Reproducibility Study of Tumor Biomarkers Extracted from Positron Emission To-mography Images with 18F-Fluorodeoxyglucose

Introduction and aim Cancer is one of the main causes of death worldwide. Tumor diagnosis, staging, surveillance, prognosis and access to the response to therapy are critical when it comes to plan and analyze the optimal treatment strategies of cancer diseases. 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) imaging has provided some reliable prognostic factors in several cancer types, by extracting quantitative measures from the images obtained in clinics. The recent addition of digital equipment to the clinical armamentarium of PET leads to some concerns regarding inter-device data variability. Consequently, the reproducibility assess-ment of the tumor features, usually used in clinics and research, extracted from images acquired in an analog and new digital PET equipment is of paramount importance for use of multi-scanner studies in longitudinal patient’s studies. The aim of this study was to evaluate the inter-equipment reliability of a set of 25 lesional features commonly used in clinics and research. Material and methods In order to access the features agreement, a dual imaging protocol was designed. Whole-body 18F-FDG PET images from 53 oncological patients were acquired, after a single 18F-FDG injection, with two devices alternatively: Philips Vereos Digital PET/CT (VE-REOS with three different reconstruction protocols- digital) and Philips GEMINI TF-16 (GEM-INI with single standard reconstruction protocol- analog). A nuclear medicine physician identi-fied 283 18F-FDG avid lesions. Then, all lesions (both equipment) were automatically segmented based on a Bayesian classifier optimized to this study. In the total, 25 features (first order statistics and geometric features) were computed and compared. The intraclass correlation coefficient (ICC) was used as measure of agreement. Results A high agreement (ICC > 0.75) was obtained for most of the lesion features pulled out from both devices imaging data, for all (GEMINI vs VEREOS) reconstructions. The lesion fea-tures most frequently used, maximum standardized uptake value, metabolic tumor volume, and total lesion glycolysis reached maximum ICC of 0.90, 0.98 and 0.97, respectively. Conclusions Under controlled acquisition and reconstruction parameters, most of the features studied can be used for research and clinical work, whenever multiple scanner (e.g. VEREOS and GEMINI) studies, mainly during longitudinal patient evaluation, are used