W. Van Elmpt

A global calibration model for a-Si EPIDs used for transit dosimetry

Electronic portal imaging devices (EPIDs) are not only applied for patient setup verification and detection of organ motion but are also increasingly used for dosimetric verification. The aim of our work is to obtain accurate dose distributions from a commercially available amorphous silicon (a-Si) EPID for transit dosimetry applications. For that purpose, a global calibration model was developed, which includes a correction procedure for ghosting effects, field size dependence and energy dependence of the a-Si EPID response. In addition, the long-term stability and additional buildup material for this type of EPID were determined. Differences in EPID response due to photon energy spectrum changes have been measured for different absorber thicknesses and field sizes, yielding off-axis spectrum correction factors based on transmission measurements. Dose measurements performed with an ionization chamber in a water tank were used as reference data, and the accuracy of the dosimetric calibration model was determined for a large range of treatment conditions. Gamma values using 3% as dose-difference criterion and 3mm as distance-to-agreement criterion were used for evaluation. The field size dependence of the response could be corrected by a single kernel, fulfilling the gamma evaluation criteria in case of virtual wedges and intensity modulated radiation therapy fields. Differences in energy spectrum response amounted up to 30%-40%, but could be reduced to less than 3% using our correction model. For different treatment fields and (in)homogeneous phantoms, transit dose distributions satisfied in almost all situations the gamma criteria. We have shown that a-Si EPIDs can be accurately calibrated for transit dosimetry purposes.

Experimental verification of a portal dose prediction model.

Electronic portal imaging devices (EPIDs) can be used to measure a two-dimensional (2D) dose distribution behind a patient, thus allowing dosimetric treatment verification. For this purpose we experimentally assessed the accuracy of a 2D portal dose prediction model based on pencil beam scatter kernels. A straightforward derivation of these pencil beam scatter kernels for portal dose prediction models is presented based on phantom measurements. The model is able to predict the 2D portal dose image (PDI) behind a patient, based on a PDI without the patient in the beam in combination with the radiological thickness of the patient, which requires in addition a PDI with the patient in the beam. To assess the accuracy of portal dose and radiological thickness values obtained with our model, various types of homogeneous as well as inhomogeneous phantoms were irradiated with a 6 MV photon beam. With our model we are able to predict a PDI with an accuracy better than 2% (mean difference) if the radiological thickness of the object in the beam is symmetrically situated around the isocenter. For other situations deviations up to 3% are observed for a homogeneous phantom with a radiological thickness of 17 cm and a 9 cm shift of the midplane-to-detector distance. The model can extract the radiological thickness within 7 mm (maximum difference) of the actual radiological thickness if the object is symmetrically distributed around the isocenter plane. This difference in radiological thickness is related to a primary portal dose difference of 3%. It can be concluded that our model can be used as an easy and accurate tool for the 2D verification of patient treatments by comparing predicted and measured PDIs. The model is also able to extract the primary portal dose with a high accuracy, which can be used as the input for a 3D dose reconstruction method based on back-projection.

Interfractional trend analysis of dose differences based on 2D transit portal dosimetry

Dose delivery of a radiotherapy treatment can be influenced by a number of factors. It has been demonstrated that the electronic portal imaging device (EPID) is valuable for transit portal dosimetry verification. Patient related dose differences can emerge at any time during treatment and can be categorized in two types: (1) systematic-appearing repeatedly, (2) random-appearing sporadically during treatment. The aim of this study is to investigate how systematic and random information appears in 2D transit dose distributions measured in the EPID plane over the entire course of a treatment and how this information can be used to examine interfractional trends, building toward a methodology to support adaptive radiotherapy. To create a trend overview of the interfractional changes in transit dose, the predicted portal dose for the different beams is compared to a measured portal dose using a γ evaluation. For each beam of the delivered fraction, information is extracted from the γ images to differentiate systematic from random dose delivery errors. From the systematic differences of a fraction for a projected anatomical structures, several metrics are extracted like percentage pixels with |γ| > 1. We demonstrate for four example cases the trends and dose difference causes which can be detected with this method. Two sample prostate cases show the occurrence of a random and systematic difference and identify the organ that causes the difference. In a lung cancer case a trend is shown of a rapidly diminishing atelectasis (lung fluid) during the course of treatment, which was detected with this trend analysis method. The final example is a breast cancer case where we show the influence of set-up differences on the 2D transit dose. A method is presented based on 2D portal transit dosimetry to record dose changes throughout the course of treatment, and to allow trend analysis of dose discrepancies. We show in example cases that this method can identify the causes of dose delivery differences and that treatment adaptation can be triggered as a result. It provides an important element toward informed decision-making for adaptive radiotherapy.

