Study evaluating different methods of tumor volume estimation on PET-CT in carcinoma lung.

Abstract

e17528 Background: PET-CT has got a very vital role in present day oncology for diagnosis, staging, radiation planning, and response evaluation of tumors. We have even started talking about PET Response Criteria in Solid Tumors (PERCIST). Unfortunately, target delineation with PET currently remains a largely unaddressed problem. PET is a functional modality and the FDG concentrations within the tumor are variable. PET may either underestimate or overestimate the actual target volume. Various thresholding techniques have been developed where we take a certain percentage of the maximum SUV for edge definition. Many studies take from 15 to 40 % of Maximum SUV. The purpose of this study is to determine the performance of different edge detection methods for tumor volume estimation. Methods: Ten cases of newly diagnosed untreated Non Small cell lung cancer that underwent diagnostic PET, which had a measurable tumor on the images, were enrolled. The PET Images were transferred to the radiotherapy treatment planning system and PET and CT images were co-registered by pixel data matching. Gross Tumor was contoured on CT scan slice by slice manually by an oncologist (GTV-CT). Tumor volume was determined by summation of manually drawn regions of interest. Tumor volume was determined on FDG-PET with a semi automated method. The PET-GTV was delineated automatically by use of various percentages of the SUVmax. GTV-50, GTV-40, GTV-30, GTV-20 and GTV-15, based on the volume covered by respective percentage of maximum SUV were estimated. These volumes were compared with CT-GTV. Results: The mean maximum SUV was 7±3. The average volume of the GTV-CT was 125 cc. A significant impact of PET-derived contours on tumor volume estimation has been shown in 7 of the 10 patients with respect to the CT-GTV. PET defined GTV- 50 and GTV-40, were less than GTV-CT. Whereas GTV-30, GTV-20 and GTV-15 overestimated the GTV volume in majority of the cases. Even though GTV 40 underestimated the GTV, it was the closest match. Conclusions: Whereas the different thresholding methods either overestimated or underestimated the target volume, GTV 40 is the closest method for tumor delineation in NSCLC in our study. Further large studies needed. No significant financial relationships to disclose.