Proceedings of the 52nd Annual ASTRO Meeting S539

Conclusions: RapidArc has been proved to be a safe and advantageous treatment modality for NSCLC with large volumes. Longterm observation of patients is needed to assess outcome and late toxicity.

Author Disclosure: M. Scorsetti, None; P. Navarria, None; P. Mancosu, None; S. Castiglioni, None; C. Bressi, None; A. Fogliata,Her housband acts as advisor to Varian Medical System, F. Consultant/Advisory Board; E. Clerici, None; R. Cavina, None; A.Santoro, None; L. Cozzi, He acts as advisor to Varian Medical System, F. Consultant/Advisory Board.

2749 Determination of a Multi-dimensional Parametric Model to Predict the Optimum PET SUV Threshold

Values in Lung Tumor Volume Delineation

K. Wijesooriya, J. M. Larner, S. H. Benedict, W. Yang, A. C. Cupino, P. W. Read

University of Virginia Health System, Charlottesville, VA

Purpose/Objective(s): Using FDG-PET in addition to CT to define the tumor volume has been shown to reduce the inter-observervariability and to effectively discriminate malignant from normal tissue. However, the PET SUV-threshold values used to definethe internal tumor volume (ITV) has a large variation. To complicate the situation, PET images, taken over multiple respiratorycycles are prone to respiratory motion effects. We therefore developed a functional form to predict how SUV-threshold is influ-enced by tumor motion amplitude, tumor volume, and maximum PET SUV.

Materials/Methods: Pre-treatment FDG-PET/CTs and 4DCTs from 25 lung cancer patients were co-registered using Velocitysoftware (Velocity AI). Tumor volume, V ranged from 0.5 cc to 440 cc while the motion amplitude, A, determined from thefull motion envelope of the 4DCT range from 0 to 1.5 cm. The, PET SUV-max ranged from 3.8 to 18.95 corresponding to a frac-tional SUV-threshold range, [SUV-threshold / SUV-max] of 0.2 to 0.86. The ITV was contoured in the 4DCT, and the tumorvolume was obtained. Using the full motion envelope of 4DCT as the gold standard, PET image was contoured and PET SUV-threshold value was changed until the PET ITV agreed to within 1 cc of 4DCT ITV, and visual agreement between PET and4DCT for all three dimensions were observed. Data for 20 patients were input into a multi variable non-linear chi squared mini-mization routine (MINUIT from CERN) to optimize a 5 parameter function that fit the optimum SUV-threshold. We assumed thatthe SUV-threshold depends on the following parameter combinations: constant coefficient, SUV-max up to its second order, (SUV-max)2/V, and A (SUV-max)2. Data from the remaining 5 patients was used to independently validate the predictive accuracy of theparameterization.

Results: The parameterization shows that the first order SUV-max term is the most significant contributor to the SUV-thresholdwhile (SUV-max)2/V term is the second most significant contributor. The convergence of the minimization showed that motionamplitude is more important for tumors with higher SUV-max values. The mean difference of tumor radii between the measuredITV and the parameterization predicted ITV for the 20 patients used for the fit was 0.1 mm (SD, 1.1mm, maximum, 1.7mm). Themean difference of radii between the measured GTV and the parameterization predicted GTV for 5 verification patients were 0.7mm (SD, 0.9 mm, maximum, 1.4 mm).

Conclusions: We have defined a model with 5 physically meaningful parameters which accurately predicts the SUV-threshold ofa PET scan for lung tumors, given the full motion amplitude, SUV-max, and 4DCT ITV volume. This model will allow investi-gators to ‘‘normalize’’ SUV thresholds across different patient populations by using our parametrically derived SUV-threshold.

Author Disclosure: K. Wijesooriya, None; J.M. Larner, None; S.H. Benedict, None; W. Yang, None; A.C. Cupino, None; P.W.Read, None.

2750 Investigation of the Clinical Utility of 4D FDG-PET/CT Scans In Radiation Treatment Planning

M. Aristophanous1, R. Berbeco1, J. Killoran1, J. Yap2, D. Sher1, A. Allen1, E. Larson1, A. Chen1

1Brigham and Women’s Hospital and Dana-Farber Cancer Institute, Boston, MA, 2Dana-Farber Cancer Institute, Boston, MA

Purpose/Objective(s): Whole body FDG-PET/CT scans are being used to aid gross tumor volume definition in many radiotherapydepartments. However, for lung tumors, respiratory motion can cause a decrease in quantitative accuracy. In this study, we assessedthe potential benefits of 4D PET acquisition in radiation treatment planning, relative to standard 3D PET.

Materials/Methods: Nine patients with non-small cell lung cancer (NSCLC) had sequential 3D and 4D FDG-PET/CT scans in thetreatment position prior to radiation therapy. Five breathing phases were used to bin the 4D data. The tumor volume and involvedlymph nodes were contoured on the PET using three different techniques: manual contouring by an experienced radiation oncol-ogist using a pre-determined protocol; a constant threshold of SUV . 2.5; and, an automatic segmentation technique. Volumeswere defined on the 3D scan and each one of the five phases of the 4D scan. For each lesion the 3D volume from the 3D scanand the ITV constructed from the five phases of the 4D scan were recorded and compared.

Results: A total of 18 distinct lesions, including both primary tumors (10) and mediastinal lymph nodes (8), were identified on thescans and were analyzed. The definition of the tumor volume was not possible with the constant thresholding technique in onepatient due to increased uptake in the spinal column and mediastinum. The 3D volume was on average 8.5cc with the manual def-inition, 18.6cc with the constant SUV threshold and 15.5cc with the automatic segmentation. The corresponding ITVs were onaverage 9.1cc, 25.3cc and 19.2cc. The difference was statistically significant for all three techniques; p value\0.001 for the con-stant threshold and automatic segmentation methods and p value = 0.025 for the manual method. For one case an approximately 1ccvolume of moderate uptake in the hilar region was picked up by all three techniques on the 4D scan (SUVmax=3.3), but not on the3D scan (SUVmax=2.3).

Conclusions: We compared tumor volume definition on 4D and 3D FDG-PET/CT clinical data, using three common techniques.Despite differences in the tumor volume estimates between the techniques, they consistently showed an increase in the tumor vol-ume definition from the 4D scans. Our results suggest that 4D PET/CT may better define the full extent of moving tumors, com-pared to 3D PET/CT and potentially reveal additional information regarding regional disease.

Author Disclosure: M. Aristophanous, None; R. Berbeco, None; J. Killoran, None; J. Yap, None; D. Sher, None; A. Allen, None;E. Larson, None; A. Chen, None.