Prediction of DVH parameter changes due to setup errors for breast cancer treatment based on 2D portal dosimetry.

Electronic portal imaging devices (EPIDs) are increasingly used for portal dosimetry applications. In our department, EPIDs are clinically used for two-dimensional (2D) transit dosimetry. Predicted and measured portal dose images are compared to detect dose delivery errors caused for instance by setup errors or organ motion. The aim of this work is to develop a model to predict dose-volume histogram (DVH) changes due to setup errors during breast cancer treatment using 2D transit dosimetry. First, correlations between DVH parameter changes and 2D gamma parameters are investigated for different simulated setup errors, which are described by a binomial logistic regression model. The model calculates the probability that a DVH parameter changes more than a specific tolerance level and uses several gamma evaluation parameters for the planning target volume (PTV) projection in the EPID plane as input. Second, the predictive model is applied to clinically measured portal images. Predicted DVH parameter changes are compared to calculated DVH parameter changes using the measured setup error resulting from a dosimetric registration procedure. Statistical accuracy is investigated by using receiver operating characteristic (ROC) curves and values for the area under the curve (AUC), sensitivity, specificity, positive and negative predictive values. Changes in the mean PTV dose larger than 5%, and changes in V90 and V95 larger than 10% are accurately predicted based on a set of 2D gamma parameters. Most pronounced changes in the three DVH parameters are found for setup errors in the lateral-medial direction. AUC, sensitivity, specificity, and negative predictive values were between 85% and 100% while the positive predictive values were lower but still higher than 54%. Clinical predictive value is decreased due to the occurrence of patient rotations or breast deformations during treatment, but the overall reliability of the predictive model remains high. Based on our predictive model, 2D transit dosimetry measurements can now directly be translated in clinically more relevant DVH parameter changes for the PTV during conventional breast treatment. In this way, the possibility to design decision protocols based on extracted DVH changes is created instead of undertaking elaborate actions such as repeated treatment planning or 3D dose reconstruction for a large group of patients.

Early variation of FDG-PET radiomics features in NSCLC is related to overall survival - the “delta radiomics” concept

41 A Hypercellular Component of Glioblastoma Identified by High b-value Diffusion Weighted Imaging Y. Cao,1,2,3 D. Wahl,1 P. Pramanik,1,3 M. Kim,1 T.S. Lawrence,1 H. Parmar2 1 Department of Radiation Oncology, University of Michigan, Ann Arbor, Michigan; 2 Department of Radiology, University of Michigan, Ann Arbor, Michigan; 3 Department of Biomedical Engineering, University of Michigan, Ann Arbor, Michigan

SU-D-110-02: Evaluation of the Differences between Locoregional Lung Ventilation Estimation Methods Using a Single Deformable Image Registration Algorithm

Purpose: Three methods of calculating ventilation from 4D CTimage sets have been explored by several research groups. This study is to investigate the differences of these three local ventilation calculations. Methods: Optical flow (OF) deformable image registration of the normal end expiration and inspiration phases of 4D‐CT images was used to correlate the voxels between the two phases. The OF was validated using a 4D pixel‐ based and point‐validated breathing thorax model, consisting of a 4D‐CT image data set along with associated landmarks. Ventilation derived from 4D‐CTs from 20 esophageal patients were retrospectively analyzed. Differences between the ventilation images generated by three methods, the Jacobian, the DeltaV, and the HU, were examined on a voxel‐to‐voxel basis. The Jacobian method uses the first derivative of the deformation field to approximate the change in volume of voxels. The DeltaV method directly calculates the volume change. The HU method uses the change in Houndsfield Units (HUs) of corresponding voxels to calculate ventilation. Results: The target registration error (TRE) for the deformable image registration was an average of 1.6±0.68 mm and maximum of 3.1 mm. Average difference between the DeltaV and the Jacobian ventilation as a percentage of the maximum ventilation value was 0.51±0.3% (range 0.33% to 1.32%). Average difference between the DeltaV and HU ventilation was 2.4±4.5 % (range 0.4% to 19.2 %). A small number of voxels show significant differences. We speculate that the larger differences were due to some image registration variances. Regions of highest and lowest ventilation matched well for all methods. Conclusions: Highs and lows in ventilation were more pronounced in the DeltaV method compared to the Jacobian. In general the differences between the two ventilation methods were small. However, the differences between the DeltaV and the HU methods were considerably larger.

