.Typically, patients who may benefit from adaptive planning are identified via a replanning process, i.e., re-simulation, re-contouring, re-planning and treatment plan quality assurance (QA). This time-intensive process significantly increases workload, can introduce delays and increases unnecessary stress to those patients who will not actually gain benefit. We consider it crucial to develop efficient models to predict changes to target coverage and trigger ART, without the need for re-planning.Methods: knowledge-based planning (KBP) models were developed using data for 20 patients' (400 fractions) to predict changes in PTV V95 coverage (∆𝑉95 𝑃𝑇𝑉 ). Initially, this change in coverage was calculated on the synthetic computerised tomography (sCT) images produced using the Velocity adaptive radiotherapy software. Models were developed using patient (cell death bio-marker) and treatment fraction (PTV characteristic) specific parameters to predict (∆𝑉95 𝑃𝑇𝑉 )and verified using five patients (100 fractions) data.Results: three models were developed using combinations of patient and fraction specific terms. The prediction accuracy of the model developed using biomarker (PD-L1 expression) and the difference in 'planning' and 'fraction' PTV centre of the mass (characterised by mean square difference, MSD) had the higher prediction accuracy, predicting the (∆𝑉95 𝑃𝑇𝑉 )within ± 1.0% for 77% of the total fractions; with 59% for the model developed using, PTV size, PD-L1 and MSD and 48% PTV size and MSD respectively. Conclusion:the KBP models can predict (∆𝑉95 𝑃𝑇𝑉 )very effectively and efficiently for advanced-stage NSCLC patients treated using volumetric modulated arc therapy and to identify patients who may benefit from adaption for a specific fraction.

Predicting personalised and progressive adaptive dose escalation to gross tumour volume using knowledge-based planning models for inoperable advanced-stage non-small cell lung cancer patients treated with volumetric modulated arc therapy

“…To facilitate the image processing within Velocity, the following process (see [28] for further detail) was followed:…”

“…The planning protocol used for this study mirrors that used in a previous study which we have previously (28) described. Where patients are able to comply, a four-dimensional computerised tomography scan (4DCT) (as well as a free-breathing (FB) scan, otherwise only the latter is used.…”

“…To facilitate the image processing within Velocity, the following process (see (28) for further detail) was followed: 1) to remove the impact of residual setup errors (30); the pCT and CBCT images were rigidly registered, using the same transformation obtained during the respective treatment fraction 2) to produce the sCT; the CBCT images were deformed to match the pCT images 3) using the deformation matrix, obtained in step 2, a secondary structure data set was produced in the sCTs from the original structures (including GTV and OARs), 4) The registration and volumes for each sCT were reviewed.…”