Evaluation of third treatment week as temporal window for assessing responsiveness on repeated FDG-PET-CT scans in Non-Small Cell Lung Cancer patients

PURPOSE Early assessment of tumour response to treatment with repeated FDG-PET-CT imaging has potential for treatment adaptation but it is unclear what the optimal time window for this evaluation is. Previous studies indicate that changes in SUVmean and the effective radiosensitivity (αeff, accounting for uptake variations and accumulated dose until the second FDG-PET-CT scan) are predictive of 2-year overall survival (OS) when imaging is performed before radiotherapy and during the second week. This study aims to investigate if multiple FDG-PET-derived quantities determined during the third treatment week have stronger predictive power. METHODS Twenty-eight lung cancer patients were imaged with FDG-PET-CT before radiotherapy (PET1) and during the third week (PET2). SUVmean, SUVmax, SUVpeak, MTV41%-50% (Metabolic Tumour Volume), TLG41%-50% (Total Lesion Glycolysis) in PET1 and PET2 and their change (), as well as average αeff (α¯eff) and the negative fraction of αeff values [Formula: see text] ) were determined. Correlations were sought between FDG-PET-derived quantities and OS with ROC analysis. RESULTS Neither SUVmean, SUVmax, SUVpeak in PET1 and PET2 (AUC = 0.5-0.6), nor their changes (AUC = 0.5-0.6) were significant for outcome prediction purposes. Lack of correlation with OS was also found for α¯eff (AUC = 0.5) and [Formula: see text] (AUC = 0.5). Threshold-based quantities (MTV41%-50%, TLG41%-50%) and their changes had AUC = 0.5-0.7. P-values were in all cases ≫0.05. CONCLUSIONS The poor OS predictive power of the quantities determined from repeated FDG-PET-CT images indicates that the third week of treatment might not be suitable for treatment response assessment. Comparatively, the second week during the treatment appears to be a better time window.

EP-1865: DCE-CT lung tumour and aorta enhancement: is it an appropriate input vessel for kinetic modelling?

Results: Texture parameters were computed in the three perfusion maps and their 3D wavelet transforms, which resulted in 945 texture features defined for each of the two tumor sites. The discretization of images using set number of bins and setintervals gave the similar number of stable texture parameters (Table 1). 40 parameters were correlated with tumor volume. Potentially standardizable factors introduced more variability into texture features than nonstandardizable. The highest variability was observed for pixel size. It caused instability in about 80% of parameters for both HN and lung tumors. Ten parameters were found to bestable in both HN and lung for potentially non-standardizable factors after thecorrection for inter-parameters correlations:

OC-0382: A novel concept to tumour targeting: inverse dose-painting or targeting the "Low uptake drug volume"

Results: Composite perfusion changes were associated with dose. Statistically significant dose-dependent reduction in regional perfusion was observed at 3, 6 and 12 months FU. Comparison of dose-response curves based on their slopes showed a dose-dependent reduction in perfusion at all time intervals (R2=0.8-0.9) except 1 month (R2=0.4). Relative perfusion loss per dose bin was 4% at 1 month, 14% at 3 months, 13% at 6 months and 21% at 12 months FU (Figure 1).

PO-0929: Dual Energy CT imaging of tumour vasculature in NSCLC: an intra-patient comparison with DCE-CT

Purpose or Objective: Quantification of vasculature is frequently performed by dynamic contrast enhanced CT (DCE-CT) or MRI imaging. However, there are some limitations to this technique: DCE-CT requires a detailed kinetic fitting procedure, a prolonged acquisition time with increased dose to the patient, has a limited FOV and is not easy to implement in clinical routine. Dual Energy CT is an evolving field in CT image analysis that allows quantification of contrast material uptake using a single acquisition, making it easily implementable in a clinical workflow. Therefore we investigated the correlation between the DCE-CT derived vasculature parameters, blood flow and blood volume, with iodine related attenuation measured on a Dual Energy CT acquisition for non-small cell lung cancer patients.